Basic Course in Biomedical Research
Basic Course in Biomedical Research
being offered by National Institute of Epidemiology and Indian Council of Medical Research.
Today, I am going to give you Introduction to Health Research. Primarily, we have to understand that
there are many dimensions to health research. Health research is conducted for theoretical purposes and
when we have some good evidence created out of that we go for applied research. There are also
prevention and therapeutic angles to doing health research as well, in the sense some of the preventive
technologies as they emerge, they have to be tested out through research methods as well as newer
therapeutic options become available they need to be tested out through research methods.
Essentially, we have many a times heard of and we have seen in the movies the bench best research,
where we have seen scientist in white coats working in the laboratories trying to do something with
animals like mice and monkeys and so on and so forth, but what they do there has no significance until
the findings or the learning’s that they have got in that particular research are carried further and they
become available to the bed side, that is they are, they can be applied to the human beings either for
alleviating diseases or for preventing diseases. And this is a process which one has to understand and one
has to really follow correctly.
Sometimes the research is of exploratory in nature, where in the sense we do not know much about it
right in the beginning, but what we try to do is employ various methods and try to figure out if we can get
some clues for further research. Sometimes, it can be confirmatory, in the sense some of the clues that
have been obtained by people can they be say sort of strengthened or can we get additional information
which could be of practical significance? That is called as Confirmatory Research.
There are other terminologies which are also employed; one which has been recently employed in a big
way is Implementation Research and Translational Research. Many of the government programs get
implemented and what is important to understand at this particular point of time is how these programs
are functioning well and it is also therefore important to make mid course corrections, if they are
necessary or to decide in which areas you need specific angles of the programs to be strengthened and so
on. Transnational research basically talks about the earlier concept of bench to bed site that I talked about.
This is a process of development of technologies for human benefit and human welfare.
(Refer Slide Time: 02:47)
There are some important principles that we have to understand when we conduct health research, but
one thing which is really critical is that the planning stage in research is absolutely critical, why? Because,
if we do not spend enough time and if we do not involve the right kind of people in planning the research
study, it is very likely that some of the issues that we could have earlier prevented or they get missed
which we could have prevented in the real sense. Hence, in my opinion team work is really critical in
research, very rarely solo research succeeds. But for any kind of research to happen it has to undergo
several layers or levels of reviews.
Scientific reviews are important because they look at, what is the novelty in the concept that is being
looked at? Is there rationality behind doing that? And what is the justification for that? And in that
particular context the rational for doing that particular study in a specified country is also critically
important. Whereas, the ethics review primarily focuses towards finding out, whether the human subjects
protection issues have been adequately taken care of or not. What is important is research definitely
means development, research definitely means advancements, but it cannot be at the cost of human
subjects. People who are participating in the research are described as human subjects or human
participants and we should do nothing that can really harm them in the long run or in the short run. So,
ethics review ensures that this does not happen.
There are certain ‘in country’ procedures or ‘in country’ reviews called as Regulatory Reviews and they
are basically there to decide about, what kind of foreign funding is being received for that particular
project? Are there any sample shipments that are going to happen? Because there is a lot of intellectual
property also which is attached to this data sharing which happens, we always should be protective of our
own intellectual properties and so the regulatory authorities in our country do take care of this and they
ensure that our intellectual property is properly protected. Some projects do also involve exchange of
visitors and several countries have their own rules and regulations regarding the visitors, who are coming
and going. These are mandatory aspects and it is important that the regulatory committees do review
that aspect as well.
One thing I would like to stress, the health research or for that matter any research is a process and it
contains multiple components and each and every component in this process is of critical importance. It
all starts with collecting data. We have to collect data with a specific purpose and for that data to be of
high quality, the data collecting instruments also have to be appropriately designed. If the data quality is
good, then we can draw meaningful conclusions based on that and then we can make appropriate
decisions.
What is important is once these decisions are made, the policy planners and the program managers of the
country , they decide, whether this is the right time to take these particular learning’s from research to
appropriate actions which can be employed at an individual level or at a mass level. This is what we call
as ‘Evidence based action’. Primarily, all this is done so that we ensure that there is a reduction in suffering
and ultimately improvement in health and well being of the people or the community.
But when we talk about the kind of information that we collect in research, it can have wide range of
breadth and depth as well. There is some information which is required to be collected with respect to
the human host for example, and they can be some times healthy, they can be sometimes susceptible to
a particular condition or they can be suffering from a particular disease, some of them may have died
because of the that particular disease also. We have to figure out, how we can get the required
information from these kinds of subjects from these various kinds of host that I just talked about.
The disease cannot occur in an individual unless and until there are many factors which come together
and that is why we in the modern times always say that there is a multi factorial origin for occurrence of
any particular disease and we all know, the environment which is surrounding us and the society in which
we live, plays a very significant role in occurrence of diseases. There could be factors like climatic factors,
there could be housing factors, housing related factors, the vectors around us, the animals around us,
various socio-cultural practices that we follow, the family structure with which we live, all these factors
can affect not only the health of individuals which could be physical or mental as well but they can lead
to an occurrence of a disease as well.
In addition to that one more dimension that comes into the picture and which is really important from
the context of health is the health care infrastructure. Sometimes, people find it very difficult to access
health. The access to health is a critical component because many of these studies have shown that if
people are not able to access health in time, the kind of complications that arise and the death rates that
result from that are unusually high in nature. So, there could be multiple angles which are require to be
collected, we call all those like study variables. We have to collect correct information on those study
variables when we carry out health research.
Primarily, when we talk about the scope of health research or the objectives of health research they could
be many. For example, most of the times when we think of research, we think of something like creating
new information but the objective could be creating totally new information but we sometimes also get
additional information on something which already exists. Say for example, are more of diphtheria and
pertussis cases reported among adults in the recent times? We used to know that these were the diseases
of childhood earlier but now, it has been seen that some rare instances the cases in adults are also getting
reported, so we need to figure out why this is happening? And where is it happening?
Similarly, in terms of getting new information somebody might want to do research to find out the full
genome structure of hepatitis B virus and hepatitis E virus. It could have lot of implications in terms of
understanding what kind of pathogenic impact it would have in human beings, also it could have
significant decision making with respect to development of vaccine against those viruses.
Another objective that we also can pursue through health research is to verify and confirm available
information. Here is where, most of the research that happens in our country is now-a-days happening.
For example, are etiologies of pediatric pneumonia, you all know, pneumonia is a very serious disease in
childhood particularly in children below 5 years of age. So, are the etiologies different in the developed
countries and a resource limited countries like India. If one wants to figure this out, this could fit under
this particular objective. Also, have the incidence and complications of diabetes changed with increased
consumption of pre-cooked or packaged food. We have seen this to change, the eating habits of people
have changed but has it got any relation with the incidence and complications we might want to study
through health research.
Many a times health research is also focused on finding out the cause and effect relationship and this
could be applied if you think about it in multiple situations. For example, there is a presence of a specific
receptor on a type of white blood cells which are called CD4 cells in the human body and they are believed
to protect against HIV infection. But whether they project against a particular type of HIV infection, a
particular sero-type of HIV infection or they have a generalized effect, this could be evaluated under a
cause and effect relationship related research. You must also have heard that in the recent times breast
implant operations are undertaken, but do these breast implant operations or do these recipients, are
they more likely to develop breast cancers, this is something like a cause and effect relationship, where
we can think of breast implant is a cause and breast cancer is a effect.
(Refer Slide Time: 12:00)
One more important area in health research is that testing of new interventions like say, new drugs, new
vaccines new tools or newer prevention treatment technologies, etcetera. For example, it has been said
that tuberculosis is one of the most common dreaded complications or diseases to happen in HIV infected
individuals and mostly this is the disease which eventually kills them. So, can we introduce INH
prophylaxis? INH is a potent drug which is used against tuberculosis, so can it be introduced among HIV
infected individuals, so that they can be protected from getting tuberculosis. This is an important
intervention which could be tried out and to reduce mortality associated with HIV because of prevention
of tuberculosis in them.
Or for example, we are nowadays talking a lot about indoor air pollution, particularly in the rural areas.
So, if we decide to introduce smokeless stoves in those areas, would that lead to reduction in mortality
and morbidity in the rural areas. This could again be a part of an intervention evaluation. Sometimes, the
government has various on going programs, Public Health Programs which are vertical in nature, which
are operated by the federal government or the central government in our country. All of you know that
the general health services are provided in our country by the state level health services but the national
level control programs are operated and funded by the central government.
So, for example, we have seen that there has been large number of women noted with anemia and this is
one reason, why many complications occur in pregnant women. There has been this oral iron
supplementation has been the main stay for treatment of this particular condition for many years, but
nowadays iron sucrose injectable iron has become avail available and it is believed that it probably can
improve or tackle anemia during pregnancy much better then oral iron. So, evaluating that as been given
in the program could be an important thing that we can think about under this part of research. Also, the
government has introduced a program for several years now called as Integrated Disease Surveillance
Project or Integrated Disease Surveillance Programs now. So, can it predict the epidemics of newer and
emerging diseases likes H1N1 influenza that has been affecting us recently.
So, these are some important objectives of research that we have to keep in mind but one important thing
is, if we want to do sound research, it is really critical that we make the absolutely a right choice about
the study design over here because some kind of a miss judgment here can lead to a futile kind of a
research or which does not help you to understand it. This can be broadly say described in terms of some
enquiries or some kind of health research can be qualitative in nature, which requires person to person
interviews or discussion in focus groups or just say free listing or observations and things like that. It mostly
is observational in nature and it is open ended and people do probe at times to find out this information.
Whereas quantitative studies are mostly based on structured questionnaires, that is where people deal
with questionnaires, which are previously thought of and only the questions with very specified options
of answers that can be made available are used or employed in the study.
One more important differentiation that is made is the observational study and experimental studies. In
epidemiological terms, observational studies are those wherein the investigators do not change the
environment in which the study participants are living and the main thing that distinguishes the
experimental studies is, this is where the participants are exposed to some kind of an intervention at the
will of the investigators. So, some kind of a change is made and that is how this kind of studies becomes
different kinds of studies. They also are described as retrospective studies or prospective studies.
Retrospective studies are classically where the information on the outcome that we are trying to study in
a particular health research is already available. But in a prospective study we only start the study with
people who are at risk or who are susceptible to a particular disease , also have a comparator arm or
control group, comparison group and then follow these two groups to find out how many people in either
of these two groups develop a particular disease, which they were free off at the beginning of the study.
So outcome happens sometimes in the future and it is called as a Prospective study design.
As I mentioned earlier, planning is a very critical stage in research and some important considerations in
this particular phase, which have to be kept in mind include, there should be adequate justification for
conducting that particular research. Always please remember research involves investment. Investment
in terms of money, investment in terms of manpower and hence, it is absolutely important that there
should be adequate rational or justification for conducting any research study. For this to happen, the
original research question should have total clarity and focus only then the study becomes a good study.
Another important point to be considered is that, the case definitions used for various study variables and
outcomes should be standard and unambiguous. For example, if you are looking at cancer cervix then
what we call as carcinoma in-situ and what we call as invasive cancer has to be known to everybody, who
is involved in that particular research. For example, if you are collecting information on oral contraceptive
pills use, what is considered to be as those women who are using oral contraceptives pills versus those,
who are not, should be adequately clarified.
Another critical aspect that has to be kept in mind is the sample and a sample size. It has two dimensions
to that, sample has to be qualitatively representative of the population in which the study is being
conducted because here this leads to what we call as external validity or generalizability. In simple terms,
it means that if we are able to just generalize the findings of our study which are based on a particular
sample adequately and comfortably to the whole population from which the sample was drawn then we
have served our purpose.
But this also has to be complimented with adequate sample size as well because this is, we should have
enough power to draw meaningful inferences and hence sample size should be appropriately decided for
any kind of a research study.
(Refer Slide Time: 19:16)
We always have to remember research can never be free from errors, but if we predict them well then
they can be minimised through appropriate study design, this is an important aspect. Sometimes errors
happen because no two human beings are equal, one person has a different haemoglobin level than the
other, one person has a different height and a different weight than the other and hence, we look at many
such people together there would be some kind of a variation which would naturally be happening in
them and this is what we call as a random error, which anyway is going to be there in any kind of a research
study.
How we minimize it is by taking a large sample size. If we take a large sample size this intra individual
variation within a particular sample gets minimised to a large extent. But sometimes, there could be
another type of error that could occur and the results could be due to a something called as bias. And this
is a variation due to some kind of distortion by some kind of a faulty procedure, it might be a measurement
error, it is may be due to the kind of way the information is collected, it might be due to the process in
which the participants are enrolled in the research study. So, one has to remember this kind of an error
can only be minimised by improving the study design and taking care of that very adequately.
(Refer Slide Time: 20:41)
There will always be challenges in designing and implementing any kind of a research study. But the
important thing that we have to understand is how we predict those, and one way to tackle this very
appropriately is to do a thorough literature research before we do any kind of a study because it gives us
an idea about two important say entities which can affect the result, they are called as Confounders and
Effect modifiers.
Well, confounders are the entities which affect both the study variable as well as the study outcome and
invariably, if you just leave out all those people having this confounding characteristics, it is quite possible
that you may land up with a situation where you do not have enough people to do research at all and in
this particular scenario therefore, it is important that we understand which are the confounders. Collect
the right kind of information on all the confounding factors because statistical analysis can take care of
this confounding, which happens in a research setting. Effect modifiers are little difficult to deal with
because they can alter or distort the true relationship between the study variable that we are looking at
and outcome by independently effecting the outcome itself. So theoretically, it is a good idea to be aware
of them and to also understand the kind of effect they are likely to make on the outcome variable, also
one strategy that could be tried is to not to include people with effect modifiers in the study.
(Refer Slide Time: 22:16)
When we do any kind of a study, it can have multiple stages. It might be a good idea to do a pilot study.
Pilot study just helps us to carry out all, what we want to do in the study in a say sort a few individuals this
kind of becomes a rehearsal. It helps us to understand what kind of difficulties we are likely to face in
collecting the information, carrying out certain procedures, doing the interviews, doing the inform
consents, peoples understanding about informed consent forms and so on which can be changed.
There have to be certain issues related to study participants that need to be tackled very well, which are
the inclusion criteria? And which are the exclusion criteria? One has to be very specific about all of them.
Why? Because, if we are very specific about inclusion and exclusion criteria, enrolment of people who are
not eligible to participate in the study can be substantially minimized. Also, every research study has to
move with very specific research targets and it should follow certain strategies, like in some studies we
might want to do the recruitments by going out in the community, whereas in some we might just want
to do it based in the healthcare facilities, this decision has to be taken up front. Data collection instruments
are really critical because if there are any mistakes the way they have been designed we cannot make any
kind of a change at a later stage. So, lot of work has to be done before hand in drawing or in deciding the
right kind of data collection tools.
Measurements, here I am referring to all those measurable items which are either using the laboratory
methods or assays. They have to be properly standardized. They should have proper internal and external
controls, here positive and negative controls for quality control. This lab also should probably be a part of
an external quality assurance program because all this ensures quality control and assurance at all levels.
It is also important that the plan for statistical analysis for any kind of a study has to be decided right up
front, right in the beginning because that gives you a very clear idea about how we are going to collect
this particular information and how the results of this particular study are going to look like eventually.
So when we are doing health research, we have to have some kind of an orientation towards either
promoting behaviour or promoting health then preventing disease and also preventing mortality and then
maybe. So, how can we do that? How can health of the population be improved? This is something like a
direction that we have to think about. How can we predict occurrence of a disease in an individual? How
can various diseases be prevented? How can we effectively cure the diseases and reduce the cost and also
the associated morbidity and mortality? What are the various societal community based and
programmatic interventions for disease prevention and control? These are the kinds of directions which
health research takes and follows.
(Refer Slide Time: 25:24)
But basically, what we try to do is to find answers or practical solutions at individual and community levels.
For example, at community levels we have to see how healthy behaviour can be promoted? What can be
the personal prevention achievable at an individual level? How can early diagnosis be achieved at an
individual level and adequate and appropriate treatment be instituted? What kind of rehabilitation be
done at an individual level? But when we think of the community level we have to think of improving the
community behaviour and practices, prevention and control programs as supporting the effected people
or stigma reduction. All in all, healthy individuals only can build healthy nations.
Lecture - 02
In the introductory talk, Dr. Sanjay Mehendale talked about, the scope and focus of health research. He
mentioned that the goal of research is to establish facts or principles through careful and systematic
investigation in a particular area. Ultimately, the result of such investigation is to improve the health of
the population. The first step in such research is formulating research question.
Today, we are going to cover 3 areas - spelling out research question, stating research hypothesis and
formulating study objectives.
(Refer Slide Time: 01:02)
Any research has a life cycle. It begins with an uncertainty or needs in a particular area and that needs is
translated into research question. Subsequently as study objectives and a plan of analysis is formulated
to guide framing data collection instruments. Using these tools, data is collected and then subsequently
is analyzed as per the plan and as per the objectives conclusions are drawn and recommendations are
formulated. Ultimately, this particular recommendation is shared with the stakeholders for whom it
matters. This process ends with another uncertainty that may begin the cycle once again. Therefore, the
first two steps clearly indicate that we need to start with a good research question. Therefore, we are
going to see what is research question?
(Refer Slide Time: 02:02)
Research question is an uncertainty about something in the population that the investigator wants to
resolve by making measurements in the study population. The uncertainty is referred to as data needs in
the life cycle. Therefore, we need a clear question to facilitate choosing an optimal study design and
identify whom to be included, what are the outcomes that we need to measure and when these outcomes
to be measured.(Refer Slide Time: 02:34)
A Research question is all about refining your ideas into systematic process of framing a question. It begins
with the general uncertainty about a health issue in the context of health research and then it is narrowed
down into a concrete researchable issue.
While translating uncertainty to research question, one frames the problem in specific terms. In health
research, it could be in clinical or public health terms. We need to focus on only one issue at a time and it
is written in everyday language. So, that everybody understands what the question is. You may choose to
use more than one operational work, if needed. It should link the question, if answered what action will
be taken and it is stated as a question that is why it is called research question.
Refer Slide Time: 03:28)
A research question sets out clearly what the investigator wants to know and definitely not what he or
she may do or what this results will ultimately contribute to. It does not just include these two aspects; it
should spell out what the investigator wants to know.
The third aspect could come from careful observation in your clinic or in your basic science work, on your
sphere of life. Teachers get enormous opportunities, while they are preparing for teaching and teaching
and interacting with students and then finally, tenacity to go to the bottom of the things. Last, but very
important, if you have a good guide or a mentor, he can help you in identifying and framing research
questions.
There are two categories of research questions that we need to remember. There are only two categories
let me put it that way. One is a descriptive question another is analytical question. So, your research
question should fall into in either of these categories. A descriptive question is one which involves
observations to measure a quantity. When I say quantity, it could be height, it could be knowledge, it
could be to what extent the problem is present in a community, it could be to what extent something is
present in a given context. There are no comparison groups, no interventions in a question that is
supposed to be called descriptive question.
I am just going to give you some 6 steps as to how one can conceive a research question. One, a review
of state-of-art information from the literature, second, you raise a question and the third you decide
whether it is worth investigating through a peer review. Forth, by defining a measurable exposures and
outcomes and fifth sharpening the initial question based on the above and defining the question by
specifying details based on all of the steps that were explained earlier.
(Refer Slide Time: 06:54)
Suppose, let us take an example, should diabetics do exercise daily? That is a very common sensitive
question; one reviews a literature as to what is the effect of exercise on human body to begin with.
Literature shows exercise reduces blood sugar level and body fat, exercise improves protection against
developing complications due to diabetes. So, definitely it is worth investigating.
(Refer Slide Time: 07:24)
Then, let us raise a question based on our review of literature. Can exercise help control blood sugar level?
Sounds better than the earlier question, but definitely this is vague. We need to refine it especially, what
do you mean by exercise? What do you mean by blood sugar level?
The fourth step is to now, tighten up the framing of exposure and outcome. For example, here exercise
could be defined as a predetermined physical activity comprising of a body movement produced by a
muscles resulting in increased energy expenditure. At least one session of 60 minutes every day for one
year, you can be specific about this exercise or physical activity by specifying whether it is walking, jogging,
cycling or aerobic or even dance. Finally, outcome, now we have specified, it is fasting blood sugar level
which means after the dinner anybody goes into sleep and the next day morning the stomach is empty
after 8 hours. At that time, if you measure blood sugar that is called fasting blood sugar and we are
defining that as our outcome.
(Refer Slide Time: 09:37)
The fifth step therefore, is to sharpen the initial question with these steps in which we have progressed
further. Among diabetics, does physical activity for one hour daily help in reducing fasting blood sugar.
You can still refine it by specifying further details.
After a framing a good research question, we need to test this question to a test called ‘so what?’ This
test comprises of five elements, it is called FINER as an acronym. Is this research question feasible to
answer? Is this interesting to answer? Is it novel? Is it ethical to do studies around this research question?
Is it relevant?
(Refer Slide Time: 11:12)
So, basically feasibility means whether we will get adequate number of study participants. Do we have
technical expertise to do this study? Do you have resources both material and manpower to do this? Is it
interesting? Does it really enthuse people to engage in this particular research? Is it worth doing it? In
terms of novelty, does it confirms, refutes or extends the previous findings? Or does it provide new
information? That is the question that we are interested in answering and the fourth test is ethical angle.
Does this research based on the research question is allowable under the ethical norms? Will an ethics
committee pass this research based on the question? Finally, is it relevant in terms of advancing science,
advancing practice and also influencing policy?
(Refer Slide Time: 12:10)
Now, let us look at a hypothesis, what is research hypothesis? It is nothing, but a specific version of the
research question that summarizes the main elements of the study that establishes the basis for statistical
test of significance. So, it is stated for statistical purposes. The main elements that I mentioned include
the sample, the exposures and outcomes. A hypothesis is stated only for analytical questions with
comparison groups. Remember, we talked about 2 types of research questions, descriptive questions
analytical questions. So, only the second type of research questions involving analytical aspect needs
statement of hypothesis.
If you have any doubts about, what is your analytical question? You check in your research questions, if
that contains terms such as greater or less than, causes, leads to, compared with, more likely than,
associated with, related to, similar to or correlated with. If these terms are contained in your research
question, this is an analytical question that needs a statement of hypothesis. Purely descriptive questions
do not require a statement of hypothesis.
(Refer Slide Time: 13:37)
An example of research hypothesis on the basis of example that we are discussing on diabetes and
exercise, among diabetics, type 2 diabetes. From an area, who do brisk walking for at least one hour daily
results in an average reduction of 10 milligram percentage of fasting blood sugar level as compared to
those who do not. This specification of what much level? What study area? What group of study
participants helps investigators in tightening the rope in terms of statement of a null hypothesis,
statement of alternative hypothesis about which we will see in the sample size section and calculating
sample size? So, research hypothesis helps in statement of specifying certain details in the context of
statistical test and sample size.
(Refer Slide Time: 14:40)
What is a good hypothesis? A good hypothesis should be simple, should be specific and should be stated
in advance. What do I mean by simple? It should be one exposure, one outcome. What do I mean by
specific? There should be no ambiguity about the study variables or study participants. It should be stated
in advance, ‘a priori’ that is a terminology used. It should be stated in advance, it should not be discovered
at a later part of the study and the hypothesis is focused around the primary objective.
Now, let us come back to the life cycle of research. We have now talked about research question and how
research question is translated in to research hypothesis for analytical questions. Now, let us see how
research question is converted into statement of objectives.
(Refer Slide Time: 15:32)
When you translate research question in to objective as compared to research question, a statement of
objective is stated in scientific and epidemiological terms. It takes the research question in only few
limited axis. It is written in scientific and epidemiological language. It should ideally make use of no more
than an operational web for each of those questions. It is ideal to sort them out as a primary objective
and secondary objective and the statement of objective should be very clear about what research
question it is trying to answer. Is it descriptive question or an analytical or experimental question? So, it
should clearly spell out, whether we are answering descriptive or analytical question.
(Refer Slide Time: 16:24)
For example, to estimate prevalence of physical activity in diabetics, that is a descriptive objective. For
analytical, you use the terms that are more powerful denotes this comparison, connotation. Testing
hypothesis has to be denoted in the verb for example, the verb, determine may be preferred for example,
in the diabetes and exercise related studies that we are discussing. Determine whether exercise reduces
blood sugar level? As you can see, the verb is also equally important and therefore, we do not recommend
use of the word study, to study in the statement of objectives at all. To study is a very poor statement of
objective.
(Refer Slide Time: 17:56)
Now, let us go back to our example, the research question that we stated was, in order to improve
management of type 2 diabetes, we wish to know whether brisk walking by diabetics for at least one hour
daily reduces fasting blood sugar level as compared to those who do not? We can translate this into a
primary objective to determine the effect of brisk walking for at least one hour daily on fasting blood sugar
level of patients with type 2 diabetes compared to those who do not. So, determine is a very useful word
especially when you have an analytical objective.
(Refer Slide Time: 18:33)
Let me show you some of the good and bad examples of study objectives. Determine importance of
sedentary lifestyle among diabetics. The word determine is not particularly suited for this particular
statement because it seems to be descriptive studies. Therefore, estimate prevalence of physical activity
among diabetics is ideally suited for this statement of objective.
Assess physical activity and diabetic complications. I think they are trying to do again a descriptive study.
Therefore, estimating the effect of physical activity on the rate of diabetic complications in a group of
diabetes is what they are doing. So, the word estimate is preferable. Evaluate depression and diabetes, it
may be preferable since it is analytical study to use a word verb determine than evaluate, but the
depression is more common among diabetics as compared to healthy individuals.
(Refer Slide Time: 19:31)
Let me sum up, asking yourself the right question may end up in answering right way. Two ways to deal
with a poor or irrelevant research question, first you may answer it, but then the answer may be of no
use or there may be no answer. Try to reframe it, if you have a poor question. If your research question is
wrong, no good hard work will save your work. If your question is correct you have an opportunity to do
a good job, all the best wishes to do a good job.
Thank you.
Dr. P. Ganeshkumar
Lecture - 03
Literature Review
Hi, welcome to Basic Course in Biomedical Research, I am Dr. Ganeshkumar from ICMR School of Public
Health, National Institute of Epidemiology. This ascent is about literature review. Literature review is an
important step in any health research because this is an important link between what is known and what
is not known. Basically, research is a systematic investigative process to increase our existing knowledge
about your concern subject of interest or to revise our existing knowledge. So, it may be basic or applied,
where basic is increasing an existing knowledge, where applied is that applying this basic research to
develop new processes, new products, new knowledge or throw your light over an unknown area.
In this context, literature review is an important step in any health research, where it is going to tell us,
where it helps and aids us to tell us to guide us, what is already existing knowledge about your concern
subject. So, here in this session the most important learning objectives are why we require a literature
review? What is a literature review? How a literature review is performed? And what are the steps of the
literature review and certain ethical concerns in a literature review?
Why we require a literature review? As we already explained in the definition of a research that, it gives
important link between what is known and what is not known. So, for to know that we require this
important step called literature review, where it saves lot of time in your research and secondly, it know
the subject matter better. So, this throws what is our existing knowledge about the concerned subject
and it suggests new research topics and questions.
For example, how it saves yourself from work? When you see; when you want to develop a questionnaire
for a physical activity. So, you review a literature, you can come across an already existing standardized,
regionalized questionnaire for physical activity. So, you can use that which you no need to spend lot of
time to develop your new questionnaire to measure the physical activity of an individual. Of course, it
paves a way and throws the light of, to know the subject matter better and third by reading the existing
articles you may find certain lacunae in the existing knowledge, which makes you to carry out your new
research. So, it will aid you to carry out a new question, it aids you to develop new research questions, it
aids you to develop new methods of an existing known subject.
Is it a just a summary? So, we are going to read, whatever it is known and we are going to summarize it
and tell, when this is what it is there in the existing literature or existing evidence. Now, it is not just a
summary, it is first step, it is basically an information seeking. It scans the literature efficiently using
manual or a computerized methods to identify a set of potentially useful articles, say like how we usually
do in literature review is that, the whatever kind of a resources, valid resources over an existing subject,
we need to retrieve the information from that. So, it may be a text book, it may be a manuscript, it may
be a published article or it may be a conference proceeding. So, from there we need to go through that
and we need to retrieve the kind of information from that and second thing is that after we have collected
all those things, what we basically do is, not just summarize it.
We need to critically appraise, whatever the article we have collected so far. So, the critical appraisal is a
most important step in literature review, which is an ability to apply principles of analysis to identify which
is useful for you. So, these will be useful for me, this will not be useful for me, this is valid for this concerned
subject, this is not valid for my point of view. So, likewise you need to critically appraise and existing
collected literature.
This is not just a summary. So, what it is then? It is the organized research, how organized it is? So,
organized research is a systematic process, say like the first important thing is that, you need to organize
the information whatever you have collected so far and you have to relate it to the concerned research
question which you are trying to develop or which you have already identify it and second thing is that it
helps synthesize the results of it, from this existing whatever information which we have collected. So, in
that collected information you have to summarize it, what is and what it is not known. This is all
summarizing that is called synthesizing the results.
The third step is that you need to identify the lacunae, when you synthesize the results, when you organize
or when you synthesize the results, the third thing there you can see is that, which is the lacunae here as
appears in the literature; that means, for example, after when you are trying to review the existing
diagnostic tools of tuberculosis and after you have reviewed that there are many tools are available and
this is the different cast of the different diagnostic tools and these are the different work easy friendliness,
quicker the results and you can identify that there is no quicker tool, which is not available for specific
type of tuberculosis. Now, this gives you lacunae of existing literature. This may be helping you and this
will lead you to identify a new research question or identify a new research path in it. So, that is how it is
an organized research, it is not just summarizing this existing articles and writing a kind of a paragraph.
When you have seen that, you can see there is an information retrieval mechanism is inculcated in the
literature review. Information retrieval in literature review what it is? It is basically identifying from a large
database and from that by a set of queries, you are going to identify what are the documents which is a
relevant for your needs. So, that is called information retrieval, that it happens from a big data base or it
may be a huge database where all these articles are archived, collected or whichever is stored into that
and from that you user by means of your set of query is seeking an information from this database. Now,
this query system is user defined, it basically according to the need which a user is developing into that
over the particular research question.
(Refer Slide Time: 07:26)
What actually is this database? And what actually is this database management? What is this a set of
queries? So, you see here, all this information about a scientific knowledge in this information era is
currently stored mostly in an electronized form. Even certain non-electronized form is also stored in term
of books or in terms of printed journals or conference proceedings. All these things are collectively called
as a database. For example, it may be e-library, for example, it may be a collection of citations.
This data base will be huge, from this, like a last page of your text book, you can see index which helps
you to identify what you are trying to search. Likewise, there are certain indexing mechanisms for this
database from where you are going to collect required information by means of set of queries. Now, that
is called indexing. This set of query is what, user is defined to retrieve the required information from this
huge data bases.
(Refer Slide Time: 08:37)
Where we have to search? Which is appropriate place to search the required information? Where we
need to search? It may be a general information about your specific health related event or a disease. For
this, already we know that Google is a huge search engine, which has a huge database of the require
information in it, but from all the Google results, we cannot rely entirely on that. So, there is a kind of an
accreditation mechanism by a non-profit organization called Health on Net, this is called HON certified
websites. This HON certified websites are those accredited websites, where you can get certain means of
an authorized, certain means of reliable information about your specific health related events that is
about general information, but by a professional, by a researcher.
What is a specifically when you want to retrieve an information, for example, I need to retrieve an
information about all clinical trials about diagnostic test on tuberculosis. So, when I am going to do this
kind of a specific query, this specific query has to happen over the specific scientific databases, which are
PubMed, which is a huge database of scientific citations, biomedical citations. It may be a Scopus, which
also have huge data base or even Google Scholar is also another URL, which is another portal where a
huge collection of citations and abstracts across literature is available. So, from this specific query, you
can retrieve the required article or the required manual script which you are trying to search.
Third, it may be from an archived full text articles, the archived full text articles are usually available in
free open access, directories of journals or it may be from your fee based libraries. And this libraries may
be e-library or may be a physical library, which is present where full text articles are there. From there,
we can able to get the required information, which we are intended to research over a concerned topic
and finally, you want to take the articles or you want to review those articles which is evidence based,
which was the highest level of evidence currently available is a systematic review and meta analysis. So,
from this collection of systematic reviews and meta analysis also you can retrieve that information, which
is available in Cochrane library and even certain database also have these collection like Map of Medicine
is also a collection of evidence based medicine articles in that portal.
These are the places, where you can do a query search specifically by an indexing mechanism and the
required information can be retrieved from these databases. The most popular among all these things is
PubMed. So, PubMed is a huge database, which has a 25 million citations is available and another
commonly used portal is Cochrane library, which is a collection of systematic reviews and meta analysis.
This popular PubMed, let see about it, what is this PubMed? So, PubMed, which comprises more than 25
million citations of biomedical literature. This PubMed is maintained by National Center for Biotechnology
Information situated in US National Library for Medicine under National Institute of Health and this
PubMed is free and open to access. Each and every PubMed, it is a collection of abstracts and full text
articles. Each abstract has a link out resource, where the full text article is available and there are certain
articles where you can freely access in PubMed, through a portal called PMC, which is called PubMed
Central. So, through this PubMed Central portal you can see the free full text articles about your
concerned topic of interest or your research question.
