Objectives: Patient and health care databases are available in many countries. These are often based on routinely collected diagnosis and prescription data. Various research questions, such as those related to pharmacoepidemiological health services or drug supply, can be evaluated on the basis of these databases. In Germany, the Disease Analyzer patient database is the largest database of its kind. Using various validity criteria, the representativeness of this database is examined with respect to variables relevant to pharmacoepidemiological and pharmacoeconomic studies.
Methods: The Disease Analyzer patient database contains data on diagnoses, prescriptions, risk factors (such as smoking and obesity), and laboratory values for approximately 10 million patients from Germany, the UK, France, and Austria. The database also contains data from various groups of specialist physicians as well as from general practitioners and specialists for internal medicine. Data from physicians' practices in Germany form the basis of this investigation. To check the validity and representativeness of the data, the distributions of several variables are analyzed. These variables refer partly to the physicians' practices participating in the study and partly to the patients in these practices. The factors observed include prescriptions for generic drugs, the distribution of diagnostic groups among participating physicians' practices, the distribution of patients according to health insurance fund, the most frequent products, the distribution of package sizes prescribed, and the age structure of patients with various incident cancer diagnoses. These factors were compared with available reference statistics.
Results: The sampling methods for the selection of physicians' practices appear to be appropriate. Prescription statistics for several drugs were very similar to available data from the pharmaceutical prescriptions report (Arzneimittelverordnungsreport). The age structures for given diagnoses in Disease Analyzer also agreed well with those from corresponding disease registries. Additional comparisons were also in good agreement with data from available sources.
Conclusion: The analyses carried out in comparison with reference statistics find no indication of lack of representativeness or validity of the Disease Analyzer database. In principle, the database appears suitable for pharmacoepidemiological and pharmacoeconomic studies. Development and maintenance of large pharmacoepidemiological databases is needed for modern health services. Such databases allow assessment of health care quality and rare adverse drug effects.