1

CHAPTER III
DECISION ANALYSIS

Bisrat H. ( B-pharm, MSc)

Pharmaceutical supply chain Management
Haramaya University, Harar
 3

PROBABILITY
If I had only one day to live, I would live it in my statistics
class – it would seem so much longer.
Quote in a university student calendar
 Basic concept of Probability
4

Definition :-
 Probability is a numerical measure of the likelihood that an event will
occur.
 Probability of an event P(E): “Chance” that an event will occur:
Notation: The probability of an event E is written P(E) and pronounced “P
of E” Scale: 0 ≤ P(E) ≤ 1
 Properties of probability:
 Must lie between 0 and 1
 “0” implies that the event will not occur
 “1” implies that the event will occur
 The sum of probabilities of mutually exclusive outcomes is equal to 1
 Why Learn Probability?
5

 Nothing in life is certain. In everything we do, we gauge the

chances of successful outcomes, from business to medicine to the
weather
 A probability provides a quantitative description of the chances
or likelihoods associated with various outcomes
 It provides a bridge between descriptive and inferential statistics

Probability

Population Sample
Statistics
 Definition…..cont’d
6

The frequentist school of statistics

 The probability of an event as the number of times the event occurs divided by the
number of trials in which it could have occurred ,n, as n approaches infinity.

 P (A) = [The number of times the outcome occurs]/[Total number of trials]

 For example, the probability that a coin will come up heads is 0.5 because, assuming
the coin is fair, as the number of trials (flips of the coin) gets larger and larger, the
observed proportion will be, on average, closer and closer to 0.5.
 Definition…..cont’d
7

The Bayesian school

 The probability of any event occurring as the personal degree of
belief that the event will occur.

 Therefore, if I personally believe that there is a 70% chance that team

A will win tonight’s game, then that is my probability for this event.
8

Assigning probabilities

I. Classical or theoretical method

II. Relative frequency method.

III. Subjective or judgment method

9

I. Classical method

 Exact outcome is unknown before conducting experiment

 All possible outcomes of experiment are known and each outcome is

equally likely

 Experiment can be repeated under uniform conditions

(e.g. Flip of a coin – outcome of H or T)

10

II. Relative frequency method

 This method is used for an experiment where it is not possible to apply

the classical approach:
Usually because outcomes not equally likely or the experiment is not
repeatable under uniform conditions.
 The probability of an event E is the relative frequency of occurrence of
E or the proportion of times E occurs in a large number of trials of the
experiment.
 Example: The case of Canadian politics:
11

 If I meet an individual Canadian

Party supported Percent support and I know nothing about this
person, the best estimate of the
Conservative 38
probabilities that the person
supports the various parties are:
Liberal 27
 P (individual is a Conservative
NDP 17
supporter) = 0.38
Green 9  P (individual supports the Green
Bloc Québécois 8 Party) = 0.09
12

III. Subjective or judgement method:

 It is an estimate that reflects a person’s opinion, or best guess about

whether an outcome will occur.

 Important in medicine  form the basis of a physician’s opinion (based

on information gained in the history and physical examination) about
whether a patient has a specific disease. Such estimate can be changed
with the results of diagnostic procedures.
 Sample Space and Events
13

 An experiment is an act or process that leads to a single outcome that

cannot be predicted with certainty or a process that generates well-
defined outcomes.
 The sample space, S, for an experiment is the set of all experimental
outcomes.
 Each of these outcomes is referred to as a sample point.
 For an experiment with n sample points, label these E1, E2, E3, ... , En.
14

 A sample space is the set of all possible outcomes in the experiment. It

is usually denoted by the letter S . Sample space can be written using
the set notation, { }.
 Experiment 1: Tossing a coin
 Possible outcomes are head or tail.
 Sample space, S = {head, tail}.

