Validating mortality in the German Pharmacoepidemiological Research Database (GePaRD) against a mortality registry

PURPOSE: The study aimed at validating the status and date of death of a large electronic health insurance database against data of a mortality registry, using a probabilistic record linkage. METHODS: Data were obtained from one local health insurance company contributing to the German Pharmacoepidemiological Research Database (GePaRD). Furthermore, data of the Bremen Mortality Index (BreMI) containing all information from death certificates were used as reference information. Both data sources were probabilistically linked. The study sample consisted of insurants dying in 2005 or 2006. RESULTS: Of 3245 deaths in GePaRD, 83.7% were successfully linked to BreMI records. The linkage success did not differ between men and women, age groups or insurance status groups. Date of death was accurate in 97.1% of all linked deaths. CONCLUSIONS: The accuracy of the status of death in GePaRD is likely to be underestimated in this study because of factors related to the record linkage procedure leading to failure of the probabilistic record linkage approach and the limited completeness of BreMI. A previous validation study comparing aggregate mortality information in GePaRD with German national statistics did not indicate an overreporting of deaths in GePaRD. Thus, a higher accuracy of the status of death in GePaRD than estimated here can be assumed. Copyright © 2016 John Wiley & Sons, Ltd.