INDIANAPOLIS — A new study from researchers at the Fairbanks School of Public Health examined how accurate healthcare settings are measuring food insecurity, housing instability, financial strain, transportation barriers and legal issues.
The study, led by researchers at the Fairbanks School and Regenstrief Institute, found that simple screening questionnaires performed better than advanced machine learning methods. But no single method was perfect, and each had gaps.
The research team evaluated four approaches: electronic health record screening questionnaires, natural language processing of clinical notes, rule-based algorithms and machine learning models. Questionnaires were most effective at identifying food insecurity, transportation barriers and legal needs, but all methods struggled with financial strain.
"Health systems increasingly recognize that social factors are as important to health as medical care, but it's not always clear how to identify patients with those needs in routine practice," said Joshua Vest, PhD, professor of health policy and management at the Fairbanks School, a research scientist at the Regenstrief Institute and lead author of the study. "Our study shows that while questionnaires are a strong starting point, we cannot rely on any single method to capture the full range of patients' social needs."
The study also revealed unfairness in how well each method worked across age, race and gender groups. For example, some tools were more likely to miss needs among Hispanic patients and older adults.
"We should be very mindful that the tools we use to identify social needs don't inadvertently widen healthcare inequities," said Christopher Harle, PhD, chair of the Department of Health Policy and Management at the Fairbanks School of Public Health, chief information officer at the Regenstrief Institute and an author of the study. "This work highlights both the promise and the limits of data-driven approaches, and the importance of combining them with human-centered engagement."
The authors said the findings underscore the need for health systems to use a mix of approaches to better understand patients' circumstances and connect them with resources. By combining data-driven tools with direct patient engagement, providers can more effectively address the social factors that influence health outcomes.
The study, "Performance of 4 Methods to Assess Health-Related Social Needs," was published in August in JAMA Network Open.
