[Selective loss of participants as a consequence of under-coverage and non-response in research among the aged: an overview].

This article presents an overview of biases in studies among the aged as a consequence of under-coverage and non-response. The question studied is whether participants and non-participants in research differed in their socioeconomic status (education; income; SES of neighbourhood), social integration (marital status; living arrangements; social participation; ethnicity) and health (general health situation; mortality in the period after study; lifestyle and risk factors, namely obesity and alcohol abuse; contact with physicians). In all, 65 studies were reviewed. The results showed that research participants had a higher socioeconomic status, were more integrated socially and in better health than research non-participants. The differences (effect sizes) were small to moderate. On the basis of these results, we can conclude that univariate distributions and prevalence estimates as well as relationships between variables in published research will frequently be biased.