Health Checks Microsimulation Model

Requirements

• Python 2.x

• Python libraries:

  • numpy (scientific computing library)
  • multiprocessing (for parallelisation)
  • random

Simulating a population

• Model input parameters: HC_parameters.py

• Main model file: HC_main.py

import HC_main as hc
H = hc.HealthChecksModel(population_size=250000,
                         simulation_time=60, # number of years to simulate
                         HealthChecks=False,
                         randseed=1,         # random number seed 
                         randpars=False,     # draw random parameter values?
                         nprocs=4  # number of CPU cores for parallelisation
                         )
H.Run()

The python object H contains all simulated outcomes.

Examples of extracting aggregate results from the simulated population are given in getresults.py.

The memory of a typical desktop computer permits a maximum of about 250000 people per model run.

Shell scripts are therefore used to simulate larger populations in batches.

Scripts to reproduce the analyses in the paper

Base case only

• Shell script basecase.sh, which calls the Python script basecase.py.

All scenarios

• Shell script scenarios.sh, which calls the Python script scenarios.py.

All scenarios, with statistical uncertainty

• Shell script unc.sh, which calls the Python script scenarios_unc.py. Requires the SLURM high performance computing facility.

Key variables

• nruns in the shell scripts specifies the number of batches.
• ps in the Python scripts gives the number of people in each batch.
• The total population size is then nruns*ps.
• run in the Python scripts is the batch number of the current run. This is passed through from the shell script as a command line argument to the Python script, and is used as the random seed to simulate the current batch.
• randpars (True or False), a named argument to HealthChecksModel(), specifies whether random parameter values are drawn from their uncertainty distributions for each batch. Used to quantify statistical uncertainty. Defaults to False.

Formatting results

• The Python library getresults.py specifies which aggregate results are extracted from the simulated population.

• The Python scripts output aggregate results from the simulated population in text files. These files are stored in the results folder.

• The R code in the file tables.r formats the information in these output files to produce the tables in the paper.