Incidences of disease are of primary interest in any epidemiological analysis of disease spread in general populations. Ordinary estimates obtained from follow-up of an initially non-diseased cohort are costly, and so such estimates are not routinely available. In contrast, routine registers exist for many diseases with data on all detected cases within a given calendar time period, but lacking information on non-diseased. In the present work we show how this type of data supplemented with data on the past birth process can be analyzed to yield age specific incidence estimates as well as lifetime prevalence. A non-parametric model is studied with emphasis on the required assumptions, and a brief outlook on the analysis of the non-stationary case with calendar trends in age-specific incidence is given. The developed methods are applied to case cohort data on treatment with anti-diabetic medications and projections are provided for both diabetes incidence and prevalence. As projection of diabetes prevalence requires estimation of the distribution of disease durations, two novel approaches for this estimation is studied, a parametric and a non-parametric, respectively.



Included in

Epidemiology Commons