 Experiment 2: Tossing a die

 Possible outcomes are the numbers 1, 2, 3, 4, 5, and 6
 Sample space, S = {1, 2, 3, 4, 5, 6}.
15

 Event: A collection of outcomes for the experiment, that is, any subset
of the sample space.
16

 Simple event: event with only 1 outcome

 Null event: the empty set F
 Certain event: S
 Event Relations
17

 Mutually exclusive
 Complement of an event
 Addition law – intersection and union
 Conditional probability
 Independence
 Multiplication law
 Mutually Exclusive Events
18

 Two events are mutually exclusive if, when one event occurs, the other
cannot, and vice versa when the experiment is performed .

A = 
S

A B
 Law of Complements:
19

The complement of an event A consists of all outcomes

of the experiment that do not result in event A. We write
AC. S

A
AC

 represents the event composed of all basic outcomes in S

that do not belong to A
20

 Law of Complements:
P(AC ) = 1- P(A)
 Example: If the probability of getting a “working” computer is 0.7,
What is the probability of getting a defective computer?
21

 Unions of Two Events “If A and B are events, then the union of A and B,
denoted by AUB, represents the event composed of all basic outcomes in
A or B.”
 The notation for the union is P(AUB).
 Intersections of Two Events “If A and B are events, then the intersection
of A and B, denoted by AnB, represents the event composed of all basic
outcomes in A and B.”
 The notation for the intersection is P(AnB).
22

AUB AnB

If two events A and B are mutually exclusive, then P(A B) = 0.

 Additive Law of Probability
23

Law of Addition
 If A and B are mutually exclusive events, then the probability that

either one of them will occur is the sum of the individual probabilities:
P(A or B) = P(A) + P(B)
 If A and B are two events, not necessarily exclusive or disjoint, then

 p(A or B) = p(A) + p(B)-p(A and B)

24

Example
 Role a six sided Die. The possible outcomes (Sampling space) are six

(1,2,3,4,5,6). Each event has equal probability of occurrence (i.e.

1/6).
 Probability of rolling an even number would be:

 p(even)= p(2)+ p(4)+ p(6)

 = (1/6)+(1/6)+(1/6)=1/2
 WEIRD PROBABILITIES!!
25

You know the number of black and white balls in each pot but can’t see
into the pot. You have one opportunity to take a ball out. If it is black you
win a prize of $50. Which pot would you choose, the red or blue pot?

5 Black 3 Black 6 Black 9 Black

6 White 4 White 3 White 5 White
 WEIRD PROBABILITIES!!
26
5 Black 3 Black 6 Black 9 Black
6 White 4 White 3 White 5 White

5/11 > 3/7 6/9 > 9/14

0.455 > 0.429 0.666 > 0.643
Choose Red Choose Red
 WEIRD PROBABILITIES!!
27
5 Black 3 Black 6 Black 9 Black
6 White 4 White 3 White 5 White

11 Black 12 Black
9 White 9 White
 WEIRD PROBABILITIES!!
28

11 Black 12 Black
9 White 9 White

11/20 < 12/21

0.55 < 0.571
Choose Blue!!!
SIMPSON’S PARADOX
 Simpson’s paradox
29

when using two-way tables – tables set up to display the relationship

between two categorical variables – an association or comparison that
holds for all of several groups can disappear or even reverse direction when
the data are combined to form a single group
 Conditional probabilities
30

 A conditional probability refers to the probability of an event A

occurring, given that another event B has occurred.

P( A  B)
P( A | B)  if P( B)  0
P( B)

 In words, for any two events, the conditional probability that one event
occurs given that the other event has occurred equals the joint
probability of the two events divided by the probability of the given
event.
 Conditional probability
31

 What is the probability that a neonate died? Clearly, this depends on

a number of factors, including maternal wellbeing, gestational age,
birth weight, the presence or absence of neonatal illness etc.

 The probability of neonatal death is conditional on other factors and is

not a single constant number by itself.

