Data linkage of German statutory health insurance claims data and care needs assessments preceding a population-based cohort study on nursing home admission

Abstract

ObjectivesWe perform and evaluate record linkage of German Care Needs Assessment (CNA) data to Statutory Health Insurance (SHI) claims data. The resulting dataset should enable the identification of factors in healthcare predicting the time between the onset of long-term care dependency and the admission to a nursing home in Germany in subsequent analyses.DesignA deterministic record linkage was conducted using the key variables region, sex, date of birth and care level. In further steps, the underlying cause of care dependency (International Statistical Classification of Diseases and Related Health Problems, 10th revision (ICD-10)) was added for a higher level of distinction. Before linkage, the suitability of the two datasets for these procedures was assessed. After linkage, the results of each stage were analysed and the resulting dataset was evaluated cross-sectionally with respect to bias generated through this process.SettingThe study comprises data from the German SHI and Statutory Long-Term Care Insurance.ParticipantsThe study cohort comprised 158 069 individuals who became care dependent in 2006. We obtained CNA data for the year 2006 including 188 935 individuals.ResultsWe could link CNAs to 66 310 individuals of the original study cohort, corresponding to 42.0%. Records from two federal states could not be matched due to missing data. Linkage rates were lower where more people shared the same attributes. The resulting dataset showed minor differences regarding age, sex and care level compared to the original cohort.ConclusionsData linkage between German SHI claims data and CNA data is feasible. Failure to link was mostly attributable to a lack of distinction between individuals using available identifiers. The resulting dataset contains relevant information from both health services provision and functional status of care dependent people and is suitable for further analyses with critical reflection of representativity.