Based on a telephone survey of elderly people (> or = 60 years) the impact of measures to optimize response rate (repeated telephone calls and converting temporary refusals) on the sample structure is analyzed. Results show slight differences between persons who can be reached easily and those who are difficult to contact as far as their sociodemographic, socioeconomic and health related characteristics are concerned. Moreover there are no significant differences between temporary refusals and cooperative respondents on these characteristics. Differences are considerable however on multivariate relationships between sociodemographic, socioeconomic and psychosocial variables on the one hand and subjective health on the other. Results indicate that nonresponse has an effect on relationships between variables which are essential for research in medical sociology. This effect clearly shows the necessity of measures to increase response rate for reducing nonresponse bias especially in surveys of elderly people.
