Abstract
BACKGROUND: Surveillance of diabetes requires up-to-date information on the prevalence of diabetes and its complications over time. For this purpose, statutory health insurance (SHI) data is being increasingly used, as the data is available in a timely fashion and case numbers enable detailed estimates also of diabetes complications. The aim of the present study was the development and internal validation of case definitions for the prevalence estimation of diabetic retinopathy (DRP), diabetic polyneuropathy (DPN) and diabetic foot syndrome (DFS). METHODS: Persons with diabetes differentiated by type 1, type 2, and other diabetes in an age- and sex-stratified sample of persons insured by Barmer SHI in 2018 (n=72,744) comprised the study popuation. Based on the central ICD codes for microvascular complications (DRP: H36.0; DPN: G63.2; DFS: E1X.74/.75), case definitions were developed including additional ICD codes for complications without direct diabetes reference. Subsequently, the case definitions were internally validated. For the validation, coding in the inpatient setting (m1S) or repeatedly in the outpatient setting (m2Q) as well as coding of specific procedures (EBM, OPS) and drug prescriptions or by relevant specialists were considered. Additionally, we analysed the documentation of the diagnoses in the previous years. RESULTS: In 2018, the prevalence of the central ICD codes was 8.4% for DRP (H36.0), 18.9% for DPN (G63.2) and 13.4% for DFS (E1X.74/.75). After inclusion of additional ICD codes in the case definition, prevalence increased significantly for DRP (9.6%) and DPN (20.7%), and barely for DFS (13.5%). Internal validation confirmed the majority of diagnoses (DRP: 96.7%; DPN: 96.5% DFS: 95.8%) and m2Q represented the most relevant criterion. When up to four previous years were considered, prevalences were up to 30% higher for DPN and DFS and up to 64% higher for DRP. CONCLUSION: The inclusion of additional ICD codes in the case definition of microvascular complications of diabetes appears meaningful, as this increases the sensitivity of the prevalence estimate. Internal validation suggests that the documented diagnoses are plausible. However, not all diagnoses are documented annually, leading to an underestimation of the prevalence using a cross-sectional study design of one year.