Background: End-stage renal disease (ESRD) incidence and prevalence are increasing in many countries worldwide. Due to the high cost of therapy, predicting future numbers of patients requiring dialysis and transplantation is necessary for health care planners. Projecting therapy-specific chronic disease prevalence is inherently problematic, and examples of suitable models and their application are sparse. When applied, rarely was the adequacy of such models evaluated.
Methods: We describe and illustrate a method for projecting therapy-specific ESRD prevalence in Canada for 1995-2005 using data obtained from the Canadian Organ Replacement Register. The projections combine the Poisson model for incidence rates and a Markov model for patient follow-up. Model adequacy is empirically validated by data-splitting.
Results: Large increases in ESRD prevalence are expected in Canada, with an average annual increase of 6.9% projected for 1995-2005. Upon validation, the projection model based on 1981-1987 data was able to predict 1994 prevalence within 1%, while projected therapy-specific prevalences closely approximated those observed.
Conclusions: Therapy-specific ESRD prevalence was successfully projected using Poisson and Markov models. Where multistate prevalence forecasts are required, the method could be augmented for application to various other chronic diseases.