Diagnostic tests

Research Skill II
Kriengkrai Srithanaviboonchai MD, MPH
Department of Community Medicine
5 November 2020
 Typical steps in taking care a patient

DDx: 1,2,3… DDx: 1,2,3…

DDx: 1,2,3…

History Targeted
physical exam Diagnostic
taking
(signs) test
(symptoms)
a patient

Prescribe
Diagnose
Treatment(s)

DDx = Differential diagnoses

 Diagnostic test
▪ A diagnostic test is a procedure performed to confirm, or
determine the presence of disease in an individual suspected of
having the disease
▪ Diagnostic tests provide doctors with needed information to come
up with proper diagnosis.
▪ Selection of proper diagnostic tests and correct interpretation of
the results is necessary for the accuracy of the diagnosis
 Forms of diagnostic tests
▪ Laboratories: CBC, U/A, Blood sugar, lipid profile,
tumor markers, serology tests, ECG, etc.
▪ Imaging procedures: X-rays, CT, MRI, Ultrasound, etc.
▪ List of criteria (questionnaire) : Apgar score, Glasgow
coma scale, AUDIT, MMSE, etc.
 Reduction of complexity of data
▪ Raw data can be Nominal, Ordinal, Discrete, or Continuous
▪ Data usually be reduced to nominal dichotomous variable in
order to easier interpret and use.
- e.g., disease/no disease, normal/abnormal, positive/negative
▪ It is done by determining the cutoff value.
- e.g., PS ≥ 140 mm.Hg. or PD ≥ 90 mm.Hg. = Hypertension
▪ Sometimes reduced to ordinal variables
Estimated GFR For Staging Chronic Renal Disease

6
 The Gold Standard test
▪ The best diagnostic test for particular condition is called gold
standard test.
▪ It is used to determine whether the patient really suffers from the
disease or not.
▪ Gold standard tests are usually complex, expensive, and may
harm the patients.
 Examples of gold standard test
Target Disorder Gold Standard Diagnostic test
Breast cancer Excisional biopsy Breast MRI
Coronary stenosis Angiography Electrocardiogram
Pulmonary
Tuberculosis Sputum culture AFB stain
Strep throat Throat culture Centor clinical criteria
HIV infection Western blot ELISA, Latex agglutination

8
 The Gold Standard test
▪ The best diagnostic test for particular condition is called gold
standard test.
▪ It is used to determine whether the patient really suffers from the
disease or not.
▪ Gold standard tests are usually complex, expensive, and may
harm the patients.
▪ Therefore, other diagnostic tests are used instead in real life
clinical practices.
▪ The characteristics of these diagnostic tests are then compared
to the gold standard tests.
 Accuracy of a diagnostic test
▪ Diagnostic tests usually not provide the results that 100% correct.
▪ 4 possible interpretations of the test results
- True positive: disease present & test positive
- True negative: disease absent & test negative
- False positive: disease absent & test positive
- False negative: disease present & test negative
▪ Accuracy of a diagnostic test is the proportion of accurate results
out of all results.
Association between test results and disease

Disease
(gold standard)

Present Absent

Positive True False

positive positive
Test
Negative False True
negative negative
 Sensitivity
▪ The proportion of patients with disease who test
positive.
▪ The percentage of sick people who are correctly
identified as having the condition
▪ The diagnostic test with high sensitivity won’t miss the
diagnosis (low false negative)
▪ Sensitivity = Probability of true positive
⁓ Power
⁓ 1 - type II error
 Specificity
▪ The proportion of patients without disease who test
negative
▪ The percentage of healthy people who are correctly
identified as not having the condition
▪ The diagnostic test with high specificity won’t missed
diagnose healthy person as sick person (low false
positive)
▪ Specificity = Probability of true negative
= 1 - type 1 error
 Sensitivity and Specificity
Disease

