Background: Selection bias is often inevitable in epidemiologic studies. It is not surprising that study conclusions based on participants’ health status are frequently questioned. Objective: This study aimed to assess whether the non-participants affected the characteristics of a general population of the very old people. Design, Setting and Participants: Prospective, cross-sectional (N=650, aged 85 years old) analysis and 1-year follow-up (n=273), in Linköping, Sweden. Measurements: We analysed data on health-related factors from a postal questionnaire, a home visit and a clinic visit at baseline and at the 1-year follow-up. We calculated the effect size to evaluate the degree of differences between the groups. Results: A greater proportion of non-participants resided in sheltered accommodation or nursing homes (participants vs non-response vs refusal, 11% vs 22% vs 40, P<0.001, φ=0.24). During the home visit or clinic visit, a higher proportion of dropouts reported mid-severe problems in EQ-5D domains (mobility and self-care) and limitations in personal activities of daily living, but the differences between participants and dropouts were very small (φ<0.2). No significant difference was found between the groups with regard to emergency room visits or hospital admissions, despite the fact that more participants than dropouts (φ=0.23) had multimorbidities (≥2 chronic diseases). Living in sheltered accommodation or a nursing home (odds ratio (OR), 2.8; 95% confidence interval (CI), 1.5-5), female gender (OR, 1.8; 95% CI, 1.1-3.1) and receiving more home visits in primary care (OR, 1.03; 95% CI, 1-1.06) contributed positively to drop out in the data collection stages over the study period. Conclusion: Non-participants were not considered to be a group with worse health. Mobility problems may influence very old people when considering further participation, which threatens attrition.
Clinical trials in geriatrics: actual status and international regulatory guidlines