 Such probabilities are known as conditional probabilities

 Conditional probability…
32

 General Notations
 If we denote one event by A, and a second event by B, then
 Pr (A | B), is given by dividing the unconditional probability that these
two events occur together by the unconditional probability that B
occurs
 P(A\B) = P(A∩B)/P(B)
 This formula is conveniently rewritten as the following which follows the
Multiplicative Rule.
 P(A∩B) = P(A\B) X P(B)
 Conditional probability…
33

 Example : Assume that the overall probability of a patient in the clinic

having endometrial cancer is 10%. Furthermore, suppose that the sensitivity
of endovaginalsonography for diagnosing endometrial cancer is 90%.
 If we let A represent the event that the patient has a positive test result and
let B represent the probability of endometrial cancer in this patient
population
 i.e. Pr(B) = 0.1 and Pr(A | B) = 0.9
 The probability that a patient in this clinic has both endometrial cancer and
positive results is
 0.1 ×0.9 = 0.09 or 9%
 Defining Independence
34

 We can redefine independence in terms of conditional probabilities:

Two events A and B are independent if and

only if
P(A|B) = P(A) or P(B|A) = P(B)
Otherwise, they are dependent.

• Once you’ve decided whether or not two events are independent,

you can use the following rule to calculate their intersection.
 Example
35

If tuberculous meningitis had a case fatality of 20%,

A. Find the probability that this disease would be fatal in two randomly selected
patients (the two events are independent)
B. If two patients are selected randomly what is the probability that at least one of
them will die?
Answer
A. P(first die and second die) = 20%  20% = 0.04
B. P(first die or second die)
= P(first die) + P(second die) - P(both die)
= 20% + 20% - 4%
= 36%
36
 37

DECISION ANALYSIS IN HEALTH

38
39

 Decision analysis is the application of an analytical method for

systematically comparing different decision options.

 Decision analysis graphically displays choices and facilitates the

calculation of values needed to compare these options.

 It assists with selecting the best or most cost-effective alternative. This

method of analysis assists in making decisions when the decision is
complex and there is uncertainty about some of the information.
40

A. Explicit
 It forces the decision maker to separate the logical structure into its
component parts so that they can be analyzed individually, then
recombined systematically to suggest a decision
B. Quantitative
 The decision maker is compelled to be precise about values placed
on outcomes
C. Prescriptive
 It aids in deciding what a person should do under a given set of
circumstances
 As a set of methods, decision analysis can satisfy five important
objectives for any economic evaluation
41

1. Structure
☞ It can provide a structure that appropriately reflects the possible prognoses
that individuals of interest may experience, and how the interventions being
evaluated may impact on these prognoses.
2. Evidence
☞ It offers an analytic framework within which the full range of evidence
relevant to the study question can be brought to bear.
3. Evaluation
☞ It provides a means of translating the full extent of relevant evidence into
estimates of the cost and effects of the alternative options being
compared.
42

4. Uncertainty, variability, and heterogeneity

☞ It facilitates an assessment of the various types of uncertainty relating to
the evaluation. Models also provide flexibility to characterize
heterogeneity across different subgroups of individuals.
5. Future research
☞ Through the assessment of uncertainty, it can estimate the value of future
research, and identify likely priorities for it.
43

 Historically, it was developed as a method to help clinicians make

decisions on how to manage individual patients about risk, benefits
and costs.
 This general set of methods was later adopted for structuring and
analyzing collective decisions in health care such as program
evaluation and economic appraisal.

