How the current models fit together

The NCD platform compiles small modules into a runnable model. A module owns one calculation or one coherent part of a calculation. The client selects interventions and edits parameters; it does not contain a second copy of the model logic.

There is one demographic population for each scenario. Disease modules describe marginal disease processes over that population. Their local disease-free values are derived residuals used inside the disease calculation, not separate national populations.

For each run, the compiler builds the same graph for baseline and comparison. The two scenarios differ only in parameter values. It then runs the model one year at a time. At the end of each year, cause-specific mortality changes are reconciled with all-cause mortality, the demographic population advances, and each disease partition is rebased to the surviving population before the next year begins.

Intervention modules own their coverage and effect calculations. Risk-factor modules own exposure changes. The shared risk-factor reconciliation module combines effects that require a common population-attributable-fraction denominator or another shared method, then sends the resulting disease-facing input to the relevant disease module.

Healthy years lived, years lived with disability, years of life lost and disability-adjusted life years are calculated after the annual model run. These metric modules calculate one scenario at a time. A scenario difference is calculated by comparing their separate outputs.

The active disease modules are asthma, chronic obstructive pulmonary disease, cardiovascular disease, diabetes and lung cancer. Their source repositories own the detailed graph, contract, assumptions and limitations.