Getty Images
What's the best way to measure health-related social needs?
Research shows no system for measuring health-related social needs can capture every need, suggesting that organizations should leverage multiple systems for a holistic approach.
When it comes to measuring health-related social needs, not every assessment is created equal.
According to a new JAMA Network Open study, there's no single measurement that can successfully measure every common health-related social need (HRSN), sometimes referred to as the social determinants of health. This means healthcare organizations should consider adoption multiple HRSN measurement tools to ensure they capture the full breadth of patient needs.
Assessing for social determinants of health has become common practice as more healthcare providers acknowledge that social needs can influence health outcomes just as much, if not more, than clinical interventions.
Screening for HRSN allows healthcare organizations to set up or refer patients to social services that help ameliorate these issues. Programs that address transportation barriers, housing insecurity and food insecurity are all popular among healthcare organizations.
But which method for HRSN screening is the best?
HRSN screening is not one-size-fits-all
In this latest study, researchers from the Regenstrief Institute, Indiana University and Cincinnati Children's Hospital addressed that question, comparing screening questionnaires, natural language processing (NLP) of clinical notes, rule-based computable phenotypes and machine learning (ML) classification models.
The researchers tested those methods on clinical notes of more than 1,200 adult patients receiving care at one of two primary care clinics in Indiana between January 2022 and June 2023. Specifically, the team assessed how the four screening methods performed in flagging food insecurity, housing instability, financial strain, transportation barriers and history of legal problems among patients.
Overall, screening questionnaires performed the best of the four methods. Notably, screening questionnaires had positive predictive value, meaning providers could reliably make patient referrals to social services if a screening questionnaire flagged them for a certain HRSN.
However, questionnaires were not without imperfections, the researchers reported. For example, screening questionnaires fell short in true-positive housing instability, financial strain and transportation barriers.
The other methods didn't perform well, the researchers said. For NLP, the researchers observed high positive predictive value of certain HRSNs, including food insecurity, housing instability and transportation barriers.
But healthcare providers don't often document HRSNs in clinical notes, in large part because patients don't talk about them unless a screening has occurred. This means NLP doesn't have much data upon which to draw insights, rendering the method somewhat ineffective.
Meanwhile, ML had poor performance, but it did tend to be sensitive, the researchers said.
"As a more sensitive method, ML classification could pair well with the high-specificity screening questionnaires in a 2-stage screening strategy," the research team remarked. "Because health care organizations need to screen large numbers of patients, such a 2-stage screening approach could be effective and efficient."
Finally, rule-based phenotype performance was poor and did not improve after inclusion of new features, the team said.
Importantly, the researchers said their assessment uncovered some issues with unfairness, which could impact some of the most traditionally underserved populations.
"Measuring fairness as false-negative rates, we found indications of unfair performance across the HRSNs assessed," with the biggest differences falling along age and racial lines, the team said. "The findings suggest organizations using this information as part of population health management strategies should implement quality monitoring processes to ensure consistent referral rates across patient populations."
Ultimately, these findings indicate that healthcare organizations should employ multiple methods for screening HRSN and social determinants of health. Although leveraging questionnaires can be an effective first step, having multiple systems in place could prevent organizations from missing the full scope of HRSN affecting their populations.
Sara Heath has reported news related to patient engagement and health equity since 2015.