 DA yields an estimate of the net value of the different decision options

in relation to each other.
 Decision Analysis
44

 Decision analysis is used when:

 There are real alternatives
 There is uncertainty
 When consequences are important or have series consequences

 Information from decision analysis can be used to

Decide on how to manage individual patients
Formulate policy recommendations about group of similar patients
45

 Provides the conceptual foundation for cost-effectiveness, cost-utility,

and cost-benefit analyses
 May utilize
 expert opinion,
 literature-based results,
 data from prospective or retrospective analyses
 Key Attributes of a decision Analysis
46

 All relevant aspects of the decision and  Alternative course of action

relevant strategies are explicitly  Novel treatments
articulated:
 Standard of care
 Base Case
 Watchful waiting
 Perspective
 Chance events
 Time Horizon
 Probability of treatment failure
 Treatment strategies
 Probability of adverse reaction
 Probabilities
 Consequences
 Costs
 Economic, Clinical or Humanistic
 Outcomes outcomes
 Uncertainty
47

 DA for economic evaluation proceeds by careful structuring of the problem

using decision tree –a graphic scheme where we begin with the decision
and trace out all probable pathways and consequences.
 Advantages using decision tree are:
 Analyst can quickly identify what data components are required (probabilities,
cost and utilities)
 Laying out the problem in this way helps to separate issues of facts from values
 Disadvantage
 Various pieces of information from different studies and population are put
together into the same model.
 Expected Monetary Value (EMV) Example
48

Estimated monetary value(EMV) is a type of decision tree where you calculate the expected
monetary value of a decision based on its risk event probability and monetary value
 Decision Analysis
49

What is a Model?
 Mathematical structure that represents the health and economic
outcome of patients or population under variety of scenarios.
 Fundamental analytic tools used in DA to display the temporal and
logical sequence of decision problem
 Various Types of PE models like
 Decision Tree
 State Transition Markov model
Uses
 to offer answer to a specific PE decision via evaluating the expected
costs and outcomes of alternatives
 to evaluate “what if” scenario via changing key assumptions
 Types of models
50

 Descriptive
 describes

 Prescriptive
 suggests

 Deterministic
 certainty

 Stochastic (decision tree and Markov modelling )

 probability
51

 The choice is dependent upon:

• The complexity of the problem,
• The need to model outcomes over extended periods of time, and
• Whether resource constraints and interactions of various elements in the
model are required.
 Decision Trees
 Markov Models
 simulation Models
 Deterministic (Mechanistic) Models
 Example deterministic, prescriptive model
52

Become ill

See a doctor Do not see a doctor

Obtain a prescription

Take medication

Recover rapidly Slow recovery

 Example stochastic, prescriptive model (decision-tree)
53

Rapid recovery Leaves

(p)
See doctor
Decision node
1-(p)
Slow recovery

Become ill

Rapid recovery
(q)
Branches
Do not see doctor
1-(q)
Slow recovery
Chance node
 Decision Analysis Models
54

Concerns about modeling

 The data that are used to operationalize the model are from diverse
sources, many of which are subject to varying degree of bias due to
confounding variables, patient selection or method of analysis

 For any modeling endeavor, a number of key assumptions have to be

made that might not meet our specific problem
 Steps in Conducting Decision Analysis
55

Step 1 • Frame the Question

Step 2 • Structure the Clinical Problem
Step 3 • Estimate the probabilities
Step 4 • Estimate the values of the Outcomes
Step 5 • Analyze the Tree (Average Out/Fold Back)
Step 6 • Test assumptions (Sensitivity Analysis)
Step 7 • Interpret the Results
 Decision Trees
56

 A decision tree is a diagrammatic representation of the possible

outcomes and events used in decision analysis

 The questions to be asked in an analysis of a question are arranged as

a series of decision or chance nodes, each node with resulting branches,
creating a tree effect

 The sequential steps proceed with each step depending on the decision
or probability outcome from the preceding step
 Example
57

Fig: Decision tree for breast cancer screening options

58

 A decision model comprises the modelling structure itself (the decision

tree), which represents:
The decision that is being made and the outcomes that can occur as
the result of each decision,

The probabilities that the various outcomes will occur, and

The values of the outcomes if they do occur.

 Decision Analysis
59

Structuring the Problem

 What are the major issues involved in the decision? Must be specified
 What is the perspective of this analysis?

