Advancing social determinants of health research and practice: Data, tools, and implementation

Chenyu Li, Jingchuan Guo, Jiang Bian, Michael J. BecichJournal of Clinical and Translational Science 2025 · DOI · OpenAlex

Social Determinants of Health (SDoH)—the conditions in which people are born, grow, live, work, and age—are increasingly recognized as the primary drivers of health outcomes. However, clinical systems have historically struggled to integrate this non-clinical data into routine care. A recent editorial in the Journal of Clinical and Translational Science synthesizes a special issue of 36 papers that aim to bridge this gap, moving SDoH from abstract research concepts into actionable clinical data through a "Learning Health Cycle."

The Three-Fold SDoH Gap

Despite the consensus on the importance of SDoH, the medical field faces three significant hurdles in utilizing this information effectively:

  1. Data Silos: SDoH data is frequently trapped in disparate geospatial databases or buried within unstructured clinical notes, remaining disconnected from Electronic Health Records (EHRs).
  2. Conceptual Confusion: There is often a lack of clarity between contextual SDoH (neighborhood-level factors), social risks (individual-level adverse conditions), and social needs (immediate patient-identified priorities).
  3. Implementation Inertia: Even when data is collected, it rarely informs real-time clinical workflows or broader health policy decisions.

The Learning Health Cycle Framework

To address these challenges, the authors categorize recent research into a Learning Health Cycle (Figure 1). This framework illustrates how the field is evolving from simple data collection toward AI-driven tools and real-world implementations designed to close the health equity gap.

The SDoH in Translational Sciences Learning Health Cycle framework, illustrating the flow from data to knowledge to performance.

The 36 papers in the special issue are organized into four distinct pillars that support this cycle:

  • Data Infrastructure: Focuses on "geo-enriching" EHR warehouses by linking patient addresses to social and environmental datasets using the FAIR principles (Findable, Accessible, Interoperable, and Reusable).
  • Tool Development: Highlights the use of Natural Language Processing (NLP) to extract social risk factors—such as housing instability or incarceration history—from unstructured clinical narratives.
  • Tool Implementation: Examines the application of these tools in clinical settings and their measurable impact on health equity.
  • Community Outreach: Ensures research reflects the lived experiences and actual needs of the populations being served.

Key Insights: NLP and Data Granularity

A recurring theme across the research is that Natural Language Processing (NLP) is the "secret sauce" for SDoH integration. While structured EHR fields often miss social nuances, NLP can mine the "hidden" data within clinical notes.

However, researchers also identified a "Granularity Paradox." Neighborhood-level data (contextual SDoH) does not always correlate perfectly with individual social risks. Predictive models tend to be significantly more accurate in high-risk areas than in low-risk ones, suggesting that geospatial modeling has limits and must be supplemented with individual-level screening.

Real-World Impact and Scalability

The editorial highlights several platforms and models that are already seeing significant adoption:

  • RocHealthData.org: A web-based tool providing access to thousands of national datasets, which has reported over 1,033 registered users, nearly 30% of whom are from the community.
  • Sociome Data Commons (SDC): A cloud-based repository for harmonized geospatial data.
  • SEDoH Informatics Maturity Model: A new standardized tool for healthcare organizations to assess their readiness to handle social and environmental data.

Furthermore, 2024 marks a regulatory turning point. The Joint Commission and the Centers for Medicare and Medicaid Services (CMS) now require SDoH data collection, shifting these initiatives from academic interests to mandatory clinical requirements.

Future Directions

While technical feasibility has been demonstrated, the authors note that seamless integration into daily clinical decision-making remains a challenge. Future work must address the ethical implications of using AI to identify sensitive social risks, ensuring data privacy and preventing algorithmic bias. Ultimately, the authors conclude that medical records alone are insufficient; sustainable progress requires deep community engagement to ensure that data-driven interventions align with the priorities of the people they are intended to help.

Summary updated

BibTeX
@article{li2025advancing,
  author = {Chenyu Li and Jingchuan Guo and Jiang Bian and Michael J. Becich},
  title = {{Advancing social determinants of health research and practice: Data, tools, and implementation}},
  journal = {Journal of Clinical and Translational Science},
  year = {2025},
  doi = {10.1017/cts.2024.657},
}
Figure