Social Determinants of Health & Patient Outcomes
Explore the critical debate: Can social determinants of health (SDoH) enhance predictive models? Research indicates that SDoH, possibly influencing up to 80% of health outcomes, could substantially improve healthcare predictions. However, the effectiveness hinges on data quality, source, and model design. News Directory 3 uncovers how patient-level SDoH data, combined with clinical data and structured information like median income and education levels, can lead to more accurate predictions for vulnerable patient subgroups.Discover how integrated datasets linking medical claims with social, physical, and behavioral factors are shaping the future of healthcare forecasting. what innovative approaches are on the horizon?
Social Determinants of Health Impact on Predictive Models Examined
Updated January 26,2024
Social determinants of health (SDoH) may influence as much as 80% of health outcomes,but whether these factors improve the accuracy of predictive models remains a topic of debate.The answer frequently enough hinges on data type, source, quality and model design.
SDoH data comes from subjective and objective sources. Subjective data includes self-reported information,clinician-collected data,and unstructured electronic health record (EHR) data. Objective data includes individual and community-level information from government, public, private and consumer behavior sources.
Research on the value of SDoH in predictive models has produced mixed results. Some studies show no significant differences when SDoH are added to models, while others report considerable improvements. These varying results often depend on the extent of reliance on customary clinical models and, more importantly, on the types and sources of SDoH data used.
Some studies indicate that SDoH predictive models can fail due to model design and unstructured, inconsistently collected EHR-level data.Dependence on EHR-derived population health databases for SDoH can also be problematic because the data is often a proxy for individual-level social factors based on assumptions rather than evidence.
Other research indicates that objectively collected or highly structured and consistent data can lead to success. One study found that adding structured data on median income,unemployment rate,and education from non-EHR sources improved a model’s health prediction granularity for vulnerable patient subgroups. Another study found that combining structured SDoH data from the U.S. Census with machine learning techniques improved risk prediction model accuracy for hospitalization,death and costs.
Change Healthcare has curated an integrated national-level dataset linking billions of past de-identified medical claims with patient-level social, physical and behavioral determinants of health. This data can help determine the relative importance of specific patient SDoH factors compared to clinical factors alone for various conditions, including COVID-19. Research indicates that economic stability is frequently enough a high predictor of the healthcare experience.
What’s next
as researchers learn more about the best types and sources of SDoH data and develop better-suited models, healthcare predictive models are likely to improve, leading to better predictions of health outcomes and potential health disparities.