There is a good tutorial which is available in the landing page of the PubMed, where you can see in the
portal itself. Apart from that, this US National Library of Medicine is maintaining their own set of defined-
predefined vocabulary called MeSH, which is called Medical Subject Headings. Medical Subject Heading is
a National Library of Medicine controlled vocabulary thesaurus and which is, you can see as called
keywords under any abstract. If your keyword is in concordance with NLM defined vocabulary called
MeSH, then your article has higher chances of getting identified by a set of systematic query mechanism.
So, that is why MeSH terminologies are very important, which you can access the entire MeSH database
in the left side of your portal, a PubMed landing portal and when you see in the PubMed portal, there you
can have a tutorial which is available.
This entire tutorial explains you, that using PubMed, how to make a search in the PubMed? How to make
a search through MeSH terminology? How to search through a single citation manager? And how to search
through an author, year wise search, article wise search, study design wise search? So, it aids you this
entire tutorial, which has a quick tour video tutorial, is also available in the landing portal and when you
access it, it gives you a good detailed form of how to perform a search in PubMed portal.
(Refer Slide Time: 14:46)
Now, I will explain you, basic search in your data base. How to do a basic searching in a database and this
kind of search is called Boolean query, which is very common in any search engine, it may be web search
engine, it may be Google or it may be Embase, it may be a PubMed. This Boolean query is standardized
one, which uses this most important connections called and, or, not queries. In these Boolean queries, I
will just give demonstration about, how actually when you are putting a query and how the information
is retrieved. For example, here I am showing you two different database, one database is about lung
another database is about infections, which encompass all infections and if I am putting a query, a Boolean
query of lung and infections and how the query retrieve the result. It retrieves the result of lung and
infection, which is those articles, those items which are specific to lung and infections.
(Refer Slide Time: 15:52)
Whereas, when I am putting a query about lung or infections, what it gathers is, it gathers information
about the entire items on lung, entire items on infections and even the items which is common that is
lung and infections, which is lung infections. So, we get a huge search of results, when you are putting a
query called lung or infections.
(Refer Slide Time: 16:16)
When you put query called lung not infections, so in this lung not infections, it totally remove this entire
information related to infections from the whole search. Again, the infections related articles and items
from the lung. So, what you get is results are the items, which is related to all items, which is related to
lung except infections. So, this is how, this Boolean query works and, or, not.
(Refer Slide Time: 16:51)
I will give a basic demonstration, similar to that how I did a Boolean query in PubMed. So, here you see,
these are all the screen shots which I have cropped. When you see, when I made a Boolean query of only
lung, what are the total numbers of items, which I have retrieved? The total number of items I have
retrieved is nearly 7.2 lakhs. Only infections, the total number of items, which I have resulted is 17.7 lakhs.
When I put a Boolean query of lung and infections, it resulted me, only 82000. So, 82000 articles are
related to only lung and infections. So, it removes all other lung related, all other infection related and it
resulted specifically to lung and infections.
Let say, this gives another screen shot, when I make a search of Boolean query of lung or infections, here
you can see, where we get information of entire information on lung, all the items about infections and
all the items related to lung and infections. So, total is that nearly 24 lakhs. So, that is how depends upon
your Boolean query, say like lung not infections. In lung not infections, it removed all the infections related
items and resulted only say like nearly 6.3 lakhs, which is only article related to lung except infections. So,
this small demonstration which gives you an idea about how a Boolean query works; however, this is not
a structured or a systematic way how to do a PubMed search. This session is not going to cover entirely,
how to do MeSH search and all. So, the tutorials are already available in the PubMed dot com landing
page. This is how a Boolean query works in any search engine including PubMed.
(Refer Slide Time: 18:53)
Now, what are the steps, how we have to do a literature review? Already, we know that after we have
organized information is that, which are the studies is related to you? How we are selecting the studies?
The step one is that by means of your Boolean query or by a structured query system, you have collected
your specific articles related to do. For example, you are searching a Boolean query of same lung and
infections; I want only RCT's about lung and infections, which is published in the past 5 years. Now, I have
organized everything. So, what you need to do is that after you have organized it you need to read all
those articles and critically appraisal it and how to do this, will be explaining you at the second step.
The second step is that, you have to select which is of those good qualities. Remember, here this critical
appraisal should be scholarly, it should not be too critical. It should be scholarly, for example, an article is
explaining about the randomized controlled trial about a therapeutic regimen over lung infections and
which you have seen is that these RCT's which you have collected is addressing over only the western
population and you have this lacunae. So, what you are going to critical appraisal is that the existing
evidence is available over the western population and Asian population is not available and Indian
population is not available. Those articles whichever is reported about Indian populations are only quasi
experimental studies, here it is scholarly critically appraisal.
You are not too critical; you are not finding a fault on it. Now, when you do a scholarly critical appraisal
on it, you will identify your new area of research, you will identify what is existing lacunae, existing gap,
which you can try to fill up by doing your new research. Third step is that, you need to compare the
methods, results, findings of all those articles which you have organizing it. When you compare it, you can
understand, what is an existing knowledge, about the different methodologies they have conducted in
this particular topic?
So, that throws you a light, more about what is the methodology they have followed? So, far, in those
kinds of your search query, for example, here when you want to do therapeutic regiments on lung
infections, what are all the methodologies they have applied, while they want to test your therapeutic
regiment for a lung infection? Likewise final step is that, you need to tabulate in your form, each study
which has to be tabulated when you tabulate it, it helps you to organize. Number 2, it helps you to
compare. Number 3, it helps you to compare the individual studies by itself and itself with other studies
and it will give you good idea about and it will helps you to critically appraisal it. This is how, your model
table looks like this.
An example table of literature, where you can see, you can put up, add and delete your required columns.
I made an example table; you can see citation, design, objectives, study population, sample size. What are
the measurable outcomes? What are the main findings? And what are the authors’ conclusions? You can
organize it in kind of your table, which gives you, helps you to critically appraise it and it is better when
you organize this table chronologically, which is from recent to past. So, your recent studies can be tabled,
can be made in their first initial rows. It helps you to select those recent evidences or recent existing
knowledge about it.
So, after you have organized all these, since you have tabulated it. Now, you need to write this literature
review. In the literature review, in the write up part, these are the other three important parts. The first
part is introduction, when you are writing a literature review, in the introduction; it gives you what is the
purpose of the review? Why you are doing this? And how you have organized this review organization of
the review? How you made the queries? How you have collected this information? And how you have
organized it? That explains and basis of the ordering, most important to least and as well as what is the
earliest to recent or recent to the earliest? How you have organized it? It gives you an introduction, how
you have written? How you did this literature review?
Second is that, where you are actually writing all your existing information called empirical literature. So,
it includes the quality studies, which is relevant to your research question or the topic and it explains each
study about it is positive and in paraphrasing manner and your critical appraisal should be a scholarly way.
So, that explaining it to that, these are all the identified lacunae, these are all the strengths, these are the
novel methods. This you need to write in this part. Finally, you need to summarize your existing this entire
literature review and when you summarize it, it should be a concised way, where you are going to tell
that, this is about your existing knowledge about your research question or your concerned topic and this
is known and this is not known and this is lacunae of an existing literature. So, this is how you have to end
a literature review write-up and there are certain ethical issues, which we need to concern when you are
doing a literature review.
Mostly, this problem is that when you are retrieving information from your specific manuscript or an
article, the content from the study should be presented honestly. It should not be distorted, that you
should not be read in between lines and taking only a part of your result as such. That is a most important
thing. Second is that, any weakness of the study, again I am emphasizing it, that it should be very scholarly.
You should not be too critical and it has to be addressed in a research point of view, that is scholarly point
of view and finally, the source should be accurately documented.
There is a way how a source should be documented. There are different styles that are available, a very
popularly followed style is Vancouver style, even Harvard style is available and based on that it has to be
cited accordingly those sources. Finally, this is how a literature review can be done and this gives you a
basic view about, what is a literature review? Why this literature review has to be done? What are the
steps in doing this literature review? And how you can write a literature review?
Dr. R. Ramakrishnan
Lecture - 04
Welcome to this session of Basic Course in Biomedical Research . In this session, we are going to
see some measurements we commonly use to measure the Disease Frequency.
Now, let us look at some example. The population at risk of developing carcinoma of cervix is female
population and in the age group of 30 to 70 years. A Population at risk of hepatitis B are those individuals
who are at risk of developing hepatitis B, but were negative.
In order to make that as, suppose whenever your population at risk is very large and you have number of
cases were small, you will get a value P as 0.001. So, in order to make into a round number you multiplied
by 1000, 10000 or 100000 depending on what the value of P you get. So in some studies, the total
population used as in approximation if data on population at risk is not available because with the
philosophy that everyone in the population are at risk of developing a particular disease.
Again, this prevalence are broadly classified into say 2 different categories. We called as one as a Point
Prevalence and another one as a Period Prevalence. By point prevalence what we mean is, this measures
the frequency of disease at a given point of time, it is like a snapshot. This applies when the data has been
collected at one point in time. It is denoted as P is equal to C by N, where C is the number of observed
cases at that particular point of time t and N is the population size at time point t.
(Refer Slide Time: 03:21)
The example of point prevalence is supposed, if there are 150 children in a school and you are screening
them for refractory errors, at a particular point of time t. And you find 15 children they require glasses,
they have problem. Then the prevalence of refractory errors is 15 divided by 150 which is equal to 10 and
called a 10 percent of the school children they have refractory errors or the point prevalence of refractory
error in this particular school is 10 percent.
(Refer Slide Time: 03:56)
The Period Prevalence is that measures the frequency of disease over a period of time. This applies when
the data has been collected over a period of time and it is denoted as PP, which is equal to C plus I divided
by N. What is C? C is the number of prevalent cases at the beginning of the time period and I is the incident
cases, that is the new cases that develop during the period of your survey and sum of these two are
divided by N which is the size of the population for this same time period point.
(Refer Slide Time: 04:39)
Example of this period prevalence is you know, you take a scenario of a population of 150 persons and
you follow up for 1 year. At the beginning of your survey the 25 had disease of interest and another 15
new cases developed during the year. When we calculate it point prevalence at the start of the period
and period prevalence over the period of time. Point prevalence is given by C divided N there is 25 by 150
or 0.17 or 17 percent. Period prevalence is 25 plus 15 there is 40 over 150 it comes 0.27 that is 27 percent.
(Refer Slide Time: 05:21)
So, now there are several factors that influence the value of prevalence, the number of new cases, the
duration of illness. Suppose, if the duration of illness is short, the prevalence is reduced. Say suppose,
somebody gets disease and immediately you know either it cures or it dies then it goes out of our
calculation, with thus when we go there are no disease persons. So, the prevalent of sudden infant deaths
cut by times would be even 0 because when you go if there are no deaths that are there, all the children
who had the disease had died. So, there are no cases at that particular point of time, so your prevalence
may be 0.
And if disease is of very long duration, the prevalence you know it goes on suppose it is in chronic diseases,
a rare lifelong diseases it can accumulate to build up a very large prevalence.
(Refer Slide Time: 06:15)
And prevalence could also increase, if there are healthy people going out of a particular region. So that
your denominator is low and your numerator is all unhealthy or disease people are there and so your
prevalence may increase. The decrease in the prevalence could happen exactly you know the opposite
causes, the shorter duration, high cure rate, high case fatality, the decrease in new cases, immigration of
patients, improved cure rates and immigration of healthy people. All these could bring down the
prevalence. So, to conclude the changes in prevalence may have many causes and are difficult to interpret.
So better we need to have a checklist of all these items and then look at them all before we try to rather
say that the prevalence has increased or decreased over a period of time in a particular region.
(Refer Slide Time: 07:56)
What are the uses of prevalence data? Prevalence of data is used to assess health care needs. It is very
useful in planning health services because it measures the burden of disease, and it measures occurrence
of conditions with gradual onset and prevalence is very useful in the study of chronic diseases.
Let us rather look at the different aspects of the incidence. Let us rather take the cumulated incidence
first. The cumulative incidence is CI, is a number of new cases divided by population at risk at the beginning
that is multiplied by a factor of 10. This is known as attack rate and it assumes that the entire population
at risk at the beginning was followed-up for the time period of observation.
(Refer Slide Time: 09:39)
Now, let us look at what do you mean by Risk? The risk is the probability that an individual will experience
a health status or change over a specified follow-up period. This assumes that the individual does not have
disease at the beginning and die from other causes during follow-up. This corresponds to a cumulated
incidence.
(Refer Slide Time: 10:01)
Now, let us look at the other measure which is called incidence density, denoted as ID. The incidence
density is the number of new cases divided by total person-time of observation and that is multiplied by
a factor of 10. This is also known as incidence rate. This reflects more exactly the person-time observed.
The dynamic of incidence and prevalence can be depicted with a diagram like this see. There are new
cases pouring in, there are cases going out because they are dying or they are getting cured. The incidence
cases are the new cases and the cases which are remaining in the tap are the prevalence cases.
(Refer Slide Time: 11:13)
The relationship between the prevalence and incidence could be the prevalence depends on the incidence
and the duration of disease and it is denoted as prevalence is equal to incidence into duration, P = I
multiplied by D. Change in prevalence from one time period to another may be the result of changes in
the incidence rates, changes in the duration of disease or could be both.
(Refer Slide Time: 11:43)
Now, let us look at the patterns of incidence and prevalence. High prevalence and low incidence, there
are disease like Diabetes Mellitus. Low prevalence and high incidence are the examples are common cold.
So, let us summarize these basic disease frequencies that we are finishing, one is the prevalence and
prevalence is a static measure taken at a point in time or over a period of time. If you take it at a point in
time it is called Point Prevalence, over a period of time it is called the Period Prevalence. Incidence is a
dynamic measure taken over a certain time and the mortality is calculated using population denominators
to reflect the burden, while the case fatality is calculated using cases as denominators to reflect severity.
These are all the measures of a disease frequency.
Lecture - 05
Descriptive Study designs
Welcome to this session of Basic Course in Biomedical Research . Today, we will be discussing about
Descriptive Study Designs.
Before we go on to the descriptive study designs, let me walk you through an outline of what are the various
designs in the health research. So, as you can see in this slide broadly, the health research studies or the
epidemiological studies can be divided into 2 categories; Descriptive studies and Analytical studies.
Further, the descriptive studies can be of 2 kinds; one is the studies which are done for individuals and other
is the studies which are done for populations. The studies which are done for populations are called
Ecological studies and there are various types of studies that can be done at the individual level that include,
Case reports, Case series and Cross-sectional surveys.
Now, in this lecture we will not be talking about the analytical study designs which are the other kind of
studies, which will be covered in the different lectures. Now, within analytical, the other lectures will walk
you through thus observational study designs, such as Case-control studies and Cohort studies.
(Refer Slide Time: 01:20)
Now, coming to the descriptive studies, in today’s lecture, we will be discussing about the 4 main types of
descriptive studies that are, Case reports, Case series, Ecological studies and Cross-sectional studies.
So, case reports are extremely useful when you want to teach others about or share your experience with
others about new diseases, unfamiliar diseases, rare manifestations, though you have seen the disease
before, but this particular type of clinical scenario for this disease you have not come across. These kinds
of case reports may be extremely useful either to generate a hypothesis about pathophysiological
mechanism or just your fellow clinicians to be aware that they may come across this type of clinical
manifestation during their clinical practice.
As you can see here, this is an example from a journal how the case reports look like. The case report is
usually published in most of the peer reviewed journals, in a separate section, wherein they would mention
what the clinical condition is and what the particular case report is covering. In most of the case reports,
will also talk about what is the existing evidence or what is a literature about this and whether this type of
manifestation has also been observed by other clinicians.
Now as a case reports cover one or two patients, case series is a step ahead of that, wherein you study a
relatively larger group of patients who have a particular disease for example, you may study 10 or more
patients with some rare type of cancer or with some rare type of tuberculosis or any other disease, how does
it help you? It helps you to understand whether these kind of findings that they are observing, are they really
due to the disease or is it due to the chance. It may help you understand or develop what would be the
clinical picture of patients presenting with a particular type of disease. The only problem with case series
is that you don’t have a comparison group.
(Refer Slide Time: 04:24)
So, just to give you an example. This is a very interesting case series that emerged on pneumocystis
pneumonia, which was observed in 5 gay men and this led to an understanding that, how is it that the
pneumocystis pneumonia, which is not a very common condition is actually occurring in same kind of
individuals, which subsequently led to the discovery of AIDS. Coming to the third type of study design.
(Refer Slide Time: 04:49)
Now, this study design is slightly different. In what way is it different? The difference here is, here
individual is not the unit of study or unit of analysis. This is a study that you do for groups, that you can do
at a country level, you can do for populations, you can do a particular region of the country and you try to
understand a particular problem for a group and you try to relate problem as well as, what could be the
possible reason, why that problem is occurring also at the group level. So in this type of study, group is the
unit of analysis.
The limitation is that you do not have individual-level data. You really do not know what the individual
level exposure is. However, you are able to look at a particular population let us say; we can talk about
what is an average intake of fat among people living in a state of Tamilnadu? Or, what is the average intake
of carbohydrate among people living in a particular state? And then, you may look at what is the incidence
of cardiovascular disease. So, both the measures you only know for the population. But you are not able to
have any data at the individual level. How does it help you? It helps you to generate hypothesis because
what you can see here is you can try to co-relate. Let us say, is it so that the states which have high per
capita consumption of fat also have higher rates of coronary artery disease or not, if you find that then you
could think of doing a further detailed study to actually understand whether there is a correlation between
the fat intake and the cardiovascular disease.
(Refer Slide Time: 06:38)
Now, I am going to come to the most important study design and I think this is a study design that many of
you might have used in your research studies and obviously, if you are going to be in research, you will end
up using this study design for sure. This is the commonest study design and I think you all are aware, you
all are aware of census, right? We all have every 10 year census is done for the whole country. Census is
nothing but a cross sectional survey, what do you mean by cross sectional survey? It is an observation of
cross section of a population at a single point of time.
Now, what happens during census? In the year 2011, somebody would have visited your house they would
have come and asked you, how many members are there in your house? How much each person has studied?
Where do they go for work? Where do you live? And, this information is for year 2011; however, it may
not be the same information in 2012. So, what cross sectional survey captures is, a particular set of
information that you want to collect from a population or an individual at a particular point of time.
Now, these kind of studies are extremely useful to know the magnitude, means how big the problem is, like
if suppose if you want to know how many people in the community have high blood pleasure? Or how
many people have tuberculosis? Or you may even not about diseases even general information, how many
people in this particular area are actually professionals? Any kind of information, any kind of health or non-
health information, when it is collected at a particular point of time, it is called cross sectional survey. Now,
for a cross sectional survey, you will be learning in other lectures about, what kind of sampling method you
can use? How you can select the participants?
But, broadly speaking, census is one example of a survey which covers the entire population; whole country
is covered in this kind of survey. However, all surveys are not done like that, most of the surveys they will
select a sample. It is not possible to survey one billion people for everything, so you might have come
across something called National Family Health Survey, wherein a sample is selected and then the data is
collected. Similarly, if you are a clinician, you may want to plan a study in your patients in the clinic.
However, it may not be possible for you to include all the patients in your study, and you can think of using
a sample. A sample allows you to understand, what is the magnitude of problem in your study population
by doing survey at one point of time.
Now, here we can collect data on various type of exposure or outcome, what does it really mean? The
exposure here means different types of risk factors for example, you may do a study, wherein you want to
know about high blood pressure; high blood pressure could be your outcome, but in addition to that you
also want to know whether people are overweight? Whether they are having any behavioral risk factor?
Whether they smoke? Whether they drink? So, these are called exposures. So, your survey may include
asking questions about their history of blood pressure or doing measurements or asking various questions
about what you think are the potential exposures.
Your study might have included people from all age groups, you may have under 5 children, you may have
adolescents, you may have older people and what you can find out from this survey is let us say, of all the
people I surveyed I may find the diarrhea is more common in children. So, how it helps the service provider?
Cross sectional survey may provide very useful information for planning your health services, at the bigger
level. What I mean is at the state level, at the country level or the global level. Cross sectional surveys help
us in setting priorities, for disease control.
Now, where should I invest my resources? Which is the most important disease? Should I invest it in a
study on hypertension? Should I invest it on a study on depression? To understand that you need to know,
what are the big health problems in your country or in your state or whatever be the unit of study? The
cross-sectional survey allows you to measure the magnitude of the health problem. Now, cross sectional
surveys are not ideal for testing hypothesis, you will be introduced to the analytical study designs, where
you can do that; however, they help you generate hypothesis.
You can think of what are the potential risk factors, if you find you did a study and you found that the
prevalence of hypertension is very high. Let us say, 40 percent people have hypertension and
simultaneously, you might have asked various questions regarding their eating habits, regarding smoking,
regarding alcohol use, regarding over weight and you could do a preliminary analysis and understand
whether which of these risk factors might be playing an important role. And this part of cross sectional
survey is called analytical cross sectional survey.
However, to test your hypothesis you need to use the other study designs. The other use of cross sectional
survey is, suppose you want to know after having introduced a particular type of intervention, it could be a
clinical intervention for example, you introduced a new drug. A new drug into your regimen of a particular
disease for example, for malaria, you are using a particular drug or for TB. You want to know before and
after, after introducing your intervention, whether the patient outcomes are better, you can do that using
cross sectional surveys. So, this cross sectional survey design is extremely useful design in trying to
measure the burden, understand the magnitude, understand the distribution, and sometime even to
understand the effectiveness of your interventions.
Now, these are the some of the examples of research questions that can be answered through surveys. What
is the prevalence of hypertension or prevalence of any disease in an XY city? The other type of study, in
many time many of you might be in the health services and you may want to do surveys to understand your
patient satisfaction or you may want to even just do a survey to know, what kind of utilization pattern is
there of your health facilities? So, you can do cross sectional survey of patients. Many times they are also
called exit surveys.
Now, thus cross sectional surveys can be done in any setting. It could be done in schools, it could be done
in health facility, it could be done at the community level, it could be done at the clinical level. So, another
example would be that you want to know how common is the physical inactivity? As you know, you may
be interested in understanding childhood obesity and you may want to know, what is the prevalence of
physical inactivity among school children? So, all these kind of questions can be answered using cross
sectional survey study design.
(Refer Slide Time: 14:59)
Now, let us come to what are the advantages of this and what are the limitations? Now advantage, I think
you all must have realized by now, they are quick you can plan and complete a cross sectional survey in a
very short period of time; may be two weeks, one month depending on how big your sample size is. And,
as you know I think the biggest cross sectional survey census even that is completed in less than a year
where the whole billion population is covered. If you are doing a smaller survey they tend to be less
expensive, unlike some other study design where you may have to keep on doing data collection for much
longer period of time.
There are few limitations of cross sectional survey that should be kept in mind while designing your study.
First of all, they are not useful to study the disease etiology, as I told earlier these are descriptive studies
and here you cannot test hypothesis. They are also not suitable in study of rare diseases, let us say if your
disease, let us say cancer; now cancer incidences 20 per 100,000 and if you want to measure the burden of
cancer, just to get those 20 cases you will have to survey 100,000 people, it is not practical. This kind of
situation, it may not be an ideal study design.
(Refer Slide Time: 16:20)
Now, coming to the most important limitations and this you should all be very familiar and keep in mind
when you design a cross sectional survey. Now, if you go and collect data on hypertension, there are patients
who will tell you that; yes, I have hypertension, I am taking treatment for last one year. Now, you do the
blood pressure measurement in addition to these old cases, you are also going to pick up some of the new
cases, who were not aware at the time of survey that they have hypertension. In a cross sectional survey,
you measure all the prevalent cases, I think in the earlier sessions you might have been introduced to these
terms, terms such as prevalence, incidence.
So, cross sectional surveys measure prevalence. Now which is fine, but what is the problem here? Now, if
you go and survey an individual and you ask, do you have hypertension? They say, yes and then you ask
them, are you a smoker? They say, yes. What it does not tell you? It is the chicken and egg story, you do
not know which happened first, whether hypertension happened first or smoking happened first. Similarly,
if you take obesity and diabetes, you go in a cross sectional survey you may find people who are diabetic
as well as overweight. How do you know which happened first? Whether diabetes happened first or
overweight happened first? So, when you do a cross sectional survey, you need to be very cautious in
interpreting your results, you need to keep in mind that these exposure and outcome cannot be linked very
well in this kind of study design and if you want to do that, you will have to use different kind of study
designs that you will be learning in the upcoming lectures.
(Refer Slide Time: 18:06)
Now, just to summarize, case reports and case series are extremely useful to document uncommon clinical
manifestations in a certain set of patients. And, these are extremely useful for the clinicians. Ecological
studies are useful when you want to relate a group level data and generate hypothesis. Cross sectional
surveys, most common study design and the very useful study design, it is useful to measure burden or
magnitude of health conditions.
Lecture – 06
Analytical study designs
Hello and welcome. My name is Manoj Murhekar and in today's lecture I will give you an overview
of cohort and case control studies. In earlier lectures, my colleagues have given you an overview
of different study designs.
Let us have a quick recap. Epidemiological studies are broadly divided into 2 categories; first is
Experimental studies and second is Observational studies. And this categorization is based on these
questions; did the investigator assign the exposure? So, in experimental studies investigator assigns
the exposure and this exposure could be in terms of new intervention, new drug or new vaccine.
These studies are further classified into Randomized and Non-randomized studies, based on
Random allocation of exposure.
On the other hand, in observational studies investigator does not assign the exposure. If there is no
comparison group in observational studies, such studies are called as Descriptive studies. And in
these studies, we described health event in terms of time, place and person. If there is a comparison
group in observational studies, the studies are called as analytical studies, which are further divided
based on the direction of the studies. Cohort studies they progress from exposures to outcome,
whereas case control study progress from outcomes to exposure and in cross sectional studies we
measure exposures and outcomes at same time.
(Refer Slide Time: 01:53)
So, in short analytical studies are the one in which investigator does not assign the exposure, there
is no randomization. So, what investigator does essentially is he carefully measures the pattern of
exposure and disease in population. There is a comparison group in analytical studies and using
this comparison group the investigator next inferences about exposure and the disease.
(Refer Slide Time: 02:20)
Let me first talk about Cohort study. The word cohort has a military origin, military roots rather
than medical routes. In Roman army, a 300 to 600 man unit was called as cohort, whereas in
epidemiology, the word cohort is a group of individuals sharing some common characteristic; one
such example could be a birth cohort, all the children who are born today will form today's birth
cohort.
(Refer Slide Time: 02:55)
Let us see, how the design of cohort study is. As we know, cohort study will progress from
exposure of outcome and in this particular example exposure is say, cigarette smoking and
outcome development of cardio vascular disease. Cohort studies begin with selection of Exposed
and Un-exposed cohort, and in this example, it would be people who are the cigarette smoking and
those are not smoking. Once we identify this cohort, these cohorts are followed in time. Some of
this exposed and non-exposed individuals will develop the disease that is cardio vascular disease,
whereas the remaining people would remain non diseased. We will then calculate the incidence of
cardio-vascular disease in exposed population and in unexposed population. And we will compare
this incidence using a measure of association called as relative risk. I will talk about this relative
risk later.
(Refer Slide Time: 03:59)
There are 3 types of cohort studies; the first is Prospective cohorts study. In the prospective cohort
study, by the time your study starts exposure and disease has not yet occurred. Whereas in case of
your Retrospective cohort study, both the exposures and disease has already occurred when you
start the study. And there is a combination of these two approaches, which is called as Bidirectional
study or Ambispective study, wherein when your study starts, the exposure has already occurred
and then you follow this exposed and un-exposed individuals, till they develop the outcome. Let
me explain these different types of studies by giving some example.
(Refer Slide Time: 04:51)
First example is of Framingham heart study, which is one of the oldest cohort study initiated in
1940s. The objective of this study was to identify risk factors for cardio-vascular diseases. This
study was conducted in a town of Framingham, which had a population of about 28,000. So, this
population; in fact, the sample of this population was then divided into 2, based on those having
risk factor and those who was not having the risk factor. And the investigator considered several
risk factors, one of which was hypertension. So, for the purpose they classified this population into
those who had hypertension and those who did not have hypertension. This cohort was then
followed up in time and the incidence of cardio-vascular diseases was compared in two cohorts.
(Refer Slide Time: 05:50)
This is an example of a Retrospective cohort study. The objective of this study was to evaluate the
role of Aniline dyes or exposure to aniline dyes and development of urinary bladder cancer. So,
the investigator for this study recruited about 4622 workers who were working in dye industry
between 1920 and 1951. So, this recruitment was based on available records in those factories.
Investigator also reviewed the death records of this 4622 individuals and essentially, they looked
about any mention of urinary bladder tumors on their death records, and then they compared death
rates in these population with that of expected number of deaths of bladder cancer using national
statistics. So, by the time the studies started, both exposure and outcome had occurred.
(Refer Slide Time: 06:58)
So, these are the 4 important components of cohort study. First is selection of study population,
second is gathering baseline information, third is following up this cohort and fourth is doing
analysis.
Once you select this study population, the next important state is collecting baseline information
from this population and the objective of this step is to have a valid assessment of exposures status
of members of cohort. And by doing this baseline information we can also collect identification
details of the study population, we can exclude those individuals, who are having the disease of
interest at baseline so that the population which remains is at risk of developing the disease and
we can also obtain the data about other risk factors or other exposure variables.
(Refer Slide Time: 08:18)
As we have seen in analytical study there is a comparison group and we have 2 options of having
comparison group in case of cohorts. One is internal comparison group and the unexposed persons
in the population is taken as an internal comparison group and example could be Framingham
cohort study. Wherein, those who had hypertension were considered as exposed population and
those who were normotensive was unexposed population.
Sometimes it is not possible to have internal comparison group. As we have seen in the case of
aniline dye example, everybody in those factories were exposed and it was not possible to have a
internal comparison group. And therefore, the investigators compared the death rates with that of
general population, so you could take an external comparison group in such situations.
(Refer Slide Time: 09:18)
Once you recruit your exposed and unexposed population, the next very important step is doing a
good follow-up of these populations. There could be 3 principles of having good follow-up, first
is having a uniform surveillance in exposed and unexposed group; having complete assessment of
exposures and outcomes and third is using a standardized diagnosis of outcomes, especially since
the cohort studies can last for a long period of time.
(Refer Slide Time: 09:50)
This is how the data in cohort study would look like this is how the 2 by 2 table in cohort study
would look like. We started the study with selecting people who are exposed, which is a plusb and
people who are unexposed which is c plus d. And we followed this people so a plus c developed
the disease and b plus d remind non-diseased. So, at the beginning of the study we know who were
expose and who were not exposed.
(Refer Slide Time: 10:22)
So, we can calculate incidence of disease in exposed population which can be given by formula a
upon a plus b and incidence in unexposed population would be in c upon c plus d and the ration of
these two incidence is a relative risk.
Cohort studies have certain strengths as well as certain weaknesses. So, what are their strengths?
They allow calculation of incidence because when we start the study, we start the study with
selecting exposed population and unexposed population and we follow them in time therefore, it
is possible for us to calculate the incidence of disease. We can examine multiple outcomes for a
given exposures. We are very confident about temporality in case of cohort study, and last is that
this studies are especially useful for rare exposures.
So, what are the weaknesses? The sample size for cohort study could be very large; we need to
follow these people for a very long time and therefore, cohort studies could be expensive and time
consuming. They are not recommended for diseases which are rare or diseases which have very
long latency, and if you do not have good follow up or differential follow loss in exposed and
unexposed population it could introduce certain amount of bias in your study.
Let us now see Case control study. Case control studies are exactly opposite to that of cohort study
when it comes to the direction or logic of the study.
(Refer Slide Time: 13:06)
Let us first see the design of cohort study. And to explain the design, I will use the example of one
of the very old case control study conducted by Doll and Hill. The objective of this study was to
test the association between cigarette smoking and lung cancer. As the name suggests case control,
we start with selecting cases and control is the one who does not have the disease in question. So,
the first step is selecting cases, so for this study Doll and Hill selected lung cancer cases, who are
admitted in hospital in about 20 hospitals in London. These cases were all histopathologically
proven cases of lung cancer. So, for each case they selected a control which was a non-lung cancer
patient admitted in the same hospital and these cases and controls are then interviewed to find out
their prior exposures.
So, Doll and Hill found out how many of the cases were cigarette smokers, they had a detailed
questionnaire to ask about history of cigarette smoking. They asked how many of them were
smoking. What is the age of starting smoking? What type of cigarette they were smoking? And so
on and so forth. So, we find out how many of the cases are exposed? How many of the controls
are exposed? And same way how many of the cases are unexposed? And, based on this data we
calculate what is called as Exposure odds among cases and Exposure odds among controls and
then we calculate what is known as odds ratio as a major of association between exposure and
outcome.
Like cohorts study, there are four important elements of case control study. First is selecting cases,
second is selecting controls and third is collecting valid information about exposures and then
doing the analysis.
(Refer Slide Time: 15:03)
Coming to the selection of cases, theoretically all people in the source population who develop the
disease of interest could be included in your study or you could sample these cases. However, one
thing you should keep in mind that, the selection of cases should be independent of exposure under
study. We need to have a clear definition of outcome to be studied. One also need to decide whether
to include prevalent cases or should include only incident cases.
