Primary Care in Germany – an International Comparison: Results of a Survery Using a Validated Instrument: Response Rate Too Low

by Klaus Koch, Ulrich Gehrmann, Prof. Dr. med. Peter T. Sawickiin volume 38/2007

The Seven Countries Study obtains its sample of physicians interviewed in Germany from classified telephone directories. 18.1% of responding interview partners are insufficient for a generalization. The statistical bias of list-based selection is associated with a low survey rate. Recruitment of the interviewed persons referred to as random and representative does not describe the selection-by-lot procedure. The methodology section refers to simple randomness, and yet the data are adjusted according to German Federal States. Conversely, it would have been necessary to define the sample sizes in advance to obtain a representative picture of state contingents.
Significance is the result of inferential statistics and not of descriptive statistics which evaluate distribution. "No adjustment was made for multiple testing" – hence, testing was multiple. A correction of the significance, e.g., a Lilliefors correction for nonparametric testing of distribution, or a Bonferroni correction for inferential testing, was not performed. Knowledge questions which would have revealed awareness of guidelines were not asked. "Very many – several – very few" was the criterion for accuracy of the permissible responses.
Inferential statistics presuppose the prior existence of a pair of hypotheses. Workload as a source of professional dissatisfaction is not proven with the Seven Countries comparison. Final thinking of the kind practiced in the scholastic sphere becomes apparent in the discussion when the authors offer self-culpable working patterns as an explanation. Presumption, i.e., working with a preconceived opinion, is the hallmark of imprecise discourse. DOI: 10.3238/arztebl.2008.0084a

Dr. med. Martin P. Wedig
Roonstr. 86
44628 Herne, Germany