Health disparities and selection bias in obtaining broad consent in a general practitioner setting.

Background: For research with electronic health records in the outpatient setting, obtaining Broad Consent (BC) is increasingly important. However, the presence of potential selection bias in this context remains unclear. Since 2020, the BeoNet-Halle outpatient database collects patient data from participating general and specialty ambulatory practices' management systems in Germany, whereby data is obtained anonymously or pseudonymously Methods: This is a single-center, cross-sectional study. From February 2021 to May 2023, patients were asked by a general practitioner or a specially trained member of the joint practice to provide BC. Within the yearly contact group of 2022, we compared patients who provided BC with the reference population (RP) of patients with at least one physician-patient contact during that period in a joint practice including eight eneral ractitioners. Data pertaining to health, morbidity, and health utilization were extracted from the BeoNet-Halle database. Results: A total of 5,034 patients were analyzed (BC-group: 439 vs. RP group: 4,595). Sex was similar distributed between the groups. In the BC group, patients were slightly older (56.2 vs. 54.1 years), had more physician contacts (15.0 vs. 9.2) and more often at least one chronic condition (76.1 vs. 51.6%) than patients from RP group. Patients were more likely to be referred to at least one other specialist (74.0 vs. 44.4%) and to get at least one drug prescription (89.5 vs. 69.6%). Conclusion: Differences between BC and the reference population (older age, higher multimorbidity, more contacts, referrals, and drug use) indicate selection processes at the point of consent. Given the single-practice design and descriptive analysis, generalizability beyond similar German group practices is limited and requires validation in multi-site studies.