The effect of changes to question order on the prevalence of 'sufficient' physical activity in an Australian population survey.

BACKGROUND Population surveys are frequently used to assess prevalence, correlates and health benefits of physical activity. However, nonsampling errors, such as question order effects, in surveys may lead to imprecision in self reported physical activity. This study examined the impact of modified question order in a commonly used physical activity questionnaire on the prevalence of sufficient physical activity. METHODS Data were obtained from a telephone survey of adults living in Queensland, Australia. A total of 1243 adults participated in the computer-assisted telephone interview (CATI) survey conducted in July 2008 which included the Active Australia Questionnaire (AAQ) presented in traditional or modified order. Binary logistic regression analyses was used to examine relationships between question order and physical activity outcomes. RESULTS Significant relationships were found between question order and sufficient activity, recreational walking, moderate activity, vigorous activity, and total activity. Respondents who received the AAQ in modified order were more likely to be categorized as sufficiently active (OR = 1.28, 95% CI 1.01-1.60). CONCLUSIONS This study highlights the importance of question order on estimates of self reported physical activity. This study has shown that changes in question order can lead to an increase in the proportion of participants classified as sufficiently active.