[Are Prevalence Rates Comparable in Survey and Routine Data? Prevalence of Myocardial Infarction in Claims Data of the AOK Lower Saxony and in Data of German Health Interview and Examination (DEGS1)]. / Sind Prävalenzen in Survey- und Routinedaten vergleichbar? Herzinfarktprävalenzen in Krankenkassendaten der AOK Niedersachsen und in Daten der Studie zur Gesundheit Erwachsener in Deutschland (DEGS1).

AIMS OF THE STUDY: This study compared prevalences of myocardial infarction between data drawn from health interviews and claims data from statutory health insurance. Previous comparative studies have drawn comparisons without having considered possible differences in the sociodemographic structure of the underlying study populations. The approach applied here aimed to match the sociodemographic structure via available information in both datasets and to compare prevalences in parallelized samples. METHODOLOGY: Data from the German Health Interview and Examination Study for Adults (DEGS1) and claims data from the AOK Lower Saxony (AOKN) were used. To match the sociodemographic structure of the two data sets, a parallelized sample was drawn from the AOKN data according to sex, age, and vocational training qualification. As part of a sensitivity analysis, additional samples were drawn and a mean overall prevalence was calculated from them. RESULTS: Data from 5779 DEGS1 respondents and 22 534 AOKN insured persons were used for the analysis. After parallelization of the AOKN-sample by sex, age, and vocational training qualification, no significant differences in prevalence of myocardial infarction could be found between claims data from the AOKN and data from the DEGS1 Health Survey. In men, there were tendencies toward lower prevalence of myocardial infarction in the AOKN data. Possible explanations include the selection of less morbid insured persons by using the vocational education degree for parallelization or memory discrepancies in survey data. CONCLUSION: Differences in sociodemographic structure may play a role the interpretation of disease prevalence from difference data sources. This can be compensated for by parallelizing the samples. Future comparative analyses should take into account characteristics of the socioeconomic status. Similar analyses of other diseases such as stroke, diabetes, and metabolic disorders would be desirable.