Comprehensive Summary
While social determinants of health (SDOH) include political, socioeconomic, and cultural factors that influence health outcomes, they are difficult to quantify due to existing gaps in the literature. AI and large language models (LLMs) offer new opportunities to analyze large healthcare datasets, thus improving risk stratification for patients and our understanding of the interactions between SDOH and health outcomes. Specifically, natural language processing (NLP) has demonstrated success in extracting SDOH information from electronic health records (EHRs). However, there nevertheless exist a number of barriers, ranging from gaps in digital literacy to racial and gender biases that may exacerbate health inequities. Low- and middle-income countries (LMICs) face additional barriers, such as a lack of standardized patient data, weak infrastructure to support large-scale AI implementation, and privacy concerns due to weak data protection frameworks.
Outcomes and Implications
AI presents several opportunities for improving research in SDOH and increasing healthcare accessibility. Multilevel modeling can combine individual- and group-level SDOH data, enhancing disease surveillance, predicting population health trends, and evaluating the effectiveness of public health interventions. AI-powered tools can democratize specialized care in both under-resourced and high-income settings through telehealth, innovative diagnostic tests, and home-based monitoring. Furthermore, LLMs have the potential to bridge the linguistic and geographical gaps in remote regions that face shortages of healthcare workers. However, critical barriers remain, including the lack of standardized SDOH data collection methods, infrastructure limitations in LMICs, and the environmental impact of AI due to rising energy demands and carbon emissions. There is also concern about privacy challenges for patients, though generative adversarial networks (GANs) may help by generating realistic synthetic healthcare data to protect patient privacy. For the future, there is a significant emphasis on digital inclusion to become a “super SDOH,” as well as efforts to bridge AI development to real-world implementation through stakeholder engagement and community-level integration strategies.