gms | German Medical Science

Background and state of the art: Health services research uses secondary data often available in annual data sets. Longitudinal analyses require a data linkage method as a prerequisite. Identifying research subjects is a preparatory step for the linkage. Analyzing the annual data sets of the G-BA hospital report cards provided as distinct hospital site data files, the hospital’s identification code and site number are often used for data linkage. However, it is unclear to what extent these variables allow data linkage considering e.g., formation of hospital networks, site closure or restructuring.

Research question and objective, hypothesis: The objective of the presented research is to analyze hospital longitudinal identity and develop a procedure of linking hospitals as a precondition for longitudinal hospital data analyses addressing the impact of minimum caseload regulations (MCR) on hospitals in Germany from 2016 up to 2020.

Method: To prepare the longitudinal hospital linkage, an analytical process is proposed for selecting key variables according to the characteristics of the data set for the intended longitudinal analyses. In a second step a probabilistic-based scoring procedure is tested for longitudinally identifying MCR-related hospitals using German annual hospital report cards data sets from 2016 up to 2020.

Results: Date linkage is established by linking annual data sets year by year chronologically. Two key variables are identified and selected for linkage: hospital identification code and address. Medical departments as a structural feature of a hospital were considered as a further identification variable but was not included in the first and main linkage procedure step since it helped linking only very few hospitals but augmented the single hospital site data preparation for linkage disproportionately due to its extent and variety. Applying this developed linkage method 92.9% - 97.8% of hospitals could be linked between 2016 up to 2020 by identical address and hospital identification code. In a second linkage step the remaining cases with changed address and or identification code were verified using medical department as an additional linkage variable, defined for each of the four types of minimum caseload interventions. And, city/town is set as a tolerance range for address linkage considering the specific objective of analyzing MCR-service provisions. This second linkage step rises the success rate of hospital linkage up to 99.24% - 99.90% depending on the minimum caseload intervention.

Discussion: A linkage procedure could be developed considering the administrative hospital identification code and the hospital address as a geographical variable in combination with provision of MCR-service as a specific variable of the intended longitudinal analyses. The achieved linkage rate above 99% provides a sound basis for the longitudinal analyses.

Implication for research: The linkage procedure with the defined hospital’s longitudinal identity variables and the established identification process can serve as a template for developing future hospital linkage procedures adapted to the specific research questions for longitudinal hospital analyses.

Funding: Innovationsfonds/Versorgungsforschung; 01VSF20032