This repository provides a simple example to use the methods described in the paper "How many Monte Carlo samples are needed for probabilistic cost-effectiveness analyses?" by Yaesoubi R..
In this simple illustration, we consider a decision problem involving four strategies (A, B, C, and D) in addition to the status quo (O). The cost and effect of each strategy are uncertain but for this illustration, we assume they follow normal distributions with mean and standard deviation provided below:
dict_strategies = {
# mean cost, stdev cost, mean effect, stdev effect
'O': [0, 0, 0, 0, 'green'],
'A': [250000, 100000, 20, 7.5, 'blue'],
'B': [500000, 100000, 10, 7.5, 'red'],
'C': [750000, 100000, 25, 4, 'orange'],
'D': [1250000, 150000, 40, 5, 'purple'],
}
First, let's run the cost-effectiveness analysis (CEA) with 200 samples (see script sample_analysis.py).
The CE plane and net monetary benefit (NMB) lines along with 95% confidence interval are shown below:
Using the approach described in the paper, we can estimate the number of samples needed as follows:
Therefore, if we would like to estimate ICERs within $2,000 of the true ICER with 95% probability, we would need to increase the number of samples to 2,000.