Prevalent cases mean those cases which already occurred in the past, whereas incident cases are
the newly occurring cases. So, if you take a prevalent case they are readily available and by
including them we can save our time and money. But, in spite of this obvious advantage it is
generally recommended to include incident cases, mainly because prevalent cases maybe related
more to the survival with the disease than the development of disease.
(Refer Slide Time: 16:13)
Where from I can select these cases? Again, there could be two important sources; one is from
hospitals or clinics, it is easier to find cases in this hospitals and clinics; however, it is quite possible
that cases which are admitted are more severe cases and may not represent the cases in community.
The other approach could be a population based selection of cases and one such example could be
cancer registry and these cases are more likely to represent the source population, primarily
because they are not biased by factors drawing patient to a particular hospital.
(Refer Slide Time: 16:52)
Where from I can select the controls? As I mentioned earlier, control is the one who does not have
the disease under investigation. Why do we need control? Controls essentially represent the
distribution of exposure in source population. So, they generally tell you the background rate of
exposure in the population from which cases have come. Like cases, they also need to be selected
independent of their exposure status.
(Refer Slide Time: 17:30)
There again could be 3 sources of controls, first is the population based controls and you could
sample from general population. Second is you could select controls from health facility and in
case of Doll and Hill study they selected control from the health facility, but we could select patient
with other diseases. And the third source of controls could be case-based controls that are from
friends or neighborhood.
(Refer Slide Time: 17:59)
Once you select cases and control, the next important step is collecting data about past exposures.
And again, there are 3 important principles collecting the data on exposures objectively so that
your measurements are reproducible, accurately and precisely.
So, what do we do? Then what we do is, essentially we calculate a measure of association called
Odds ratio. This odds ratio is the odds that cases was exposed is given by this formula; probability
that case was exposed, upon probability that case was not exposed. And we know that probability
a case was exposed is a upon a plus c and probability that the case was not exposed is c upon a
plus c and the ration of these 2 probabilities is a by c. Same way we also calculate the odds that
control was exposed, which comes out to be b by d and the ratio of these 2 odds becomes odds
ratio which is ab by bc, which is nothing but a gross product ratio.
(Refer Slide Time: 20:03)
How to interpret this odds ratio? Again like relative risk there could be 3 scenarios, one is odds
ratio equal to 1. If odds ratio is equal is to 1, it means that odds of exposure among cases and
controls are same and we can conclude that exposure is not associated with disease in such
situation. Odds ratio could be more than 1, it means that your odds of exposure among cases are
higher than that of controls and we can conclude that exposure is positively associated with the
disease. If odds ratio is less than 1, it means that odds of exposure are among cases are lower than
that of controls and we can conclude that exposure is negatively associated with the disease.
(Refer Slide Time: 20:51)
Case control study also has certain strengths and weaknesses. These studies are especially good,
if the outcome is rare or the diseases have the long latency period. They are fairly easy or quick to
conduct and hence inexpensive. Requires relatively less subjects than that of cohort studies and
multiple exposures or risk factors can be examined at the same time. The weaknesses of cohort
study include that they are susceptible to several biases, the recall bias one of the most important
bias. Sometime selection of control could be a problem, selection of an appropriate comparison
group may be difficult and we cannot calculate incidence or disease in these studies.
Thank you.
Lecture – 07
Experimental study designs – Clinical trials
Hello. In the course of Basic Course in Biomedical Research, today I am going to discuss the Experimental
study designs or Clinical trials.
As has been discussed earlier, the various types of epidemiological study designs are classically described
as observational designs and experimental study designs. Observational study design is where the exposure
is not manipulated by the investigators and there are two different types of observational studies. There
could be descriptive studies like a series cross sectional studies or ecological studies and analytical studies
include case controls studies, which are retrospective in nature and cohort studies, which are prospective in
nature. But, considered as more advanced are the experimental study designs, where the exposure is
manipulated by the investigator and a classical example of this is clinical trials.
(Refer Slide Time: 01:05)
These clinical trials, these kinds of studies were one of the main scientific advances in the last century, they
are considered as the methodological standard of excellence, are also described as gold standard for
scientific experiments.
These are planned experiments, primarily designed to assess the efficacy of prophylactic, diagnostic or
therapeutic agents. It also helps to test out the new devices or different types of drug regimens or new
procedures that are being introduced including the investigative procedures, etcetera, particularly in human
subjects. Basically, it involves comparing the outcomes in two groups of individuals and when we talk
about doing a therapeutic trial, we talk about people suffering from a particular disease, who are grouped
into two different groups.
Wherein, one group requires a new kind of a treatment and the other group requires the standard treatment
that is being available at that point of time. Then, what is done is over a period of time, they are evaluated,
which we call it as a prospective study and we find out how many of them get effectively cured for example.
So, this is the prospective nature of the study, all the participants are essentially followed for a certain period
of time and that is why it is a planned experiment. Why? Because here is where, as we have discussed
earlier in the definition, the environment has been modified, manipulated or modified by the investigator
because the investigator has decided that some people will be placed in one particular arm of the clinical
trial, wherein the other group will be placed in the other arm of the clinical trial and that is what is the
process called randomization, which we will be discussing sometime later.
Clinical trials can be broadly used for a variety of reasons as we briefly discussed. New drugs, new
treatments, new medical or health care technologies, new organizational or delivery systems of health care,
new methods of primary prevention or new programs for screening of early detection of diseases. They
could be employed in a variety of scenario.
(Refer Slide Time: 03:44)
Normally, clinical trials are equated with drugs and pharmaceuticals. So, this essentially have intent of
therapeutic benefit. So, if we just think about how this could be done is? We have a large number of
individuals, who are actually suffering from a particular disease. What we do here is? We take out a sample
of people, who are, say eligible to apply eligible to, say participate in a clinical trial. Here is where the
problem is, that these are the people who should be willing to participant in the study. So, we cannot say
that, this particular sample is essentially generalized or it is generally representative of the total population
that we are dealing with.
Now, in a process of randomization, which we will talk about they will be sent out to two different arms of
the clinical trials. Whereas, one arm will receive the new treatment or one type of treatment, whereas the
other arm will receive the standard old treatment or the second type of treatment. Then all these individuals
will be prospectively followed up for a defined duration and with a defined frequency. Maybe it is 1 year,
once in every 3 months or 2 years, once in every 6 months or whatever and then we assess them again to
find out because it is a therapeutic trial. What we want to figure out is how many of them have actually got
cured of the disease? And how many continue to have the disease?
The reason why we are testing out a new treatment, we are having an expectation that the newer treatment
will have more benefit. So, the number of people being cured of this particular disease would be more
compared to that in the other arm and this can be statistically analyzed.
In case of a prevention scenario, say if we talk of new vaccine for example, this can also be tested in a
clinical trial design. Here, the type of population will obviously change. Here, we will talk of people or the
population, which is susceptible or at risk of developing that particular disease. Here, again we will take
out a sample of people who are willing to participate in the study. Like in the therapeutic trial, these will be
randomized into two arms, one arm or one group of people in them will receive the vaccine and the other
will receive a placebo. This is a critical decision which needs to be really discussed at length, whether it is
a placebo or some other type of a vaccine or what can be given in this particular scenario as a comparator
arm.
Again, these two groups of individuals are followed for a certain period of time with a certain periodicity
and what we try to judge is how many of them in either of the two arms actually acquire the disease? The
expectation is that if the vaccine is effective, less number of people in that particular arm who are receiving
the vaccine would acquire the disease compared to those who have not received the vaccine in the
comparator arm or the placebo arm. This can also be statistically analyzed and so we decide, whether this
particular vaccine has been effective in preventing occurrence of that particular disease in susceptible
populations.
The most critical step in a clinical trial is what we call as randomization. Randomization is a process, where
we say that the participants have an equal chance of being assigned to any study group. There may be one
study group, there may be two arms in the trial and there may be multiple arms in the trial. Accordingly,
the original numbers of participants that are found eligible that get placed into various arms in a pre-decided
manner. In the number, in each arm is decided and candidates get assigned or the participants get assigned
to different kinds of arms here. What is important here is, the participants do not decide that he or she or
they want to participant in a particular arm. It is done through a neutral process. Also, the investigators do
not decide, whether a particular participant goes in arm A or arm B or arm 1 or arm 2. It is a process of
randomization, which is a third party procedure which decides that.
So, one group gets the most widely accepted treatment, which we call it as the standard treatment or the
gold treatment. Whereas, the other group gets the new treatment and here, the hope and expectation is that
the newer treatment is better than the standard treatment. This is the process of randomization. So, it ensures
that we can test the effectiveness of new agent optimally here because all groups are most likely to be
similar, as similar as possible and confounding, co-interventions and bias in outcome ascertainment can be
minimized.
Similarly, blinding is another quality improvement technique, which is often used in clinical trials and
blinding is a procedure when participants, if they do not know, what they are getting, whether they are
getting drug a or drug b? This is called single blinding that is here participants are not aware as to what kind
of drug is being given to them, investigators may be aware.
Sometimes participants as well as investigators do not know what kind of drug allocation is being done
because the type of treatments that are been given, the nature, the color and the appearance of the
interventions is exactly identical. This is a double blinding procedure and even when the analysts also are
kept blinded, while they analyze the results, this is called as triple blinding. Basically, this stepwise blinding
from single to double to triple blinding that eliminates the kind of subjectivity in judgment of outcomes in
a large number of instances.
There are two problems that generally arise, one is a co-intervention. What happen is participant, if they
realize that one particular group is getting a particular type of drug, the other group is getting the other type
of drug. They try to do some kind of an exchange and decide something between themselves. Sometimes,
this is also facilitated by some kind of study staff or medical providers or family or friends, etcetera and
hence this kind of a co-intervention that takes place can vitiate the results of the study and hence blinding
helps to eliminate that.
Also, if for example, if a patient or a study participant knows what he is getting, if he is getting a new drug,
may be some of the minor elements he might want to exaggerate as compared to the, if he knows that he is
getting the standard treatment, he may not want to report that. Similar thing may happen in case of those,
who are actually evaluating that study participant and hence these kinds of biases in outcome ascertainment
that can come in, can be minimized in blinding.
A phase III design, essentially is a large trial, which looks at the efficacy of a particular intervention and
under controlled clinical conditions, generally they involve thousands of individuals and who are more at
risk of acquiring that particular disease or who are suffering from that particular disease and we try to look
at in a therapeutic scenario, how that particular drug is effective in curing or in prevention scenario how
that particular prevention measure is helpful in preventing the occurrence of that particular disease. Once,
all these phases I, II and III are completed, the product goes for licensures in the country, it gets a license
and once it gets the license, it gets marketed in the country and then phase IV trials are undertaken, which
are considered as post-marketing surveillance and they are done in again thousands and thousands of
individuals, but it is generally collection of information in a community based manner not in a thorough
clinical kind of a clinical trials set up. This is how clinical trials are conducted.
Somebody comes up with the new drug regimen, which probably is a cheaper regimen. Probably where the
number of tablets required are much smaller than the standard HAART regimen and so, one wants to
evaluate whether this is an optimum regimen to be used in the country or not. So, how would we go about
it? We will identify HIV infected individuals and define the inclusion and exclusion criteria for
participation, whether they are eligible to be put on antiretroviral therapy. Then they will be randomized,
all those who are found to be eligible in screening will be randomized into two groups or arms. One will
receive the new drugs regimen, the other will receive the gold standard that is available now; which is the
highly active antiretroviral therapy will follow them up periodically.
The expectation will be both these sets of drugs will effectively lower the viral load and help to improve
the CD4 counts. This will be evaluated periodically, once in every 6 months for example and if the total
period of study is 1 year or 2 years, we just see, how these two groups have reacted or responded to these
two different types of the regimens and statically we will try to find an answer, whether this new therapy
that we are talking about has provided any kind of an additional benefit. This can be done using standard
statistical methods.
(Refer Slide Time: 15:15)
In case of a prevention trial, new vaccine candidate has been developed and there is a generally data
available in the laboratory, also in the animal studies which you have, which support that, it is a safe vaccine
candidate and it is able to generate some kind of a immune response against a particular disease and it is
felt that it can be used as a candidate in humans for prevention against that particular disease. So, how we
will go about this particular thing is, first is essentially to decide at the national, regional or local level,
whether this vaccine is appropriate for the country and whether we want to evaluate this or not. Then
develop a phase I trial design, here is where because we are talking about a phase I study now. A new
vaccine is being developed; we have to start with a phase I trial design. We will have to find healthy adult
volunteers, then depends on what type of vaccinate it is. If it is a primarily a disease in children, sometimes
children could also be recruited in the study.
With then, we carry out screening to find out whether, who are the people who are willing to participate in
the study are actually eligible for enrollment and then we randomize them into two groups. One group will
receive the vaccine, the other will receive the placebo and then we will follow them up to find out the safety,
how will the safety be evaluated? We will see what kind of reactions are occurring immediately and in the
long run, say for a period one to one and half years. What kind of adverse events are reported in both these
kinds of individuals and then try to compare between the people who are receiving vaccine verses those
who are receiving the placebos. Also, we will do some blood test, to find out what kind of immunogenic
response has been stimulated in either as these two groups and compare them using standard statistical
methods. So, that is an example of a prevention trial.
So, even if there are impediments here, the clinical trials are the only way for making a progress in medical
science because, if there are no clinical trials, no new drugs will come. If there are no clinical trials, no new
technologies will be tested, no new vaccines will be tested and so they must be supported and the adequate
information about clinical trials must be disseminated. The advantages of a clinical trial include that this is
the only effective method known to control the selection bias of participants. Also, it controls confounding
bias without any adjustment, facilitates effective blinding in some trials and maintains advantages of cohort
study or a prospective study, but there are certain disadvantages of this. It is a very complex procedure, it
requires thorough training of all the investigators and also any clinical trial is very expensive.
As I mentioned earlier, we essentially use a sample from the total population that is eligible to participate
in the trial and hence, there is some level of problem with respect to generalizability of this particular
finding. But, that is something which one has to accept and there are ethical challenges which are immense.
So, we know that this is a difficult study to do, but we also know that progress will essentially be made if
we have new drugs, if we have new therapies, if we have new technologies and if we have new
interventions.
Lecture – 08
Validity of Epidemiological Studies
Hello, welcome to this session of Basic Course in Biomedical Research. Today, we are going to talk about
Validity in Epidemiological Studies.
What if you come across one day, a headline in the newspaper that a study says, that coffee drinking doubles
the risk of heart attack? What is going to be your reaction? In order to further go into indepth into the study,
we will actually need to look at how the study was done? And how valid are the results of the study?
(Refer Slide Time: 00:42)
If you look at any epidemiological study, the basic goal of any epidemiological study is twofold; one is to
obtain an accurate estimate of whatever is being studied, whether that is the frequency of a disease or the
effect of an exposure on a health outcome. And all of this, we study in a certain sample of the population.
Now, this aspect of any epidemiological study is known as internal validity of the study. How valid are the
method that are being used to either estimate the frequency or determine the effect of an exposure?
Now, in the long run what we would want is we would want that this estimate is generalizable to the relevant
target populations among which the study is being done. Now, this aspect of any epidemiological study is
known as external validity, so that the results of the study can be extrapolated to the whole population.
(Refer Slide Time: 01:40)
When we talk of accuracy of the estimate, what accuracy actually means, it consists of two things, precision
and validity. If we look at say, a bull’s eye and we want to hit the mark, what we would want to be, is to be
precise and as well as valid, so that we try to hit the bulls mark as many as times as possible. So, similarly
every epidemiological study can have results, which are both precise and valid, which is what we would
actually want in every study. However, there could be studies, where the results may be precise which
means that every time the study has been done you get the similar results, but it may be that the methodology
was not correct and so they are not valid. It may happen that the results are not precise, but sometimes they
may be valid or in the worst-case scenario, either result may be neither precise nor valid. So, when we are
looking at any epidemiological study, we need to be wary of both precision as well as validity.
(Refer Slide Time: 02:45)
As I told you in epidemiological studies all that we are doing is, basically estimating. We are estimating,
either the frequency of a disease or a health outcome or we are estimating what is the effect of an exposure
on an outcome?
Now, when we are doing these estimations there are bound to be errors that may happen in our studies.
There are two kinds of errors that we come across when we are doing epidemiological studies. One are
called random errors or errors that happen due to chance, which is basically the variability because of any
unknown or uncontrollable causes such as errors in sampling or errors in doing measurements. However,
the more problematic error that we may face in any study, are called systematic errors or biases. These are
the errors that are basically a threat to validity of any epidemiological study.
And, how do these errors happen in any study? Basically, the way we do the study, the methodology that
we use to do an epidemiological study and if it is done in a certain way that tends to produce results, that
are not the true results, then that leads to errors which are called biases. Ultimately, what we would see is
that, either the estimates or the associations that we are trying to assess between the exposure and the
outcome in the study sample may defer from the true causal associations between the same exposure and
outcome that may be there in the source population.
(Refer Slide Time: 04:16)
So, let us look at the various kinds of biases or Threats to validity in epidemiological studies. There are
essentially three kinds of biases that may occur in epidemiological study. These are called as selection bias,
information bias and confounding.
Now, how do all these things happen in epidemiological studies? Remember that, we are selecting our cases
and controls and these may happen through either, we are using a surveillance mechanisms from which
there is a systematic notification of cases and if we are taking more of exposed cases from this surveillance
mechanisms that could be one way in which selection bias could occur. We will be screening and doing
diagnosis more systematically among those, who are exposed. If we know their exposure history before
hand and then that can artificially create biases. Again, selection biases can occur in, if we select our cases
and controls from health care facilities…hospitals and where if, it is likely that more of the case patients
who are exposed are admitted or the other way round that can lead to selection bias.
Another common way in which selection bias occurs is, when we select those cases who are alive, the cases
of the disease who are dead would not be part of our studies and it may be that, the reason why these cases
are alive, may have to do with the exposure status and hence, selective selection of survived patients can
actually lead to selection bias.
In cohort studies, selection bias usually occurs when there is a lost to followup. Remember that, we have
to follow up people over a period of time in cohort studies and if it is likely that people, who are less exposed
or more exposed, are more likely to be lost or people who are at more risk or at lesser risk are more likely
to be lost to follow up, that eventually can lead to results that are biased and that would be attributable to
selection bias.
How do we deal with selection bias? We can deal with selection bias at any stage of our study. Ideally, we
would want to make sure that the way in which we design the study is free from selection bias. So, one way
would be to use incident cases and not prevalent cases because prevalent cases have the issue of survival
bias. Especially, case control studies are more prone to this, to selection bias and various ways in which, to
deal with selection bias in case control studies is to use population based design rather than hospital based
design, such that the cases and controls are actually selected from the community or the population and not
from few or a particular health care facilities.
We need to make sure that we apply the same eligible criteria, when we are selecting cases and controls
and we are not leaning towards a particular exposure among the cases and controls. Again, both the cases
and controls should undergo the same diagnostic procedures and the same intensity of surveillance in order
to identify them as cases and controls. So, that we are not biased at the time of their selection.
(Refer Slide Time: 08:12)
Now, at the time of data collection what we need to ensure is to minimize non-response, to minimize non-
participation and make sure that we do not lose many people, especially in cohort studies over a long follow-
up period. Even if, we should anticipate actually that we may lose people and so it would be a good idea to
actually keep a record of all these losses, at least some basic socio demographic characteristics of these
people. So, that later on, at the analysis stage, we can actually compare people who were lost to follow up
versus those who remain in the study and see if there were any major differences in these two populations,
which could lead to selection bias.
We also need to make sure at the time of data collection that the diagnosis of disease is not affected by the
exposure status, which means at the time of selecting, who the cases and controls are, the person who is
selecting the cases and controls should not be aware of what the exposures status of this population, of the
cases and controls are and this one way in which we do, this is called blinding.
(Refer Slide Time: 09:19)
Now, even at the analysis stage, what we can do is, as I told you before, we can compare those who
responded or those who did not respond. Those who are dropouts compared to those who are left in the
study with respect to the base line variables and see if there are any large or small difference between these
two groups. If we find large differences, it is suggestive of selection bias; however, small differences do
not rule out selection bias. So, we need to be wary of that. Again, another way to assess whether there may
be a selection bias may have occurred in our study, is to do what we call as sensitivity analysis, in which
we try to do an analysis assuming how much bias could have happened, and what direction it could have
gone and try to see how it affects the study results. If the study results are affected in a major way then we
can assume that, yes selection bias has occurred.
(Refer Slide Time: 10:19)
Moving on to the next threat to validity and that is called information bias. Information bias is essentially
a bias that can occur when we are measuring the characteristics of study participants. Now, what do we
measure, we measure exposures, we measure outcomes and we measure other variables which may
influence the exposures and the outcomes, which are called as third factors or confounders or modifiers.
What we need to make sure that the measurements that we are doing accurately, represent what it actually
is, the level of exposure is accurately measured whether there is an outcome present or absent is accurately
measured and other variables, such as socio demographic, age, gender, education, income all those variables
are also appropriately measured.
(Refer Slide Time: 11:11)
Now, how does this happen? In case control studies, information bias can happen if we are collecting
exposure information which is leaning towards a particular exposure status, if we are trying to collect more
of people, who are exposed compared to the unexposed or the other way round. This can lead to information
bias. One of the very common ways in which information bias occurs in case control studies is through the
process of recall. Remember, that we have cases and controls and we are trying to, we are asking them to
recall the past history of exposures and it may be, it may so happen that those people who are diseased or
who have a certain health event may be more likely to recall certain exposures compared to those people
who are healthy and this is what we call as recall bias.
It may also be possible that better exposure data is available on cases compared to the controls and that
again can lead to information bias. In cohort studies, information bias can happen, if we collect information
leaning towards a specific outcome status. If we follow the exposed population much more rigorously
compared to the unexposed population, that is something that can lead to information bias in cohort studies.
It may also be possible that better outcome data is available among the exposed and then again compared
to the unexposed, which can again produce information bias in the study.
Information bias can be introduced in a study both, either by what the investigator does in the way in which
the investigators collect the information about the cases, about the controls, about the exposure, about
whether they get the disease or not get the disease and if there is a systematic way in which this is being
done irregularly that can lead to information bias and last, but not the least, of course, remember that in
general, in observational epidemiological studies, we are dependent on what our study participants tell us
and if there is any systematic distortion of the true facts by the study participants, that is anyways going to
lead to information bias.
Now, how do we deal with information bias? First of all because we are measuring the exposure variable,
the outcome variable and the other variables, we need to set up precise operational definitions of what we
are going to measure and how much is it going to be. We need to have detailed measurement protocols in
the way we are going to measure each of these variables.
Sometimes, it is also good way to do repeated measurements on key variables say, for example, blood
pressure and we know that blood pressure can vary from time to time. So, we may take more than one
readings of blood pressure and then take an average of that reading, in order to say, what the actual blood
pressure of that individual is at that particular point of time. It is very important, that the investigators are
trained and certified in the way in which they follow the study protocol and all the methodology that needs
to be done to collect information. There we can do data audits, both of the interviewers and of course, of
the data management centers where the data is stored to, in order to make sure that the way in which the
data is collected, the data is retrieved, the data is stored is done correctly and there is no information bias
happening because of the same.
Once, the data is collected, we need to make sure that the data is clean. We need to go through the data both
visually as well as through computer programs, softwares and make sure that we getting clean data. It is
also good practice to actually rerun all your analysis before you are trying to do give say, send your paper
for publication, just to make sure that there is no possibility of any information bias occurring because of
the way the analysis was done.
Now, we are going to look at the next threat to validity, which is called confounding. Confounding comes
from a French word, which actually means confusion of effects. Now, what effects are we talking about
here?, remember what we are doing in epidemiological study is looking at the effect of an exposure on the
outcome, whether if you are more exposed are you more likely to get the disease or vice versa. Now, what
we want to know is the effect of this exposure on a particular outcome. This effect can be confused with
the effect of a third factor, which can have an influence both on the outcome as well as the exposure and
this is what leads to the phenomenon of what is called confounding.
(Refer Slide Time: 16:07)
Now, what does confounding do? Actually, confounding is probably the most, the biggest threat to validity
in any epidemiological study because confounding can actually simulate, can show you an association even
when it does not exist. Confounding may hide associations that is actually there or confounding may
actually increase or decrease the strength of the association. So, you may say that an exposure is more
associated with the outcome or less associated with the outcome than what it actually is, and in the worst
case scenario, confounding can actually change the direction of an effect. If an exposure, say, causes an
outcome, because of confounding you may see that the exposure is preventing the occurrence of that
outcome and that is the most dangerous threat to validity in any epidemiological study.
(Refer Slide Time: 17:06)
So, how does confounding happen? So, diagrammatically what we represent that confounder is a third
factor, is a variable, which influences both the exposure and the outcome and when we are trying to
determine what is the association between the exposure and outcome, this association is influenced by this
third factor. Now, we can deal with confounding, both at the design stage and at the analysis stage. It is
always better to deal with it at the design stage than to take care of it at the analysis stage. So, at the design
stage, we can do several things.
One, we can do what is called restriction; we can restrict our study participants to only those people who
are in one stratum of the confounders, so that the confounders cannot play a role in the association between
exposure and outcome.
Secondly, we can do, what is called matching. If we already know what the potential confounders could be
in a particular study we can match our cases and controls on those particular confounders and which will
negate the effect of the confounders and then the association that we see between the exposure and outcomes
would be without the influence of the confounders. Of course, remember that, if you do matching you have
to do what is called matched analysis.
In experimental studies, we do what is called randomization and that is something that actually
automatically takes care of the confounders and makes sure that the two arms in a randomized trials are
similar in all ways in terms of the confounding variable. Now, at the time of analysis, what we need to do
in every study, is to actually first test whether there is any confounding or whether there are variables,
which could be acting as confounders, which need to be taken care of at the time of analysis and this is
where we do what is called stratified analysis. And we stratify our data in various stratum of the confounder,
and then try to find associations and which helps us to identify whether there is confounding or not.
Now, in order to take care of these confounders, we can do, what is called a multivariate analysis, where in
we do, we use regression techniques, whether it is logistic regression, linear regression or other advanced
methodologies in order to take into account the effect of confounding and then the associations that we get
between the exposure and outcome are without the influence of the confounder or as we say adjusted for
the confounders.
So, how do you evaluate associations? Whenever you see a study, whenever you see a risk ratio or an odds
ratio, what you see is a crude association. Now, how do we make sure that this crude association is actually
the true or the causal association, that is the true relationship between the exposure and the outcome? What
we need to make sure is, we need to go through this spiral. We need to make sure that it is not because of
chance. We need to ensure that there is no selection bias. We need to check if there could be any information
bias. We need to understand if there could be confounding and if that confounding has been taken care of.
Only after going through this process, we would be able to say that whether the crude association is actually
the casual association or not.
(Refer Slide Time: 20:46)
So, coming back to our problem, does coffee really increase the risk of heart attack? Well, let us analysis
this. What we wanted to do is to look at all population, all adults in the population, who are drinking coffee.
Now, in the study what we get is a sample of people who agreed to take part in the study. Now, these people
could be people who are more likely to drink coffee or less likely to drink coffee. These people, may be
hospitalized patients, and if you are doing a study in a hospital and it maybe that these patients are
hospitalized for say, gastric ulcer and that is because of coffee drinking. So, the way in which we select
these participants can actually lead to a bias and that is what is called selection bias.
Now, what are the exposures that we are trying to assess here, is the coffee, actual coffee intake of the study
participants and what we get from the study participants is actually what they report? Are they reporting
the true coffee intake? Do they actually remember how many cups of coffee they have had in the past?
What is the average number of coffee they drink? Whether they drink coffee with milk, without milk? What
is the strength of the coffee? All of these issues can actually influence whether the coffee intake that we are
measuring is actually the true coffee intake and that can lead to information bias.
Again, remember that we are also trying to see whether the people really had a heart attack or not and it is
possible that they maybe have may have been a misdiagnosis of a heart attack. There could be other, the
chest pain that the study participants may report as heart attack, may actually have been may be due to other
causes and that is reported as heart attack. So, actually what we may be seeing is not heart attack, but some
other causes for chest pain and that is again the study results would then be influenced by information bias.
Then of course, there is confounding. Could it be that this association that we saw between coffee and heart
attack, what we call myocardial infarction in medical terminology, could it be confounded by smoking? Is
it possible that , we know that those who are smokers are more likely to, is a known risk factor for heart
attack. It is also known that those who are smokers are more likely to be coffee drinkers and it is possible
that because we may see more of smokers among the coffee drinkers and more of smokers among those
who had an heart attack the association that we are seeing between coffee and heart attack is not due to the
actual coffee, but it is actually because of the effect of smoking on heart attack. So, the result of this
association between coffee and MI could have just been confounded by the effect of smoking.
(Refer Slide Time: 23:49)
What we need to understand is that, there are various threats to validity in any epidemiological study and
these biases can occur in all epidemiological studies, more so in observational studies, such as case control
and cohort study and less so, in randomized trials. Biases can occur during all stages of the study, when we
are designing the study, if the study is not designed appropriately, if the study is not conducted appropriately
or if the analysis is not done appropriately, all of which can lead to one or the other biases. And we know
that biases threaten both the internal and external validity.
Remember that, the study which has no internal validity cannot be generalized and so it does not have any
external validity. So, we need to keep in mind is that, when we are designing a research study, we need to
be thinking of all the possible ways in which these various biases could creep-in into our study and design
it appropriately and try to prevent as many biases as possible at the time of designing and implementing the
study.
However, we should also remember that, there could be some biases which cannot be avoided. What we
need to understand is at the time of analysis of the results, we need to be aware of what these biases could
have been and state these biases in the form of limitations of the study. So, it is critical that whenever we
look at the results of any epidemiological study, we need to be wary of what possible threats could be to
the validity of these studies and make sure that the investigators have taken care of these various threats.
Thank you.
Basic Course in Biomedical Research
Dr. Tarun Bhatnagar
Lecture – 09
Welcome to this session of Basic Course in Biomedical Research. Today, I am going to give
you a glimpse of another method of research and what we generally call as Qualitative Research
Methods.
I am going to give you a snap-shot of what these methods are. In the paradigm of health research or any
research for that matter, we generally divide methods into quantitative and qualitative. Qualitative
methods generally find their origin in the Science of Anthropology, Sociology, and Psychology, wherein
which is more about dealing with human beings understanding their behaviors. When we look, sort of try
to compare the quantitative with the qualitative methods some of the basic difference is that you would
see is that the qualitative is always concerned with words, with text. So the data that we get in qualitative
methods is actually text and not necessarily numbers that you usually see in quantitative.
In qualitative research, what we try to do is to interpret the social reality more from the participants’ point
of view and their experience rather than measuring it objectively from the investigators’ experience. So,
this is what is known as the emic perspective. In terms of the logic of inquiry, again, qualitative methods
are more inductive and they are used in the way to understand the processes that are derived from the
data compared to the quantitative methods, wherein which is more deductive and where we try to test
our formal hypothesis using the data.
In terms of the research design, qualitative methods are mostly about interpreting the responses of the
participants, rather than trying to see whether what they are saying is what the investigator feels.
Whereas, quantitative methods ensure repeatability of the data and the methods are such that if the
same studies are done by somebody else we would try to except similar kind of results.
Qualitative methods, the validity is more in terms of credibility of the responses that we get from the
respondents and also the credibility of the investigator, who is doing this. Again, in terms of the
generalization of the qualitative data that we get from these methods, it is more abstract and it is
something that transits boundaries of cultures and we can understand different cultures in terms of what
the participant wants to say. Whereas, in terms of quantitative methods, it is more application of the same
methodology to different cultures trying to understand different cultures using the same methods.
(Refer Slide Time: 02:56)
Now, when do you use qualitative research methods? Well, we would want to do when we really want to
understand, why the events happened? How did they happen? And what circumstances made them to
happen? We would like to get more understanding of why these health phenomenons are occurring? We
want to provide insight into the meanings of the way people behave. We would like to know, why people
are behaving in a certain way.
Sometimes the subject matter that we are studying is insufficiently researched and qualitative research
methods help us to understand these methods. Many a times, if we get into new areas, may be the subject
matter vocabulary is something that we need to understand as an investigator and that is again where
qualitative research would help us to do that. In terms of social sciences, qualitative research methods
actually help us to view the social phenomenon more holistically in terms of how the participants see it
through their eyes.
(Refer Slide Time: 04:02)
Now, What are the various methods that we use in qualitative research? The various methods that we
use are little different from the methods that we tried to use in quantitative research. In qualitative
research, the methods are more interpretative and open ended, they are more iterative rather than fixed.
The methodology emerges as the research goes on rather than using a pre-structured format. In fact,
qualitative research methods are done so that it becomes a partnership between the participants and the
investigators and the investigator actually becomes himself or herself, an instrument in the whole
research process and it is more like a co-interpreter of what the participant wants to say. The main
methods that we use in qualitative research include in-depth interviews, focus group discussions and
participant observations. We will go through one by one through each one of them.
(Refer Slide Time: 05:01)
So, in-depth interviews are basically open ended interviews, which are done one to one between the
investigator and the participant and these are done to discover the interviewee’s own framework of what
they mean in their language. It is useful to obtain rich contextualized information and it also avoids
imposing the investigator’s assumptions over what the participants wants to say. In terms of the
techniques that are used in-depth interviews, although it is open ended and unstructured, we still follow
what we call as an interview guide, which is a set of items that we would like to discuss and get information
from the participants.
