[Response rates and non-response bias in a health-related mailed survey]

Abstract

A health-related mailed survey was conducted to investigate the effect of follow-up mailings to response rates. In addition, the answer distributions among early and late respondents were compared to check non-response bias. Approximately 3,000 persons aged 40-64 were randomized into two groups; a questionnaire with four pages (twenty-two questions) was assigned to the first group, and a questionnaire with eight pages (thirty-five questions) to the second group. Both questionnaires contained questions of current health status, health-related practice, smoking status, etc. Follow-up mailings were sent twice to non-respondents. Response rates were increased from 38% to 62% by the first follow-up mail, and to 71% by the second follow-up mail. Although the length of the questionnaire did not affect response rates, response rates among the older subjects was higher than the younger subjects. Positive response to current smoking status was 10% lower among early respondents than late respondents, collected after the second follow-up mailing, and response to regular participation in physical/medical checkup was 15% higher among early respondents, whereas there were few differences for answers to other questions. Odds ratios between current health status and several health-related questions may not be biased by late response. However, increased response rates are needed to prevent non-response bias, because there were some differences in responses from follow-up mailings.