Presentation (CTI)

OBJECTIVES: Social Determinants of Health (SDoH) data plays a crucial role in predicting health outcomes in the real world, primarily by providing a more holistic understanding of patient health or at-risk population. Incorporating SDoH information enables health systems and professionals to assess patient complexity, tailor interventions to address diverse needs, and enhance care delivery through integrated services. Our study investigated the impact of incorporating SDoH data into risk prediction models.
METHODS: A comprehensive search was performed in the PubMed and Embase databases. Keywords and Medical Subject Headings (MeSH) terms related to SDoH, and risk prediction were used. We analyzed the existing literature to highlight the successes, challenges, and opportunities with the use of SDoH.
RESULTS: We identified a total of 750 articles through database searches. After removing 182 duplicates, we screened the remaining articles and included 29 studies for detailed analysis. The addition of individual-level SDoH data into prediction models improved the predictive capability in various health outcomes, including prognosis, medication adherence, hospitalization, hospital readmission, length of stay, mortality, etc. However, the inclusion of area-level SDoH data in risk prediction has not shown conclusive results. Integration of SDoH to machine learning models can considerably mitigate disparities in prediction across population sub-groups. Challenges identified include the lack of standardized SDoH measures, which necessitate novel approaches to capture and utilize SDoH data effectively.
CONCLUSIONS: Incorporating individual-level SDoH data into risk prediction models enhances their accuracy and utility, supporting better healthcare interventions and outcomes. This approach aims to improve health outcomes, reducing health disparities and costs. Future efforts should focus on standardizing SDoH data collection and developing integrated tools to further improve patients’ health.