Exploring-Social-Determinants-of-Health-A-Study-of-Housing-and-Economic-Barriers-in-MIMIC-III-

Social determinants of health (SDoH), including housing and economic factors, significantly impact patient health risks and outcomes. However, the majority of SDoH information is embedded in unstructured clinical text within electronic health records (EHRs), making it difficult to access and utilize for patient care and research. To tackle this challenge and harness the potential benefits of incorporating SDoH data into clinical decision-making, we propose a comprehensive natural language processing (NLP) methodology. Our approach explores both unsupervised and supervised learning techniques, finding that supervised approaches perform better. We then compare various supervised approaches, including ChatGPT 3.5, to determine the most effective method for extracting SDoH information from EHRs. By implementing this robust NLP methodology with a focus on the best-performing supervised learning technique, we aim to unlock valuable SDoH insights, enabling improved patient care, early interventions, and data-driven policy decisions that contribute to enhanced health outcomes and health equity