Sociodemographic Representativeness in a Nationwide Web-Based Survey of the View of Men on Involvement in Health Care Decision-Making: Cross-Sectional Questionnaire Study

Abstract

BackgroundBeing able to generalize research findings to a broader population outside of the study sample is an important goal in surveys on the internet. We conducted a nationwide, cross-sectional, web-based survey with vignettes illustrating different levels of patient involvement to investigate men’s preferences regarding participation in health care decision-making. Following randomization into vignette variants, we distributed the survey among men aged 45 to 70 years through the state-authorized digital mailbox provided by the Danish authorities for secure communication with citizens. ObjectiveThis study aimed to investigate the sociodemographic representativeness of our sample of men obtained in a nationwide web-based survey using the digital mailbox. MethodsResponse rate estimates were established, and comparisons were made between responders and nonresponders in terms of age profiles (eg, average age) and municipality-level information on sociodemographic characteristics. ResultsAmong 22,288 men invited during two waves, a total of 6756 (30.31%) participants responded to the survey. In adjusted analyses, responders’ characteristics mostly resembled those of nonresponders. Response rates, however, were significantly higher in older men (odds ratio [OR] 2.83 for responses among those aged 65-70 years compared with those aged 45-49 years, 95% CI 2.58-3.11; P<.001) and in rural areas (OR 1.10 compared with urban areas, 95% CI 1.03-1.18; P=.005). Furthermore, response rates appeared lower in areas with a higher tax base (OR 0.89 in the highest tertile, 95% CI 0.81-0.98; P=.02). ConclusionsOverall, the general population of men aged 45 to 70 years was represented very well by the responders to our web-based survey. However, the imbalances identified highlight the importance of supplementing survey findings with studies of the representativeness of other characteristics of the sample like trait and preference features, so that proper statistical corrections can be made in upcoming analyses of survey responses whenever needed.