Lecture 6 - Decision Analytic Models Flashcards
Decision Analytic Models
Translate knowledge about clinical effectiveness into an estimate of the costs and benefits of different treatment strategies
Always comparative
Generally require a range of extrapolation and modelling strategies to make effectiveness estimates relevant over time
Should reflect how think the disease looks like in the real world
Markov models
People transition between, and stay in, different mutually exclusive health states over discrete time intervals
Markov Models - health states
Circle
Markov Models - transitions
Straight arrow
Markov Models - transitions possible on both sides
Straight arrows going each way
Markov Models - staying in the same state
Curved arrow going back into health states
Discrete Time Interval
Each time a transition probability matrix is applied is called a cycle
Memoryless
Where you go in time t+1 has nothing to do with where you were in t-1
Cohort simulation
Comparing imaginary groups of patients ‘travelling’ through the model
Tunnel states
Series of states you only pass through linearly
Sometimes used to avoid the memoryless property: you can only ‘escape’ after a certain number of cycles
Absorbing states
A state you cannot leave
e.g. death, metastases
Calculating transition probabilities
For each state at time t, the transition probabilities for time t+1 must sum to one
Sending as rows, destination as columns
Transition probabilities
Can vary over time and by group
Markov model to ICER
Assign each health state a cost and a utility value
Sum up all the hypothetical people’s cost-and-utility journeys
Compare the hypothetical intervention group cohort with the hypothetical control group cohort
Uncertainty
Can be statistical
Several assumptions made for models to run
