MODELLING

Question 1

Definition of Economic Evaluation?

Answer 1

A comparative analysis of alternative courses of actions, in terms of both the costs and outcomes (consequences)

• Economic evaluation is the process of measuring cost-effectiveness [Goodacre, 2002]

Question 2

Definition of Decision-Analytic Cost-Effectiveness Model?

Answer 2

A statistical method by which to inform a decision process that incorporates cost-effectiveness analysis

Question 3

What is it important to be aware of in Decision Analysis?

Answer 3

UNCERTAINTY

Decision Analysis looks specifically at decision making under uncertainty, using theories of probability and of expected utility

Question 4

Objective of Economic Evaluation?

Answer 4

Obtain an estimate of the new treatment’s ICER (treatment effect) compared to the existing treatment

Question 5

What is the purpose of Modelling in Economic Evaluation?

Answer 5

Use modelling to ‘draw’ patient pathway through treatment intervention or through a disease … in order to obtain estimate of new treatment’s ICER (treatment effect) compared to existing treatment

‘Models are used routinely to guide public policy decisions in many areas that affect human life and health’ (Weinstein et al., 2001)

‘Models ability to synthesise data from multiple sources and estimate the effects of interventions can be invaluable, especially in areas where primary data collection may be infeasible’ (Bennett et al., 2012)

Question 6

Relationship between Modelling and Clinical Trials?

Answer 6

Modelling allows us to understand what is happening after a clinical trial ends … predict what will happen to the patient (potential effect of the intervention) in the future

• Need to know what will happen to patients when finish trial

Question 7

Main purposes of Modelling in Economic Evaluation?

Answer 7

1. Follow on from clinical trials … predict the potential effect of the intervention/treatment
2. Help deal with uncertainty

Question 8

What is a Clinical Trial?

Answer 8

Clinical Trials are an investigation of a new treatment/intervention on people… an experiment is carried out under controlled conditions to determine in a scientific way which of a number of possibilities is the best
- capture data on costs and effectiveness of a new treatment/intervention

• Randomised clinical trials - “gold standard” method of assessment

Question 9

Why is drug modelling important in England?

Answer 9

There are no clinical trials on certain drugs … can use information from other settings to create a model-based economic evaluation and estimate potential cost-effectiveness of that intervention

Question 10

How does modelling help with Uncertainty?

Answer 10

Model the future effects and costs of an intervention
- Probability of something happening to a patient is not 100% … there is a level of uncertainty … but modelling can help to identify the probabilities

BUT: To construct a model based economic evaluation involves using uncertain information … because colliding information from multiple sources

Question 11

What are the processes of modelling?

Answer 11

1. Inputs - Information sources
2. Model uses natural history (process) of intervention or disease … capture all the potential steps a patient will follow
3. Estimate of ICER … Base Case: model that has the best assumed structure and data

Question 12

What should a good health model look like?

Answer 12

• Be populated w/ the most appropriate and good quality of clinical data (e.g. from meta-analyses)
• Reflect a realistic and credible picture of current clinical practice
• Use the appropriate comparator(s) … to avoid bias
• Be run for an appropriate time period
• Be valid, transparent and reproducible
• Explore uncertainty
• Be easily interpreted

Question 13

Definition of Decision Trees?

Answer 13

A flow diagram showing the logical structure of the problem to help people to make better decisions about interventions and treatments

Question 14

Layout of Decision Trees?

Answer 14

“time flows from left to right” with the branches depicting all the possible patient pathways contingent on particular events

1. DECISION nodes [square nodes]
2. CHANCE nodes [circle nodes]
3. TERMINAL nodes [triangle nodes]

Question 15

What is a Decision Node?

Answer 15

Represents the alternative CHOICES for decision maker

• Alternatives must be mutually exclusive

Question 16

What is a Chance Node?

Answer 16

Represent UNCERTAIN EVENTS that are not under the control of the decision maker – have a probability of something happening or not happening (risk)

• Each of these events has an associated probability and these should sum to 1.0 (100%) for all events associated w/ a particular chance mode
• mutually exclusive and collectively exhaustive [sum to 1]

Question 17

What is a Terminal Node?

Answer 17

Represent the ENDPOINT of the model and are assigned value(s) or payoff(s)

• These payoffs can be utility, costs, or anything else thought to be a critical outcome for a particular scenario

Question 18

Steps in a Decision Tree?

Answer 18

1. Structure/formulate problem by constructing a mathematical model of decision making as a series of connected events
2. Identify decision alternatives by listing the potential outcomes of the alternatives and specifying the sequence in which events might take place … Identify what we want to capture in our model
3. Quantify uncertainty by assigning probabilities to chance events [estimate joint probabilities of events]
4. Quantify preferences by assigning values to all possible outcomes of chance events
5. Estimate expected values of each alternatives and summary measures
6. Interpret the results by finding ICER of decision and then will compare that against a threshold to see if adopt treatment or reject treatment
7. Test results by using sensitivity analyses

Question 19

How do you Quantify Uncertainty of Chance Events in Decision Trees?

