[Nonparticipation as a factor influencing the value of follow-up studies. Results of a telephone 5-year follow-up interview of 55-74-year-old participants of the Augsburg 1989/90 MONICA Survey].

: In prospective cohort studies losses to follow-up are the major source of bias. This article describes the results of the investigation whether there are systematic differences in socio-economic and health-related factors between participants in a follow up study and those who are lost to follow-up. Subjects included were 1,030 men and 957 women aged 55-74 who participated in the second MONICA Survey Augsburg, F.R.G., 1989/90 (MONICA = Monitoring of trends and determinants of cardiovascular disease). They were reexamined in 1994-95. In 1994/95 910 men and 912 women of the baseline study were still alive, 1305 persons (663 men, 642 women) took part in the follow up by telephone. Altogether 120 men and 45 women had died since 1989/90. 144 persons were not eligible for the study (15 moved away, 5 severely ill persons, 124 without telephone). 373 persons were classified as eligible non-responders. The response rate was 77.8% concerning 1678 eligible persons (men 79.1%, women 76.4%). Non-response has led to an underrepresentation of the lower social class (workers, less than 10 years of education; household income less than 1500 DM) and of persons living alone (single, divorced, widowed; single-person households). In females, non-responders were overrepresented by "never employed persons" (odds ratio 3.02; 1.89-4.85). We found that the odds of being non-responder for single, divorced or widowed men was 3.40 (95 %-CI: 2.20-5.23) compared to married men. Men "without chronic diseases" at the baseline study compared with ill men had an odds for non-response of 1.54 (95%-CI: 1.00-2.37); the odds was 1.38 (95%-CI: 1.02-1.85) among women who had described their health status as "not so good/bad" in comparison with women with good self reported health. The reported distortions by non-response could be important in analyses of social conditions of morbidity and mortality. The underrepresentation of men without chronic diseases and women with bad self-assessment of their health is relevant for the estimates of morbidity as health outcome. Both aspects have to be taken into account in the interpretation of the results.