PRM226 - Incidence and Prevalence Estimations Based on Claims Data – New Methodological Considerations

Abstract

Scientific analyses with claims data such as burden of disease analyses are often based on incidence and prevalence estimates. Latest methodological considerations indicate that the diagnosis-free observation period should be extended as much as possible to not overestimate the incidence in chronic diseases. Aim of this study was to evaluate the impact of expanding the diagnosis timeframe for the incidence as well as the prevalence estimates. This methodological analysis focused on the chronic diseases diabetes mellitus (DM) and multiple sclerosis (MS) in 2013 in Germany and was based on anonymized data from the Health Risk Institute Research Database. Patients continuously insured for six years (2008-2013) were included in the study (n=3,026,154). Incidence changes due to different diagnosis-free intervals before a diagnosis in 2013 (1 to 5 years) were assessed. Correspondingly, the prevalence estimation for 2013 was varied by expanding the timeframe for diagnosis from 1 year up to 5 years, as it was assumed that chronic diseases identified in previous years persist until 2013. Moreover, disease-specific validation of the diagnosis codes was applied as sensitivity analysis. DM incidence was 24% higher when a 1-year diagnosis-free observation period was applied compared to 5 years (25% in MS). When expanding the prevalence timeframe up to 5 years, the prevalence estimation increased by 14% in DM and 21% in MS, respectively. The relative proportion of incidence to prevalence also changed by varying the utilized timeframe. Out of the prevalent diabetes patients in 2013 10.1% were incident when a 1-year timeframe was considered, whereas 7.3% were incident when 5 years were applied (11.7% and 7.9% in MS, respectively). The methodological concepts should coincide when estimating both the incidence and the prevalence of chronic diseases in claims data. Estimates may be biased especially when only short timeframes are utilized.