[Stroke occurence in Germany - on the comparability of insurance data and registry-based data]

This article presents epidemiological data regarding stroke frequency in Germany based on nationwide statutory health insurance data (Deutsche BKK) and aims to analyse them in the context of current research. The comparability of the most important resources of stroke frequency data - stroke registers, DRG data and insurance data - is initially discussed in order to assess the presented data adequately.

The study cohort comes from a population of about 1 000 000 people insured with BKK and consists of all persons who were treated for a stroke in an acute care hospital in 2007 (n = 4,843). Data were subjected to statistical secondary analysis including uni- and bivariate statistics and t tests. Reference studies for the observation period include data from GEK and AOK health insurances, from quality assurances Hessen and Bayern, from the ADSR, and hospital DRG data. The different study types are compared regarding their inclusion/exclusion criteria and the resulting effects on reported prevalences.

Different inclusion criteria and accordingly different operationalisations of "stroke" impede the comparability of existing German data resources regarding stroke. The inclusion of TIA, non-traumatic subdural haemorrhage (I62), and the frequency of unspecified strokes (I64) is especially inconsistent. In addition, recurrent strokes and the definition of first-ever strokes are treated differently. The study cohort reveals no major discrepancies regarding aetiological subgroups compared to previous results, only the percentage of women (60.3 %) seems exceptionally high.

The gender effect is attributed to the BKK member structure, and especially the high proportion of women in the older age groups. Discussion of stroke frequency in Germany needs to take structural differences between study types into account. There are two vulnerable groups that tend to be underrepresented: TIA patients with a high risk of recurrent strokes, and high-risk patients who have already had a stroke and are care-dependent, which are often unspecifically coded. In the future, study designs should include the whole range of stroke coding, thus enabling differentiated analyses.