Participation rates, response bias and response behaviours in the community survey of the Swiss Spinal Cord Injury Cohort Study (SwiSCI) (BMC Medical Research Methodology)

Background: Surveying persons with disabilities is challenging, as targeted subjects may experience specific barriers to survey participation. Here we report on participation rates and response behaviour in a community survey of people with spinal cord injury (SCI) in Switzerland. The cross-sectional survey was implemented as part of the Swiss Spinal Cord Injury Cohort Study (SwiSCI) and represents the largest population-based SCI survey in Europe including nearly 2000 persons. Design features to enhance participation rates included the division of the questionnaire volume over three successive modules; recurrent and mixed-mode reminding of non-responders; and mixed-mode options for response. Methods: We describe participation rates of the SwiSCI community survey (absolute and cumulative cooperation, contact, response, and attrition rates) and report on response rates in relation to recruitment efforts. Potential non-response bias and the association between responders’ characteristics and response behaviour (response speed: reminding until participation; response mode: paper-pencil vs. online completion) were assessed using regression modelling. Results: Over the successive modules, absolute response rates were 61.1, 80.6 and 87.3 % which resulted in cumulative response rates of 49.3 and 42.6 % for the second and third modules. Written reminders effectively increased response rates, with the first reminder showing the largest impact. Telephone reminders, partly with direct telephone interviewing, enhanced response rate to the first module, but were essentially redundant in subsequent modules. Non-response to the main module was related to current age, membership of Swiss Paraplegic Association (SPA) and time since injury, but not to gender, lesion level and preferred language of response. Response speed increased with household income, but was not associated to other sociodemographic factors, lesion characteristics or health indicators. We found significant associations between online completion and male gender, younger age, higher education, higher income, SPA membership, tetraplegia, longer time since injury, higher quality of life, and more participation restrictions. Conclusion: In this sample with little non-response bias, recurrent and mixed-mode reminding and mixed-mode options for response were key features of optimizing survey design.