We use probes to get more in-depth understanding of the responses from the participants. In general, we
collect the respondents’ perspective and the data as I told you before is more of the words and that which
are recorded or noted down by the investigator. The level of structure, how we go on with the interview,
varies depending on how the response comes from these participants and as I said this is more an
emergent kind of methodology rather than a fixed pre-structured method.
When would you like to use in-depth interviews? Well, when the subject matter is complex and we want
to know more about it from the respondents, when the respondents probably are more knowledgeable
about the study that we are doing. Many a times for highly sensitive subject matters like sexual behaviors
or family planning issues, drug use, alcohol use, things like that, in-depth interviews are a good way to
interact more freely and get more in-depth information from research participants. Many a times, the
research participants may be geographically dispersed and it is good to talk to them one by one in different
time and place and to get more information. Of course, each method has it is own advantages and
disadvantages.
Advantages, the in-depth interviews as per name are most in-depth. We really get to understand, why
certain behaviors are practiced by individuals and why not. It gives us data on how people think. How do
they conceptualize their own behavior? What is their context in terms of the way they do or think? We
are able to get the exact words and language that people may use about the subject matter that we are
trying to understand and it really gives us an insider’s perspective of the subject matter that we are trying
to research on.
In terms of the disadvantages, since we do few in-depth interviews and we do not do it on many people,
like 100s or as we may do in quantitative methods, so there is no specific sampling. The sampling is more
convenient and purposeful and we try to get people who are more knowledgeable and who will be able
to give us information. This makes the findings from these interviews not generalizable in the strict
quantitative sense. Many a times, the interviews usually go very long, may be 30 minutes, 40 minutes,
even go, sometimes may go to more than an hour or so and you get a lot of data which means a lot of
words and it could be time consuming to actually analyze these words. Since, the analysis is more
interpretative, sometimes there is a possibility that what the data means the interpretation could depend
on how the researcher feels and interprets this data rather than how the participant would may have
wanted it to be known. So these are some of the pros and cons of doing in-depth interviews.
The other method is Focus Group Discussions. Again these are open-ended interviews, but instead of
individuals, now we have a small group of people who discuss a certain topic amongst themselves. Usually,
we have 6 to 8 participants, who are a homogeneous group, which means they are supposed to be similar
in terms of various characteristics, like age or gender or socio economic status or education, occupation,
and this helps us so that the discussion goes on in a more cohesive manner.
Usually, in a focus group discussion apart from, you would have a moderator, who moderates the whole
discussion and another individual, who works as a note taker in order to take down the notes that are
being discussed. We could also have recordings, video or audio recordings of these focus group
discussions, which could be used to analyze it later. Again, as in in-depth interviews we use an interview
guide, which is basically a set of items that we would like the participants to discuss and we make sure
that all the items are discussed in the way that we want them to be. However, the interview guides are
flexible in the sense of, we do not have to follow question one, question two, question three, it basically
flows as the discussion flows and the moderator is able to guide the discussion in order to obtain all the
information that is there to obtain.
When do we use Focus Group Discussions? These are used in areas where group interactions give us a lot
of rich information. Again, cost and timing may be issues if you were to do in-depth individual interviews,
we could get information from a lot many more participants in smaller time compared to in-depth
interviews. It helps us to generate ideas because we have a group of people, who, more than one person
whose able to respond to the various ideas that we are trying to generate. It helps to identify problems
and define goals. Many a times, depending on what you want, if you want to understand local
terminologies and the vocabularies, then these focus groups are a good way to understand these issues.
And if you want to again look at, if there has been an intervention that has been put into place and we
would like to evaluate whether it has worked or not worked, then focus group discussions are a neat way
of understanding these issues.
Of course, the pros and cons. We know that there are some people, who are more comfortable talking
openly in group sittings, and it may be a natural way for some people to talk and so it is a nice way to
understand their problems and personal issues. It is good way to collect information on social norms,
where people can discuss these issues amongst themselves. And as I said earlier it provides a lot of data
in a limited amount of time compared to doing individual in-depth interviews.
However, the disadvantages are that since we are talking to a specific group of people, the data that is
generated will depend on the actual make-up of this group and it may be difficult to access the practice
of some very personal or sensitive behaviors because people may not like to talk about them in groups.
Again, since these are groups of few people, the data may generally not be generalizable to the bigger
target population because what we get in the information is more subject to dominant personalities may
be people who are more talkative their views may be heard more than people who do not like to talk as
much. Of course, as with in-depth interviews there would be a lot of data that is generated, many a times
it runs into pages and transcribing this can be time consuming, and of course it gives a challenge in terms
of how we are able to analyze this data.
The third another common method that is used in qualitative research is what is called Participant
Observation. Now, this comes more from ethnographic and anthropological domain, wherein the
researcher himself becomes a participant in a social event or group that we are trying to study. The good
thing is that we get very deep and detailed data because you as a researcher, you are part of whatever is
happening in that group. However, it sometimes becomes difficult to systematically collect the data
because you are right in there and it may be hard to take notes or do recordings and again the analytical
methods that are used for participant observation data are still evolving and that can again be a challenge
in terms of the analysis.
Now, coming to analysis. there are various ways in which qualitative data can be analyzed and as I told
you earlier, qualitative data means text. The basically there are two ways in two methodologies that have
been known to be used in terms of the analysis, one is called a Grounded theory approach and then there
is a Content analysis approach. The grounded theory approach goes from understanding the data and
then developing theories based on that. Whereas, content analysis is when you start of with the
theoretical framework and then you try to analyze the data to understand the theory.
What is common in both of these methods is that, first of all you need to transcribe the interviews into
text, maybe from local language into English or whatever is the language of the investigator. There is a
coding of themes and categories and then relationships are built based on these themes and categories
and these are used to understand the perspective of the participants. And then codes are used, just as we
have tables with data in quantitative research, in qualitative research, if you may have seen certain papers
from that, you would usually see codes from the interviews which are used as examples to illustrate the
main theme that is generated from this data.
(Refer Slide Time: 15:31)
How do we use qualitative research methods? We could use qualitative research methods in different
ways, it could just be used as a tool to generate ideas and basically we try as a preliminary step in
developing a quantitative study. For example, if say you are trying to understand why people go, for say,
open defecation, you could, you may be developing a survey to understand this process. But in order to
understand the various categories of questions, we may do a small qualitative study to understand what
could be the probable reasons and then use it to modify the questionnaire for the quantitative survey on
this subject.
(Refer Slide Time: 16:14)
It could be used to help understand the results of quantitative studies. So you have done a quantitative
study first and then you get results, which you need to get an in-depth understanding of why you are
getting these kinds of results. And again qualitative methods could help you to do that. Or you could use
qualitative research methods as the primary data collection method and this would be where, what the
objective of the study is to actually understand behaviors, get an in-depth understanding of why people
do or not do certain things, why certain behaviors are practiced or not practiced. And you get a lot of rich
data in terms of using qualitative data itself.
(Refer Slide Time: 16:31)
Sometimes, but not necessarily, it can be done along with the quantitative study also.
There could be a triangulation of analysts or theory, which helps us to control for any selective
interpretative biases. We could use multiple methods and then focus on what information that, what is
the same kind of information that we get from these methods or we could use different data sources for
the same study and then try to understand the differences is that emerge to these data sources and then
put all of these together to actually get comprehensive understanding of the research topic that we are
trying to do.
How are these methods useful? In general, qualitative research methods are very good to actually identify
the determinants of health, why do people behave in a certain way? What are their attitudes? What do
they perceive in terms of what they are doing or not doing? If there have been certain interventions
already done qualitative research methods are a good way to understand, why these interventions have
been successful or not successful. Qualitative research methods are again a good tool to explain may be
various problems that may have arisen in terms of, why people make certain choices? Or why people use
certain services or not use certain services? And in general overall again they are a good way to
understand the context in which certain decisions are made whether at the policy level or a social or a
legal level.
So, just to sum up, we have the 2 paradigms of research, qualitative research and quantitative research.
Qualitative research is a very good way to explore, to understand why and to seek a more depth in
understanding of the topic that we are trying to study. While we are doing that it is always good to use
the advantages of both qualitative and quantitative methods to enrich our research process. And of
course, triangulate the various methods to check whether the analysis that we are getting is something
that can be enriching and fruitful.
Thank you.
Lecture – 10
Welcome to the session of Basic Course in Biomedical Research. We often hear statements like I
have data on 100,000 Leprosy patients, I have data on climatic change in Chennai city, I have data on road
traffic accidents and so on. In all these statements, there is a word “data”. In this session, we are going to
rather see what this data means, what are the different types of data? And how to convert these data
into pieces of information?
Data can broadly be classified into Qualitative and Quantitative data. Qualitative data as the name
suggests, we cannot quantify them, it is on some sort of a quality. Again, this qualitative data could be a
Nominal data or an Ordinary data. The nominal data, the examples are the color of eyes, the different
regions of a city and so on and the ordinary data are the data which can be arranged in a sort of an order
like examples you know stages of disease condition. Quantitative data again are of two categories, one is
a discrete data which essentially is a full number, a number of siblings, family size, etcetera and the other
one is a continuous data, where is a continuous measurement like height and weight. So, these are all
different types of data which requires different type of analytical skill.
Now, our aim is to get some information out of data. A large set of data, it is very essential but still looking
at just the data, you can’t rather get any information. So, we need to summarize them. One of the ways
of summarizing the data is to get a value of, a sort of an average. Now, average you mean, the first average
that comes to our mind is the Mean.
(Refer Slide Time: 02:37)
A mean which is also called an Arithmetic Mean, this is a most commonly used and simply it is called
Mean. You add all the observed values, we call that as sum which is sigma Xi in a mathematical notation
and mean is nothing but divide this sum by the number of observations you have used in your calculations,
which is n. The sample mean is denoted by x bar (x̅), line on top of x and population mean is denoted by
mu (µ).
(Refer Slide Time: 03:13)
Let us rather see an example. Suppose, there are 10 pregnant patients who had visited an ANC clinic and
their ages are 26, 31, 25 and so on and what is a mean age of these pregnant woman? The mean is got by
summing up all the ages which comes to 260. There are 10 observations, so divided it by 10. It is 260 by
10 which is equal to 26. We say the mean age of pregnant woman who visited the ANC clinic is 26 years.
(Refer Slide Time: 03:50)
Now, one of the problem with this average mean is, some extreme values either big or small even one or
two, if they are present in your data set, that could influence on their average because you are adding all
and one big value, if we add the whole mean becomes an over estimation. In order to control this or in
order to avoid this, we have another measure which is called Median.
(Refer Slide Time: 04:22)
The median is, literally the middle value of the distribution. It divides the distribution exactly into two
halves that is 50 percent of the data will fall on either side. This is a very useful measure, especially when
you have extreme values.
Mode is the value that occurs most frequently. In fact, mode is the only location statistics which we can
use for nominal data, which are not measurable. In epidemiology, we do use more quite often. In an
epidemic curve with respective time, we look for the modal class and then that gives an idea of the
incubation period of the pathogen.
(Refer Slide Time: 06:17)
The example for a mode is the color preference and the number of persons, 354 people they prefer green,
852 prefer yellow, 310 prefer white and 474 prefer red. So maximum number of people they prefer yellow,
answer the mode of class is yellow. As you could rather see as in the respect of rather mode, there can be
multiple modes, there cannot be a mode at all in a sequeal (Refer Time: 06:42). Suppose, if all the values
are 354 here, then there is no mode. So, mode can exist, there can be multiple modes in a data set.
Now, we have seen Mean, Median, Mode are 3 good measures of summarizing your data to get an average
value.
(Refer Slide Time: 07:08)
So it is not enough you just rather know the average value. Say for example, you go to a swimming pool
and you do not know swimming and you are 5 feet 7 of inches and then, if the pool managers says the
average depth of the swimming pool is 4 and a half feet you feel very comfortable and you jump, and
suppose a place where you jump is 9 feet then you know the thing that you missed to ask is, yes the
average is 4 and a half feet, but what is the variability? There maybe you know place where it is as shallow
as 3 feet and as depth as 9 or 10 feet. So, you need to rather ask, what is the variability? One of the
measures that comes to our mind is the Range.
(Refer Slide Time: 07:57)
The range is the difference between the minimum and then the maximum value of the observations. An
advantage of this measure is it is very quick and easy indicator of dispersion. But, as I had said about the
mean, the range also is influenced by extreme values and also we consider only two values, the first and
then the last and in between we are not using the data at all and that is the great disadvantage of range.
There is another value which is called Inter-quartile range.
This to a large extend take care of this extreme values in the sense, we divide the data sets into 4 quarters.
(Refer Slide Time: 08:38)
And we try to remove the first quarter and then the last quarter and consider only the middle 50 percent
of values and this inter-quartile range is the Q3 minus Q1 and a great advantage of this is, this value does
not rather get affected by extreme values. But again the disadvantages is, it covers only the middle 50
percent of the value and then the same disadvantage that we had for range that uses only two values and
in between values are not made use of and that is a great disadvantage of this value.
(Refer Slide Time: 09:15)
Another measure of variability is Mean deviation from mean. What you mean by that? Say for example,
from your data set, every data point we try to subtract a mean and then we try to take average of these
mean deviation, which is called Mean deviation from mean.
So, in order to get over that we have another measure, what we do is? We do take the difference of each
observations from mean and instead of ignoring the sign, we square them because square takes care of
even the negative and then the positive everything becomes positive and then we take an average of that,
that value is called Variance.
(Refer Slide Time: 10:59)
And since variance is, we are squaring and then the measurement also squares we take a square route at
the end and that is called Standard Deviation. The standard deviation, which is denoted as SD is a square
route of the average of the squared deviations of the observations from the arithmetic mean. The square
of the standard deviation is the variance. Advantage of standard deviation is most important measure of
distribution, while the variance is in unit square, the standard deviation is expressed in the same units of
the measurement and it is suitable for further analysis. So, standard deviation together with arithmetic
mean is useful for describing the data and these two measures are extensively used for further treatment
of your data set.
(Refer Slide Time: 11:51)
I am going to introduce to you one more measure which is called Coefficient of Variation. The purpose of
this measure, suppose if you have a different groups, different data sets to compare and then you want
to rather compare the relative variability in different groups. So, the coefficient of the variation is the
standard deviation expressed as a percentage of arithmetic mean because the standard deviation by
arithmetic mean, what happens is, they both are the same units of the measurements, so units of
measurement get canceled, so what you get is a pure number and that number expressed in terms of
percentage that is multiplied by 100, you get coefficient of variation.
(Refer Slide Time: 12:38)
In summary, we have to choose an appropriate central or dispersion values. The Mean and Standard
Deviation are the most appropriate central and dispersion values especially, if there are no extreme
values. If there are extreme values, there are methods of still using mean and standard deviation using
some transformations of your data that requires (Refer Time: 13:06) expert handling of your data.
Otherwise, you go in for median and inter-quartile range and these two measures median, inter-quartile
range do take care of extreme values. The mode and range is normally used for qualitative variables, time
distributions in epidemic curve. The mean and standard deviation, as I said are the most used measures
of variability and the summary statistics.
Thank you.
Lecture - 11
Sampling Methods
Welcome to Research Fundamental Modules, the NIeCer course one. Today, we are going to talk about
some aspects of Sampling Methods.
Sampling is really required, whenever you are dealing with a very large population and you want some
quick information. Now, look at the definition of sampling. The sampling is a procedure by which some
members of the population are selected and they are supposed to be the representative of the entire
population.
(Refer Slide Time: 00:47)
See, if you have a population of people like this and you are looking at a portion of them and that is called
a Sample.
Say for example, how many injections do people receive each year in India? The study population in this
case is the entire population of India. Suppose, your research question is, how many needle-sticks health
care workers experience each year in India? Then the study population becomes health care workers of
India. Suppose, if your study question is how many hospitals have a needle sticks prevention policy in
India? Then your study population in this case becomes hospitals of India.
(Refer Slide Time: 01:55)
The sample which we select should be representative of the population for which we require an answer and
this representation should be in accordance with , seasonality, the day of the week, the time of the week.
So now, there are different ways how or why we should do the sample populations? If you have enough
resources, you can probably study the entire population. But still, even if you have the resources, it is not
wise to study the entire population because often the population is very large and a large population when
you are going to collect information one of the major constraint could be the time. You may require lot of
time to collect information and you may say I will employ lot of people to do that, but what would rather
probably happen is if you have lots of people collecting information there could be lot of inter observer
variations which could rather add on to a tremendous amount of error and unfortunately you cannot measure
the amount of such errors. So, it often rather happens that by doing a sample survey you often get accurate
information. The information that you get from sample surveys are more accurate than the information you
do on a large scale population studies.
(Refer Slide Time: 04:36)
So, a population could be an entire universe, whereas a sample could be as selected a small regions.
Let us look at a practical example. Suppose, the ministry of health of a country X wants to estimate the
proportion of children in elementary schools, who have been immunized against childhood infectious
diseases. You could just imagine, you know the proportion of children of all elementary schools, who have
been immunized against childhood infections of a country. So that is a task but one of the conditions that
he has put is the task must be completed in one month. So, the objective is to estimate the proportion of
immunized children and you want the results in a month’s time.
Now, let us rather look at the different ways; how you can rather get this information? Or in other words
what are the different types of samples that could rather group. See, broadly speaking the sample could be
a Non-probability Sample or a Probability Sample. What do you mean by a Non-probability Sample? Non-
probability sample is the probability of being selected that is a sample, the probability of being selected for
your study is not known. It could be a convenient sample or purposing sample, you just rather convenient
whatever the region that is convenient to you, close by to your place, you can rather go and rather see first
100 people that you come across that could be a convenient sample.
What could rather happen? That sample could be biased or it can rather give either a best or a worst scenario,
people you know it is a convenient location. You may get rather the results very different from a location,
which is not very convenient or which is very remote and difficult to approach. And also, some of these are
all very subjective samples and to derive some objective criteria from a subjective sample is always
difficult. But, nevertheless these non-probability sampling methods still are useful and that is being
extensively used mainly to generate hypothesis or to prepare for more systematic probability samples. Now
let us look at, what do you mean by Probability Samples?
In a probability sample, every unit in the population has a known probability of being selected. What is the
advantage? This is only sampling method that allows to draw valid conclusions about the population.
(Refer Slide Time: 07:30)
It removes the possibility of bias in selection of subjects and also ensures that each subject has a known
probability of being chosen. It allows application of statistical theory because many of the statistical text
that you do it insist on a random sampling and these tests are valid only if the samples are a random sample.
Now, let us look at some of the popular sampling methodologies that are employed in sample service. Let
us rather look at the first Simple random sampling.
(Refer Slide Time: 09:11)
As a name suggests, it is a very simple sampling procedure, very easy to understand in which every
individual sampling units have got an equal chance of being included into a sample. How do you do that?
We number all the units and we randomly draw units. The advantages as I mentioned, it is very simple and
sampling error is also very easily measured. Major limitation of this is, you need to have a compete list of
all units. Many times it may not be available and also some times you may get a sample, which is very
different from the whole population may not be very representative of the population.
(Refer Slide Time: 10:07)
See, an example of a simple random sampling could be if you have the list of all say about these 48 names,
you pick a random numbers of 9, 18, 32, and 40. So these are all the names that are selected as your sample.
What are the advantages? It ensures representatively across the list. It is easy to implement. You gave a
worker that you say, you start from this house and every 10th house you go on rather see and cover all the
houses, it’s very easily been done. If there is some sort of a cycle of some specific characteristic that you
are studying then you might probably get a sample, which is very typical in a systematic sampling and also
some of the statistical measures that you are going to compute, it is difficult when you are going to have
systematic sampling, where you do not have an exact formulas, you may have to use some approximate
formulas.
The example of a systematic sampling is you see, in the first, the red house is selected and then every eighth
house from that is selected and all the red houses in these houses are your selected samples.
(Refer Slide Time: 12:13)
There is a sampling method called Stratified Sampling. The principle of it is, who classify population into
homogeneous subgroups, which are called 'strata' and you draw sample from in each strata combine the
results of all the strata to get an idea of the whole population. The advantage of it is its more precise, if
variable associated with strata and all subgroups are represented, allowing for separate conclusions about
each one of them. Suppose, a natural strata could be male and female, so you have an estimate for male and
you have an estimate for female and you can have an estimate for a combine male and female for the whole
population. But the disadvantage is, sampling error is difficult to measure, and that could be loss of
precision, if you are going to rather have a lot of strata and for each strata you have small numbers in it.
(Refer Slide Time: 13:13)
Example of a stratified sampling is, suppose if you want to estimate the vaccination coverage in your
country. One sample drawn from each region north, east, south and west and the estimate calculated for
each of the stratum and at the end you can weight the stratum according to the size of the regions.
The sampling unit is not a subject, but a group or a cluster of subject. The assumptions here it is, that
variability among the cluster is minimal. The variability within each cluster is what is observed in the
general population.
(Refer Slide Time: 15:00)
Now, how these clusters sampling is usually done. It is done as two stage approach. In the first stage, a
probability proportional to size, that is select the number of clusters to be included, compute a cumulative
list of all the population in each unit with a grand total, divide the grand total by the number of clusters and
obtain the sampling interval, choose a random number and identify the first cluster, add the sampling
interval and identify the second cluster and so on and by repeating the same procedure, identify all the
clusters.
Once your clusters are identified, then in the second stage in each cluster, you select this random sample
using the sampling frame because as I had mentioned you earlier on simple random sampling when you
want to do a simple random sampling you need to have all the list of the your sampling frame. So, in a
small cluster it is possible for you to formulate the sampling frame and you can select people from that
sampling frame on a random basis.
(Refer Slide Time: 16:14)
Another important sampling methodology employed is called a Multistage Sampling. In this multistage,
especially in a very large, you want some estimates for at the national level, you need to do a sampling in
several chains samples and several statistical units are there. The advantage is, there is no complete listing
of the population is required and it is most feasible approach for large populations. The disadvantage is,
there are several sampling units and sampling error at times, it is very difficult to unless you follow certain
very specific methodologies for selecting at each stage.
(Refer Slide Time: 17:02)
Some of the key issues that I would like to bring to you is we cannot study the whole population so we
sample it. Whole population studying could impact a result in inaccurate results so taking sample leads to
sampling error, but which is easily measurable and we do not have a measure for non-sampling error,
whereas we have a measure for sampling error. Good design and quality assurance ensure validity and while
appropriate sample size will ensure precision. The probability samples are the only one that allows the use
of statistics as we know them and so it is always advantage to use a probability sample so that you can have
a valid conclusion, a precise conclusion and also you can employ statistical test on them.
Dr. R Ramakrishnan
Lecture - 12
The usual question most of the investigators they have in their mind when they want to rather start doing
a research study is, how much subjects that I should recruit in to my study? How many patients I should
see? How many households that I should cover? And a simple answer for this question is there is no simple
answer. This requires a little bit of logical thinking like, and usually it depends on some of the information
that we already should know before we start our study.
So in this module, we will try to understand, what’s the relationship between sample size and I am going
to introduce you, to a concept called power. And, we also try to rather determine the sample size, which
is absolutely essential or necessary to achieve a given level of power for estimating may be a simple
proportion or any other measures of effect.
(Refer Slide Time: 01:58)
As I was mentioning to you, there is no simple answer for estimating a sample size. We need to go in a
systematic manner and let me rather take you step by step in the process of estimating sample size. First
of all, you will have to identify, what is a major study variable you are planning to study? In any
investigation, you will be seeing a number of things, you will have to identify which among them is the
most important variable, which you want to rather study about. Say for example, when you are studying
on may be scrub typhus in a community; your aim may be to estimate a prevalence of scrub typhus in
which case, the variable whether a person has got scrub typhus or not is a major variable.
Suppose, if your interest, is not on the prevalence but on some of the associated factors, whether a person
has been exposed to a forest or something like that, in that case that becomes the major study variable.
Then the second step is to determine the type of estimate. Are you going to rather study at mean or a
ratio or a percentage or proportion because accordingly we need to rather re-frame or have a formulae
for computing the sample size.
Then one of the important things that comes out it is, you need to indicate the expected frequency of
factor of interest, common sense says, suppose if you are going to rather study something very rare you
need to rather have a large sample, unless you see a very large number of people you may not probably
get sufficient number of people with the factor of your interest. On the other hand, if you are going to
rather study something which is very common, in that case you do not need a large sample even and a
small sample you may be able to rather give fairly a good sufficient and a precise estimate of your factor
of interest.
Then the next factor is, the decide precision of the estimate. How precise you want your estimate to be?
You want your estimate to be within 5 percent this side that side or within 10 percent this side that side.
As you want your estimate to be more precise then naturally you need to rather have a larger sample. If
you are willing to rather give say plus or minus 20 percent, then probably your sample size will be small
as compared to plus or minus 10 percent.
Then the next point is, I want plus or minus 10 percent but how sure I want that my estimate is plus or
minus 10 percent? What is the amount of rather risk that I am willing to accept? Whether a 5 percent risk
or a 10 percent risk. So These are all some of the elements that are essential to compute the sample size
and invariably these are all the elements that has to be rather given by the investigator to whoever is
computing the sample size. The other three items that I have rather given are, you have to adjust what
population size. Are you going to take your sample from a very large population? Or you are going to
rather take from a small population? Because usually the sample size formulae assume that you are taking
a sample from a very large population. If you are going to rather take your sample from a small population,
you need to do some adjustment factor into it.
The next is, adjust for estimated design effect. See in my earlier lecture on sampling I talked to you about
a clustered design effect, wherein you are going to rather select not individual subjects as your sample
but you are going to rather select cluster of subjects as your sample. There could be a correlation between
the subjects in the same cluster. So, in order to get over it you need to multiply your sample size by a
factor called Design effect so that you have a larger sample which takes care of this correlation within the
subjects in a cluster.
Then the last bullet point in this, it is adjust for expected response. You have to rather you know, you
decide that you want to do 300 and if you just go and study 300 maybe you know 10 percent of them they
did not turn up and you have only 230. In order to adjust for that you have some may be 10 percent extra
as your sample size so that assuming a non-response, you still have sufficient sample to answer your
question.
(Refer Slide Time: 07:31)
I am going to introduce you now to some concepts, which are essential to understand the computation
of sample size. The first one is the alpha or the Type I error, the significance level of a test. What do you
mean by that? It is the probability of rejecting the null hypothesis when actually it is true. In the statistical
parlance, it is called Type I error. And the confidence level is the complement of that, that is 1 minus alpha
and that is naturally the probability that an estimate of a population parameter is within certain specified
limits of the true value.
(Refer Slide Time: 08:23)
The next are beta and Power. Beta is nothing but the probability of failing to reject the null hypothesis
when actually it is false. So, if something is false you have to reject it, but you accept it and the probability
of making this is called a Type II error. And the complement of that is 1 minus beta is commonly denoted
as power and which is a correct decision and that is nothing but probability of correctly rejecting the null
hypothesis when it is false.
(Refer Slide Time: 09:01)
Another important concept that goes into the computation of sample size is the Precision. By precision
what you mean is? It is a measure of how close an estimate is to the true value of a population parameter.
It may be expressed in absolute terms or relative to the estimate. They say plus or minus 10 percent or
plus or minus 10 percent of the estimate.
(Refer Slide Time: 09:27)
Now, let us look into a scenario where you need to compute a sample size and your interest is to estimate
the Mean of a Population. You have a sample, you have to compute a sample mean and you want to
estimate the population mean from your sample mean. And what should be the sample size that you
need? So, the general idea of the computation of sample size is, it is always a reliability coefficient into
standard error is called d and through d, we can estimate the sample size n. Say, the d is a formula given
here is z into sigma (σ) by square root of n. How do we get that? We get that mainly using the concept of
a sampling distribution.
What do you mean by sampling distribution? Suppose, I take several samples of the same size and each
of the sample given estimate of the population mean and if I have a distribution of all those sample means,
theoretically, it is proved that, that distribution is at normal distribution and also the standard error which
is the standard deviation of that state distribution is given by sigma by root n. So, what we normally we
do it is? We try to rather have see the using the principles of normal distribution 2 sigma limit of the
sampling error 95 percent of the values they lie. So The z in the formulae is nothing but the standard
normal deviate for a particular level of significance. And suppose, the idea of sample size is you fix that
and then when you fix that you have only one unknown namely, the n in the denominator and if we can
solve for the n, which is nothing but n is equal to z square into sigma square by d square, then you get an
idea of what the n is.
(Refer Slide Time: 11:47)
I think you know these concepts will get clearer, if you see an example. A health department nutritionist,
he wishes to do a survey among population of teenage girls to determine the average daily protein intake.
So, that is the research problem. And what information is needed to estimate the sample size? So the
nutritionist must provide three items of information. The first one, the desired width of the confidence
interval, the next one, is the level of confidence desire and you should rather give a rough magnitude of
the population variance.
(Refer Slide Time: 12:36)
Assume, that he gives them all, say the nutrition feels that the 10 units this side and that side is what he
is expecting, which means now 10 units on the whole, so 5 units this side and 5 units that side and the
confidence coefficient of 95 percent is decided upon, and from his past experience from the literature
review, the nutritionist feel that the population standard deviation is probably about 20 grams. Now, we
have the information z is 1.96 because 95 percent confidence interval has got a corresponding z value of
a normal distribution 1.96. Sigma is already given as 20 and then the desired length d is 5 units this side
or that side. If you plug in all these value into the formula, it become n is equal to 1.96 square 20 square
multiply these two, divided by 5 square it comes to 61.47. Which means you need to have at least 62
teenage girls in order to get an estimate of the mean protein intake and the estimate that you gave 95
percent of the time will be within 5 units this side or that side of the true mean population mean of protein
intake.
(Refer Slide Time: 14:08)
In this formula, we have sigma that is the population variance. And in most scenarios you may not
probably know a value of sigma because this is study you are going to rather do sigma may not be available
and how to get a sigma? One of the way that you can get this variance is do a pilot survey, preliminary
survey. Of course, you can even use this observation used in the pilot for your final sample tool. An
estimate available from the pilot survey could be used or you can use an estimate which is available from
previous studies, and suppose you know you have a large data available and you have the range of the
data, assuming that it is normally distributed, you can get an approximate value of sigma as the range
divided by 6. So, these are all ways of getting the value of sigma in your formula.
(Refer Slide Time: 15:17)
Suppose, we are going to estimate a Proportion not a mean, the formula is more or less similar but what
you need to give is you must rather have knowledge of p that is the proportion of the characteristic or the
factor of interest in the population. This also may not be, see you are going to do a study to estimate a
proportion and invariably when I ask this question to the investigator he says, Sir, I am going to do a study
to find it, how can I have an idea of p, can probably as I had rather mentioned earlier he can probably do
a pilot study to get an idea of p or you can get from the literature, what could be the value of p and if it is
impossible then the best thing is to estimate to get a value of p as 0.5, so that it is the maximum value of
n.
(Refer Slide Time: 16:19)
So the formula for that is very simple the n is the z square and d square in the denominator are common
and here instead of your sigma what you have is the pq, where q is nothing but 1 - p.
So naturally the q must be 0.2 then z alpha is 1.96. And when we plug them all into a formulae what we
get is 384.
(Refer Slide Time: 17:36)
So, you need 384 subjects to get an estimation of the immunization coverage, within 4 percentages this
side or that side and you have 95 percent confidence that the true value lies in this particular interval.
In similar sort of scenario, suppose we are going to rather do a case-control study or a cohort study and
how you should go about in estimating the sample size. What you need is the desired value of the
probabilities of alpha and beta and the proportion of base line or there is a control or non-exposed
population. In the case of case control studies, the proportion of exposed or in the case of cohort studies,
the proportion of disease and you need to rather have some idea of them. These are all often based on
previous studies or reports and also you should have some sort of an idea of the magnitude of the
expected effect that is the magnitude of relative risk or odds ratio.
This again is based on a previous studies or reports. And what is the minimum effect the investigator
considers? What is detecting? These are all some of the information once it provided then there are very
easy formula available, the different formula depending on the study design, research question and the
type of data.
Now, let us take few examples and this example should rather give you some idea of, how the sample size
is computed in different scenarios? Take for example, a cohort study of oral contraceptive use in relation
to the risk of myocardial infarction among women of childbearing age. Previous studies have indicated
that the proportion of non-OC users who are at risk of disease is 0.15 that is 15 percent of non-OC users’
women of childbearing age have a risk of myocardial infarction. So, proportion of OC users who are at risk
of disease is 0.25 and the conventional alpha is 0.05. Suppose, the beta is taken as 20, 0.20 that is you
want 80 percent power to detect the difference if it truly exists and assume that you are going to rather
have equal sample sizes for your users and non-users.
(Refer Slide Time: 21:18)
The formula is obtained using these following parameters you know p nought which is nothing but
proportion of non-OC users who are diseased, which is given as 0.15. p1 is proportion of OC users who
are diseased, which is given as 0.25 and your q nought which is a complement of p nought is 0.85. Your
q1 is a complement of p1 which is 0.75. Z alpha is 1.96 we saw in the last example. Z beta is 0.84.
Now, let us take an example of a case-control design. How do you go about? In the case control study of
oral contraceptives use in relation to the risk of myocardial infarction among woman of childbearing age.
Previous studies says 10 percent of woman use OCs and OR of MI associated with current OC use is 1.8.
Then the other thing as conventional alpha as 0.05, conventional beta is 0.20 and assuming equal size for
case in control.
