Item non-response in a population-based mail survey of physical activity

Background. This study assessed item nonresponse (INR) in a population-based mail survey of physical activity (PA). Methods. A questionnaire was mailed to a random sample, with a 57% response rate (n = 2532). The magnitude and type of PA INR and the association with sociodemographic variables was examined using logistic regression. Results. Among survey respondents, 28% had incomplete PA data; 11% missed 1 item, 11% missed 2 items, and 5% missed all 3 items. Respondents missing 3 items tended to be female, less educated, low income, in poor health, and current smokers. The walking item was missed by 8% of respondents, and 18% and 23% missed the vigorous-intensity and moderate-intensity PA items respectively. These groups were sociodemograpically different from those without INR. Incomplete PA data was also associated with sociodemographic INR. Conclusions. Mail surveys may underrepresent individuals insufficiently active for health, in particular those of low socioeconomic position