38.9 PartSA and Other Analyses

PartSA models within can take advantage of other analyses within TreeAge Pro - cost-effectiveness analysis, sensitivity analysis, etc. Note that PartSA models are by their nature cohort-level models, so you cannot run Microsimulation.

Cost-Effectiveness Analysis

We will use the Partitioned Survival Analysis Example Model PartSA-Decision.trex to demonstrate PartSA models for decision analysis via cost-effectiveness analysis (CEA).

The model above is based on the PartSA-Example.trex model we looked at earlier in detail. However, this model has a decision node with two strategies. The second strategy is a clone of the first one. Two parameters differ between the strategies - the rate/distribution for the PFS state and the cost of chemotherapy.

You can run the PartSA graphs and reports at either PartSA node, but we will focus on CEA. To run CEA, select the Decision node then Analysis > Rankings.

The cost and effectiveness for the first strategy are 613870.53 and 6.5596, respectively, just as they were for the earlier example model. However, the second strategy has a higher cost and a higher effectiveness, reflecting the higher chemotherapy cost and the lower rate of progression from PFS.

Sensitivity Analysis

You can also run all forms of sensitivity analysis on PartSA models - including 1-way, tornado, probabilistic, etc. We will not examine all of these here as those techiniques are described in detail in other chapters. (Deterministic Sensitivity Analysis and Probabilistic Sensitivity Analayis on CE models.) We will, however, run a tornado diagram here.

The following ICER tornado was generated with simple 5% adjustments above and below the base case values for 6 of the parameters.

Note that to run probabilistic sensitivity analysis (PSA), you would need to add parameter distributions to the model.