PERT distribution parameters

Hi everyone,

I am conducting a microsimulation (TreeAge 2023.1.2) and use probabilities that depend on the number of previous hospitalisations and the time since the last hospitalisation. Therefore, I use tables with the rows representing the time since the last hospitalisation and the corresponding rate in the value column. Now I want to conduct a probabilistic sensitivity analysis. I have already found the article about PSA on values that change by cycle (Probabilistic Sensitivity Analysis (PSA) on values that change by cycle – TreeAge Software (zendesk.com) and want to use the second method described there.

I understand that I need to use a PERT distribution and add two columns with high and low values to my tables. If I enter 2 for "likeliest" in the distribution parameters, then the values given in the recent value column of my tables (i.e., the rates) are chosen for the base case analysis. First question: Can I use the PERT distribution for the rates or should I replace the rates in the tables with probabilities? Second question: I am still not sure, what to use for the low and high values in the tables. The rates/probabilities are obtained from literature. Is it reasonable to use the lower and upper limits of a 95% confidence interval as low and high value of the PERT distribution? Third question: Is it reasonable to use the default value of 4 for the shape of the PERT distribution?

I would be grateful for any advice.

Kind regards,

Hannah