And see you have all these parameters p nought is equal to the proportion of controls who are current
OC users which is 0.1. p 1 is equal to proportion of cases, who are current OC users and that is 0.18. Q
nought is 0.9 and q1 is 0.82. Z alpha 1.96, z beta is 0.84.
(Refer Slide Time: 23:27)
If you plug them on in a formula then you get 291.06 which indicate that you need to rather have 291 or
292 cases and 292 controls in order to get an estimate of your OR.
(Refer Slide Time: 23:43)
Now, this slide gives you the required sample size for various OR. Say for example, you want to detect an
OR of 1.2 then you need a sample size of 3834. Whereas you have to estimate an OR of 3, it is enough you
have 59 in each group. So, what it means it is, if you want to detect a very small difference then you need
to have a large sample to identify that small difference. If you want to detect the large difference then it
is enough you have a small sample, then you know you will be able to get an estimate of your OR which
is 3 or more.
(Refer Slide Time: 24:32)
Then see in any of this analytical studies, when you are looking for an association of one variable, there
could be a third factor which could be effecting the values of this association, which is in the
epidemiological parlance, we call that as confounders. There could be one variable or two variables, which
are confounders in a particular association. The general rule is, if you have some confounders in your
studies you hike your sample size by 10 percent for every confounder variable that you have.
(Refer Slide Time: 25:09)
Now, having seen different scenarios, where the sample size is computed and then the concepts behind
it, I am going to introduce you to 2 softwares, which are free softwares in the open source, which can be
very easily used to compute sample size of different study designs. One is called OpenEpi and this software
is supported by the CDC Atlanta and the website for that is www dot openepi dot com and it is very, very
simple software to use. There is a tutorial in for this particular software which gives you some examples
and depending on what sort of a study design that you have, you can plug in the values that the software
ask and you will get the desire sample size.
Another one, which is called PS that is Power and Sample size calculation, this is by the Department of Bio
statistics Vanderbilt University, this is also an open source software. This is also fairly user friendly
software where you can rather compute to your sample sizes. So, wherever you do an investigation and
when you compute the software, you give the software in your reference saying that you use this
particular software and these are all the assumptions or these are all the values that you had rather
plugged-in in this software, so that this is my sample size. This should be reflected in your method section.
So, to recapitulate the module, sample size there is no magic number as sample size available. Sample
sizes have to be computed using various parameters that are supplied by the investigator. Investigator
may have some idea of it ready made, if he does not have those ideas you may have to probably do a pilot
study to get this kind of an idea. And then you know it depends on, how much risk that you are willing to
take? How much precision that you want on your estimates? And this is usually there is no fixed number
we can always negotiate depending on the resources that are available in terms of money and time.
Suppose, you know I say you need to rather have 300 and you do not have that much resource and you
do not have that much time to do, you can always rather they know reduce the sample size, but you
should know what price that you are going to pay for reducing you may have to compromise on the
precision or the risk that you will be taking on this sort of estimates.
Lecture - 13
Selection of Study Population
Hi! I am Dr. Ganeshkumar from ICMR School of Public Health, National Institute of Epidemiology. Today,
we are going to have a lecture on Selection of Study Population.
A good choice of study subjects serves the vital purpose of ensuring that the finding in the study accurately
represents the population of interest. So, that is an important part, how the selection of a study population
gathers the right information about the health research over the population of interest.
(Refer Slide Time: 00:43)
See, when we are going to a select a study population or a selection of subjects from the population of
interest. There are 3 important things which we have to remember particularly; first number one is that, we
have to select the study subjects at an acceptable cost in terms of time and then money. Second, we need to
have an adequate size of the study population so that it controls the random error, the adequate size is most
important. Number three, your study population; the study subject should be representative enough to the
population of interest so that your findings can be generalizable to the population of interest.
That is how in this video lecture we will be seeing, how to select a study population with a good
representativeness? And we will be also discussing about what are all the issues and what are all the
solutions of selecting a study population? And what are the recruitment strategies in a clinical research?
Whereas, how to achieve your adequate sample size and the process of sampling in a research will be dealt
in the separate lecture. So, this lecture will not be covering about what is adequate size? and As well as,
what is the techniques or sampling techniques in selectioning the study population?
(Refer Slide Time: 02:14)
Ok, In a health research, in a clinical research or a public health research, when you are going to select a
study population, these are all the terminologies which we need to understand, which ideally happens over
the process of doing a research. For example, in a layman term, a population is a larger area for example,
population of India or the population of Tamilnadu, a larger geographical area. From that population by
setting up or for defining certain clinical under demographic characteristics, what you are trying to derive
is called a target population. So, your target population is determined by certain clinical and demographic
strategies, characteristics.
By applying certain geographical and as well as temporal characteristics over there temporal target
population, what you are trying to derive is a called accessible population. So, if from these accessible
population by means of certain subset of this accessible population, what we are trying to derive is called
Study sample. So, this is how a process of deriving a study sample from the population. So, from this step
you can able to understand that two important terminologies; one is called Target population which is
defined by clinical and demographic characteristics, second is a Accessible population which is defined by
geographical under temporal characteristics of the subset of the target population and from the subset of the
accessible population what we derive is a study sample.
I will give one example for these; for example, our target population which based on the research question,
if the research question is like I want to study about what is a low dose of metformin to reduce the
dysmenorrhea among PCOS females in reproductive age group. So in that here, when you specifically see
that the clinical and demographic characteristics are those PCOS females in the reproductive age group and
they are sufferings from polycystic ovarian disease and they should be having a clinical feature of
dysmenorrhea. And, the geographical and temporal characteristics are defined by the accessibility of the
population; this is the subset of the target population.
For example, when I am going to conduct in a city and those patients who are attending my OPD are my
accessible population and I am recruiting them for this study. So, this defines the geographical and temporal
is that from January 1 to December 31st of a year, a specified year. So, all the patients who are attending
my OPD's with this clinical and demographic characteristics in this period will be recruited for my study.
And from the, I estimated sample size, I will be recruiting those subset of accessible population called
study samples.
So, let us see how the journey of choosing the study subjects that represent the population happens. So, I
am just throwing a kind of an algorithm here which will explains you certain two important terminologies
called External validity and Internal validity. For example, by means of a phenomena of interest; that means,
by your research question your setting up certain clinical and demographic characteristics what to derive is
a target population, and from that you have an intended sample and that intended sample you are measuring
certain intended variables and what you derive is a actual subjects, that is study subjects, and where you do
your actual measurements. So, same how you do is that, you do your research, you choose your research
question of the phenomena of the interest and you have a study plan of conducting the study and measuring
it with an actual measurements and that is your actual study here.
So, how it happens in reverse? Say, from the findings of your study, from those actual subjects what you
infer is called a truth in the study and this truth in the study is what you are trying to infer over the truth in
the universe, so this goes in a reverse. So, you are trying to generalize your study findings over the
population of interest and that is where you have depicted as truth in the study and truth in the universe. So,
here you can clearly note that there are certain errors, which may happen when you are deriving the subjects
from the target population to the actual subjects. And previously, I explained that these kinds of random
errors can be handled by an adequate size of the sample population.
Now, you can clearly see here that, where this external validity and internal validity lies. So, the internal
validity is a kind of a degree that how far the dependent variable influenced by the independent variables
and this is very much consistent within the study subjects. If it is very much consistent and that has a good
internal validity and same, when this derive, mean this findings or we are going to generalize to the
population the generalizability is called that external validity, where the findings of the study are applied
over the population of interest. So, internal validity is very, very essential. So, that is a most important part
when we are trying to generalize the results.
The internal valid results, are not necessary be generalized to target population. So, that is how and I am
trying to explain here that there is always a trade-off between internal validity and generalizations. So,
generalizability hardly a categorical answer of yes or no, it is very rare. It has many trade-off, it is kind of
a mix where scientific unpractical decisions has been taken when we are choosing the study subjects and
when we are going to infer the study findings and applicable to the population of interest. So, it depends
upon the design, it depends upon the method and also it depends upon the specific research question and
where we are applying to derive the findings.
In general, generalizing the study findings can be very valid can be easily happen in clinical studies and
clinical trials and analytical studies. So, generalizable of these study findings is very common. So, very
popular example is say Framingham Cohort Study. So, in that Framingham cohort study what they have
selected is from your individuals from your county called Framingham. From that, when they are selecting
it whether the findings, whichever the findings of the study can be generalizable to the whole population of
United States of America? No, it depends upon the study findings.
What they have identified in Framingham is that, the association between those cardio-vascular risk factors
or the association, the strength of association of hypertension has a cardiovascular risk factor is consistent
not only to the population but also to other kind of genetically different ethnical group called Americans,
Africo-Americans and all those things. Whereas, when we are trying to generalize certain other findings
the prevalence of hypertension, which is identified in this Framingham study may not be generalizable to
other ethnicity.
So, that is how descriptive studies always have an issue that, because there the study results cannot be
generalizable as such. It is specific to the population which it has been studied and because descriptive
studies mainly study the distribution of the characteristics of the population which may not be generalizable
as like that when we are doing in analytical studies or a clinical trails. So, now, you can understand that
generalizing the study findings is based on even the study design and the type of method and how the
findings depends upon the findings, whether it is an analytical study findings or it is a descriptive study
findings, it is very important.
Let us see that, the other important part of selecting the study population with health research is called
selection criteria. So, what is that selection criteria? So, this flow chat which you have understood already,
which we have discussed already is about that how we derive a target population by a set of clinical and
demographic characteristics and how we derive the accessible population, which is the subset of the target
population by means of geographical and temporal characteristics. So, the terminology called inclusion
criteria and exclusion criteria is based on this, how? See, inclusion criteria is the main characteristics of a
target population that pertain to the research question. So, that main characteristics means here the clinical
characteristics, demographic characteristics and often the selection criteria in terms of geographical and
temporal characteristics are defined in the inclusion criteria, who has to be included in my study. So, it
depends upon these characteristics.
Now, what is that external exclusion criteria. So, in exclusion criteria we are deciding in our study that who
are those subjects, which we are not going to include because they might interfere in the success of the
follow-up efforts or they might interfere with the quality of the data and they may interfere with the
acceptability of the study or even there will be interfering with the ethical concerns. So, those study subjects
or those groups of people with those characteristics, we do not want to include in our study is defined as
that exclusion criteria.
So the step as such is that; first, we need to define a specific inclusion criteria based on the specific inclusion
criteria, what you derive is a accessible population and by means of an exclusion criteria you are neglecting
or you are avoiding to include certain individuals, so that what we are trying to derive is a subset of your
accessible population called study samples, with an estimated sample size in it.
(Refer Slide Time: 13:52)
So, more in that inclusion criteria and exclusion criteria, here I am giving an example what we have
discussed already. See, the study, designing a selection criteria say for a clinical trial of low dose metformin
to reduce dysmenorrhea in females with the polycystic ovary. Here, let us see what how we can fix up our
inclusion and exclusion criteria? So, the inclusion criteria, as we discussed already that the main
characteristics of the population those who have to be included in my study, which is relevant to the research
question and which will be very efficient to the study. So here demographic characteristics, I will include
females in the reproductive age group of 15 to 44 years; yes, it defines that particularly and the clinical
characteristics are those reproductive age group females with polycystic ovary and suffering from
dysmenorrhea.
And the geographical characteristics, patients attending the OPD of the hospital where the study is going
on in the particular region and temporal characteristics is that, between the period January 1 to December
31 of the specific year. So, temporal is that time. With this set of inclusion criteria, now in this study what
are the exclusion criteria I am fixing up, a subset of the population which should not be included, which
cannot be included in my study. See, when I am concerning with those subjects who will interfere with the
loss to follow up or chances of moving out form the study area may have a higher chance of getting drop
out, loss to follow up or some amount by means of marriage, those who are potential, in this particular study
period the loss to follow up.
Number two, interfering with the quality of the data, how? For example, here in this example patient who
are already in metformin therapy for some other cause may be due to diabetes. So, they may not be included
in my study and what is in terms of being on high risk of possible adverse effects with this kind of a design
feature, I will exclude those individuals who are hypersensitive to metformin therapy and contraindication
for metformin is called renal dysfunction. So, those individuals with the renal dysfunction will also be
excluded from this metformin therapy. So, that is how this gives you an idea about, in terms of selection
criteria, how to fix our inclusion criteria and how to fix our exclusion criteria in the selection of study
subjects as selection criteria.
With this understanding, what will be your source of the population, Clinical versus Community
Population?
Say like, clinical population where we will include those studies where we have a research question
involving the patients, but the previous example those are all the PCOS reproductive age group, females
with PCOS with dysmenorrhea. If it is so, then obviously, thus source of study population obviously will
be from hospitals or from the clinic based patients. So, it is preferred that best when we need to have a good
internal validity and as well as, so that to be generalize these findings to the population of interest it is
preferred with good quality of data the source of clinical population is from the primary health care clinics.
For example, when we are studying a true population; true population based sample which are of a huge
geographical area, a community based populations and which will be a very good source for healthy
subjects, when we are trying to study over a healthy subjects, over a healthy individuals. For example,
Vaccine - efficacy studies obviously, it has to be from the community based studies.
So, in this term, what is a problem here in terms of true population based study is that, it is very difficult.
House to house of enumeration have a more difficulty of including them and getting the study subjects
within the specific time period is another difficulty and it is very expensive to recruit also. But, in terms of
deriving a good public health and clinical decisions among the community then obviously, population based
sample has to be done, which I have given the example called vaccine efficacy studies. So, what will be
your source of population? It depends purely upon your research question.
So, let us see, what are the recruitment strategies? And what has to be considered when we are doing the
recruitment of study subjects in your study? So, what is that recruitment basically? Most important factor
whenever we are going to recruit the study subject is that feasibility, because feasibility is an important
factor when we are to consider to choose those accessible population and when we are trying to do a
sampling procedure. Feasibility decides both, what is our sampling procedure? And how we are going to
choose this accessible population from the target population? Or how we are going to choose this subset of
samples from this accessible population when the feasibility is an important factor?
And, there are two important goals in term of recruitment, what are the goals of the recruitment? Number
one goal is that the subject should be adequately representing the target population, because you are trying
to select a subset of a target population by means of geographical and temporal characteristics with
predefined clinical and demographic characteristics. So, these subjects should adequately represent it,
whatever the kind of findings you are trying to infer and apply is over this target population. And second
important goal is that there should be enough subject to meet the sample size requirement because
previously we discussed that it should have an adequate size to counter the errors the random error generated
in the study so that adequate size should be there, that are the 2 important goals.
Let us see that one by one, how to achieve a representative sample? Achieving a representative sample, it
happens at the beginning of the design phase and ends at the implementation phase. How it happens in the
design phase? And the design phase itself, we will decide, what is the choosing the population accessible
population from the target population and choosing those sampling methods wisely from the target
population can happen at the designing phase itself, Whereas, in implementation phase how it ends?
Number one is that, when you are trying to guard the errors by applying certain criteria over a period of
time; that means you have these selection criteria.
By means of selection criteria what you can do in the implementation phase is that, you can achieve this
representative sample and you can exclude those individuals, who cannot give a consistent or representative
finding to the study. And second is that, unit monitor this entire study; monitor out this study subjects are
adhere to this entire period, there should not be any loss to follow up and if there is a loss to follow up how
it has to be guarded? And, that is what we are trying to discuss in the next slide.
So, as an overall what you can see is that, when you are achieving a representative sample which is begin
at the design phase and it ends in the implementation phase.
But, how we can able to tackle these non-responses in the study? Then say, at the initial instance where you
could able to get the adequate size and the person is not responding or not available, then repeated contact
attempts may give you or may give you an access to the recruit them again into your study. And second is
that, design that avoids discomfort to the participants by after they got recruited in between, if they have a
loss to follow-up how to tackle this, that.
You can play some kind incentive mechanism, you can take away certain discomforts like, there should not
be too sensitive an instrument should be, which should not be invasive. So, you can try to not disturbing in
your other technique in the study or methodology in the study, you can still think of designing certain study
instruments which has to be in the local language of them, which it has to be understandable and those
interviewers should talk in their local language to the participant. So, the bilingual staff with the translated
questionnaire may reduce the discomfort of the study subjects.
So, by these mind of a mechanisms, it is where we could able to tackle of recruiting this adequate samples
at the beginning itself and as well as where we could able to tackle to stop them loss to follow up while the
study is going on. And finally, I am trying to give a summary of how this selection of study population in
a step by manner can happen.
(Refer Slide Time: 24:11)
So, in the first step what happens is that, we define a target population by a specific set of inclusion criteria
by means of a clinical demography, geographical and temporal characteristics. Second is that, this
accessible population is again by specific set of inclusion criteria as said by geographical and temporal
characteristics. Third step is that, what we are deriving is a subset of this accessible population, which we
are excluding by setting up some kind of exclusion criteria because we are eliminating the subjects which
are unethical and also inappropriate to study.
Fourth step is that, we are doing a sampling procedure here, so by defining the sampling technique and by
then estimated sample size which is large enough to control the random error will be done in the fourth
step. And finally in the fifth step, is a recruitment strategy where we are trying to recruit those adequate
subjects and as well as those subjects by those recruitment strategies to reduce the non response rate and
as well as last to follow-up. So, that is how this is where the important considerations has to be think of
about selecting the study population for your health research.
Thank you.
Hello. In our course of Basic Course in Biomedical Research, today we are going to discuss Study plan and
project management. It is really critical that lot of energy and time is spent by all who are involved in the
project or in the research study plan development in planning it really well.
Because, only a well planned study succeeds. Most of the times if we have not thought of all the
eventualities, will land up in certain situations where there are difficulties which are faced while
implementing the study and also while interpreting the study. And today, what we are trying to discuss is
what kind of systematic process or approach can be taken to ensure that the project is implemented
appropriately or the study is implemented appropriately.
What is important in project management is to ensure that the defined objectives are adequately met
and there are certain deliverables, which are defined right at the beginning of the study that this is what
we want to achieve, which is our definition of objectives, and so they should be reached within the defined
time frame and with the available budget not compromising the quality at all. So, all this is achieved
through effective project management. Primarily, because what we do is the project if it ends successfully
the result should be able to provide direction for future implications, basically what it means is whatever
research we do it should help us to do something better tomorrow.
It is important that we understand that any kind of implementation process involves some kind of
underlying principle, with a primary objective of achieving a specific goal. Let me take an example of the
process of resource allocation and resource management. Well, the principle here is appropriate time
management. The resources have to be allocated in a timely manner, if it is going to be a long term study
it will be on multiple occasions that the resources will have to be mobilized. So, that timing really becomes
very critical because only then we are able to achieve the goal of efficient, we can progress efficiently
towards achievement of our goals.
Another critical aspect in the process is planning and scheduling the activities. This is the important
principle which undergoes behind that and to ensure that, this happens is the monitoring and supervision.
Every single detail has to be planned out appropriately so that when we reach the goal, we reach it with
the best possible quality standard and hence it is important to keep this whole process in mind and the
principles in mind.
(Refer Slide Time: 03:03)
Ad hoc decisions which are taken or ad hoc approach which is taken while conducting research study often
becomes non-productive. So, there can be lot of confusion at the beginning of the study in the minds of
people. So, like I want do a study, but I am not clear about what the objectives are. I have prepared a
questionnaire, but I am not clear about exactly what information I need. I also feel that I will be able to
collect the data, but I am not clear how will I use it. All these kinds of confusions arise because the
investigator himself has not understood what the research is all about.
What is going to happen as a result of this? This is going to result in a disaster, because this will be a
situation which will lead to production of data or compilation of data which is difficult to analyze or then
the analyzed data becomes difficult to interpret and even if the interpretations are made probably they
are of no use to the program or also in the policy making. So, eventually what it means is any kind of an
ad hoc approach, ad hoc approach which is taken without proper thought being given to that is not likely
to succeed.
(Refer Slide Time: 04:21)
Any research process typically starts with identifying the need for that particular research, then correctly
verbalizing the research question or spelling out the research question, formulating the study objectives,
planning the analysis, then preparing data collection instruments, then collecting data, analyzing data,
drawing appropriate conclusions, making the specific recommendations to the concerned people and
eventually again accessing, whether our needs that we had initially identified have been fulfilled or not or
whether there is any need to do anything else.
If we look at this whole process, in terms of identifying data needs and spelling out the question this all is
the planning stage or the initial stage even while the study is being conceptualized. This is a pre-planning
phase. The steps of formulating the objectives and the analysis plan and deciding about the study
instrument methodology, where the way in which data will be analyzed and then the way it will be
interpreted, this all is a part of analysis plan. And, what we do after that is the dissemination of these
findings to the concerned stake holders so that they can use it for appropriate programmatic absorption
or policy making.
(Refer Slide Time: 05:48)
So, basically the road map to study planning and management involves multiple steps. It all starts with
formulating appropriate objective for the study, then choosing the right design to determine the key
indicators. Please understand, I am going to walk you through this particular thing but proper decision
about what kind of study is required to answer the objectives that we have framed is very critical and
important one.
(Refer Slide Time: 06:37)
We have to also identify the parameters that are needed to estimate the indicators that we find, that we
have decided, that are important for this particular study and then prepare the analysis out line. Also,
important to estimate the sample size before the study is initiated, because the study conducted on a
small sample also will be not generalizable. When we talk about the study objectives, the basic principle
is fewer the better. Most of the studies with a long list of objectives often become very confused studies
because many of these objectives remain unfulfilled, because they become complex, the data collection
tools increase, there are lots of variations that come in while collecting the data and in general there is a
chaos. So, fewer the better is the principle.
They can be described. Objectives can be described as primary objectives and secondary objectives. It is
important, the primary objective is important because that generally decides the sample size for that
particular study. Often in any study, sample size is calculated based on the presumption that we should
be able to achieve the primary end point at least. Secondary end points are the analyzable issues, which
are the additional information, pieces of information that we obtain in any research study. But it is
important that the objectives are clearly phrased.
Normally, they could be of 2 types, they could be more of a exploratory type or what we call it as aimed
at testing a hypothesis and here is, where we use the word determine. So, determine whether a
contaminated well caused an outbreak that is an example. Or they could be say sort of confirmatory in
nature or estimating in nature, say to actually decide the prevalence of a particular condition, for example,
diabetes in a population. So, we have to keep in mind and use the appropriate verbs while defining the
objectives.
There are different types of study designs, which are adopted to answer various research questions.
Whenever we talk about or whenever we think about descriptive objectives, whenever we are exploring
the acute conditions like for example, pneumonia case is occurring in children, the right kind of designs or
diarrheas occurring in children and etiology.
The right kind of designs would be to do cohort studies or surveillance studies which are may be hospital
based surveillance or community based surveillance. And normally, the major that you derive out of it is,
incidence; this is true in case of acute conditions. But when we talk about chronic conditions, once that
conditions occurs it persist for a long time either on treatment or not on treatment. So, the major that we
normally get out of such kinds of studies is prevalence. And here, the right kind of designs to be used are
either the cross-sectional studies or the cohort studies.
Well, many of the epidemiological studies deal with comparing 2 groups say for example, people who are
exposed to a particular condition not exposed to a particular condition, people suffering from a particular
disease not suffering from a particular disease, and so whenever we are moving from the variable or the
exposure to the outcome, we call it as cohort study. As we know it is a prospective assessment and
whenever we try to look at the exposure after the outcome has already developed it is called as a
retrospective approach and often the commonest study design which is employed is a case control study
here.
But, what is important to understand and remember here is a cohort study can be undertaken only when
the outcome is more likely to occur frequently, because if it is going to take a long time to happen then
probably the study will be of enormously long period and the adequate number of outcomes may not be
achieved. So, for a frequently occurring outcome which, wherein from exposure to the outcome the
length is likely to be minimal, cohort study is a good approach to take or alternatively a cross sectional
analytical study can be under taken; but in case of rare exposures and where the duration between the
exposure and the outcome is likely to be very long, then it is better to go for a case control approach and
in these situations the relative risk and odds ratio. In case of cohort studies, it is the relative risk which we
obtain which is a more definitive say indicator of relationship and odds ratio is also is a strong indicator of
association.
But the basic principles that we have to follow while we collect the information elements is that, we must
use the variables which will be actually analyzable, this information we can obtain by reviewing the
literature fairly scrupulously because it provides us a lot of evidence of which variables co-variates are
important. It is important to also use validated or standardized methods because then the chances that
this particular study will be accepted globally are maximum. We must adopt standardized case definition,
for example, when we are going to talk about pneumonia; what is pneumonia? We should define it
properly.
If we are going to talk about smoking, this as an exposure variable, what is that we are going to consider
as smoking? Is it the frequency of smoking? Is it, yes or no? Or the number of cigarettes smoked per day?
We have to have clarity on these matters. Sometimes, we also use laboratory criteria and so we have to
have also well defined definitions there, for example, if you have to define anemia; how do we define
anemia here? Does it depend on age? Does it depend on gender? All these have to be specified right in
the beginning; we have to then decide which is the most reliable and accurate way of collecting that
information. Sometime it could be just the observation or it is a questionnaire through which we collect
this information or it could be actually laboratory assay through which we get this information.
For example, if we are going to talk about as an outcome, whether there is an evidence of chronic iodine
deficiency. The way to look at it is, we would look at what is the say prevalence of goitre in a specified
community, and how we will do it is by actually doing physical examination. But if our objective is to find
out what is the current exposure to the iodine here? What we would try to do is, at an individual level try
to estimate the urine iodine excretion and for this we would require some laboratory methods to actually
estimate this.
But sometimes, it also becomes important to go one step behind and find out what are the dietary
patterns? Is there an adequate iodine being provided through the diet? And so, what one would want to
do is to test the household salt for iodine and this would involve some kind of field level spot test which
are done to figure out, whether the salt which is consumed in the various households actually has enough
iodine in them.
I did mention about the risk factors, about the confounding factors. There are certain risk factors like
income and the community which are related to access for example, or the level of literacy, the practices,
cultural and social which are observed by the community, the dietary patterns all these also influence the
outcome. Then they also have to be appropriately analyzed when we do the interpretation of our results.
Similarly, with age the risks sometimes vary with gender, the risks vary the residence, they also vary. They
are considered as confounding factors, if they affect both the exposure as well as the outcome variable.
There is no harm, even if the confounders are there, provided we have collected information on all of the
confounders as a part of our study and in the questionnaire. We can always analyze the effect of the
confounders.
(Refer Slide Time: 16:06)
It is important to make the analysis plan because it helps to focus on the objectives of the study. This all
thing can start by once, if you have clarity in our mind, what study we are doing. We can also prepare
dummy tables right in the beginning of the study because then we know what we must do and what we
must not do. We also know what data we should collect and what data we no need to collect because it
saves time. It can result in to quick publication and quick dissemination of findings and early policy
feedback. So, this is important to make a good analytical plan.
(Refer Slide Time: 16:44)
Sample size is really critical because it is decided by what exactly is the type of outcome assessment that
we are doing; whether it is by measurement or by testing. What kind of study design we are doing?
Whether it is a cohort study, case control study or a survey and whether it is a descriptive study or an
analytical study. This is in itself is a sort of a big lecture point and so what is important to understand is
when a study is being planned, it is important to involve a statistician, who would help you to analyze,
help you to determine the sample size for the study.
(Refer Slide Time: 17:30)
Often it happens that the studies fail, why do the studies fail? It is because, either they are badly defined
research questions or objectives are not correctly defined. The time skills that have been decided either
are too short or too long. Sometimes, the staff is inappropriate, incompetent. This might be because of
lack of correct direction, lack of motivation or lack of training. So, all these have to be taken care of. Or
sometimes, it can also result from everything else is right, but there is no adequate monitoring and there
is a failure to respond to contingency situations and carry out mid-course corrections that is where
monitoring and supervision becomes important.
(Refer Slide Time: 18:12)
So, for success of any kind of a study there are certain attention points one has to look at. And they start
with human resource management, a very critical aspect. The study staff has to be carefully chosen,
appropriately trained and with appropriate communication there should be a good dialogue. Only,
generally the observation is a team succeeds, but the individuals fail. And so, between the various
members of the team, there should be good bonding, extremely good communication and the leader has
to ensure that this often team meetings take place and this rapport between individuals builds strongly.
Time management essentially is the responsibility of the leader, and one has to ensure that this is taken
due care of. Time management in terms of appropriately scheduling various activities ensuring that they
are done in time, this is really important. Financial management is also critical. Sometimes it so happens
that the study starts well with the funding being given, but suddenly some kind of glitch develops by which
the finances are not being granted or given in a continuous manner, suddenly the activities of the project
stop. And hence, this also is an important part of the planning. This has to be planned well in advance at
what stage what kind of money will be released and it must be ensured by the researchers that targets
which are defined well in advance are appropriately met. So, the financier does not find it difficult to
release the money what is earlier decided on.
Quality management at all levels is critical. Quality management in data collection, various clinical
procedures, data management, various laboratory procedures, in supervisory visits, every single aspect of
a study that we can think of quality is really critical and if that is maintained then often the studies are
very successful.
Data management is important. Often the studies that take care of data management in a timely manner,
where concurrent data management is planned they are able to give away the results in a timely manner.
If the researchers have not planned it well and then they decide to do the data management at the end
of this study, often it is disaster because if there are some issues that are happening in the way the data
is being collected, if the data is being managed in a timely manner, somebody is looking at it, finding the
faults in it, there is a possibility to make a change, do necessary corrections this opportunity gets lost if
we are handling this whole issue at the end. I did talk about team work and coordination, which is really
critical and important.
And monitoring the progress and target, here it is to be decided depending on the budget whether it is an
internal mechanism that is set up for the supervision and monitoring of the study or an external monitor
is brought in to take care of what is happening in this particular study. But, this is again a very important
step. Whether it is a research project or it is any other project, there are various aspects that we have to
think about; it is a teamwork, it is about communication, it is about human resource management, it is
about time management, all these factors are important and we succeed as a team.
Lecture - 15
Designing data collection tools
Hello and welcome to this session of Basic Course in Biomedical Research. Today, we are going to talk
about designing data collection tools or the instruments that we use to collect data in health research.
(Refer Slide Time: 00:22)
Generally, in health research, the kind of information that we would like to collect can be basically divided
into three areas. We want to know facts such as, the characteristics of our study participants, the
environment that they live in and their behaviors or practices. Secondly, we might want to know their level
of knowledge for things such as, risk factors for getting disease or knowledge about healthy lifestyle, so as
to prevent diseases and thirdly and very importantly we might want to collect information on what we call
the domain of judgments, basically what are the opinions of the research participants on a certain issue,
may be such as quality of health services.
We would also may, we may want to know about the respondent’s attitudes towards certain things for
example, could be something as wearing seat belts, use of open defecation and so forth. Now, how do we
collect data on all these different kinds of information that we want to collect.
(Refer Slide Time: 01:31)
For this purpose, we can make use of a variety of tools depending on what kind of data we would like to
collect. One, we have what are called abstraction forms, which is basically doing a review of records of the
study participants, which could be their personal records, the forms. If you want to get information on their
disease conditions, signs and symptoms, a treatment given, then we could look at the clinical records. We
could look at data in general, the data that is collected through disease surveillance and we could also look
at registers, wherein some information, which may be there and all of this information can be culled out
into what we want, in the form of a data abstraction form.
Secondly, another tool that we have for collecting data is a structured observation guide. This is very useful
when we would like to document certain processes, whether they are happening, the way they are
happening, that they are happening in time or not, are the objectives met and for this purpose we may use
a check list of items that we would like to collect data on which could be the textual or figurative and third
and most importantly, the tool that is used most commonly is a questionnaire, wherein we would like, we
talk to the person and get information.
Now, again questionnaires can be divided into two kinds - it could be interviewer administered, where the
data collector actually administers the questions to the respondents, which could be done either face to face,
it could be done on a telephone and now even have computer-assisted technologies to do face to face
interviews. The questionnaires could also be self administered, if the study participants can read and write
and they are knowledgeable enough to understand what the investigators want. These could again be either
paper based or now we also have computer assisted self administered questionnaire which can help the
respondents to directly put their information into a computer database.
Now, whatever the tool of data collection we use, what we want is that we need a valid response from the
study participant. The response should make sense and we should be able to use that information effectively
for our health research needs. In order to do that, every data collection tool needs to make sure that there
are some key elements that make up this data collection tool. Some of these elements are the clarity of
words that you use in a data collection tool, the balance of phrases and sentences, the length of sentences;
how long are these questions, the comprehensiveness of the responses in terms, if you are giving categories
of responses that you expect from the study participants.
These categories can actually even post constraints in terms of what information could or could not be
collected through a data collection tool. Of course the utility of the data collection tool and specially the
utility of the instructions; which I will elaborate in the next slides. The order in which the questions are
asked is a very important element which can decide how your respondents answer your questions and of
course, the context in which you frame these questions and the tools are used.
(Refer Slide Time: 05:18)
So, if we look at any data collection tool generically it has 4 components, there is the introduction part at
the beginning and a conclusion at the end. There are what are called identifiers, then the each question may
have linked instructions for the data collectors and of course, the whole body of the instruments which is
basically the question items. In terms of the introduction, the introduction is used to present the study to the
study participants, state out the objectives and probably get informed consent and so forth. It is also good
to always have a concluding statement at the end of your questionnaire and thank the study participant for
their time and effort that they have put in to answer your questions.