Answer 19

Multiply probabilities backwards along the branch

Question 20

How do you Quantify Preferences of Chance Events in Decision Trees?

Answer 20

Add costs along the branches … multiply costs by expected probability

Question 21

How do you Estimate Expected Values of each alternatives in Decision Trees?

Answer 21

Weighted average of each decision found by adding all the results

Question 22

Advantages of Decision Trees?

Answer 22

Accounts for uncertainty in health outcomes - probabilities of chance events

Reliable and effective decision-making technique

It forces the consideration of all possible outcomes of a decision and traces each path to a conclusion

Question 23

Disadvantages of Decision Trees?

Answer 23

There is no time dimension incorporated … describes a fixed process where a fixed sequence of events leads to an outcome

Can become very big if there are many decision options and outcomes

No way for patients to go back to original state – i.e. get ill, get better, then get ill again

Question 24

Definition of Markov Modelling?

Answer 24

A decision-analytic technique which assigns group members to a fixed number of health states and then models transitions among the health states

Question 25

What is a Markov Cycle?

Answer 25

1 Markov Cycle = 1 health cycle | - being in one of the health states for a fixed period of time

Question 26

What are Transition Probabilities?

Answer 26

Given probability of moving health states At the end of each Markov Cycle ... person transits from one health state to another (or remains in same state) based on transition probabilities

Question 27

What is the Absorbing State in Markov Models?

Answer 27

Terminal state - final state of the model - No arrows out of the absorbing state – individuals cannot move out of this state - At the end of the process – most of the patients should be in the absorbing state

Question 28

Steps in a Markov Model?

Answer 28

1. Specify the mutually exclusive health states which should be reflecting the clinical history, be associated with the disease and treatment over time - There is an initial distribution of the population amongst the given health states 2. Define directions of movements (i.e. health → sick → death [absorbing state]) 3. Choose the Markov cycle length [length of one health cycle] 4. Set of transition probabilities should be established which govern the movement of patients between the states 5. Assign utility values and costs to each of the states to create cumulative costs and cumulative utilities 6. The process ends when all individuals are absorbed into the terminal state 7. Calculate ICER using the cumulative utilities and cumulative costs

Question 29

How do you calculate Cumulative Utility in a Markov Model?

Answer 29

QALYs accumulated (total utility) in each cycle = the [sum over the different health states of the number of people in each state] * by the utility value for that state

Question 30

How do you calculate Cumulative Costs in a Markov Model?

Answer 30

Total cost in each cycle = the [sum over the different health states of the number of people in each state] * by the cost value for that state

Question 31

Advantages of Markov Model?

Answer 31

Accounts for uncertainty in health outcomes Incorporates a time cycles … therefore can model how patients change health states over time Markov models are particularly useful for diseases in which events can occur repeatedly over time, such as cancer recurrence, or for depicting predictable events that occur over time, such as screening tests

Question 32

Disadvantages of Markov Model?

Answer 32

Once entered a state, the model “forgets” where the patient came from - Markovian assumption = the transition probabilities depend only on the current health state, independent of historical experience Complicated to work out and time consuming

Question 33

What is the key challenge with health care modelling?

Answer 33

Situations being modelled are changing in ways not explicitly included in the model (Weinstein et al., 2001) - Technologies being modelled, and those to which they are being compared, may change between the time the model is developed and the time data become available - Population characteristics may change as a result of environmental changes or changes in a disease entity E.g: A striking example of the latter was the onset of resistant strains of HIV after initiation of use of antiretroviral therapies such as zidovudine, thereby invalidating the previous model

Question 34

Does modelling remove uncertainty?

Answer 34

Modelling itself comes with uncertainty - Parameter uncertainty: uncertainty about the true numerical values of the parameters used as inputs’ as inputs in an economic evaluation model - Methodological uncertainty: comprises model structure uncertainty [concerns how the elements of the model are fitted together] and modelling process [results from decisions made by the analyst]

Question 35

How will Brexit affect modelling and clinical trials?

Answer 35

[Forster, 2017] - More than 600,000 patients a year could be denied access to potentially life-saving clinical trials after Brexit, medical research organisations have warned. - New regulations are set to make it significantly easier for drug companies to test pioneering new treatments in EU countries – sparking concerns the UK will be “bypassed altogether” when it leaves the bloc. - Many patients, including those with cancer, diabetes and rare diseases, benefit greatly from access to drugs in late stages of testing that are not yet available on the NHS.

Question 36

Hypothetical example for essay question?

Answer 36

Breast Cancer modelling - Petrou and Gray, 2011 Decision Tree example: Breast Cancer Screening Policy - use a decision tree to examine whether it is cost effective to screen for breast cancer every two years compared to not screening Markov Model example: Hypothetical breast cancer intervention - There are three health states: well, recurrence of breast cancer, and dead - Loops indicate the possibility of remaining in a health state in successive cycles - Dashed line indicates the possibility of backwards transition from recurrence of breast cancer to the well state after successful treatment - The cycle length is set at one year