gms | German Medical Science

Background: Nearly 90 percent of the German population (about 70 million people) were covered by one out of more than one hundred statutory health insurance (SHI). Information of their prescribed and approved drugs are available as standardized claims data. It is obvious to use this enormous amount of data (with an annual total of more than one billion individual prescriptions) for health services research. Besides to data protection requirements, some methodological challenges must be overcome when using these data.

Materials and Methods: To begin with, these claims data (in German also called secondary data) must be strictly validated prior to their scientific use considering that they have only been checked by the SHI with respect to their primary purpose. Also, the raw prescription data only contain information on the pharmaceutical number, so scientists need an appropriate drug index, which allows a classification of the prescriptions according to the anatomical-therapeutic-chemical (ATC) code. Further, although the outpatient claims data contain specific diagnostic and performance data (ICD- or EBM-coded), a direct link to the prescriptions is missing. In the case of unspecific drugs and those with a broad range of indication, this fact requires internal validation of the data, especially in the case of diagnosis-specific analyzes.

It should also be noted that only prescriptions were documented which were reimbursed by the statutory health insurance. Over the counter (OTC-) drugs or private prescriptions are not covered by claims data. Prescriptions also don’t permit any conclusion on the intake of drugs. So, analyzes of drug adherence or the extent of continuous medication in chronic diseases require certain assumptions. For health economic analyzes, it is relevant that the data usually do not contain any SHI-specific discounts. Finally, it should be noted that there are significant structural differences between SHIs. These differences are relevant in assessing the external validity of the claims data when using data of one or only few SHI’s. When looking on claims data of private insured persons: their prescription data can be assumed as incomplete because usually not all bills are submitted for reimbursement of costs.

Results: The intended use of claims data in drug-related questions therefore should weight up its potentials also with its limitations taking into account possible alternatives. In 2013 a data base with prescriptions of all SHI’s from the morbidity-oriented risk-adjustment sceme has been set up at DIMDI for scientific use, which, however, is subject to certain restrictions. As an alternative to claims data analysis, primary data from health surveys or epidemiological studies are available. But, these data also have limitations depending on the specific collection mode (subjective information of the participants or presentation of the drugs actually taken).

Conclusion: Taking these facts into account several studies try to combine synergistically the potentials of primary as well as claims data by individual data linkage, thereby to overcome the limitations of both individual data sources.

Finally, when evaluating claims data-based studies, general methodological standards for planning, implementation, analysis and reporting are to be considered (e.g. good practice of secondary data analysis, standardized reporting of claims data analysis).