TreeAge Pro supports four types of Monte Carlo simulation.
- Individual Patient Simulation (Microsimulation, Trials)
- Probabilistic Sensitivity Analysis (Samples)
- Samples & Trials (Both 1 & 2)
- Expected Value of Partial Perfect Information (EVPPI)
Individual Patient Simulation (Microsimulation, Trials)
This analysis technique is used to run individual trials through the model rather than evaluating the model at a cohort level. Running trials expands your modeling capabilities by allowing for data to be associated with individual patients.
This data tends to take two forms.
- Individual patient characteristics - allows for a heterogeneous cohort.
- Patient events - allows for event-tracking.
Both types of data can affect cost, utility and probabilities within the model.
Probabilistic Sensitivity Analysis (Samples)
This analysis technique is used to measure uncertainty in the model. A subset of the model parameters are represented with distributions. For each simulation iteration, all parameter distributions are sampled. The sampled values are substituted into the model and the model is reevaluated.
The resulting data set includes a large number of model recalculations, each based on a different set of parameter samples. It is common to look for the percentage of the iterations that confirm the base case results. The higher the percentage, the more confident we can be in our conclusions.
Samples & Trials (Both 1 & 2)
This analysis technique combines the first two techniques described above. This is the analysis to run when you want Probabilistic Sensitivity Analysis (PSA) results, but the model also requires Microsimulation.
In an outer simulation loop, parameters are sampled like in PSA. Once the sampled parameters are in the model, an inner loop runs a trial set through the model via Microsimulation. The resulting data set takes the same form as regular PSA results that do not also require Microsimulation.
Expected Value of Partial Perfect Information (EVPPI)
This is a special form of PSA that includes two separate PSA loops. By running two loops a specific subset of the parameter distributions (often just one) is isolated from the effects of the other distributions. This allows you to determine the uncertainty caused by that distribution subset within the context of additional uncertainty.
It is possible to run Microsimulation with EVPPI resulting in three loops (outer PSA, inner PSA and trials).