Now, every data collection tool will also have what we called identifiers, which could be either the actual
identifiers, the information such as the name and address of the study participant, which can identify who
the person is, it is always a good idea from an ethical and a human subject protection perspective, if you
are collecting this data, to collect this in a separate sheet of paper and which can be referred to later on, if
need be.
On the other hand, in order to maintain confidentiality investigators also use coded id numbers to give
identifiers to the study participants and these id numbers could be composite in such that they could have
numbers which denote say, the state which the person belongs to the district the village and then the
household and then the individual id. So, it could be a mix of all these numbers, all these codes and then
you get a composite code looking at which you can actually identify who the respondent is, but would not
be able to actually get an exact identification and which is good from the confidentiality perspective.
In terms of the instructions, there could be general instructions for the data collectors such as prompts in
terms of, for example, do not read out all the responses, tick only the one that the study participant mentions.
There could be instructions for skip patterns, now the questionnaires may have skip patterns in the sense
that there may be some questions which, based on the response to those, the subsequent question may not
be relevant and then there is an instruction, which says that you skip this from question 2, you may go to
say question 19 and skip rest of the questions because they are not relevant to this study participant.
It is also always a good idea to may be used different fonts, so as to make it clear that what is the actual
question and which part of the item is actually an instructions for the data collector and then of course, we
have the whole body of the instrument which is basically the question items. Now, these question items
could be of various types, we could have what are called open questions, we could have closed ended
questions and we could have somewhere, something in the middle.
Of course, when you generate a lot of responses, open questions, the inconvenience is that it may be difficult
to code and analyse. You may have a long list of responses and then to categorize them later may be an
issue and sometimes, if it is open, the responses may be unfocused or incomplete and that can pose a
challenge in terms of the analysis. Now, to overcome this problem, we can have open questions, but then
we can have then with closed ended answers. Although, there is a category of answers given for those that
question, but the data collector does not suggest an answer from these categories to the study participant.
So, when the answers are freely mentioned by the respondent the interviewer will spontaneously tick those
that are specified from the list of categories of responses given in the question here. So, it is expressed as
an open question, but you finally analyse this as a closed ended question.
Then we have Closed questions with multiple options, so more than two options. Now here again, they can
be two kinds of Closed questions with multiple options. We could have questions where although there are
multiple options, only one of the option is acceptable. So, depending on what the respondent says, one of
this option is ticked by the data collector. On the other hand, we could have closed ended questions with
multiple options, wherein even multiple responses by the study participants may be acceptable.
The important thing to note is that, while you are designing the questionnaire you need to put a clear
instruction for the data collector, whether only one answer is acceptable for this question or more than one
answer may be acceptable. So, we have a large choice of answer options; again depending on how the
question is framed and what the actual question is, sometimes it may become inconvenient and difficult to
choose only one option, if there is a possibility of more than one option, but that possibility is not provided
in the questionnaire.
So, we have to be mindful of this when you are designing your questionnaires. And thirdly, we could have
Closed questions which have quantitative answers, where the respondent has to provide a number, such as
age, such as may be one example here, if you see; how many times did you visit the clinic in the last 12
months? These kind of questions allows the creation of continuous variables and then measuring and doing
the analysis for continuous variables. If you need we can always categorize these variables later on in the
time of analysis if needed.
However, sometimes it can become inconvenient to give a quantitative answer because some quantified
answers may be limited in the way they can be handled as continuous variables and where the number itself
is difficult to interpret. So, we need to be careful in how we are framing these kinds of questions.
(Refer Slide Time: 13:33)
Then there we could have something called, which are called semi-open questions, where you basically
have a question with several responses, the answers here are suggested, but there may be one option where,
which is kept open and the most common one that we see in data collection tools and questionnaires is
others. For example, did your child have complication following measles? There could not have been any
complication, that could be pneumonia, diarrhea, eye problems or there could be some other complications,
which may not be so common so as to be put in a category, but then you give an option to the respondent
to even say things other than what is in the list of categories of responses. So, it leaves the door open for
unplanned answers; however, if there are too many of these responses, it may sometimes become difficult
to analyze.
(Refer Slide Time: 14:34)
Now, let us look at, how some of the principles and do’s and don’ts of formulating the questions. It is
always the good idea to actually have short and precise questions, say for example, if you want to know the
age of your study participant, just writing age is not a good idea. You should always use full and complete
phrases, so, what is your age? So avoiding ambiguities. It is a good idea to use simple words and not use
very complicated academic language, use everyday language in terms of questions because again remember
that your are respondents are lay people.
When you are formulating questions, again it is good idea to avoid negatives; especially double negatives.
So, one example that we have here is, do you sometimes care for patients without washing hands. If you
see carefully, there is a negative connotation here, and there are sort of two parts of the questions, one is
caring for patients and the other one is washing hands or without washing hands. So, a better way to phrase
this question could be to ask it directly and more positively, so, do you systematically wash hands before
caring for each patient?, which makes it clear and unambiguous.
(Refer Slide Time: 15:57)
Again when you ask a question, it should be only one question at a time; say one example here that we have
here, so, did you refuse treatment because you feared side effects?. Now, here actually we have two
questions; one is asking: did you refuse treatment? And the other is trying to find out the reason of why, if
the person refuse treatment, why did they do so. It may be that the respondent may not have refused
treatment then how does that respondent answer this question. So, it is a good idea to actually split such
questions into two questions, where in one could be first; the first part say could be, did you refuse
treatment? Depending on yes or no, if the answer is yes, the following question could be; was this because
you feared side effects? It makes things very clear.
Again, the questions need to be specific and not vague, so an example that we have here is say, you want
to know from the people about how HIV is transmitted and the question is: Are you aware of the modes of
transmission of HIV? which is sort of an open ended, it leaves space for people to answer whatever they
may want to answer. But if you really want to know whether HIV is transmitted through sexual route,
through heterosexual route, homosexual route, blood transfusion, drug use, etcetera, then it is better to
actually put these as categories of responses and then phrase the question as, among these practices can you
tell me those that could expose you to HIV.
So, you know that you are going to get the proper answer to the question in which you wanted to have. It
is also a good idea to always use a neutral tone and avoid judgmental tones, which can influence the
response of the study participants. Remember that you are there as a data collector to just collect data and
not be judgmental of what the respondent is telling you. One example that we have here, so if you want to
know about say the sexual practices of people, instead of asking them have you been promiscuous in the
last 6 months? Again, the words promiscuous has a negative connotation , so instead of that it could be
more neutral and more academic kind of a question, wherein you just be direct and specific and ask about:
How many partners have you had in the last 6 months? without being too judgmental.
Now,
the next thing that you would want to do, when you are designing your questionnaire is, to actually sort
out what would be the order of the questions that you have. Remember that, the way in which you ask
questions should be such as if your having a conversation with a study participant and it should be a
smooth flow of questions one after the other linked to one another.
Some general principles to keep in mind is that always ask simple questions first and keep the complicated
questions for the later part. You can ask more general questions, socio demographic characteristics, like
that, and then go on to the specific questions of what your study is all about. Ask more casual questions in
the beginning, which the respondent will be easy to answer; more of factual questions and then more
intimate questions, sensitive issues, questions about attitudes and opinions could be something that could
come later on.
It is always again a good idea, remember, to actually group together all the questions which are related to
the same topic of inquiry instead of having them spread across different parts of the questionnaire, which
can actually confuse the respondent. In terms of asking the identification questions either they can all be
asked at the end in terms of the name, age, gender, say the address and so forth or they could even be asked
at the end so as to get to the specifics right at the beginning.
If you are trying to collect information about a sequence of events, then your questions should be in that
proper chronological order of how things would have happened in real time, which will actually help the
respondent to recall the responses in a better way and also in a more logical way. Again, if your
questionnaire is complex, there are lots of questions, it is always good to give a break in the middle and
may be have some simple questions and then come back to your complex questions. Many a times, we may
ask the same kind of question in different ways in the same questionnaire and then that is used, if that is the
subject matter that is really important for the study and you really want to know what the respondent is
telling you make sense is valid is true and then. So, the multiple questions on the same topic could be asked
at different places in the questionnaire and then when you are trying to analyze it, you can triangulate these
responses to get to what information you would like to extract from this questions.
Once you have sorted out the order, now what is needed is to actually lay out all these questions in the
questionnaire. Again remember, laying out the format, the structure of the questionnaire is again critical
because the way the respondents look at the questionnaire, the design and look of the questionnaire
influences or can influence the response of the participants. So, if you have different sections in the
questionnaire, it is a good idea to split the sections, have one section, may be have line and then have the
next section.
Do not try to cramp questions all together, have spaces between the questions, so that it is readable clearly.
Try to use large fonts, not too small fonts so may be font size of 11 or 12 would be ideal. Again, do not
split questions across pages. If you have a question and then if half of it goes to the next page, it becomes
difficult for the data collector to actually read the question, you will have to turn the page and so forth. So,
if that is happening, make sure that you sort of bring the whole question on one page. In terms of formatting
and aligning, alignments are again, gives you a nice, neat look for the questionnaire. So, it is a good idea to
actually align your questions on the left hand side and your answers and codes on the right hand side, which
gives a neat two column kind of a look to your questionnaire, which makes it more appealing.
Do not forget to number your questions starting from 1, 2 and whatever that you have. Again, one very
important thing to keep in mind is coding. Remember that ultimately what you are going to do is use this
data and to enter it into software, give it codes to analyze it. So, it is always a good idea, to actually have a
coding system inbuilt in the questionnaire itself. So, you need to standardize your coding. So wherever say,
the simplest example I could give you is, a yes-no. So, you could have a code of 1 for yes and 0 for no, for
example. Make sure that every question, every response where you have yes, no; you have coded it as 1, 0.
Say you have male, female. So, you always code the male may be as one and the female as two or something
like that.
Another way to actually simplify this coding is to use what we call as auto-coding. So, the numbers that
you give to the categories, say you have four response categories. So, you would number them 1, 2, 3, 4
and if say the response is number 2, then you use the same number 2, as a code for this question item. So,
these are some of the ways in which you make sure that the layout of your questionnaire is neat and it is
presentable and it something that helps both the data collector as well as the data entry operator and the
person who is going to enter and analyze the data at the later stage.
(Refer Slide Time: 24:44)
When you are finalizing your data collection tool, make sure that the questions that you have are something
that is relevant to the study that you are doing. So, as may have been mentioned in earlier sessions, the
investigator needs to be a slave of the study objectives and the analysis that is already pre-planned for the
study. So, make sure that the questions that you have are relevant to answer those study objectives. Do not
put unnecessary questions just because you are going out in the field and doing a study does not mean that
you can ask anything and everything and if you feel that there are certain missing questions make sure to
add them.
Once you have done all this, it is also a good practice to actually review your instrument before you take it
to the field. The reviews could be done by your colleagues and experts in the field, you could also give it
to the statisticians, to actually review look at the coding; whether that is going to be something that is going
to be useful for them and then you could even the field workers or the data collectors, who are going to
actually collect the data, they can be your key informants to actually go through the questionnaire and tell
you whether the flow is appropriate, whether the questions make sense, is there any ambiguity or is there
something that is not understood and so forth.
Keep in mind that the language of the instrument; the questionnaire or the data abstraction form or so forth
has to be in the language in which you are going to interview the study participants. So, if your study
participants speak Tamil, then the questionnaire should be in Tamil, if it is Hindi, it should be in Hindi.
Generally, as investigators, English is the common language, so may be your initial formulation of the
questionnaire would be in English. Then what you need is a translation, you need to translate it into the
local language and then very importantly have somebody else do a back translation into English, so as to
make sure that the translated version in the local language makes the same sense as you wanted it to be and
when you frame those questions in English.
Before going to the field it is always important to pilot test your tools. You need to make sure that your
study instrument is clear, the questions that you have asked are understandable to the people and they are
acceptable, people are not wary of answering those questions. You need to check the flow and the skip
patterns. Make sure that the coding works and it also gives you a sense of how much time it is going to take
for you to actually finish the questionnaire. All this can be done by doing a pilot testing by actually
administering this questionnaire to a few volunteers who are similar to the study population that you are
going to do, but remember that these people on which you pilot test your questionnaire, should not be
included later on in the main study.
(Refer Slide Time: 28:57)
So, when we are designing the health research tools, we need to keep certain principles in mind. You need
to first make sure, what is it that you want to measure, remember epidemiology is all about measuring.
Then you need to relate these concepts to your study designs and the study objectives. You need to match
the scales, how you are going to measure these to and then how you are going to do the analysis. Make sure
that the scales, the questions, the questionnaires, the data collection tools that you are using are reliable and
valid for the population that you are going to apply them to. Taking all these things in mind, choose the
most appropriate method of data collection; whether it is a data abstraction form or a structured observation
guide or a questionnaire and the type of questions that you are going to put in these data collection
instruments.
Keep in mind your study participants, in terms of the language of the questionnaire and also the way in
which you are trying to measure the concepts that you are doing and then decide finally, how best you are
going to ask the actual question in the study questionnaire. Remember, a study questionnaire can make or
break the study; this is something, once you have collected data you may not have the opportunity to go
back. So, it is essential that the data that you collect is valid and reliable and in order to do that, it is key
that the data collection instruments that you develop are totally valid and appropriate to the study that you
are trying to conduct. That is it for today.
Thank you.
Basic Course in Biomedical Research
Dr. Prabhdeep Kaur
Lecture - 16
Principles of data collection
Welcome to this session of Basic Course in Biomedical Research. Today, we will be discussing about
Principles of data collection. As you all might have gone through various parts of this course, wherein we
had been teaching you about, how you develop your research proposal? How you write your protocol? And
now you have gone through all that and you are all set to collect your data. Let me tell you that, this is one
of the most important components of your research study because this is what is going to determine, what
you get out of your data.
(Refer Slide Time: 00:40)
So, today as we go through this lecture let me discuss about two main principles of data quality, which is
reliability and accuracy. So, reliability, what does it really mean? It means that your study should have
repeatability and precision. What it means? If this same study is repeated by different investigators or the
same measurement is done b at different time points, you should still be able to get more or less similar
results. It also refers to stability and consistency of the information. Now, reliability does not ensure that
your data is accurate. Accuracy refers to ability of a measurement to be correct. It could happen that both
the attributes may not happen at the same time or they may happen. So, let me give you an example of what
would be the various scenarios in terms of data quality.
(Refer Slide Time: 01:42)
As you can see in the slide, your study might give you reliable and accurate result, which means your data
is giving the measurement that you really want, the measurement is accurate and repeatable. This is an ideal
case scenario. The worst case scenario is that your measurement may not be reliable at all and may not be
accurate also. Now, there can be other scenarios, where your measurement is reliable, that is, if you repeat
this measurement again and again, you get the same measurement, same kind of result repeatedly. However,
this could still not be accurate and there can be another scenario that you may get the accurate result, but
when somebody tries to repeat this experiment you may not be able to repeat it at all. So, an ideal study
should ensure that whatever outcome you are measuring is repeatable as well as it is accurate, now to ensure
that, we can follow certain principles of data collection.
(Refer Slide Time: 02:44)
In today’s lecture, we will be going through six broad principles which are listed here. I will be walking
you through each of these principles. So, the first is, as you have your questionnaire ready and you are all
set to start your data collection. One more step you need to do is, you need to develop a guide, guide wherein
for each of the questions you can have a small explanation as to how this data actually need to be collected.
The second component that we will talk about is how you could do a good training of your staff members
to ensure that the data quality is good. The third component is how do you ensure data quality, when the
work has already been initiated. It could be in a clinic setting or it could be in the field and the next
component is how you need to do periodic reviews to ensure good quality of your data.
During this process, it may happen that your staff will come across various difficulties in data collection
and you need to have periodic debriefing for that and finally and very, very important component is how
you validate that your study results are actually correct. So, let us go over each of these components one by
one.
(Refer Slide Time: 03:58)
So, what do you really mean by a question by question guide? It is nothing but a short, simple document,
which is a guide for your field worker or your clinic research staff or whoever is involved in the data
collection. Now, why this is an important component? Everybody may not understand the question in the
same way. Different individuals or your staff members may have their own interpretations as to how this
questions need to be asked and how it need to be explained to the respondent. The q-by-q guide ensures
that all the investigators uniformly ask the question and they all understand the question in the same way.
What you do here? You just take your questionnaire and write a short explanation under each of your
question, as to how this question should be asked and what could be the possible responses and in certain
situations you may have some probes. You would like the investigator to probe little more on a particular
question. If you want them to do so, you would like to make a mention of that.
Similarly, it allows you to explain, where they should skip? Where they should give more emphasis? All
this guidance can be given through this small document. This not only provides a guide as to how data
should be collected, but it is your road map and at any point of time, whenever there is a doubt, whenever
there is a lack of consistency with in investigator you can always ask them to go back to this document to
ensure that the data collection is being done in a uniform way. Now, it could happen that as you start
collecting data, you may come across certain difficulties in the field and you may have to revise this guide,
which is all right and what is important is whatever changes are done, they should be documented in this
q-by-q guide at that point of time.
(Refer Slide Time: 05:52)
This is just to show you an example. For example, you have a question on, what is a type of house like we
could give several options, this is a kutcha house, it is semi pakka house or a pakka house. Everybody may
not understand, what is a kutcha house? So, you may want to give an explanation. Observe the house and
note, if it is made of mud or bricks and then you could write an explanation. If you find that the house is
made of mud then mark it as a kutcha house.
Similarly, another example when you are asking a question regarding household income? Now, it could
happen that in a particular house hold there is more than one member, who are earning members of the
family. And in this situation, you would like your investigators to first of all inquire, how many earning
members are there in the family? Then they need to find out each one of them, how much they earn? And
what is their monthly income? They need to add up all, that amount and then write the answer as what is a
total household income. So, similarly for each question you can have the detailed explanation.
(Refer Slide Time: 06:58)
Now, once your q-by-q guide is ready, you are all set to start your data collection. Before you get on the
ground, whether it is a clinic, whether it is a field base study, it is a household study, whatever will be the
type of study, training, good training is essential to ensure that the quality of data is good and in that, the
first and foremost step is choosing right kind of people. You need to understand, what your study
requirements are, and depending on the study requirements, you need to recruit or select right kind of
investigators. Now, this could be depending on, suppose you are doing a study, which is on clinical research.
Then you need to make sure that your investigators know how to ask the clinical questions. Are they
knowing, how to use the clinical terms, when they are interacting with the patient? If it is a field based
study, you need to make sure that your investigators are familiar with the local language. They know how
to culturally interact with the people at the community level. So, selecting the right kind of investigator,
who will be suitable for your data collection, is the most important step.
Once you have done that, you need to have a classroom session with them. This is nothing, but walking
them through your, first of all you introduce them to what your study is all about, why are you doing this
study? What are your objectives? And what is that you would like them to know the basics of, what are the
various definition you are using in the study? What are the various component of the study? What is the
kind of data that needs to be collected? For example, your study may have different component, there could
be a questionnaire, there could be different measurements and maybe you want them to measure something
like weight, height. So, you need to familiarize them, what are the various kinds of a data collection that
they need to do?
Once you have done that and given them the overview then you need to share this q-by-q guide, question
by question guide. Walk them through each question, allow them to ask questions and let them go through
it and see whether they are able to interpret those questions, whether they are able to understand those
questions. Once, they have done that, you need to tell them what exactly they need to do in the field? How
they need to ask those questions? And what should be the explanation that they should understand before
they ask the questions?
Now, as a first step having gone through this exercise, you would like them to probably do this interviews,
first of all in the classroom, wherein one of the investigator could act as the interviewer and the another
investigator could be the respondent and they should simulate this in front of you, which allows you to
guide them, to tell them, how the question need to be asked. Whether the question is being understood by
the respondent, whether question is being asked in the appropriate language that is given in the
questionnaire. Having done that you are all set to start your data collection, but there is one more step.
Now, having done something in the class room does not mean you will be able to do it on the ground
because the real situation may be very different. Some of your investigators may be working in that kind of
a situation for the first time. For example, you may hire a person who is trained in clinical research, but
they have never been in a clinical setting. Similarly, you may hire a field investigator, who is a graduate in
social work or any other background. However, they may never have been in the community setting. So,
once they get on to the ground, they may not feel the same level of comfort. So, before you get on to the
main data collection, you need to have pilot. A pilot, where they do the interviews exactly in a similar
setting, what your study setting is, but it may not be the same place, it may not be the same clinic or it may
not be the same village.
In this situation you will see that, they may come up with different kind of questions. They may not have
anticipated, how the respondent will respond. They may not have anticipated what difficulties might be
there in making the respondent understand the questions. So, as you through this pilot, you explain them
the process, clarify their doubts and this procedure you could continue unless everybody is very, very clear
about what the questions are and how the questions need to be asked.
Now, having gone through this, you are all set now to do your final data collection. Here, what is very
important is that, you need to plan your team structure. So, there has to be a supervisor, who could be
present all the time or who could be present periodically along with the investigators. Ensure that the study
forms are verified by supervisor every day. Now, why is this important because once you have left the clinic
or once you have left the village or the place where you are doing data collection, some of the errors cannot
be corrected, during, as the data collection starts they may come up with different queries. So, you need to
be available to answer those questions. This could be over the phone; this could be over messages or
whatever way you can use.
Now, once the data collection is ongoing, the next step that as a principle investigator you need to do is few
on site visits, to ensure that the data collection is done as per the protocol and the questionnaire is being
actually used in the field the way you had perceived it. I think one of the important thing you need to keep
in mind is very often what happens is there are time pressures. You want to finish your study quickly, you
want to be done with it, you may pressurize them, you may say, I would like you to do ten questionnaire
every day. However, you need to make sure that the quality is not diluted because you are pressing them to
work under very tight time lines. Having done this, having your data collection, as data collection goes on
at various time points you may want to review.
(Refer Slide Time: 12:52)
Suppose, let us say you have already collected now, 50 forms. At this stage, you may want to make sure
that the data collected is of good quality and the data is complete. How can you do that? The first step can
be done in the field itself. As your supervisor finishes the day, he can collect all the forms from various
team members and check them on that day itself. The next step is the forms will reach you as a principle
investigator and you need to go through them to make sure that you are satisfied with the quality of data.
(Refer Slide Time: 13:35)
Now, what kind of checks you can do, when you have got the forms in your hand? I think the first and
foremost and the easiest one, but the most important one is ensure there are no blanks. It could happen that
when they are asking multiple question, since your questionnaire may have about 20 questions or 25
questions, they may just forgot to ask a few questions or they may have asked the question, but they did not
mark it in the questionnaire, but in either case you will see them as a blank. So, you need to ensure
completeness of the data.
The second is readability, sometimes the way it is marked it may be hard for you to even understand and
some of the answers could have some actual explanation to be written. Some kind of clinical symptoms or
some kind of narratives have to be written, ensure that those things are readable. The next is the consistency,
does the answers make sense? Do you feel that this could be the way the people would have answered or
you got this answer because may be your respondents were not understanding the questions.
(Refer Slide Time: 14:41)
Periodic reviews are essential, in case of any study, whether it is a clinical study; it is a field study or any
type of research study. Now, these regular meetings, one it could be done by the supervisor, which could
be in the field itself or after coming back or you, can have a periodic review once in a week, once in a
month, depending on what is duration of your study. What is that these review meetings will be useful for?
First of all to clarify, if there are any queries they have about the questionnaires, if they understand the
questions well, if they came across something which they had not anticipated. Now, it may happen that
during these meetings you may end up doing some changes in the way your answers are drafted or in the
way questions are asked. If you are doing that ensure, that these are well documented and they are added it
in your q-by-q guide.
(Refer Slide Time: 15:36)
Now, coming to the last, but not the least very, very important component of data collection, which is
validation of your data. How can this be done? So, for any study a small sub sample, it could be as small as
even 5 percent should be selected and an independent second interview should be done. This is to ensure
that the data collected by your investigator is actually valid. By comparing the results, you will be able to
find out, if there are any discrepancies, if somebody has made any major errors or if it could be that a
particular investigator might have been repeating certain errors or it may happen certain errors are repetitive
across the team. So, depending on whatever the problem is you may want to discuss this with the individual
team member or with all the study team.
(Refer Slide Time: 16:27)
So, just to sum up, what we have gone through today. Before you get on to your data collection, understand
the concepts of data quality, good training both in the classroom as well as on site, in a similar setting where
you are going to conduct the study is essential for a good research study and supportive supervision, team
work are key to good quality data collection.
Lecture - 17
Data management
Hello friends. Welcome to this session, the course Basic Course in Biomedical Research. We are in the last
stages of our course. We have already covered important topics such as Importance of data collection,
Importance of validity, Importance of measurements and now I think we are in a position to understand
the Importance of the Principles of Data Management and Analysis.
(Refer Slide Time: 00:41)
Today, we are going to cover 2 areas; one, database management; the other one, data analysis strategy.
Data management includes the following; Defining variables, creating a study database and data
dictionary about which I am going to expand, entering the data and correct the errors, create a data set
for data analysis, backing up the data and archiving the data set.
(Refer Slide Time: 01:10)
Today, we are going to cover the key elements of data management, including Data structure, Data entry,
Individual and aggregated databases, Mother and daughter databases.
We need to initially formulate the entire plan of how the data is going to be managed. That can come in
the form of what is called data documentation. It can talk about the structure, meaning the name, number
of records and other relevant information about the structure. The variables in terms of the name, what
values that are assigned to the coding, etcetera and the history of this database in terms of when it was
created? When it was modified? The storage related information, in which media it is going to be stored?
Where and how it is going to be backed up? And any other relevant additional information is recorded in
terms of the structure.
(Refer Slide Time: 02:37)
Let us look at the first of the important elements in database, this is called Identifier. This identifier has to
be unique, that is why it is called unique identifier. It is maintained by a computerized index and this
particular unique identifier has to be secured by a quality assurance procedure that guarantees that each
of this data has its own internal validity.
(Refer Slide Time: 03:09)
The code can comprise information that will talk about that particular individual. For example, it can have
7 digits; each of this digit or set of digits can refer to a specific identifier information about that particular
individual, about which, about whom the data was collected, for example, in this example the first and
the second digit can denote the village or area, the next set of 2 digits the third and fourth may denote
street, the digit number 5 may indicate the house or you know flat or residence, door number, the last 2
digits can denote the persons sequential number. Therefore, the 7 digit may represent about that
particular individual and by parts it can give information.
(Refer Slide Time: 04:08)
You need to specify certain things about variables. The variable as you might have seen in the lecture on
measurement, there are different types of variables that do exist and therefore, it requires your attention
in the beginning itself. You need to specify whether the variable will be entered as digits or if it is a
numeric, whether the number of decimals are important, the variables can be entered as length, in which
case you need to specify the length and it is preferable that when you are entering as both text and
number, you turn all the letters into capitals to avoid errors, which can cause a lot of problem in data
analysis. And finally, you can have dates in specific format; when I say specific format you need to specify
whether it is entered as the Indian format; date, month and year in 4 digits or month, month, date and
then year; this has to be specified in the structure of the variable.
(Refer Slide Time: 05:19)
While creating the variable names, which pertain to your data collection instruments about which you will
have a clear idea later part of the course; you need to be very clear. The name, variable name should refer
to an item in your data collection instrument. It has to be understandable format for example, if the
questionnaire item is about exercising whether the individual exercises daily or not. The variable name
could be EXERDAILY; exercise daily that clearly denotes what questionnaire item it refers to. The second
important criteria is in terms of keeping it short, leaving no space between the letters of the variable
name. Most softwares may require less than 10 characters therefore, you have to be very choosy at the
same time it is self explanatory.
The third important aspect is, Be consistent for example, for different types of response to a question on
how frequently somebody exercises, exercising daily in the past can be denoted by EXERPAST, exercise
past that clearly know self explanatory about the particular questionnaire item. If it is currently daily then
it is coded accordingly, if it is past occasionally those words are given in the variable names so that by
looking at the variable name the investigator can easily identify; this is a variable, this is the questionnaire
item it refers to.
And finally, you may have variables collected as such these are called crude variables for example, it may
refer to number of times one exercises; it could be 3 times, 4 times, 5 times a day. And finally, you may
decide to regroup them into 2 categories; exercised or not, in which case you can denote that variable as
exercise in the crude variable. As such, when the data was collected you can change into EXERCISE
underscore 12, which denotes it is dichotomized. It is dichotomized into exercised or not. So, you have to
be consistent in the pattern by which you create variable names. And finally, it is very important that you
assign a variable name, otherwise if the software is left to assign a variable name by itself, it can create
lot of confusion including duplicates of similar items within the questionnaire.
When you are designing a data collection instrument, about which my colleague is going to expand in the
later part of this course; it is important that the design itself you are very clear about broad sections of
the questionnaire, so that when it is converted into database you know that there are sections that you
have to enter for example, there is a section called identifiers; there is a section called you know
demographics, which means you know one talks about age, gender, community and family related issues;
and then the third section called outcome. It is about the problem in the question and or disease if it is
related to typically clinical related information.
And finally, another section called exposure in which you can talk about all the variable that you are going
to measure including, what Dr. Tarun might have already talked to you about third factors including
confounders. Finally, the instrument should allow an auto-coding, if you collect information on exercise
daily; Yes - No; if it is written already written as 1 2, we need to enter into the database 1 or 2, we do not
have to code it again. So, that is what is meant by auto-coding. So, the data collection instruments should
be designed in a way that it facilitates a data entry design easy.
An important aspect of data entry is all about coding. It is always preferable to have numerical coding. Of
course, you would have seen, if it is textual information in the form of qualitative there is a different way
of dealing with it but with reference to this section we are talking about quantitative data analysis. So, it
is preferable to have numerical coding. In particular, you need to decide on how you will code missing
values; it could be in the form of a dot or a depending on the field, you may choose to enter as 999, triple
9, be careful you do not enter you know a missing value for age is 99; that can mean differently. And, if it
is not applicable you enter with a consistently with the particularly coding in the data collection
instrument. For beginners, with inadequate experience in handling databases it is advisable do not create
cumbersome codes. It is equally advisable for senior researches.
For example, if you have a field do you walk everyday? Walking, as a variable. Do you cycle everyday? As
another questionnaire; you have 2 variable names walking and cycling. Do you do both? And then there
is a coding, and somebody very innovatively thought walking and cycling if somebody is engaged we will
give a coding as 12, which is basically 1 and 2 combined, but that is not going to be helpful when you
analyze information.
And last, but very important this is very critical because most of the times you may be dealing with
dichotomized variables. So, you need to be very clear where you are going to give 1 for Yes, 0 for No or 1
for Yes, 2 for No or 1 for present and 2 for absent, as a base line for all the gradients. Some of the softwares
have a different understanding of this 1 and 0. So, when it comes to analysis you need to be careful about
your software related details also.
Finally, when it comes to data entry you need to have what is called the catalogue. Before the data entry
is made, you create what is called data dictionary or variable catalogue in which you talk about each of
these variables which questionnaire item it refers to? What are the values there will be assigned to this
variable? What is the meaning of the each of these values in a particular format?
Some of the softwares generate on their own, this data dictionary as a variable catalogue, but then it is
preferable that you develop your own data dictionary for your study, in which you refer first to the
question item, the variable name that you have given, the type of variable, the format in which data is
collected, the values that are assigned to and some logical checks, if any. This is written so that if this
database is shared with others, the person can make use of the data dictionary and can do the analysis
on his or her own. It is equally important for you, if you after some time as lapsed you go back to the
database, it gives an idea what you have done and what is it all about for each of this variables and it helps
you later when you want analyze your data again.
Before data entry, one makes sure that there are checks and balances. This is also very important from
ensuring internal validity. You specify minimum and maximum values that can enter into a particular field
for example, if your study is about children up to 5 years, the age column will not entertain anything more
than 5 years at the time of data entry itself. So, that it minimizes the errors that can come in even at the
time of data entry, so that these will be acceptable at the time of entry. You may specify skip patterns for
example, if you ask a question do you exercise and that person says, no; then you can skip a lot of
questions about type of exercise, frequency, nature, intensity and things like that. So, skip patterns are
very useful even at the time of data entry.
And then of course, we talked about automatic coding, when you enter that code it automatically denotes
something that is refered to in the data collection instrument and that can be analyzed immediately. There
may be certain times, need for copying data from the preceding record you know for example, lab results
if they have to be carried forward to another section it can get copied by itself; this can be specified in the
database. And finally, some calculations for example, you may collect height and weight data but you may
not calculate body mass index by your own in the data collection instrument. You can ask the database to
do it when you enter height and weight it automatically calculates BMI. So, these specifications are
necessary before data entry is made.
While you enter the data, it is important that we see these as an opportunity for cleaning the database.
For example, you enter a data and you find that there are some notes, you need to write, there are some
clarifications that you need because you do not think that there is appropriate. So, it serves a purpose of
cleaning, the data entry person refers this back to the investigator for additional inputs to clean up the
data. You can use checks, while entering the data which we also discussed as an automated check within
the database. You have to mark each paper as and when the data entry is completed so that the duplicates
are not entered and after the data is entered you may have to validate by different means. So therefore,
data entry is one step in the data cleaning aspect.
(Refer Slide Time: 15:42)
We talked about individual and aggregated databases. We showed you a database that is an individual
database, each record in the horizontal line is an observation. There could be instances in which you may
have aggregated databases, where you may enter counts in each of the records. If you enter only one
count by record that it is ideal, that is called normalize database because normalize database in which
each of the rows contain only one count for that particular record. It facilitates aggregation by a (Refer
Time: 16:20), I will show you what it is.
