36.4 Reporting - Dynamic Microsimulation Analysis

We will use the same model as the previous sections: the Healthcare tutorial example tree Dynamic Cohort - Markov Cancer.trex. This model can be run as a cohort model, as in the previous sections. We will use the model and run as a Microsimulation model to demonstrate the differences in dynamic reporting for a Microsimulation model. This will obviously not demonstrate heterogeneity and event tracking, but the methods are still the same.

To run a Dynamic Microsimulation model:

  • Select the root node and then select Analysis > Monte Carlo Simulation > Microsimulation or use the toolbar to select microsimulation.

  • The dialogue for running Monte Carlo simulation will appear. See options below the figure.

  • Select the 'Value of trial set(s)'. Select Sum to return the value for the whole population.

  • Select Begin to run the Microsimulation. (Results are in the figure below)

Consider the elements of the Monte Carlo Simulation dialogue for a Dynamic Microsimulation model:

  • 1st-order simulation trials (microsimulation): This value is fixed as 1 for a Dynamic model because the non-coherent branch probabilities within the model determine the population size.

  • Value of trial set(s): The two options are Average and Sum. The Average option will give you the model rewards (cost and effectiveness) as the average over the population - similar to usual microsimulation results reporting the average value for a population. The Sum option will add together the costs and effectiveness for the whole population and report the total value. This option is likely more appropriate for a dynamic cohort model. For this analysis, choose Sum.

  • Parallel trials options: Select this option to run trials in parallel. See Microsimulation with Parallel Trials Chapter. This defaults to checked because this is a dynamic cohort model.

  • Trial sets options: Select this to run more than one set of trials and set the number of trials to run. For this analysis, leave this unchecked.

The model has the Time Reporting feature turned off because there is a large population and this would slow the model down to report all the outcomes.

The Microsimulation outputs

The Microsimulation output's primary output is the sum of the entire trial set, so the top row shows very large cost and effectiveness values. If we had instead chosen to show Average instead of Sum, the primary/top results would be much smaller.

The calculations for both Tx 1 and Tx 2 are using the same method, so considering the Rankings Report from the Actions on the right hand side will still generate an ICER which we can use to recommend the optimal treatment. Note that the full "Sum" values are used for Rankings in this case.

Note that Dynamic and Parallel trials run together, so this analysis actually runs the trials in parallel. This does significantly slow down the simulation because it is run on a single processor thread.