 Perspective could alter the structure . who bears costs and benefits?
 What are the real alternatives?

 Include realistic alternatives that should be exhaustive.

 Even if “doing nothing” is not a viable clinical alternative, it is useful to
include such a strategy as a baseline check of the model’s ability to
predict the outcomes of the natural history of untreated disease.
60

 Determine analytic horizon?

 It indicates the period of time over which the specific strategies are
chosen and the relevant outcomes occur.

 Short-term vs. long-term (effects of various interventions to alter

cardiovascular risk)
 Complexity versus simplicity
 Decision Analysis
61

 To structure a problem, Decision Tree is usually constructed that

 Guided by a number of conventions

 Built from left to right

 Consists of
 nodes which can be decision or chance nodes
 branches and
 outcomes
 Decision Node

- By convention depicted as square

- Points where the alternative actions are under the control of the decision
maker.
 Decision Analysis
62

Chance Node
• Refer to one or more several possible events that are beyond the control
of the decision maker

• Defines events that are mutually exclusive and jointly exhaustive

• By convention depicted as circle

• Chances for the similar events should line up horizontally on decision tree
 Decision Analysis
63

• Outcomes
– Consequence of the final event depicted in the tree

– May include life or death, disability or health, positive or negative,

utility measures or any of a variety of risks or outcomes
– Depicted as triangle

• Branches
– Conventionally drawn at right angles to the nodes

– They connect nodes with the nodes and nodes with the outcomes
 Decision Analysis
64

• Decision tree is analyzed by a process called averaging and folding back.

• Can use specialized software programs or simple spreadsheet
• Calculation of Expected Value (EV):
– Averaging out could be made as
• [(probabilility of event A) * (outcome value A)] + [(probability of event
B) * (outcome value B)]
• Weighted average= sum of products/sum of probabilities at a given
chance node
– Folding back means conducting rollback calculation
 Decision Analysis
65

Stepwise rollback calculations are

 After quantifying chances and outcomes, expected value for each strategy
is computed for comparison
Steps:
1. The value of outcome for each branch is multiplied by its respective
probability
2. At each chance node, the products for the entire branch (calculated in
step 1) are summed
3. Continue the process of averaging-out and folding back to the left side
of the decision tree
66

Sensitivity analysis

 Systematically examining the influence of uncertainties in the variables

and assumptions employed on the estimated results
 E.g. change in a unit cost value of 10% lead to change in result of >10%
(sensitive) or <10% (insensitive)?

 When estimating costs and outcomes, you typically have a range of

possible values.
 Decision Analysis
67

Sensitivity analysis

• Always included in decision tree analysis

• The overall goal of SA is

– To assess the stability of the conclusion made in the analysis
– Identify crucial areas of information deficiency and
– To guide future research
 Decision Analysis
68

 Information gathering for DA almost always uses one or more of the

following
 Literature review including meta analysis
 Primary data collection
 Consultation with experts
69

Once :

 The analysis has been completed,

 The stability of the model has been tested with sensitivity analysis, and

 A modeler is convinced that the model represents the clinical and

pharmacoeconomic characteristics of the problem adequately,

 Then the results must be interpreted and summarized.

 Example
70

This lady lives in a big city. Each day she

has to get up early in the morning and rush
to work. She has to make decision on
which modes of transport she has to make.
Lets apply the steps for DA for her decision
problem
71

• The lady has a choice among 3 options for commuting to work:

– Buy a monthly pass and take the train every day

– Ride your bike and take the train only on bad weather days

– Drive your car

• Goal: to minimize commuting costs ( NB: other goals may also be

considered)
72

Decision Node
73

Going to work Example

• Chances for the same

events should line up
horizontally on decision
tree
 Outcome
74
75

Chance Events
• p Rain = 0.20
• p Street = 0.50
• p Vandal = 0.004
Outcome of interest (cost per day in our example)
• Costs per day:
– Monthly train pass = $1.80
– Daily train pass = $2.00
– Bike = $0.25
– Gasoline = $0.60
– Space rental = $2.00
– Garage = $5.00
– Vandalism = $100.00
76
77