(Refer Slide Time: 16:23)
For example, on the left hand side you see an example of individual data about people in whom place,
age, gender and onset. So, each of these records indicate an individual and at the same time, we can have
aggregated data for example, by place one can also get number of people affected, number of people
having problem and things like that. So, this is an aggregated data by place, whereas the red color
database shows the individual data.
(Refer Slide Time: 16:56)
There can be an instance in which you may have what is called, Mother and daughter database. You may
collect information at various levels; you may collect information at the village level, you may collect the
information at the household level, you may collect information at the individual in the household, you
may collect within an individual information about several episodes of illness or different problems in that
individual using different questionnaire.
So, in essence you may have information about different levels, that does not mean you know you repeat
the information in all the levels for that particular individual, for example, for that individual you will write
information about the house; for that individual you write information about the village, not necessary
you can keep them at their level. And then, at the time of data analysis you can link the database that
comes from village with database that comes from household, you can link the database that comes from
household to the individuals so that you can sensibly analyze, you do not have to worry about keeping
everything together and get confused at the time of data analysis.
(Refer Slide Time: 18:08)
For example, this is a household level data where you have information about the house ID, location, the
house has such its community status and its income. The individual in the house may have information
pertaining to whether they have a disease or not or they are exposed to particular factor or not. So, you
can see here the house ID is repeated here, the person ID for that particular first household is indicated
here and diseased or not or exposed or not are here. These are 2 different databases entered differently;
one is a household database, another is an individual house database. We can link them as and when
necessary using this connection called house ID, which is common to both the databases. There is a
procedure by which we can do these in softwares, you can even merge this files if needed.
(Refer Slide Time: 19:04)
Summing up on the data management, you need to code database numerically, you need to enter data
using quality assurance procedure which I outlined, need to store information at the level, where it needs
to be stored and we can relate or merge files when needed and as required.
Lecture – 18
Overview of data analysis
Hello friends, welcome to this session, the course Basic Course in Biomedical Research.
(Refer Slide Time: 00:13)
I am going to look at Data Analysis. What are the key objectives of data analysis? The objectives are, to
plan an analysis, program analysis; deal with chance, biases and third factors, to assess causality that is
what Dr. Sanjay Mehendale talked in the first session. The essence of research is to link an exposure to an
outcome and finally, to measure the impact it has in clinical or in domain of science.
(Refer Slide Time: 00:50)
The data analysis strategy has its sequence and I am going to talk about these 7 steps. The first step is
identifying the study type, which is essential before you venture into data analysis. The second is in terms
of identifying main variables. The third, becoming familiar with the data. Fourth, characterize the study
population and fifth, examining the association between exposure and outcome based on the study type and
sixth, in terms of a creating additional tables and finally, to conduct an advance analysis.
(Refer Slide Time: 01:30)
Identifying a study type is the first and foremost step before you enter into data analysis, because it
establishes the main frame work. I am going to repeat again that you need to know whether you are dealing
with the descriptive question or descriptive study or you are dealing with an analytical study. If a descriptive
study is involved, you need to measure a quantity and estimate appropriate indicator, which I am sure you
have seen in the measurements lecture and if it is an analytical study, you need to test a hypothesis using
statistical text.
For that clarity, you need to be familiar with the study, you need to look at the protocol especially for study
objectives and study type, you need to review the entire questionnaire, you need to look at the analysis plan,
you need to look at the data collection procedures, you need to obtain the electronic data base and you may
have to decide on software for analysis about which we will see at the end.
(Refer Slide Time: 02:28)
This analysis depends on the statement of objectives and study type. Therefore, the nature of research
question has to be very clear, whether it is descriptive involving, describing one group measuring a quantity
or is this analytical involving some intervention comparing two groups and hypothesis.
If it is analytical and again you will ask question what is the study outcome? You may decide on acute or
chronic outcome and then choose an appropriate measure of indicator, it could be relative risk or risk ratio,
if a cohort study or it is odds ratio based on case-control study or prevalence ratio or prevalence odds ratio.
(Refer Slide Time: 03:29)
So therefore, your analysis plan depends on your review of the study with reference to; what is the main
frame work? What is the research question that they are trying to answer?
.
The second thing you need to identify the main variables. What are the outcomes? What are exposures?
What are the potential third factors? And what is the variables that be need to consider for subgroup
analysis?
We need to be familiar, intimate with the data. How do you do that? You need to perform a frequency
distribution of all the variables. Look at the frequency of all the variables in your data set and then you look
at descriptive statistics which helps you to describe the study population. This will give you a fairly good
idea of what this data set is all about. And secondly, you review the observations by its status in the database,
are they duplicates, cross check, looks for missing observations, check the ranges and legal values against
the specified in the data dictionary, check for consistency in the pattern of the data. So this is the crucial
step, the third step which you need to spend a sufficient time.
(Refer Slide Time: 04:36).
Fourth, you need to characterize the study population with reference to the basic characteristics of the study
population. These could be in terms of you know socio-demographic and economic variables by age,
gender, what income groups and things like that. And then if it involves analytical study you look at in the
two groups or three groups, the comparison groups these variables. You may want look at frequency of
clinical features as well in characterizing study population.
(Refer Slide Time: 05:10)
Fifth, is in terms of examining the outcome and exposure association, this is the most interesting part. This
is based on a priori hypotheses or hypothesis, where you compare the groups for frequency of exposures
using appropriate measure of association which we talked about earlier. This has to be based on prior
knowledge; it has to be based on study design because as you of seen earlier, each of the study design has
its own measurement of association for the specific exposure and outcome. Therefore, this is a very critical
step. And then you may also apply the principles that you know you might have learnt from Dr. Tarun,
when he talked about the biases and confounding that can complicate this association apart from the
biological and knowledge that we have about particular exposure and outcome.
(Refer Slide Time: 06:07)
The sixth step is in terms of creating additional two-way table for analyzing new variables that you may
detect on the basis of findings that have already generated.
The seventh step could be in terms of a dose-response, stratifications and multivariate modeling analysis.
(Refer Slide Time: 06:34)
Therefore, these are the steps that you will conclude by analyzing the data for generating information that
you think will be useful to improve the health of the study population. Some practical tips that you may
want to keep in your notes in terms of a prior plan of analysis. It is very useful, it is has to be done. The
data analysis has to be done; plan has to be prepared well in advance.
We will recommend you to use empty tables to prepare analysis. In your reading material you will see such
empty tables for each of the study designs. We have prepared an empty table shell, so for your study you
can prepare a plan of action for analysis, this is how my table will look like and then at the time of analysis
your attempt is to fill up that empty shells. The second step is in terms of analyzing by stages go by stages.
You first do the recoding that is necessary, you may create new variables, you may dichotomize, you may
look at the descriptive information, you may decide to change the way you cut the two groups, you may
decide on three or four groups for measuring those response based on your understanding from the data,
you may do the next step of descriptive analysis and then finally, analytical analysis.
So this has to be a sequential process, it should not jump. People have the tendency to jump the steps; I
think it has to be in the measured manner one by one. And finally, please avoid any analysis that is driven
by the data that you analyze without any plan, this is called Post Hoc analysis. You did not have a plan, you
finally find something and then you tried to make news out of it. Last but very important, you do not look
squeeze of the data because you want something.
As an example, we talked about exercise and diabetics you remember. In the initial stages of the analysis,
if this question has to be answered through data analysis, we will recode the data, we will recode in terms
of creating an outcome data. Here, we talked about blood sugar levels so we have to group the outcome
into positive outcome, yes, reduce blood sugar level, negative outcome, no, not reduced, blood sugar is not
reduced, so that is the outcome. Then we may recode some key variables such as, age-groups can be cutoff,
income level can be used to cutoff as you know ok income, not so ok income, or below poverty line or
above poverty line and things like that. This recoding we may do for number of variables including for
examples, in terms of exercise, moderate, heavy, mild or no exercise, we can create groups during this
recoding stage. And then in the descriptive stage we calculate the frequency of the outcome by each of
these groups.
(Refer Slide Time: 09:42)
Analytical stage, we do it in three steps. One is Univariate analysis, where we look at one upon the other.
We look at the outcome which is in terms of reduction of blood sugar level by age, gender or income or
such similar variables that are collected in the data base. Then, if you want to do a stratified analysis or dose
response effect you can examine the outcome by categories of a particular variable. For example, if there
are income categories quartiles or levels of income, low income group, middle income group, high income
group, what is the frequency of outcome? You can examine. And then you may do a stratified analysis, for
example if you want to look at the exposure that is exercise and its relationship with reduction in blood
sugar level among income groups stratified by age, gender and income. You can look at all of this in the
next step. And finally, a model, logistic regression model would tell us whether exercise can predict
reduction in fasting blood sugar level in the diabetes.
(Refer Slide Time: 10:53)
Finally, for beginners or people who have no exposure handling databases, you may wonder what data base
to use. The usual trap that we get into is using spreadsheets for data management and analysis of any type
in size. We think that it is a small data, let use it in we know as spreadsheet but I think, we recommend that
you better avoid that temptation because spreadsheet are not meant to be data management or data analysis
tools. It is preferable to use softwares that can give you both the capabilities data management and as well
as data analysis.
For example, we are just suggesting that this is one of the softwares that we find where we using free
software call Epilnfo, which has the capability to for example, create a data collection instrument format,
enter the data, analyze using the even you know for some the latest visual analysis and then it has the
capability to map the information if that is part of your analysis. And finally, it can also be used for a fairly
a good amount of statistical analysis. So, such softwares may be very helpful. We wish you good luck with
your database a management as well as data analysis because that leads to report generation.
Hello, in this course of Basic Course in Biomedical Research today we are going to talk about and discuss
about some of the ethical issues in health and biomedical research. It is important that, we as researchers
know what these issues are and how to protect the welfare and safety of the research participants that
are going to be as a part of our research.
It is always important to remember that ethical foundation is considered as implicit for conducting any
kind of research and it is applicable not only to health research, but to any research in general. There are
certain international guidelines that have been set in, there are certain international standards and we
have to be within those standards. But we also have to remember that, there are Indian national standards
that are available, which have been developed by Indian Council of Medical Research and they are not at
all less exacting and we have to follow these guidelines as well.
Ethics review is expected in situations sometimes there is a feeling that ethics review is important only in
cases where there is a significant risk involved, some invasive procedures involved, but that is not true.
Even when we are using available data, where we say that there is no risk involved to human participants,
ethics review is required. Also, there are sometimes some situations when minimum risk is involved say
like only questions are asked to people were no samples and specimens are collected. Even in these
situations ethics review is considered important and mandatory.
In the past so many decades, many international documents have been developed and that have helped
for improving the practice of ethics in biomedical research. This all started with some kind of
experimentation that was done in the World War II, among the captives and after that whatever
happened, the people felt that lot of atrocities got committed and human subjects where used for
research in an improper way and lot of discussions started on that. One effort in this direction which began
very early was the development of the Nuremberg Code in 1947, which started some discussion related
to the rational and justification of research, basically whether that particular research is necessary? What
is the risk benefit analysis? How it is important in deciding whether that particular research is important
or not? Also, looking at the competence of investigators and also initiated some discussion on the
voluntary consent in any kind of research.
Thereafter, for the first time many countries came together and met and they signed on a document,
what is called as Helsinki Declaration in 1964, which got revised several times after that and the latest
revision came up in 2013. Helsinki Declaration, basically talks about commitment to individual rights to
make informed decisions, but at the same time also emphasizes duties of investigators. Also, talks about
patients rights, research participants welfare and in addition talks about certain groups that are
considered as vulnerable and it is necessary that certain steps are taken to protect their interest.
(Refer Slide Time: 01:44)
In the United States, in 78-79 Belmont report was published, which described the basic ethics principles
of autonomy, justice and beneficence and we are going to cover this during this particular session a little
later. And it also reemphasized the importance of informed consent in research and here was for the first
time the importance of review by ethics committees, which are called as Institutional Review Board in the
west, it was emphasized. Again in 1992-93, the Council of International Organization on Medical Sciences
and WHO called CIOMS, they developed the document called as CIOMS guidelines and which was revised
in 2002. This is another important international document, which provides guidelines regarding reporting
of adverse events and safety of research participants. This is particularly significant in case of clinical
research and clinical trials, where experimentation is done, in case of new drugs and new vaccines and so
on.
It also talked about benefit and risk balance and need and principles of pharmacovigilance. So, it was
stressed for the first time that by doing phase 1, 2 and 3 studies our responsibility does not end, but
probably we need to do a continued surveillance in the population to figure out, what is happening on as
far as the long term safety of these interventions are concerned.
In 1996, the International Council on Harmonization, ICH as it is popularly called as developed basic
guidelines for good clinical practice. Subsequently, taking the basic clues from this particular document,
good clinical laboratory practice document has been developed; good clinical epidemiological practice, a
document has been developed. So, it found lot of applications in different spheres of health related
research.
NIeCer 101
National Ethical Guidelines for All institutions in the country which carry out any form of
Biomedical and Health Research biomedical research involving human beings should follow these
Involving Human Participants; guidelines in letter and spirit to protect safety and well being of
Chennai
National Institute of Epidemiology
ICMR, 2017. Available from all individuals.
https://www.icmr.nic.in/sites/def
ault/files/guidelines/ICMR_Ethica
l_Guidelines_2017.pdf
There are several other national Genome Policy and Genetic Research [2000], Indian GCP [2001],
guidelines available Amendment of Drugs and Cosmetics Act [2002], Assisted
Reproductive Technology [2005], Stem Cell Research and Bio-
banking [2006]
nie.gov.in
India was not far behind in developing its own documents. The Indian Council of Medical Research in 2000
for the first time introduced Ethical Guidelines for Research on Human Participants. This was a major
consensus document, which was produced and got revised in 2006 and was again revised in 2017. This
document is available on the website of Indian Council of Medical Research. Basically, it gives the guidance
for all the institutions in the country, which carry out any kind of biomedical or health research, which
involves human beings and provides the guidelines that people have to follow, the researchers have to
follow to protect the safety and well being of all individuals.
But in addition to that there are several other national guidelines, which are also available which include
the document on genome policy and genetic research. There has been an amendment to Drugs and
Cosmetic Act in the early 2000, which is available. There are guidelines available for stem cell research,
assisted reproductive technologies, bio banking and researchers working in these areas have to be aware
of the guidelines, which have been provided and stick to those.
I earlier mentioned that there are certain important basic ethical principles. Among them respect for
autonomy, justice, beneficence and non-maleficence, these are the ones, which have to be considered.
Autonomy is considered as basically is a Latin word for self-rule, which means we have to respect
individuals for what they are and this is like respecting human dignity and so we must not interfere in
what people feel like doing or what peoples thought processes are. But at the same time also indicates
that all those people who are not adequately aware have to be empowered to understand, what this is
all about as well. Basically, what it means is people should be clearly informed that they have a right to
decide to participate in research or not to participate in research.
The next principle, which is of justice, it emphasizes that we have an obligation to provide all with
whatever they deserve. Basically, what it means is the participants or if there is an obligation to treat all
people equally, fairly and impartially. So here, what is required is the benefit of research should be
extended to everybody and except in certain situations there are certain groups like say, for example,
condition like pregnancy, when women cannot participate in research. So, unless contraindicated, all
groups should have an opportunity to participate in research, but this should never be imposed on
anybody.
The next two principles, like they just go hand in hand. Beneficence means we must do everything which
is fair and which is correct and we should be correct in our actions and in our deeds also and we should
take only positive steps to prevent any kind of harm, this is an important thing. Non-maleficence is the
other side of it, we must do everything to do things to help people, we should also not cause any harm to
others. So, do no harm is the principle or the explanation of what we called by non-maleficence. So,
whenever harm is evident, see whenever we talk about any new drug trial for example, there is always
expectation that some kind of a side effect would always be there. What we have to really ensure as
researchers is that we take appropriate steps to ensure that this harm would be minimum and if at all it
occurs appropriate care is taken care of. So, whenever we conduct any kind of a research, we have to
ensure that these basic ethic principles or ethical principles are followed.
(Refer Slide Time: 10:00)
And one way to assure this is through a process is what is called as informed consent. This is a process as
I very specifically mentioned because sometimes it is not a onetime event, it is not a tick mark. It is a very
systematic way in which the proposed research is explained to the potential participants in a systematic
step by step process and the potential participants are empowered to take an informed decision to
participate in the research study. What does it involve actually? So, that the participants have to
understand, what the study procedures are and what are the risks and benefits of their participation?
They can ask, they can have the liberty to ask all kinds of questions and raise their concerns which have
to be appropriately answered and then finally, the participants take a very appropriate learned and
informed decision to participate or not to participate in the study.
So, there could be several sessions that could be involved in completing this process. A certain individual
may understand the whole process in one single sitting. For an individual, it may require multiple sittings
and multiple sessions, but the researchers have to be persistent and perseverant to take the potential
participants through this process meticulously. In certain situations, like when we work in a certain tribal
populations, institutional setups, it is important to take a group consent or consent from the concerned
authorities, but one has to remember that this kind of a consent does not replace the individual consent.
(Refer Slide Time: 11:41)
The informed consent, it is classically a document, which explains various things and so the document has
at its header, the name of the project and the agency that is conducting this particular research and the
main body of this particular document talks about various things like, it describes the research study in
brief, it talks about the potential risks, it talks about the benefits. Sometimes, it is possible that when
research is conducted, benefits are not necessarily individual oriented. The research might cause benefit
to the whole community and that has to be explained correctly. Their alternatives have to be explained in
the sense that participants have to be explained that, even if they decide not to participate in research, it
is just fine and they will continue to get the services as they would otherwise get even if they decide not
to participate.
In addition, the researchers have to commit and give the assurance of confidentiality, keeping their
records confidential because sometimes some sensitive information gets collected and participants may
be worried about the information that their sharing with the research team and so it is important that
they have to be given this assurance of confidentiality. Sometimes, some harm that results as a part of
research participation has to be appropriately compensated and that clause also is essential as a part of
the informed consent processes. We have to also give some basic important contact information as to
whom the research participants can contact for any additional information, for any concerns that they
may have during participation of the study and this information should be clearly included.
In addition, one of the most important clauses that gets added is about voluntary participation. The
document has to emphasize that the every person has a right to decide whether to participate in the
research study or not. Also, during the part of the research study and this becomes particularly important
when it is a long term follow up study. A person may decide to drop out of the study at a certain point of
time and it is perfectly within the rights of the individual to do so and this has to be explained. So, all these
constitute the body of the informed consent and towards the end of this particular form, the signature of
participant and if the participant is illiterate, then the signature of a witness who is not a part of the study
team has to be obtained.
There are various stakeholders in this informed consent process. In research, please understand that this
informed consent procedure is considered as a very important procedure and it is a very sacrosanct
procedure. We as researchers have to understand that we and our institutions have an obligation to
provide the required information, which is good in length and also in-depth to the participants. They
should be given a chance to discuss their issues, their problems. They should be provided with adequate
explanation. We have to also ensure that they have understood the information that is being given to
them with respect to the research study appropriately and then also ensure that they take a voluntary
decision; there is no coercion or coaxing on part of the researchers.
On part of the participants, who are the other kinds of stakeholders in this whole process, they have to
themselves ensure that, they have understood whatever has been told to them. They just should not sign
the informed consent form without understanding what goes behind that or what is included under that.
They have to understand their rights, they have to understand various provisions, the researchers or
scientists are going to make as a part of the informed consent document. Basically, they should sign the
form freely, independently and without any coercion or force.
The third kind of stakeholders in this process includes the sponsors, the monitors or regulators here. These
are the kind of agencies that assess the fairness of the consenting processes at various levels. This starts
right from the beginning of the research study before it gets approved by the institutional ethics
committee right up to monitoring throughout the procedure by the monitors and regulators as well. They
also have an authority to verify the consent documentation of research participants. So, every person who
is involved in research has a combined responsibility to ensure that informed consent is appropriately
taken.
There are certain issues around that, whom does the informed consent benefit? It benefits everybody
really; I talked about some stakeholders here. So, from the research participant point of view, it gives him
lot of information and it therefore it is important for him, for the investigator it is a documentation that
this particular process has been completed in the best possible way in the most appropriate way.
What is important to understand is whether the research procedure and various other aspects of research
are adequately explained in the informed consent form. So, the language has to be simple. There has to
be lot of clarity on various issues, it should be in the local language. So, it is also advised that the consent
form which is developed in the local language is back translated and then either it is certified or checked
with the original English consent form for its accuracy. Some investigators have tried doing test of
understanding also and this is considered as one of the good practices. Once, the informed consent form
is done, some kind of a small objective type of test is quickly given to the research participants to assess,
whether the understanding on the informed consent form has been adequate or not.
The issue of witness can become a critical issue. This becomes particular important in scenarios, where
the people are not literate and they cannot sign the informed consent form. So, here is where, there is a
need to have an impartial witness. It is important to ensure that the principal investigator, researcher
himself does not sign as the impartial witness. It should be somebody, who is not connected with the
research process.
There has been a lot, which has been talked about, whether the oral consent is valid or not. Typically,
there should be two informed consent forms to be signed and then one has to be returned back to the
participant, which the participant can keep with him for record because it also provides lot of answers to
the questions that might arise in the mind of the participant subsequently. There has been some
discussion going on with respect to audio consent and video consent, well the regulatory authorities in
India have now made it mandatory to record the consent procedures in case of investigation in new drugs
in India. So, this is an important regulation which we must keep in mind.
(Refer Slide Time: 19:11)
Different types of reviews have to take place before any kind of study becomes ready to be undertaken
there actually. There is a scientific review and regulatory review, which takes place before the actual ethics
review happens. The scientific review looks at the novelty, the rationality and relevance of the study.
Basically, what is the justification of doing this particular study? What are the scientific merits? Whether
appropriate study procedures have been done, have been taken into consideration? What are the
inclusion exclusion criteria? Whether the sample size has been calculated appropriately? How will the
endpoint or the outcomes going to be assessed? So, these aspects are normally looked at in the scientific
review. Why this is important is because scientifically well planned research studies are often likely to be
correctly addressing the human subject issues and also the ethical issues. So, if the science is good often
it takes care of ethics which is behind that.
(Refer Slide Time: 20:12)
Similarly, regulatory reviews look at the various aspects like the pre-clinical trials data that has been done.
This is particularly important in case of newer drugs, newer vaccines, various animal data that is there,
toxicology data that is there. then there are certain in country regulatory assessments for various drug
vaccine or product imports, they have to be assessed and this is important if there are certain trials or
experiments which are being done using the products that have been developed abroad. There could be
certain special situations, where genetically engineered products are being used or there is a stem cell
research, research on the reproductive technology, organ transplantation etcetera and so the concerned
regulatory agencies have to ensure that all the necessary guidelines have been followed here.
The issues around sample shipment and transfer as well as transfer of raw data is looked at very seriously
by the Government of India because there are issues around the intellectual property rights in this area
and the government is very protective about those. So one, the researchers have to know these issues
fairly well and the regulation in this regard as well. There are certain sort of caveats; there are certain kind
of restrictions regarding exchange of scientists and visitors. The funding particularly, if there is foreign
funding coming in for a project to be done in India and during such situations, where the research is to be
done in the border or high security areas and the researchers have to be aware of the regulatory
requirements in such situations. Again, like scientific review, careful regulatory review also results in
answering some of the ethical concerns.
(Refer Slide Time: 22:00)
Now, coming to the various ethical issues that need to be addressed; well, it is important that the
competence of researchers and a research team has to be accessed appropriately. Then whatever are the
provisions made for protection of human rights and vulnerable populations, in particular they have to be
accessed. Then measures for protecting confidentiality, they have to be seen. Appropriateness of the
informed consent form, the correct completeness of the informed consent form has to be assessed.
Mechanism for reporting and management of adverse events and serious adverse events, this is
particularly important in case of drug trials. Then care and support mechanism for participants, is the
support going to be extended to the participants after the trial? Will the post trial benefits be given to the
community after the trial is finished and is proved to be useful? These aspects also have to be looked at
from the ethical angle.
The reimbursements and compensation are important issues, which the ethics review looks at. Because
one has to ensure that reimbursement is for the time lost and also the expenses paid for traveling to the
clinical research site. But the compensation or the incentive given or the money that is given should not
be as much as to course the participant to participate in the research trial. So, this decision is also made
by the ethics committee. And it is also important to continue the review of the progress of the study till it
is completed. So, the committee which really looks at all these things is what is called as Institutional
Ethics Committee in India.
(Refer Slide Time: 23:41)
It is called as Institutional Review Board abroad. Basically, it looks at whether the study has potential
benefits or community benefits? Whether the rights of the participants are adequately protected?
Whether the benefits of this particular study out with the risks that are involved here? Whether the
participants and communities have access to study findings? Whether they are eventually going to benefit
from the research participation? And what kinds of mechanisms are built-in to provide the safety and care
and support to research participations even during the study and after the study?
(Refer Slide Time: 24:29)
So, there are growing expectations about accountability from the researchers now. Various researchers
are being questioned about their responsibility and governments' responsibility as well in fairly conducting
research. There is a growing public awareness also; all this is eventually going to improve the quality of
research in our country and also, the practice of medicine in our country over a period of time. There is a
collective demand for health benefits, people are demanding. So, universal right to health care which
emanates from the principal of health for all. So, this which demands that more and more research will
have to be conducted in making more and more benefits available to the common man of this particular
country.
It is also important to understand that it is not only the researchers who have the complete responsibility
to follow ethics. It is also important that the research participants follow certain or fulfill certain
expectations. So, whatever has been described in research they have to follow appropriately. It is
explained in the informed consent form and so there is a great need for including bioethics in the medical
curriculum which is being stressed.
(Refer Slide Time: 25:47)
There is certainly a hope because ethics in the practice of public health and in health research is being
increasingly addressed. We know that there are challenges as well because the public expectations and
demands are continuously increasing, but we will continue to find solutions and this has to be an ongoing
process. For this, various public health stakeholders like public health system, policy makers, researchers
and program managers, they should show adequate sensitivity and realize that we always can improve
and a practice always can improve.
Lecture - 20
The scenario of clinical trials in India has changed over the last 25 to 30 years. In the late 20th century,
what was happening was, mostly the clinical trials that were undertaken were by the research institute
and also by the pharmaceutical industry, which were limited in nature. But, towards the beginning of 21st
century, in the first decade in particular, the clinical trials industry in India grew and expanded very, very
rapidly. But in the present decade, it has been facing some kind of challenges because of various
regulatory reforms that have happened recently. But overall, the international community particularly the
international investigators and sponsors, who want to undertake high quality clinical trials perceive that
there are certain challenges for conducting clinical trials in India.
They feel the important ones among them include, the approvals are often delayed; approvals in terms of
the ethical approvals, by the ethics committees as well as the regulatory approvals by the regulatory
authorities. There are also concerns about the quality of ethical review that takes place because it has
been only in the recent times, a very systematic effort is being taken to improve the performance of
various ethics committees of research organizations in particular.
The shipment of samples and transfer of data has always been an issue for international studies in
particular because government has some specific restrictions and regulations about transfer of the data
as well as samples. It is also felt that, there is an overall lack of duly certified and trained investigators as
well as centers in India and again a very systematic effort in this direction is ongoing in the country. What
has happened as a part of the regulatory reforms are two important changes suggested now and there is
a specific clause of compensation, which has now come up for clinical trial participants and the office of
the Drug Controller General of India has given very specific recommendations about how this
compensation has to be calculated and up to what period these laws are applicable, that has put in some
kind of pressure on the people, who are conducting the trial and there is some kind of haziness around
this particular area.
Also, in the area of audio visual recording of informed consent process, all of us will certainly agree that
to make ourselves very accountable, to make ourselves very transparent, probably this is the best
procedure. If we protect the confidentiality of the trial participant adequately, but carry out audio visual
recording that is possibly the best proof of how the consent procedure was actually undertaken here. It is
currently applicable to mostly the IND trials, which are Investigational New Drug trials, but again people
are confusing this. They are trying to do it in all different kinds of scenario and some kind of thinking,
rethinking on this particular matter is happening at the center on this particular thing in India.
.
But, for any clinical trial to happen or to start, certain approvals are mandated and they have to be taken
in. It all starts with the scientific review; it is a very systematically planned experiment, very meticulously
planned. Every step is outlined in detail and basically, what we ensure through the scientific review is
whether the research question is sound and usually it is the Institutional Scientific Advisory Committee
that decides, whether the study designs that is being proposed here is appropriate. All the methods that
are being described under that are accurate and correct.
Then, it goes to the next stage of what is called as the Ethical review. It is done by the Institutional Ethics
Committee; sometimes at the national level also it can be done by National Ethics Committee as well. We
do it, in our set up of Indian Council of Medical Research and what is really looked at here, is whether the
safety and welfare of the research participants is adequately protected or not? This is a very significant
aspect of a clinical trial. This review is considered very critical, why? Because, the basic principle is, we
should do no harm to research participants by making them participate in a research study and that is a
very critical component, which we have to keep in mind.
Another kind of review that happens is that of regulate, which we call it as a Regulatory review. Sometimes
it is new drugs, so the setup in our country is the previous Drug Controller General of India, which is now
the Standards Control Organization, that we have CDSCO. Sometimes, if the project is getting funded
internationally, then there is a health ministry screening committee, which looks at various regulations
around that particular issue. If there are genetically modified or engineered products being used then
there is a Genetic Engineering Approval Committee, which looks at that. It is the responsibility of the
sponsors and investigators to find out, what kind of regulatory approvals are required for the kind of
clinical trial that is being undertaken and all those clinical trial approvals must be taken before initiating
the study. Appropriate certification has to be kept on file for ready review by the external monitors or by
any other agencies that are authorized.
(Refer Slide Time: 05:57)
But one of the most important things that has to be kept in mind is because it is an experimental study,
because we are manipulating the environment, we have to take care of a lot of ethical issues that can
arise in this particular scenario. Primarily, the most important thing is have all the approvals been sort for
this particular thing, which include the scientific, the ethical as well as the regulatory approval, which is
currently the norm in the country at any point of time. Then what? Just getting the initial approvals is not
enough; to ensure that the patients or the participants are protected all throughout the trial is also
critically important and so some role of ethics committee to monitor this particular process is also
important.
Is there an adequate community engagement and support for this? Somehow, it has so happened that
sometime when the people were not made aware of some of the interventions that where being tried at
the national level, there where backlashes because there where something which where believe to be
culturally not acceptable and hence it is always important to engage the community and face of the
community is visible through many groups. They may be the private practitioners and doctors, they may
be the program managers of the country, they may be the doctor attending to the patients, they may be
the political people who are taking say, sort of care of the interest of the people and so on and so forth.
One of the important things is to be remembered is every single participant has to sign an informed
consent form before participating in a clinical trial. Any trial without an informed consent form is a
problem and we have to ensure that this particular process is completed. Also, after the trial gets
completed, probably the responsibility of the investigator does not necessarily end. There has to be some
kind of a mechanism for providing follow-up care to or post-trial support to the trial participants and that
maybe all well thought of right ahead of time and planned in the research protocol.
Use of placebos has raised some questions in the past and it is a topic in itself, but wherever there is a
vaccine which we are talking about, where there is no comparable vaccine available earlier of that
particular disease, doing use of placebos is considered as justifiable. One of the important things which is
an ethical issue, is the confidentiality of the study participants. The study participants sometimes are
worried that nobody other than them, even in their own family should know about their research
participation and it is our duty as investigators or researchers to protect this interest of the participant
and to actually keep all the information totally confidential.
As I said, the whole process of clinical trials starts with the informed consent process procedure. I have
deliberately used this word, it is a procedure and it can be multiple step procedure also. Some patients
and some participants may not be able to finish this particular process in one single sitting. You may
require multiple sittings to explain the things. The word ‘informed consent’ here is to be critically looked
at, informed is where the duty of the investigator is to explain all the study procedures to the study
participant very carefully and also ensure that any question or doubt the person might have is
appropriately answered. This has to be now documented, all of us know this. And now, there is also an
insistence that for clinical trials it should be, a duplicate copy should be signed and because it is a
document, which give some information about the study also. One copy should be kept by the patient
and the other should be kept with the investigators team.
The process of clinical trial participation is generally two step: There is a screening protocol first, wherein
those who are interested in participating in the trial undergo some level of screening that is the interview,
followed by medical examination, followed by some samples collection and it is ensured that they fit into
the eligibility criteria and because as investigators, we have to adhere to the inclusion and exclusion
criteria very specifically. It is important that the investigators follow good clinical and laboratory practice.
This is absolutely important, this has to be followed all throughout the trial and what this means? This is,
although this looks like something like a common sense. Do the enrollments as exactly as have been
defined in the protocol. Do the study procedures as they have been defined in the protocol. Collect the
volume of blood as has been defined in the study protocol; ensure that all the visits are made. So, these
are the basic procedures, there are responsibilities for the sponsors, there are responsibilities for the
investigators. All of them have to follow this.
In clinical trials, one of the most important things, which we have to ensure, is adherence to the
intervention. If it is a drug trial, we have to ensure somehow that the patient is taking the drug regularly.
If it is adherence with respect to follow-up evaluation, also we have to ensure that the person comes for
follow up at the define time intervals. Say for example, we are asking a particular participant to participate
in a vaccine trial and if we want to study, periodically up to 2 years from giving that particular dose of
vaccine, till what length of time, the vaccine immunogenicity or the antibody levels to that particular
vaccine candidate are maintained and if it is decided that once in every 3 months, this evaluation will be
done. If patients miss some of these visits, it is very likely that we will miss some of the important data
because we may not be able to decide up to what level the antibodies were maintained and beyond which
they were not maintained.
