A Bayesian approach to modelling the natural history of a chronic condition from observations with intervention

To assess the costs and benefits of screening and treatment strategies, it is important to know what would have happened had there been no interventio n. In today's ethical climate, however, it is almost impossible to observe this directly and therefore must be inferred from observations with interve ntion. In this paper, we illustrate a Bayesian approach to this situation w hen the observations are at separated and unequally spaced time points and the time of intervention is interval censored. We develop a discrete-time M arkov model which combines a non-homogeneous Markov chain, used to model th e natural progression, with mechanisms that describe the possibility of bot h treatment intervention and death. We apply this approach to a subpopulati on of the Wisconsin Epidemiologic Study of Diabetic Retinopathy, a populati on-based cohort study to investigate prevalence, incidence, and progression of diabetic retinopathy. In addition, posterior predictive distributions a re discussed as a prognostic tool to assist researchers in evaluating costs and benefits of treatment protocols. While we focus this approach on diabe tic retinopathy cohort data, we believe this methodology can have wide appl ication. (C) 1999 John Wiley & Sons, Ltd.