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

Area of Science:

• Epidemiology
• Health Informatics
• Data Science

Background:

• Electronic health insurance databases are valuable for epidemiological research.
• Validating the accuracy of mortality data within these databases is crucial for reliable research outcomes.
• The German Pharmacoepidemiological Research Database (GePaRD) is a large source of health insurance data.

Purpose of the Study:

• To validate the accuracy of death status and date information in a large electronic health insurance database.
• To compare GePaRD mortality data against a comprehensive mortality registry (Bremen Mortality Index - BreMI).
• To assess the effectiveness of probabilistic record linkage for data validation.

Main Methods:

• Probabilistic record linkage was employed to connect records from GePaRD and BreMI.
• Data from a local health insurance company contributing to GePaRD were used.
• The study focused on individuals who died in 2005 or 2006.

Main Results:

• A successful linkage was achieved for 83.7% of deaths recorded in GePaRD.
• Linkage success rates were consistent across different demographic groups (sex, age, insurance status).
• The date of death was accurate in 97.1% of the successfully linked records.

Conclusions:

• The accuracy of death status in GePaRD may be underestimated due to linkage challenges and registry completeness.
• Previous studies suggest GePaRD does not overreport deaths compared to national statistics.
• Higher accuracy for death status in GePaRD is likely than estimated in this validation study.