There can be a lot of issues which can arise in case of multi-centric trials. Sometimes, trials are conducted
in 5, 6, 7, 10, 50 centers in the country and here is where, this is done with the basic purpose is sometimes,
it becomes very difficult for a few sites to gather the required number of study participants. So, multiple
large number of centers have to be used, but the compromise that we make here is the quality can come
down particularly, variations can occur within the procedures that are adopted in various centers and
hence, one of the key elements here is the standardization of the study protocols. Training of the people
right in the beginning, periodic training of the people and also then using very standard well defined
protocols is something which is absolutely important.
Also in a clinical trial, one of the important prerequisites is to make some kind of an arrangement for an
independent monitoring mechanism. This is an agency, which you generally employed by the sponsor and
that carries out independent monitoring, looks at for the adherence to GCP, dispensing of study products,
how the records are maintained, how informed consent procedure is happening and so on and so forth.
This particular agency has to have a total third party view in doing that particular assessment and should
not be influenced either by the sponsors or by the investigators. As we know, as I have been mentioning
right from the beginning, safety is one of the most important thing as far as clinical trials are concerned
and maintaining or ensuring safety of research participants is of paramount importance and hence, the
safety assessments are built in all clinical trials. What is also important is, there are very well defined
reporting mechanisms for management of various kinds of adverse and serious adverse events as they
happen in, among the clinical trial participants and they have to be timely reported to the regulatory
authorities as well as sponsors as well as ethical committees.
We have to remember that all the adverse events may not necessarily be clinical. Sometimes, they can be
odd laboratory values or they could be even social or familial problems that are happening as a part of
clinical trial participation. All these pieces of information have to be appropriately trapped. Generally,
there is an agreement that some kind of reimbursements should be made to the trial participants. One is
for the kind of time they spent in coming over there and the loss of daily wages that might happen as a
result of that and also for the travel cost involved, plus maybe some kind of food expenses, which are
made particularly, in case of trials which require long term participation of the study participants.
So, but here is where a clear understanding has to be there as to we should know that we are doing
reimbursements and we are not doing incentives because sometimes incentives can become coercive. In
the sense, the incentives can persuade patients to participate in the trials, even if they are not interested,
if the amounts are really large. Recently, some kind of laws and regulation has come in with respect to
compensation. We have to be all aware of what is prevalent regulation in the country with respect to the
clause of compensation.
It is also important that we have a mechanism for grievance redressal, some of the research participants
because it is an experiment and some kind of a experimental intervention is being tried and the some of
the people might come up with some kind of complaints, which the investigators may feel are not related
to the intervention that is being tried, but the patients keep insisting that it is essentially related to that
and hence, there has to be a kind of third party body and it is always believed that grievance redressal
team should be there, which is easily available to address these issues in real time.
Every single trial has to have a defined trial stoppage rule right in the beginning itself. Generally, 3 serious
adverse events which are defined as maybe; deaths or very serious complications on the previous clinical
condition that is what happening, hospitalization due to any cause, these are considered as serious
adverse events. So, after any 3 serious adverse events, it is normally decided whether they are related or
not related to the trial by a third party body called as Data Safety Monitoring Board, which is an
independent entity, which is also predefined earlier and when they allow us to move forward, we move
forward in the clinical trial, but these rules have to be defined up front.
Another important thing in a clinical trial is documentation archival. Various sponsors have various
expectations with respect to this; some of the trial sponsors require this. All the trial related documents
to be stored for a period of 5 years, sometimes 10 years and sometimes 15 years, whatever are the
regulations we have to comply with that particular expectation.
There could be some impediments in clinical trial participations. At the patients’ level, many times patients
do not know about the clinical trials. Sometimes, they do not have access to clinical trials. There is
mistrust, suspicion or people are afraid of being participating in clinical trials. Sometimes, they feel there
are fees attached to this particular thing. If clinical trials have to succeed, patients have to agree or
volunteers have to agree to participate and hence, disseminating this particular information becomes
extremely important.
Sometimes, there are some issues at the level of health care providers also. There is a lack of awareness
of appropriate clinical trials that sometimes they are unwilling to lose controls of person’s care, say for
example, a trial is going on in cancer patients, there could be some kind of a fear among health care
providers that they would lose their patients to this particular trial once it is referred. Sometimes, they
believe that the standard therapy is best and could be also concerned about the administrative burdens
that this may add. So, even if there are impediments here the clinical trials are the only way for making a
progress in medical science.
Because, if there are no clinical trials, no new drugs will come, if there are no clinical trials, no new
technologies will be tested, no new vaccines will be tested and so they must be supported and the
adequate information about clinical trials must be disseminated. Randomized controlled clinical trials
have to be carried out in the best possible manner, with the highest possible quality adherence.
Lecture - 21
Dear friends, welcome to this session on Basic Course in Biomedical Research. All of you by now, have
gone through the basics of research in terms of conceptualizing the ideas, study designs, ethical and
scientific contact of research. I am sure now you are ready, to give life to your research ideas in the form
of concept paper. I am going to introduce to you, how to prepare a concept paper for research projects.
At the end of this session, you will be in a position to write a concept paper for research project.
(Refer Slide Time: 00:51)
Now, let us revisit the seven steps of a successful protocol. These are logical, sequential and essential
steps. The steps include, identifying research topic, framing research question and objectives and then
moving on to outlining a one-page concept paper, preparing dummy tables as per the analysis plan,
writing a detailed drafted protocol, preparing instruments and annexes including information sheet,
consent form and other things related to the study protocol; submit this to your competent peer review
committee and finally, seek review by an ethics committee.
Let us look at the life cycle of research and how the steps are important in relation to the life cycle of
research.
(Refer Slide Time: 01:44)
We will see that, we started with identifying data needs and we end with requiring more information that
necessitates for the research. So, therefore, in the concept paper has to capture all of these elements in
a miniature form of a protocol.
The time, for everybody is very precious for you, for your faculty, for your guides, for your reviewers’, time
is precious. Equally, the funding agencies are also hot pressed for time. So, a one-page in a shorter version
forces you to be focused. Many of our ideas just are not born, they remain as ideas. One page, to write a
one page as compared to a long detailed protocol, may help us in overcoming the inhibition that we have
in writing the detail protocol. Therefore, the ideas may be born rather than getting aborted, that is the
reason for one-page. I am going to now outline this one-page, what it requires in bullet styles in each of
the sections of the concept paper.
(Refer Slide Time: 03:11)
It has these sections, which is essentially, as I said earlier, miniature of the protocol. It has section on
background and justification, statement of objectives as a separate section, method section, expected
benefits, key references and budget. I am going to elaborate on each of this section. Let us begin with
looking at background and justification.
(Refer Slide Time: 03:42)
The background and justification essentially will have 3 bullets. One on the importance of the study
problem, second what is known and unknown about the problem in relation to the literature and in
relation to the local context, lastly the information that is needed to address the problem effectively.
The next section is on methods. It has to give by bullets an outline of all the aspects of methods covering,
for example, study design, study population, key operational definitions, the sampling methodology by
which you will select your study participants, the calculation of sample size and key considerations for
sample size. The data collection procedures, who will collect? What will they collect? And an analysis plan
around the key objectives. Finally, this section will end up by saying, what are the ethical issues and which
ethics committee will review this protocol, if developed.
And next section is on expected benefits. As you have listened from the earlier session, the research
question is to be answered. There are two aspects of benefits that we look forward in answering the
research question. One, what action will be taken following the results? Second, what is a future research
or planning or policy agenda as a result of this finding? This is very important from the concept paper
point of view, you have to set out this in this section and finally, you need to show that you have mastered
the literature.
These key references may be referred to in the introduction section and may be needed to refer to some
operational definitions in the method section, citation of a reference. We suggest that you need not put
in more than 5 references. However, we recommend strongly, you follow internationally acceptable
standard guidelines for writing these references. We recommend, you to adopt and use International
Committee of Medical Journal Editors guidelines, which is accepted globally by many bio medical journals.
(Refer Slide Time: 06:39)
The last section is on budget. You may wonder, why budget is important in this concept paper, but it is
equally important to technical and other aspects of preparing a concept paper. This is an indicative
budget; it need not have detailed justification. It can cover key items like salaries/per diem, travel,
equipments and supplies and miscellaneous, whatever is applicable to the conduct of the research that
you are proposing. Many research agencies insist on having an indicative budget. So, this is also equally
important.
(Refer Slide Time: 07:15)
Let us look at some of the aspects of application of this concept. If you look at Indian Council of Medical
Research, the premier medical research agency in the country, ICMR on its home page offers funding for
researchers, like you and me. There is a provision called extramural funding on its home page and you will
see on the home page there is a provision to submit, what they call as pre-proposal format, this is similar
to concept paper. This pre-proposal format covers the following, a title, an introduction, a novelty,
applicability and description of the project.
(Refer Slide Time: 07:48)
You will see that they have specified word count for each of the section. Now, if you see the outline that
be provided, most of this can be taken from the concept paper that we just now, generic concept paper
that we just now provided you, except may be, novelty you may have to write it as, in a fresh section.
In fact, some of the research funding agencies even fund and support developing a concept paper, if it is
meritorious to a full pledged protocol through their own funding. So, ready, handy concept paper is very
useful to submit to such research funding opportunities.
Therefore, what can you achieve with one-page concept paper? It helps you to organize your ideas, find
just right words in a brief, but succinct manner that helps you to stand out in the competition; you may
receive rapid and positive response. For your benefit in the reading section, you will find a template
concept paper, one pager, a word document which you may use for writing your concept paper. Second,
we have also provided a sample concept paper.
(Refer Slide Time: 10:25)
Now, here we look at two templates of concept papers, one for writing observational studies another for
interventional studies. I am going to show them and go through the parts of these two templates. The
first one is on observational studies, as you can see you can type your title of the project, you can type
your name, the first section is about background and justification and we have provided tips. You can
actually, overwrite on these 3 bullets; the first bullet is about the contexts of the study problem in a
quantified manner with linked references. The second bullet is about the local contexts, which what we
know and what we do not know is stated. The third bullet, in terms of what information we need to
manage the problem effectively, you can just overwrite on these 3 bullets.
The next section is about statement of objectives. We have some tips here, you need to state them
quanitatively for the primary objective and you may make it clear, whether you want to propose to
estimate a quantity or test hypothesis depending on the nature of the research question and statement
of the objectives, which we discussed earlier. And last, but very important, you need to clearly distinguish
the secondary objectives from the primary objectives.
(Refer Slide Time: 12:00)
The next section is an elaborate section on methods. As we discussed, it should be in bullets and you can
derive from tips for the type of study design from this widely accepted guidelines available in this website,
equator-network.orgYou can state the study population; you can state the study design in one bullet. It
could be in terms of observational study design, it could be a survey, it could be case control cohort studies
or ecological studies or sometimes, it could be you know, even case report or case series depending on
the study design that you choose, based on the objectives.
Then the operational definitions, you need to provide key definitions only and here is where, as I discussed
earlier, you can make a reference to the standard definition for the literature criteria. You can even state
control recruitment strategy for the case control study. Sampling procedure, you can state in one bullet,
sample size you can briefly mention the sample size and the assumptions. Data collection you need, as I
said, you need to spell out, who will collect the data? What kind of data they will collect and within what
time frame? And briefly, the quality assurance that is going to be used, finally, the analysis plan, we need
to summarize the type of analysis based on the primary and secondary objectives.
You may mention laboratory analysis, if they are part your study. Human subject production, you may
have to mention about key measures that are taken to ensure the protection of the human participants
and which ethics committee will review this proposal and last, but important section is on expected
benefits. You need to describe the report that will be generated and the outcome of the study in order to
manage the problem, study problem effectively in the area where the studies conducted.
(Refer Slide Time: 13:56)
The references, the 5 references, you can even make it small font size, but adhere to the standard
guidelines, which we recommend that you use ICMJE guidelines. Finally, in the budget section, in a brief
or 4 bullets. So, this is on observational studies.
So, we will now look at concept paper for intervention studies. For intervention studies, the background
has to state, what is known and unknown for the drug or interventions or management of the specific
study problem that you are talking about and you need to say why the information currently available is
insufficient. You need to state the objectives clearly here in terms of primary objective, which will look at
the primary outcome.
In terms of methods, you may benefit from using the guidelines that are available for writing a clinical trial
protocol. So, you may use the elements from that guideline, which is called SPIRIT. The method section is
pretty much similar except that you need to state in the clinical trial concept paper, specifically
interventions, the drug or interventions with dosage, frequency, nature and all the other forms of
interventions. Operational definitions, you may have to state the primary and secondary outcomes. You
may have to state the participant recruitment strategy.
The next important section is on randomization sequence allocation and allocation concealment. You have
to briefly mention the type of randomization, methods used to generate and implement the allocation,
and masking has to mention as to the level and type of masking.
(Refer Slide Time: 15:37)
The other sections are pretty much similar to what we saw in the concept paper template for
observational studies.
We have a sample out there in the reading section; please makes make use of it while writing your concept
papers. We hope you will benefit by making use of it and make writing concept paper a habit and we wish
you give birth to your ideas.
Wish you all the best.
Thank you.
Lecture - 22
Elements of a Protocol for Research Studies
Hello and welcome to this session of Basic Course in Biomedical Research. Today, we are going to talk
about the Elements of a Protocol that you would write for a Research Study. You have gone through all the
aspects of how to start designing a research study, starting from thinking about a research question, doing
sampling, selecting the research participants, doing measurements for exposures and outcomes, looking
after the human subject protection and so forth.
Now, these are all the structures of a research project and when all these structures for a research project
come together in a written format and that is what is known as a protocol, which is basically a written plan
for a study.
Remember that, when you first start out writing the protocol, you are basically writing it as you are thinking
about it and if you already made a concept paper, you are in a good position because you can then just use
the concept paper as an outline to actually write in more detail about various aspects that you have already
mentioned as bullet points in your concept paper and develop it into a protocol. Usually, the objectives
section of your protocol, you would take it as you have stated out already in your concept paper, but the
other elements like background and justification, the methods, the expected benefits, these are some of the
sections that you would have to now give more elaborate details in your research protocol.
It is also generally a norm that usually the first draft should not to be too long, generally about 1500 to
about may be 2000 words with a small introduction, less than one- fifth of the total protocol, which will
keep you focused. And of course, remember that, you would given a few references in your concept paper
and now you can increase a references, depending on how much detail you are adding to your protocol and
so you have may be 5 or 10 or even more key references as is needed for your protocol.
(Refer Slide Time: 03:09)
So, as I said, we need much more details and the major part of the protocol would actually be your methods
section, wherein you give the whole detailed plan of how you are going to conduct the study, starting from
the study design, describing the interventions, if you are doing an experimental study, describing a study
population, writing out your operational definitions, what was the sampling method that you used, how did
you calculate the sample size, what would be the data collection methods, what would be the tools to do
that? What is your analysis plan? And then, a detailed project implementation plan to maintain the quality
and finally, the last section on human subject protection. We will walk through each one of these steps,
little bit more in detail to see, how you can develop your protocol.
(Refer Slide Time: 04:03)
So, in terms of the first paragraph would be the paragraph on study designs. Now, this is where you explain,
how the objectives that you have stated at the end of the introduction section, how they would be measured?
How would you frame indicators? And what design would be used in your study to fulfill these objectives?
Remember, your study design could be experimental or observational; in term it could be a cohort study, a
case control or cross sectional study, again, depending on what your study objective is. You need to define
again, a little bit more in detail in the protocol, whether your study would follow a prospective design or a
retrospective design, in terms of how you are going to collect data and how your study is designed.
(Refer Slide Time: 04:57).
In the next paragraph, you would now want to describe the interventions, assuming that, if you are doing
an experimental study, whether it is a clinical trial or a community based intervention. Accordingly, you
will describe your intervention, who is going to be intervened, What exactly is your intervention, give it in
more detail. What would be the time period over which this intervention is going to be applied to your
intervention group and how would, that be done? So, the whole procedure of whether, you are doing a
clinical intervention or a community based intervention has to be described in detail in this paragraph, if
you are doing an experimental study.
(Refer Slide Time: 05:41)
The next paragraph would be detailing about your study population. What is the inclusion and the exclusion
criteria that you are going to be using to select the population either, which may be hospital based,
community based or population based into your study. You would use over what time period you are going
to select this people. What would be, say, the geographical location from where you are going to select this
people? And what would be the individual characteristics of the people that you going to select into your
study or keep them out of the study, which would be written in the form of an exclusion criteria.
Remember that, study population is different from the actual study sample that you are going to take from
this population, or the study population is general population among which you are going to be doing the
study and a part of them, you are going to be sampling based on your sample size and sampling strategy
into your study.
(Refer Slide Time: 06:46)
Next paragraph gives the details on the various operational definitions that you are going to be using in
your study. You spell out, how you are going to define and measure your key outcome measures. What
would be your key exposure measures? And again, what would be the operational definitions of these
measures? Remember that, you need to be very clear and specific in terms of these definitions. And many
a times, there may be standard definitions, standard ways and means of measuring the outcomes or
measuring certain exposures and then it is always good to use these standard definitions and again for which
you would need to provide references, if you are using any of the standard methodologies to define your
variables that you are going to be using in the study.
(Refer Slide Time: 07:40)
The next paragraph is about sampling. You need to give in detail, what type of sampling you are going to
be using, whether it is a random sample or a non random sample, using systematic sampling, simple random
sampling, cluster sampling, multi-stage sampling or whatever the case may be, as is applicable to your
study and how are you going to select this sample in actual practice. So, all the steps in doing this sampling
have to be detailed out in this section of your protocol. Again, if you are using standard methods of
sampling, it is also a good idea to provide references of the standard ways in which this has already been
applied in other settings.
If you are doing randomized clinical trials then you would be using randomization and as was discussed
earlier there are several ways of doing randomization. So, this would be the paragraph, where you would
explain which kind of randomization technique you are going to be using and how are the study subjects
going to be randomized into your control group and your intervention group.
(Refer Slide Time: 08:55)
Of course, what follows is the paragraph on sample size, where you will explain how you have calculated
the sample size. What are all the parameters that you have used to estimate the sample size? Remember
that, the sample size will depend on their study objective as well as the sampling methodology that you are
going to employ for your study. You further need to give in detail, how this estimate is generated, what
formulas were used, whether you will be using any softwares for doing that? And again, you need to provide
references for these things.
(Refer Slide Time: 09:32)
Following the sample size is the paragraph on data collection. There would be 2 sections to this paragraph:
one part is where, what is all the kind of data that you are going to collect? Whether it is socio demographic
characteristics, the individual characteristics of the study participants, their clinical histories, the science
and symptoms and so forth. Secondly, you would also need to specify, how you are going to collect this
data. Who is going to do that? What data is going to be collected by the principle investigator or is it going
to be some other study investigators, who are going to collect it? Is it going to be a staff nurse? Is it going
to be an outreach worker and so forth.
And of course, how you are going to collect this data. What study instrument are you going to be using to
collect this in data? Are you going to be doing interviews? Are they going to be face to face? Is it going to
be computer based? Are you going to extract data using data abstraction forms? Or if you are doing
qualitative methods, would you be doing focus group discussions or in-depth interviews? And what would
be the guides for collecting this kind of data?
(Refer Slide Time: 10:48)
Once, you have spelled out your data collection methods the next paragraph will be explaining, how you
are going to analyze this data. Once you have collected the data what would be your mode of data entry? Is
it going to be manual? Is it going to be computer based? What software are you going to be using to analyze
this data? You also need to give in detail various stages that you may go through in doing the data analysis,
if you are going to be recoding certain variables that you are going to collect during your data collection.
What are you going to estimate? What parameters? Whether it is prevalence, incidence or if you are doing
analytical studies, what kind of analysis are you going to go through and the various steps that you are
going to go through that.
So, all of this has to be mentioned in this analysis plan paragraph. It is not just enough to say that suitable
analysis will be carried out for this study. That seems quite vague and it is not a good idea. It does not give
a good impression; it shows that maybe you are not planned your study well enough. So, it is always a good
idea to actually lay out what all analysis you are going to be carrying out to fulfill your study objectives.
(Refer Slide Time: 12:09)
The next paragraph is quite important and it is actually quite a lengthy part of your protocol and which is
basically, detailing the steps that you are going to use to actually implement the study to ensure data quality
at all stages of the study, in the way you are going to collect the data, in the way you are going to do data
entry and analysis, how you going to report these study results. This is also the place, we are going to give
in detail the various roles and responsibilities that the investigators in the study have. It is also good place
to provide, what would be the project governance? Who would be responsible for this project, both in terms
of administrative responsibilities as well as the technical responsibilities? Who is the PI, the primary
investigator? Who are the co-investigators? What their roles are and how various project activities are going
to be coordinated?
Basically, all the logistics of actually implementing the data collection phase in the field and of course, you
need to also provide, what is going to be the timeline for your study? What is going to be your study period?
How are different phases of your study starting from say, designing the instruments, getting approvals from
ethics committees, then doing pilot testing and then actual collection of data followed by data entry, analysis
and so forth? So, you need to give a timeline over the period of your study or for various activities that you
are going to be doing as part of implementing your study.
(Refer Slide Time: 13:52)
The last paragraph in the methods section is the paragraph on human subjects protection and this is where
you explain in detail, how you are going to make sure that your study participants are protected from any
harm and here, this would usually have sub-headings in terms of what are the potential risks that you foresee
your participants may be going through, when you are collecting data? What are the procedures that you
are going to use to minimize these risks? How you going to maximize the benefits? Are you going to give
them any compensation?
Again, this is the place, where you give in detail of what kind of informed consent are you going to take
from the study participants. Is it going to be a written consent, oral consent? Who is going to take the
consent? How long will it take? What would be the various elements of this consent? And then what would
be the various procedures that you would need to get approval for your study from the ethics committee of
your institution.
(Refer Slide Time: 14:58)
Now, once you have put together all these elements in the protocol, you also need to append to it, all your
data collection instruments depending on your study design, your study objectives and so forth you would
have decided what data collection instruments you are going to be using, whether these are questionnaires
or abstraction forms, a structured observation guide or if you using qualitative methods and either an in-
depth interview guide or a focus group discussion guide, whatever be it. This is the place where you are
going to append all of these instruments to your protocol and it is always a good idea to have these
instruments in local languages as well because that is how in the field you are going to be collecting data.
So, you would have both the data collection instrument in English as well as in the local language.
(Refer Slide Time: 15:52)
And then at the end, what you need to do is to put annexures, which gives in more detail various different
components of actual study implementation. These could include things like forms, which state the various
standard operating procedures of how you are going to be collecting information, whether this is field based
or laboratory based or clinic based and so forth. You may want to provide, how are you going to train your
field workers or train your study investigators in collecting data. What would be the training module and
the framework for the same? If you have participant recruitment material, that is something that gets
appended into the annexes.
Other forms, such as adverse event management forms, study management forms, again would form a part
of your annexures and most importantly would be your consent forms, the participant information sheet as
well as the consent statement form, wherein the study participant, who would be signing off for agreeing
to participate in the study. Again, remember that wherever participants are involved and wherever you have
forms that involve participants, you may need to have forms both in English as well as in the local language.
(Refer Slide Time: 17:14)
And of course, once you have put this altogether, you need to firm it up. Remember that, writing a protocol
is usually not a one step process. It goes through several drafts and it is always a good idea to actually get
it reviewed through your peers, who could be your colleagues or even subject matter experts and get
feedback from them and then revise and revise your protocol to get to a final stage, where it can be submitted
for review to the ethics committee.
Now, the ethics committee may themselves have certain suggestions or they may want some things to be
changed or revised in your protocol. So, you will have to be doing that and so again, this will lead to another
draft of your protocol. By the time you are ready with the final draft, you would actually have a multiple
drafts and it is always a good idea to archive all these drafts so that you know how your study is evolved
and what are the different ways in which you have planned you study.
(Refer Slide Time: 18:24)
Now, as a resource material, I would like you to take a look at this website, which is called equator hyphen
network dot org. And this is a website, which would be very useful for you to develop study protocols. It
has a lot of links and a lot of resource material. One of the important one, which is right there on it’s home
page is what they called the reporting guidelines for main study types and if you look, you have guidelines
for observational studies, for experiments, for clinical trials, for qualitative research diagnostic studies and
so forth.
And you even have one section on study protocols, so there is the SPIRIT protocol; the SPIRIT guidelines,
which basically gives you guidelines on how to write a protocol for clinical trials in particular. And then,
the other reporting guidelines, whether it is the CONSORT statement or the STROBE for the observation
studies, PRISMA for looking at meta analysis and the systematic reviews, although these are called
reporting guidelines, but you could actually use these templates to even make your protocols because
ultimately, what you are going to report is going to based on how you have written your protocol.
(Refer Slide Time: 19:47)
Life cycle of research, which takes you through various steps in which you would do a research study and
please keep this in mind that following this life cycle is vital to have a study that is logical, that is focused
and that is efficient in the way that you are going to conduct the study. So, whenever you are thinking of
writing a study protocol, designing a study, make sure you do not jump across the steps of this life cycle
and you follow the step one after the other and I ensure you that you are going to end up, with a very good
protocol for your study design.
Thank you.
Lecture - 23
Publication Ethics
Hello! I am Sirshendu from ICMR National institute of Epidemiology, I will talk to you on Publication
Ethics.
(Refer Slide Time: 00:20)
At the end of the session I expect the participants will be able to recognize various ethical issues related to
publication and make use of guidelines available from various national and international organizations for
publication ethics.
Unfortunately postgraduate medical research in India often lack of relevant research question, may not
address the local needs, or sometimes have inappropriate design and methods and sometimes not accessible
for others. Often you don’t publish it and so it is difficult to find out the full report. But there are instances,
the postgraduate research can contribute to clinical practice and you can change your practice in field
whichever you are working in.
(Refer Slide Time: 02:54)
With this note let’s come to the main slides of today topic. I have enlisted various components of publication
ethics, but make sure that this list is not final one. There are other issues related to publication ethics and
sometimes, some of these components are overlapping with each other. So, they may not be the separate
entity. So, let’s take a look at the components, 1. Ethics review and breach of confidentiality, 2. Fabrication
and Falsification, 3. Authorship, 4. Plagiarism, 5. Ethics related to submission and conflict of interest.
(Refer Slide Time: 03:37).
The first, Ethics review and Breach of confidentiality: When you are planning for a research you must take
the human and or animal ethics committee approval whichever is applicable as per the national guidelines.
Also, you must take the informed consent or age appropriate accent whenever required. Because unless you
have these two none of the journals are going to accept your manuscript. If you are doing a trial, then you
must register this trial with clinical trial registry of India, and you make sure that you are not disclosing the
data other than people from your institution, or whoever have access to it based on your ethics committee
review. There is a common habit that without institutional permission some PGs used to share their work
or the research work with people from other institution working may be the friend, may be with somebody
else for data analysis. But you must remember that, these are not a good practice and that these are unethical,
because unless you have your institutional permission, you cannot do this, you cannot disclose your
information with outsiders.
(Refer Slide Time: 04:53)
The second one, Fabrication and falsification. If the research results not generated from the study, we call
it as fabrication or if it is generated by manipulating data, we call it falsification. These are extremely serious
misconduct. If your research comes under any scrutiny for any reason, to defend that you must preserve
your data for sufficiently long duration either in paper form or in electronic form.
(Refer Slide Time: 05:33)
I will share an example of falsification and fabrication with you. A final year postgraduate student came to
me requesting to analyse his thesis data. He also asked if I could make some significant results, though they
weren’t initially so. Participants! This is case I felt, the student has done the study appropriately but failed
to get a statistically significant result he was expecting, so he wanted me to manipulate the data to get his
desired findings. This is falsification. Then, curiously I asked how he has collected the data. To my surprise,
he revealed that the data was taken from his senior’s thesis. Now the whole game changed, I understood
that he didn’t even do the study. Either copied completely from someone else or generated the new data
without even doing the study. This is fabrication. It is the responsibility of all the postgraduate students to
be faith full with the research works, while the mentors and the guides should rigorously encourage the
students to conduct the study as per protocol. Make an analysis plan during the designing stage itself and
maintain the same. Further report the same results, whatever you find.
(Refer Slide Time: 06:55)
Falsification and Fabrication happens not only at PG level or at our country, it happens everywhere across
the world. But the most important issue what is there is breach of trust of common people, because they
trust you, you are going to deliver something they trust on it. And also you are losing time and other
resources which you cannot do and that is absolute ethical breach.
(Refer Slide Time: 07:21)
The third one ‘Authorship’, First credit implies responsibility and accountability of your published work.
Each author should be clear about their responsibility, it is mandatory to declare the contribution of each
authors, and this is mandatory nowadays before you sur manuscript to any journal. Decide on authorship
on while writing protocol itself. That gives clarity, and there is no ambiguity, when we do, when you do
right at the beginning, the international committee on medical journal editors recommend that the
authorship should based on four criteria.
(Refer Slide Time: 08:07)
The criteria are number one, substantial contributions to the conception or design of the work; or the
acquisition, analysis, or interpretation of data for the work AND drafting the work or revising it critically
for important intellectual content AND final approval of the version to be published, AND finally
agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy
or integrity of any part of the work are appropriately investigated and resolved. For ex ample if your editor
is asking for some queries, make sure that the responsible person is accountable for answering for these
questions.
(Refer Slide Time: 09:01)
The fourth one ethics related to submission. There are issues with such things, for example when one author
is submitting the same manuscript simultaneously at different journals, we call it as simultaneous
submission. And that should not happen at any point of time, you submit a manuscript to a journal, the
journal gets back to you, the editor reply back to you, and you see, what is the response from the editor. If
it is accepted then fine, if it is not accepted, they may ask a review for it again, review and submit. To all
those things and then you decide whether to be with that journal or choose a different journal. A duplicate
submission is defined submitting a new manuscript with same hypothesis, methods, data discussion and
conclusion which you have published earlier. You cannot do that. Also, it happens that when you are writing
your manuscript, you unnecessarily refers to your self - written manuscript published earlier. That should
not happen, you also be careful about submitting your manuscript to a predatory journal. These predatory
journals hardly peer reviews any manuscript. A list of predatory journal has given by University Grant
Commission Consortium for Academic and Research Ethics (UGC). Follow those list to choose your
journal.
(Refer Slide Time: 10:32)
The fifth one, plagiarism is a very serious issue and I think the most commonly faced by everybody of us.
It is defined as use of previously published manuscript by someone for his or her manuscript or unreferenced
use of others published or unpublished ideas without consent, credit or acknowledgement. As I said, if you
again look back the definition carefully, sometimes people think that is only about copying and pasting. It
is not just copying and pasting, in addition even the idea if you take from someone without acknowledging
without his or her consent that is a serious issue.
(Refer Slide Time: 11:15)
There are various types of plagiarism, like direct plagiarism when you copy from someone else work or
self- plagiarism when you duplicate your own previously published paper. And there is something called
redundant publication, when with the same objective with hypotheses and research methodology and you
are using it for publishing as different paper that is called as ‘salami’ publications. It should not happen of
course. You publish it as a single paper.
(Refer Slide Time: 11:48)
To avoid plagiarism, avoid copy-pasting, in addition write the concept in own words, you spend more time
with your papers, with your references and literature. But don’t copy paste. Acknowledge the original
sources even it is unpublished work, cite the reference accurately, you may use the reference manager for
this, avoid writing articles of same type that is ‘salami’ publication. You can’t do that. And to avoid the
plagiarism finally what you can do is use anti-plagiarism tools like ‘URKUND’ ‘Ithenticate’ or ‘Turnitin’.
But as per it should be as per the university norms, university will have their own criteria and own software
what they are using or subscribing.
(Refer Slide Time: 12:47)
The last one ‘conflict of interest’, financial, personal, social and other interest that directly or indirectly
influence the conduct of the author with respect to the manuscript. For example, a PG or researcher is
conducting a drug trial which is funded by a pharmaceutical company, of course the output can be
influenced by the pharmaceutical company, it is not just financial conflict of interest in addition they may
change the output or outcome of interest. So, make sure that you are identifying those conflict of interest
what you have to do is you have to disclose those conflict of interest during submission. The readers will
decide the influence of such conflict of interest on conclusion of the paper.
(Refer Slide Time: 13:41)
Consequences of any research misconduct can be dealt with seriously. Committee on publication ethics
describes the consequences, it depends on the type of misconduct which is major minor, if it is found to be
a major one then the author can be blacklisted by all the member journal, also they can inform your institute
and the institute can take necessary action against it.
(Refer Slide Time: 14:07)
I have summarized the guidelines related to publication ethics for the academicians in Indian institutions.
What should be the index agencies, what should be the type of articles and the authorship criteria, which is
required for your promotion and career progression.
(Refer Slide Time: 14:31)
So, I have summarized the checklist for publication ethics, when you are planning for publication this starts
right at the beginning when you start your work. Make sure you are getting the approval from your ethics
committee and get the consent or accent, when you are collecting data you make sure that the data accuracy
you have maintained, you have not done any falsification or fabrication of your data, to check plagiarism,
you check with your university or the institution what they are using, and you do accordingly. But make
sure that plagiarism is not there or as minimum as possible.
Thank you.