Questionable Validity of Data Sources
The conclusion of the study (1), that relevant differences between German federal states exist in the lifetime prevalence of important cardiovascular disorders, which can only partly be explained by variations in age, sex, social status, and size of municipality, is not validated by the data sources. The basis of the conclusion is anything but solid, for the following reasons.
The basis of the survey is a random sample derived from landline telephone numbers; households or persons without a landline are therefore not included. The proportion of actual interviews was between 23.9% and 34.5%, the cooperation rate between 51.2% and 76.6%—in other words, only 12–17% of the target population provided data. The question of whether this population is actually representative would require further study. Fundamentally, in a sample based on landline telephone connections, a bias towards a sick, immobile population cannot be ruled out. Such a sample bias cannot be balanced by representative weighting.
The information on cardiovascular events was based on medical diagnoses reported by lay persons. Was this information validated?
Figure 2 implies that the non-reported confidence interval for Germany would have had overlaps with all federal states—this would mean that no significant difference exists between the states and Germany overall. In the absence of statistical significance, it is pointless to reflect on differences.
Data from the cause of death statistics of the Federal Statistical Office were used to capture death rates. The death statistics are based on death certificates, which—according to several studies—are barely, if at all, valid. Furthermore, the data were coded at state level, which means that the data possibly reflect differences in state-specific completions of death certificates and coding behaviors. The section on limitations points out the weakness of this data source, but it states that it would be unlikely to have any effect—which should have been proved.
PD Dr. med. Kurt Bestehorn
Institut für klinische Pharmakologie
Technische Universität Dresden
Dipl.-Math. Dr. rer. nat. Maike Bestehorn
|1.||Dornquast C, Kroll LE, Neuhauser HK, Willich SN, Reinhold T, Busch MA: Regional differences in the prevalence of cardiovascular disease—results from the German Health Update (GEDA) from 2009–2012. Dtsch Arztebl Int 2016; 113: 704–11 VOLLTEXT|