Non-response bias in a postal questionnaire survey on respiratory health in the old and very old

Abstract

Aims: The authors wished to describe non-responders and reasons for non-response and determine the magnitude and direction of non-response bias in connection with a postal questionnaire study of respiratory symptoms and disease among the elderly. Methods: An eight-page respiratory health questionnaire was mailed to subjects in a sex-age stratified random sample of the general population of Bergen, Norway, 70 years or older (n=2,871). A reminder with repeat questionnaire was sent out to initial non-responders after three weeks. A random sample of non-responders was interviewed by telephone. Results: There were a total of 1,649 respondents (57.4%); 1,356 to the initial mailing and 293 to the reminder mailing. The response rates were highest for the age group 70 - 74 years (76.6%) and fell dramatically with age (27.1% in age group 95+years). Men responded (63.7%) better than women (52.0%) and this was consistent at every age level. Late and non-responders had slightly lower frequencies of respiratory symptoms but higher frequencies of current-smoker status. Conclusions: All found, differences between responders and non-responders were small and would have only minimal effects on the final prevalence estimates for respiratory symptoms and disease. In comparing the results based on initial responders to results based on all responders, associations between predictor variables and outcome variables were essentially the same, showing that a reminder had little effect on the results of the study.