Pathway framework 5 of 8
How to get a decision model by combining the evidence on costs and outcomes
This pathway shows how a health economic model is built, filled with evidence and tested. It ends with how sure a decision maker can be of the answer and whether more research is worth paying for.
12 steps
This framework organises the concepts covered at each step of the pathway on the Knowledge Map.
Concepts from other branches of the Knowledge Tree that this pathway builds on.
Step 1
What a health economic model is, and the main ways a model can be built to represent a disease and its treatment.
7 concepts
Why start here
Every model in the later steps is one of these types, so the pathway starts with what a model is and the choices that shape its structure.
Step 2
The simplest model, where branches show each choice and chance event, and each path ends in a cost and an outcome.
7 concepts
Why this comes next
A decision tree shows the logic of expected values most plainly, so it comes first among the model types.
Step 3
Models where patients move between health states in fixed cycles, with costs and health added up for each state over time.
24 concepts
Why this comes next
Decision trees grow unwieldy when events repeat over many years, and state-transition models handle repeating events in a compact way.
Step 4
Where the numbers in a model come from, and how rates, probabilities and life tables are turned into the values a model needs.
31 concepts
Why this comes next
With the main structures in place, the next task is filling them with probabilities and other values drawn from the evidence.
Step 5
How time-to-event data from trials are analysed, and how survival curves are extended beyond the end of follow-up.
81 concepts
Why this comes next
Trials rarely follow patients for a lifetime, so a model often needs survival curves fitted and extended well past the observed data.
Step 6
Models that take the share of patients in each health state directly from survival curves, often used for cancer treatments.
2 concepts
Why this comes next
A partitioned survival model is built from the fitted curves of the step before, so it follows survival analysis.
Step 7
Models that follow simulated patients one at a time, so each person's history and the timing of events can shape what happens next.
14 concepts
Why this comes next
Some questions need each patient's past to matter, which cohort models capture poorly, so individual-level methods come once cohort models are familiar.
Step 8
Models where the number of people infected changes the risk for everyone else, as with infectious disease and vaccination.
8 concepts
Why this comes next
For infectious disease, treating one person changes the risk for others, which needs a model where the population feeds back on itself.
Step 9
Checking that a model works as intended and matches observed data, and adjusting unknown values until the model fits.
16 concepts
Why this comes next
Once a model is built, its results can be trusted only after it has been checked against the evidence and tuned where data are missing.
Step 10
Testing how much the results change when inputs or assumptions change, one at a time or all together through simulation.
45 concepts
Why this comes next
Every input carries some uncertainty, so after validation the model is run again across plausible values to see how stable its answer is.
Step 11
Putting a money value on reducing uncertainty, to show whether further research is worth paying for.
17 concepts
Why this comes next
Value of information builds on the simulation results of probabilistic sensitivity analysis, so it follows the uncertainty step.
Step 12
The numerical tools that sit behind models, such as matrix algebra, numerical integration and optimisation.
34 concepts
Why this comes next
These methods support calculations in every earlier step, so they close the pathway as a toolkit to dip into when a model calls for them.
Branches that take these methods further.