Objectives: A computerised model of the Swedish two county trial of mammographic screening was built to explore the applicability of deterministic group modelling for health policy analysis and to examine trade offs between screening interval and population coverage on breast cancer mortality.
Methods: Powersim system dynamics modelling software running on a PC was used. Model inputs were published data on the populations and screening regimens used in the trial, a Swedish female population life table, incidence of breast cancer in Sweden, 95% confidence intervals (95% CIs) for mean sojourn time and screening sensitivity, and survival after diagnosis.
Results: The model's output--cumulative mortality from breast cancer--agreed closely with trial results. This was robust to uncertainties in key input variables. Furthermore, with hypothetical screening regimes that had a fixed total number of mammograms over a fixed period a positive association was found between more even distribution of mammograms among a population and reduction in breast cancer mortality. For example, screening 50% of a hypothetical population annually produced a 33% reduction in breast cancer mortality, whereas screening 100% of the population every 2 years produced a 48% reduction.
Conclusions: Deterministic group simulation modelling can be used to build reliable, evidence based quantitative models for policy analysis. This opens health policy simulation modelling to epidemiological researchers and will assist them in identifying important information needs--such as breast cancer survival according to sojourn time (the time between cancer being detectable by screening and becoming symptomatic). Scenarios examining reductions in mortality for a given number of mammograms showed that the more equitable the distribution of screening mammograms, the greater the reduction in deaths from breast cancer.