Present Absent

Positive a b a+b
Test
Negative c d c+d

a+c b+d a+b+c

+d

Sensitivity = a / a+c Accuracy = a+d / a+b+c+d

Specificity = d / b+d
Example: results of new test and gold standard test on
diagnosis of Dengue hemorrhagic fever
Gold standard test
รวม
+ -
+ 250 100 350
New
test
- 50 600 650

Total 300 700 1,000

Sensitivity = a / a+c = 250 / 300 = 0.83 = 83%

Specificity = d / b+d = 600 / 700 = 0.86 = 86%
Accuracy = a+d / a+b+c+d = 850 / 1,000 = 0.85 = 85%
 How to choose a proper diagnostic test
▪ Choose high sensitivity test:
- In early steps of the diagnosis
- To eliminate suspected diseases (negative high sensitivity test
rule out the disease)
- In severe but treatable disease (don’t want to miss the
diagnosis)
 How to choose a proper diagnostic test
▪ Choose high specificity test:
- In later steps of the diagnosis
- To confirm the diagnosis (positive high specificity test confirm
the disease)
- In severe untreatable disease
- In case false positive would
• cost a lot of money for treatment
• affect the patient psychologically or socially
Sensitivity and specific of diagnostic tests are usually fixed
 Predictive Value
▪ When the results are already come out, sensitivity and
specificity are no longer useful.
▪ What the doctor wants to know now is the
probability/chance that the result is correct.
- Positive predictive value (PPV): the proportion of patients
with positive tests who have disease
- Negative predictive value (NPV): the proportion of patients
with negative tests who do not have disease
Predictive Value of a Test Result
Disease

Present Absent

Positive a b a+b
Test
Negative c d c+d

a+c b+d a+b+c

+d

Positive Predictive Value (PPV) = a / a+b

Negative Predictive Value (NPV) = d / c+d
Example: results of new test and gold standard test on
diagnosis of Dengue hemorrhagic fever

Gold standard test

รวม
+ -
+ 250 100 350
New
test
- 50 600 650

total 300 700 1,000

PPV = a / a+b = 250 / 350 = 0.71 = 71%

NPV = c / c+d = 600 / 650 = 0.92 = 92%

 Predictive Value and Prevalence
▪ Predictive
value varies according to the prevalence of
disease in the population.
- If the prevalence increase: PPV will be increase and NPV
will be decrease
- If the prevalence decrease: NPV will be increase and PPV
will be decrease
▪ Sensitivity and specificity are not affected by prevalence.
 Disease
Low prevalence
OGTT + OGTT -

+ ve 90 90 180
Urine
sugar
- ve 10 810 820

100 900 1,000

Sensitivity 90/100 = 90%

Specificity 810/900 = 90%
Accuracy (90+810)/1000 = 90%
Prevalence (90+10)/1000 = 10%
PPV 90/180 = 50%
NPV 810/820 = 98.8%
 Disease
High prevalence
OGTT + OGTT -

+ ve 450 50 500
Urine
sugar
- ve 50 450 500

500 500 1,000

Sensitivity 450/500 = 90%

Specificity 450/500 = 90%
Accuracy (450+450)/1000 = 90%
Prevalence (450+50)/1000 = 50%
PPV 450/500 = 90%
NPV 450/500 = 90%
 Likelihood ratio (LR)
“How likely that a given test result would be
expected in a patient who have the disease compared to
how likely that a given test result would be expected in a
patient who do not have the disease”
 Positive likelihood ratio (LR+)
▪ LR+ = Probability of a person who has the disease testing
positive divided by the probability of a person who does not
have the disease testing positive.

Sensitivity
LR+ =
1 - specificity

Good LR+ if > 10

 Negative likelihood ratio (LR-)
▪ LR- = Probability of a person who has the disease testing
negative divided by the probability of a person who does not
have the disease testing negative.

