Nonresponse pattern and bias in a community-based cross-sectional study of cognitive functioning among the elderly

The demographic, health, and mental functioning characteristics of nonresponders to a community-based cross-sectional study of cognitive functioning among the elderly in Amsterdam, the Netherlands (Amsterdam Study of the Elderly (AMSTEL), October 1990 to May 1991), were examined and compared with responders. The randomly selected age-stratified (65-74, 75-84 years) sample was drawn from nonresponders listed with a subsample (n = 8) of general practitioners whose lists served as the sampling frame for the main study. The general practitioners approached and interviewed the responding nonresponders using the same standardized questions that were used in the main study. Nonresponders (n = 115) and responders (n = 999) from the same medical practices were compared by means of chi-square and odds ratios. Compared with responders, these nonresponders more often reported a history of psychiatric illness, heart attack, stroke, and diabetes, and were more likely to be unmarried, to have a lower education, and to do poorly on the cognitive test (odds ratio = 1.6, 95% confidence interval 1.0-2.6). Most significant physical and mental health differences by response status were seen among the persons aged < 75 years and not among those aged > or = 75 years. The odds for poor cognitive test performance associated with age and stroke were relatively more biased than those associated with other risk factors. These results suggest that the characteristics of young-old and old-old elderly nonresponders to cross-sectional studies of cognitive function may differ, and that there may be selective nonresponse that could bias, to a different degree, estimates of risk for poor cognitive functioning. Studies should investigate the possibilities for nonresponse in their own setting.