How to run 20+ PSA scenario’s automatically and combine results

Dear all,

We are modelling a medical intervention using a decision-tree / Markov model. We wish to run 20+ scenario’s which differ in the predefined decision tree probabilities. We have the probabilities in an excel-table, there is no relation between each probability set.

How can we avoid changing the probabilities manually between each scenario? Is there a way to automate this process? Is it also then possible to combine the results of each scenario into one CE-plane?

All help is welcome.





  • PSA generally recalculates the model using different sampled values from distributions (sampling rate per EV). However, you can instead pull in data from a table.

    1. Add a scenario column index 1, 2, 3 as a column to the left of your parameter data in Excel.
    2. Load your parameters including the index from Excel into a TreeAge Pro table.
    3. Create table lookups that pull the parameter data from the table using the keyword _sample for the row and the appropriate column for each parameter.
    4. Run PSA. For each model calculation, the table lookups will pull the appropriate values from the table as _sample increments.
    5. The results will be presented in standard PSA simulation format.

    Note that 20+ seems like a very small number of parameter samples for PSA.

    Comment actions Permalink
  • Hi Andrew,


    Thanks for your input. However I'm afraid it's not exactly what we are looking for. 

    Our process would look like this: 

    Scenario 1 : Upload scenario 1 variables in the decision tree -> Run 500 simulations (drawing from distributions present in the markov models) -> Capture outcome of scenario 1

    Scenario 2 : Upload scenario 2 variables in the decision tree -> Run 500 simulations (drawing from distributions present in the markov models) -> Capture outcome of scenario 2


    We want to either fully/partially automate this procedure (because we evaluate 20+ scenario's) or make it easier to input the scenario variables in the decision-tree. Because now we need to manually adjust 15 variables in the decision tree per scenario, which is very time-consuming process and error prone. Seeing that if we need to implement a minor adjustment into model (e.g. a cost adjustment), all the scenario outcomes have to be generated again.

    Thanks in advance! 



    Comment actions Permalink
  • The approach I gave you will do that. The issue is that you will only get the mean values from each calculation after the use of the selected inputs for that run and the distribution samples.

    If you need all the data from every model run (for PSA graphs, for example), then you will need to instead run each PSA analysis separately. In such a case, you would not want to lookup the data from your table by _sample keyword. Instead, you could manually change the row of data to run, then run the PSA. Repeat for the next row of input data.

    Comment actions Permalink

Please sign in to leave a comment.

Didn't find what you were looking for?

New post