Participant characteristics of non-consent for health data linkages: results from the BHIS 2023

Abstract Background Linking self-reported health survey data with administrative health records is an opportunity to improve the quality and scope of public health research, but requires participant consent, especially under GDPR regulations. Understanding the profile of respondents declining consent for data linkages is important to mitigate representability issues in data linkages. This study investigates the characteristics of Belgian Health Interview Survey (BHIS) 2023 participants who declined data linkage with health registry databases. Methods The BHIS 2023 collected health data through face-to-face interviews and self-administered questionnaires from 1 January 2023 to 31 December 2024. Participants could opt in for linkage of their responses to administrative health data from the Inter-mutuality Agency (IMA-AIM). We examined sociodemographic and health-related characteristics associated with consent behavior using multivariable logistic regression. Results Of 7,001 participants, 2,025 (28.9%) declined data linkage. Mean age was slightly lower among non-consenters (43.7 years) than consenters (47.8 years). In the logistic regression model, no significant associations were found with age, sex, or civil status. However, residents of Brussels (OR 0.20, 95% CI 0.17-0.24) and Wallonia (OR 0.28, 95% CI 0.24-0.32) were significantly less likely to consent than those in Flanders. Lower educational attainment was associated with lower consent: no diploma (OR 0.58, 95% CI 0.48-0.69) and secondary education (OR 0.76, 95% CI 0.66-0.88) vs higher education. Good subjective health was positively associated with consent (OR 1.97, 95% CI 1.65-2.36). Conclusions Consent for health data linkage varies by region, education, and subjective health status. Future survey waves should prioritize transparent communication on data use to increase consent from underrepresented groups. Weighting and imputation techniques should be employed to address selection bias in analyses based on linked data. Key messages • One in three BHIS participants declined data linkage; lower consent was linked to region, education, and subjective health. • Data linkage bias must be addressed through communication and statistical adjustment strategies.