(0.20*2.00)+(0.80*0.25)
= $0.60

(0.004*100.60)+(0.996*0.60)
= $1.00
78

= $0.60

= $1.00

= $3.30
(0.50*1.00)+(0.50*5.60)
79

=$1.80
=$0.60

= $2.60
 Quiz
80

1. List methods to assign probabilities?

2. Describe what mutually exclusive event mean?

3. Write the 5 objectives that a decision analysis intends to address?

 ii. Markov Modelling
81

 Used when disease progresses over time

 Patients grouped into a finite number of (Markov) states and Time

progresses in equal increments (Markov cycles)

 All events or progression are represented as transitions from one state

to another with a certain probability
 Transitions (probability of improvement or deterioration) calculated from
epidemiological and/or clinical data
82

 A Markov process is simply a mathematical representation of the

health states in which a patient might find him- or herself and the
likelihood of transitioning between those states.
83

 Spending one cycle in a given state is associated with a certain cost

and a defined outcome
 Decision trees, which represent sequences of events as a large number
of potentially complex pathways,

but Markov models permit a more straightforward and flexible

sequencing of outcomes, including recurring outcomes, through time.
84

 Patients are assumed to reside in one of a finite number of health

states at any point in time.

 Patients are assumed to make transitions between those health states

over a series of discrete time intervals or cycles.

 The probability of staying in a state or moving to another one in each

cycle is determined by a set of defined transition probabilities.
85

 The definition and number of health states and the duration of the
cycles will be governed by the decision problem:
☞one study of treatment for gastro-oesophageal reflux disease used
one month cycles to capture treatment switches and side effects,

☞an analysis of cervical cancer screening used six monthly cycles to

model lifetime outcomes.
86
 Fig: Markov state diagram and transition probability matrix for hypothetical breast
cancer intervention. Source: Petrou S , Gray A

87
88

 There are three health states:

Well, recurrence of breast cancer, and dead.
The probability of moving from the well state at time t to the
recurrence state at time t+1 is 0.3, while the probability of moving
from well to dead is 0.1.
At each cycle the sum of the transition probabilities out of a health
state (the row probabilities) must equal 1.
89

 In order for the Markov process to end, some termination condition

must be set.

 Could be a specified number of cycles, a proportion passing through

or accumulating in a particular state, or the entire population reaching
a state that cannot be left (in this example, dead); this is called an
absorbing state
90

 An important limitation of Markov models is the assumption that the

transition probabilities depend only on the current health state,
independent of historical experience (the Markovian assumption).

 In this example, the probability of a person dying from breast cancer is

independent of:
 The number of past recurrences and

 How long the person spent in the well state before moving to the recurrent state.
91

 The final stage is to assign values to each health state, typically costs
and health utilities.

 If the time horizon of the model is over one year, discounting is usually
applied to generate the present values of expected costs and
outcomes.
 Danny’s ear infections
92

 If well this week, 80% chance well next week

 If infected this week, 50% chance infected next week

 Probabilities remain constant from week to week

 Danny’s ear infections….
93

 Suppose well this week. What is:

p (well next week)?
 p (well in 2 weeks)?
 p (well in 10 weeks)?

 Long-run averages:
 how many infections annually?
 length of infected episode?
 % time infected?

 How might prophylaxis alter results?

