Normalizing complementary probabilities during sensitivity analysis
Bryony Simmons
I am trying to conduct a one-way sensitivity analysis on my cost-effectiveness model, but I am having some issues. My first node has four branches corresponding to the different genotypes of the disease - I want to vary the prevalence of each genotype in sensitivity analysis, but this inevitable takes the probability of the node either above or below 1. Is there a way to automatically adjust the remaining probabilities to account for the increase/decrease in prevalence of the genotype? I can't use the '#' complementary probability, as I want all the remaining probabilities to adjust. Thank you!
NormalizeProbs.trex
Comments
You can normalize all the probabilities via formulas as the selected probability parameter changes. In the attached model, the adjusted probability values reflect changes to any single parameter by dividing each individual probability by the sum of all the probabilities. As you run sensitivity analysis on any of the three parameters p1, p2 or p3, the adjusted probabilities p1_adj, p2_adj and p3_adj will be normalized.
This approach will work. However, you need to be aware that probabilities will not always be used as you might expect. It is clear when all the values are at the base level of 0.2, 0.3 and 0.5 respectively as no adjustment is needed. However, when p1 is set to 0 (for sensitivity analysis), the adjusted values will be 0, 0.3/0.8 and 0.5/0.8. When p1 is set to 0.1, the adjusted values will be 0.1/0.9, 0.3/0.9 and 0.5/0.9.
Note that with payoff 1 active, results are generated to show you the value of p1_adj. Change the active payoff from 1 to 2 or 3 to see the impact on the adjusted values p2_adj or p3_adj.
Please sign in to leave a comment.