1 - Sensitivity
LR- =
Specificity

Good LR- if < 0.1

Application of LR
Pre-test Pre-test PrT prob Pre-test
=
probability odds 1 – PrT prob odds

Post-test
= PrT odds X LR
odds

Post-test Post-test
=
PoT odds Post-test
probability probability PoT odds+1 odds
 Application of LR+
Serum ferritin and iron deficiency anemia
If the patient’s chance of iron deficiency anaemia prior to doing the serum
ferritin was 50%
Pre-test odds = Pre-test probability / (1 – Pre-test probability)
= 0.5 / (1-0.5) = 1
If the sensitivity of serum ferritin of 60 mmol/l = 90 % and specificity = 85 %
LR+ = sensitivity / (1-specificity) = 0.90/(1-0.85) = 6
Post-test odds = pre-test odds * LR+ = 1*6 = 6
Post-test probability = post test odds / (1 + post test odds) = 6 / (6 + 1)
= 86 %

The post-test probability of the patient having iron deficiency anaemia is

increased from 50 % to 86 % with positive serum ferritin test
 Application of LR-
Serum ferritin and iron deficiency anemia
If the patient’s chance of iron deficiency anaemia prior to doing the serum
ferritin was 50-50
Pre-test odds = Pre-test probability / (1 – Pre-test probability)
= 0.5 / (1-0.5) = 1
If the sensitivity of serum ferritin of 60 mmol/l = 90 % and specificity = 85 %
LR- = (1-sensitivity) / spec = (1-0.90/(0.85) = 0.17
Post-test odds = pre-test odds * LR- = 1*0.17 = 0.17
Post-test probability = post test odds / (1+post test odds) = 0.17 / (0.17 + 1)
= 8.5 %
The post-test probability of the patient having iron deficiency anaemia is
decreased from 50 % to 8.5 % with negative serum ferritin test
Aortic dissection: high-risk clinical features.
Credit: Kirsty Challen, B.Sc., MBChB, MRes, Ph.D., 31
Lancashire Teaching Hospitals, United Kingdom.
 Multiple tests
▪ Parallel testing (Simultaneous testing)
▪ Serial testing (Sequential testing)
 Parallel testing (Simultaneous testing)
▪ Use and interpret at least two diagnostic tests at the same time
▪ Conclude as positive if either one of the test or both test give
positive results
▪ Used this algorithm when need a prompt result such as at
emergency room or at mobile clinic
▪ increase Sensitivity and NPV decrease Specificity and PPV
 Serial testing (Sequential testing)
▪ Use diagnostic test one at a time, use the second one only when
the first test is positive
▪ Conclude as positive only when both diagnostic tests give positive
results
▪ Use this algorithm when not in rush and want to make sure with
the diagnosis that the patient has disease
▪ increase Specificity and PPV decrease Sensitivity and NPV
Determining Cut off value for continuous lab results
Diseased
Non-diseased

True
True positive
negative

False negative False positive

Negative Positive

Cut off
 Move cutoff point to the left

Diseased
Non-diseased

True
True positive
negative

False negative False positive

Negative Positive

Cutoff
Increase sensitivity / Decrease specificity
 Move cutoff point to the right

Diseased
Non-diseased

True True
negative positive

False negative False positive

Negative Positive

Cutoff
Increase specificity / Decrease sensitivity
 Receiver operator characteristic (ROC) curve
▪ A diagram used to determine proper cutoff point for a
diagnostic test
▪ Plot sensitivity (True positive rate) and 1- specificity (False
positive rate) that the diagnostic test could provide at different
cut off in Y axis and X axis respectively
▪ Area under graph shows probability that the diagnostic test
could provide accurate results.
▪ Better diagnostic test would have larger area under graph and
its peak would be nearer to the left upper corner.
Receiver operator characteristic (ROC) curve
▪ Generally, the cut off on the graph which is nearest to the
upper left corner would be the best cut off
▪ Move cut off to the right along the X axis will give a diagnostic
test with higher sensitivity
▪ Move cut off to the left along the X axis will give a diagnostic
test with higher specificity
 Receiver operator characteristic (ROC) curve

1.0

A
0.8 B

Sensitivity 0.6
(True positive rate)
0.4

0.2

0
0.2 0.4 0.6 0.8 1.0

1-specificity
(False positive rate)