 Assessment of prognosis
A critical factor in decision making
94

☞Decisions must be made over long time horizons

☞Risks are continuous over time
☞Important events may happen more than once
 Estimates are required of:
☞Life expectancy
☞Event-free survival
☞Frequency/length of hospitalization
 Sensitivity to alternative interventions
 Markov models Central idea
95

 States of health
Finite
Mutually exclusive
Collectively exhaustive
Capture disease characteristics (prognosis, cost, quality of life, etc.)
 Transitions
Probabilistic
One each cycle
 Markovian Assumption
96

 Probability of transition depends only on current health state and not

on past health states (memoryless property)

 Health state definitions should include all relevant history

history of precancerous lesion

history of acute myocardial infarction

time since HIV seroconversion

 Building a Markov Model
97

 Determine health states

 Determine transitions

 Choose cycle length

 Estimate transition probabilities

 Estimate state utilities and costs per cycle

 Calculate

 Sensitivity analysis
 Health States
98

Well Sick
 Building a Markov Model
99

 Determine health states

 Determine transitions

 Choose cycle length

 Estimate transition probabilities

 Estimate state utilities and costs per cycle

 Calculate

 Sensitivity analysis
 State-transition diagram:
100

W S
 State-transition diagram:
101

WELL SICK

DEAD
 Modeling devices:
Initial, transient, and absorbing states
102

W S R

D
 Modeling devices:
103
Tunnel states

Pre- Post Post Post

Op
op Op 1 Op 2 Op 3

D
 Building a Markov Model
104

 Determine health states

 Determine transitions

 Choose cycle length

 Estimate transition probabilities

 Estimate state utilities and costs per cycle

 Calculate

 Sensitivity analysis
 Cycle
105

A brief time interval during which patients within a cohort may make a
transition into another health state or remain in the current health state.

 Length of cycle chosen to represent a clinically meaningful time

interval
 If time horizon is patient lifetime, then cycle is usually one year

 If events occur more frequently, cycle can be a month or even a week

 Building a Markov Model
106

 Determine health states

 Determine transitions

 Choose cycle length

 Estimate transition probabilities

 Estimate state utilities and costs per cycle

 Calculate

 Sensitivity analysis
 Transition Probability
107

 The chance that patients in a particular health state will transfer to

another health state during the course of a cycle.
 State Transition Diagram
Pij = transition probability from state i to state j
108

P 11

P22
P12
WELL SICK

P13 P23
DEAD
1. WELL
2. SICK
P33 3. DEAD
 Building a Markov Model
109

 Determine health states

 Determine transitions

 Choose cycle length

 Estimate transition probabilities

 Estimate state utilities and costs per cycle

 Calculate

 Sensitivity analysis
 Health State “Utility”
110

 The value of occupying a particular health state for one

cycle.
A value of 1 will calculate life expectancy
A value of 1/(1+r)cycle will calculate discounted life
expectancy
A utility value (between 0 and 1) will calculate quality-
adjusted life expectancy
A cost value will calculate lifetime costs
 Quality-of-Life Adjustments
111

1.0 0.5 0.0

WELL SICK DEAD

 Building a Markov Model
112

 Determine health states

 Determine transitions

 Choose cycle length

 Estimate transition probabilities

 Estimate state utilities and costs per cycle

 Calculate

 Sensitivity analysis
 Common Methods of Evaluation
113

I. Cohort simulation
► Hypothetical cohort of patients transition through the model
simultaneously

II. Monte Carlo simulation

► First order simulation randomly selects a patient from the
hypothetical cohort and they transition through the model one at
a time
114

III. The Markov-cycle tree

►Uses a tree representation of clinical events and may be
evaluated either as a cohort simulation or as a Monte Carlo
simulation
115
 Cohort simulation
116

 Hypothetical population cohort is assigned to initial

states

 Transition probabilities re-distribute the population

at each cycle

 Can compute average patient-cycles for each state

Cohort Simulation can Estimate Survival Curves

Time

t WELL SICK DEAD

t+1 WELL SICK DEAD

 Quality-of-Life Adjustments

1.0 0.5 0.0

WELL SICK DEAD

 Transition Probability Matrix:
119

WELL SICK DEAD

WELL 0.75 0.20 0.05
SICK 0 0.70 0.30
DEAD 0 0 1

Suppose everyone starts in the WELL state

 Cohort Simulation
120

Cycle Total
Cycle Well Sick Dead QA life QALE

0 1 0 0 0 0
1 0.75 0.20 0.05 0.85 0.85
2 0.56 0.29 0.15 0.71 1.56
3 0.42 0.32 0.26 0.58 2.14
4 0.32 0.31 0.38 0.47 2.61

