A Multistage Incidence Estimation Model for Diseases with Differential Mortality

According to the World Health Organization, surgically removable cataract remains the leading cause of blindness worldwide. In sub-Saharan Africa, cataract surgical rate targets should ideally be set based on cataract incidence (the number of new cataracts developed each year). Unfortunately, the longitudinal studies necessary to measure incidence have not yet been feasible in these areas. Our research instead proposes a method for estimating incidence based on available cataract prevalence data. We extend a method proposed by Podgor and Leske (1986) to estimate age-specific incidence from age-specific prevalence in single diseases with differential mortality. A two-stage disease extension is created in order to differentiate between unilateral cataract and bilateral cataract. The new model, along with a numerical simulation method to generate confidence intervals, is implemented in the statistical programming language R. The model is then applied to Rapid Assessment of Avoidable Blindness survey data from parts of Eritrea, The Gambia, Kenya (two regions), Mali, Rwanda and Tanzania. Our results suggest significant geographic variations in cataract incidence, a hypothesis to be further investigated as the RAAB survey expands and improves. We also show how the model can be further extended to model any n-stage progressive disease with differential mortality.