Participation rates in epidemiologic studies conducted with postal questionnaires have steadily declined since 1970s. This can lead to an increased risk for selection bias. The aim of this study was to examine cause and effect of non-response in a large cross sectional study assessing respiratory health in western Sweden.

The study sample was 29,218. The response rate to the initial postal questionnaire was 33%. The response rates to subsequent postal reminders were 15%, 7% and 7% of eligible participants totalling a participation of 62%. Of those who did not respond to the postal survey, a random sample of 400 subjects were identified and contacted for interview by telephone.

Non-responders did not differ significantly in prevalence of airway diseases or symptoms when compared with responders. Male sex, young age and smokers were underestimated among non-responders. No clear trends in prevalence of respiratory symptoms and report of asthma were found with delayed response to the postal survey. The proportion of smokers and men increased with increasing number of reminders. Letters reminding subjects about the study did increase the participation rate but did not alter the risk estimates.

We conclude that with a response rate of 62%, our estimate of disease and symptom prevalence was not biased in this Swedish population. However, smoking was underestimated. No general trend for late-responders was seen and therefore we conclude that extrapolation of results to non-responders is not possible in our study. Causes of non-response were mainly due to circumstantial factors.