N 0 0 1 0 4.33
 Monte Carlo simulation
121

 Considers hypothetical population one individual at a time

 At each cycle, a random number generator simulates transitions

and used together with transition probabilities to determine to
which state the patient will transition

 Repeating this process for large enough sample produces point

estimates and variances
 Monte Carlo Simulation
122

Well Sick Dead

Well Sick Dead

Well Sick Dead

WWSD
Well Sick Dead
 Markov Approximation to LE

1
Survival Probability
0.8 Life Expectancy = Area Under the Survival Curve

0.6
0.4
0.2
...
0
0 5 10 15 20
Time (years)
123
 The DM analysis will model the cost-effectiveness of using a specified diet and exercise plan to
increase the length of time that prediabetic patients (impaired glucose tolerance [IGT] is plasma
glucose >140 to <200 mg/dL 2 hours after glucose challenge) avoid the transition to DM
124

Rascati et.al
 Example Calculations:
125

Cycle 1 to Cycle 2
 70% of 100% stay well = 70% well

 20% of 100% get sick = 20% sick

 10% of 100% die = 10% dead

Cycle 2 to Cycle 3
 70% of 70% stay well = 49% well

 20% of 70% (14%) get sick plus 60% of 20% stay sick (12%) = 26%

sick
 10% of 70% (7%) die plus 40% of 20% (8%) die + 100% of 10%

(10%) stay dead = 25% dead

126

 Cycle 3 to Cycle 4
 70% of 49% stay well = 34% well
 20% of 49% (10%) get sick plus 60% of 26% (16%) stay sick =
26% sick
 10% of 49% (5%) die plus 40% of 26% (10%) die + 100% of 25%
stay dead (25%) = 40% dead
127

 QALY Calculations
 Cycle 1 = 100% * 1.0 QALY = 1.00 QALY
 Cycle 2 = (70% * 1.0 QALY) + 20% (0.5 QALY) + 10% (0 QALY) =
0.80 QALY
 Cycle 3 = (49% * 1.0 QALY) + 26% (0.5 QALY) + 25% (0 QALY) =
0.62 QALY
 Cycle 4 = (34% * 1.0 QALY) + 26% (0.5 QALY) + 40% (0 QALY) =
0.47 QALY
 Benefits of a Model in CEA

Incorporates the benefits and costs beyond time horizon

of existing data

Considers all relevant clinical strategies

Incorporate data from multiple sources

Evaluates “what if” scenarios

 Extend Beyond Time Horizon
129

 Can “translate” an intermediate endpoint (e.g., increase in CD4 count,

decrease in blood pressure) into clinical endpoints such as life years
saved or quality-adjusted life years gained

 Incorporates important long-term effects of treatment

 Consider All Relevant Strategies
130

 Comparison of interventions often not compared head-to-head in a

trial

 pharmaceutical vs. educational interventions

 Clinical trials do not always consider those strategies which would be

most relevant in a cost-effectiveness analysis

 Too many plausible strategies to be feasible for a clinical trial

 Data from Multiple Sources
131

• Clinical trial data from primary sources

• Existing databases
• Studies reported in the medical literature
• Expert opinion
 “What If” Scenarios
132

 How great would the diagnostic accuracy of a noninvasive imaging

test for coronary arteries have to be to replace the tests used in
current practice?

 How effective does a treatment need to be, and at what duration of

treatment effect, to replace current therapy?

 Identifies important gaps in our knowledge

133

Bisrat H