Key points are not available for this paper at this time.
Background: Electronic health record (EHR) data are widely used in public health research, including HIV-related studies, but are limited by potential bias due to incomplete and inaccurate information, lack of generalizability, and lack of representativeness. This study explores how workflow processes within HIV clinics, among data scientists, and within state health departments may introduce and minimize bias within EHRs. Methods: Using a constructivist grounded theory approach, in-depth individual interviews were conducted with 16 participants purposively sampled in South Carolina from August 2023-April 2024. A focus group with 3 health department professionals with expertise in HIV disease surveillance was also conducted. Analysis was conducted as outlined by Charmaz (2006). Results: To reduce bias in EHR data, information entry forms should be designed to expansively include patient self-reported social determinants of health (SDOH) information. During data collection, healthcare providers should create a supportive healthcare environment, facilitate SDOH information disclosure, and accurately document patient information. Patients should have access to their EHRs to confirm that SDOH information are correctly recorded. During data curation, data scientists should inspect datasets for completeness, accuracy, and educate public health researchers on dataset limitations. During data management and utilization, health department professionals should crossmatch data across the state, customize data collection systems to reflect local needs, and provide community-based data education and stigma management. Conclusion: Study results suggest that future research is needed to understand how healthcare systems can be incentivized to create and implement EHR bias reduction strategies across all workflows and between stakeholders.
N’Diaye et al. (Mon,) studied this question.