Increasing response rates to postal questionnaires: a randomised trial of variations in design

Abstract

Objectives: Low response rates to postal questionnaires can threaten the validity of studies by reducing the effective sample size and introducing bias. The identification of methods with which to optimise response rates could, therefore, improve the quality of studies. In an attempt to identify such methods, we undertook a randomised trial of two simple variations in questionnaire design. Methods: Using a 2 × 2 factorial design, we conducted a randomised trial to test two variations in questionnaire design; the questionnaires were printed on either single-sided or double-sided paper and had either a single- or multiple-booklet layout. Using equal random allocation, 3836 women were randomised to receive one of these questionnaires as part of a study investigating risk factors for osteoporotic fractures. Results: One thousand eight hundred and seventy questionnaires were returned, giving an overall response rate of 48.7%. There were no significant differences in the overall response to each of the four questionnaire designs. When the number of responders who completed at least 50% of each of the three sections was identified, it was found that single-booklet questionnaires had a better response than the multiple-booklet questionnaires and that single-sided questionnaires had a better response than double-sided questionnaires. However, these results were not significant at the 5% level. There were no significant differences in the response to questions on the odd (left-hand side) pages for the single- compared with the double-sided questionnaires. Conclusion: As the most cost-effective use of resources, we would advocate the use of double- rather than single-sided questionnaires, and use of a single- rather than multiple-booklet design.