Models for forecasting chronic disease processes in adult and elderly populations: effects of stochasticity

Forecasting the population health burden of chronic diseases requires models consistent with the relation, over time and in an uncertain environment, of risk factors and diseases at the individual level. There is now sufficient longitudinal data, and scientific understanding, of some chronic diseases to construct detailed process-models to better predict their population health burden and more realistically describe the effects of interventions. A crucial clement in constructing models is the way in which stochastic influences are described, e.g. are they allowed to interact over time with deterministic model features?

A review of statistical and forecasting models aimed to establish what ancillary data and scientific insights are necessary to describe multivariate stochastic health processes and their response to interventions. For circulatory diseases and cancer there exists sufficient longitudinal data and biological insight to construct stochastic multivariate process models. For other diseases, biological knowledge is less complete and there are fewer data sets where multiple risk factors are assessed longitudinally. Forecasting models for those diseases will then rely more heavily on theoretical assumptions about disease behaviour.