The research team, led by Miaoqing Jia and Yongkang Zhang of Weill Cornell Medical College together with Kevin Fiori of the Albert Einstein College of Medicine, linked patient-reported social needs screening data from 18 primary care clinics in the Montefiore Health System to longitudinal electronic health records from the INSIGHT Clinical Research Network, a repository spanning five major New York City health systems. Montefiore’s ambulatory network serves more than 460,000 patients in the Bronx, the borough with the highest poverty rate in New York State, where roughly 85 percent of patients are covered by Medicare or Medicaid. Because the INSIGHT network captures encounters across competing health systems, the investigators could track hospitalizations and emergency department visits even when patients sought care outside Montefiore itself, a methodological advantage that most single-system studies lack.

The screening tool, adapted from the Health Leads toolkit and embedded in the electronic health record since 2018, asks patients about nine distinct social needs: exposure to interpersonal harm, problems with their residence, legal assistance needs, housing stability concerns, cost barriers to care, childcare problems, food insecurity, threatened utility shutoffs, and transportation-related delayed care. Of 27,235 patients screened in 2023, 20,337 met the study’s inclusion criteria, and 15.9 percent of them screened positive for at least one need. The most common problems were food insecurity at 5.9 percent and problems with residence at 5.6 percent, followed by housing stability concerns at 4.6 percent and transportation barriers at 4.1 percent. Exposure to hurts, insults, threats, or screams was rare, reported by just 0.8 percent of patients.

What happened over the following twelve months revealed a striking hierarchy. After adjusting for demographics, comorbidities measured through the Elixhauser index, and eleven census tract-level social indicators from the American Community Survey, transportation-related delayed care stood out as the single most consequential need. Patients who reported that transportation problems had delayed their care faced a 6.8 percentage point higher probability of at least one hospitalization, an 8.4 percentage point higher probability of at least one emergency department visit, a 1.2 percentage point higher probability of a preventable hospitalization, and a 4.2 percentage point higher probability of a preventable emergency visit. Each of these estimates was statistically significant, and the pattern held across multiple sensitivity analyses, including Poisson models that counted repeated encounters and generalized estimating equations that accounted for clustering within neighborhoods.

The technical definition of preventable matters here. Preventable hospitalizations were identified using the Agency for Healthcare Research and Quality’s Prevention Quality Indicators, which flag admissions for ambulatory care-sensitive conditions, illnesses that well-functioning outpatient care should typically keep out of the hospital. Preventable emergency visits were classified with the NYU Billings algorithm, which labels a visit preventable when the combined probability of it being nonemergent, primary-care-treatable, or preventable reached at least 0.75. In other words, the patients most affected by transportation barriers were not simply using more care in general; they were arriving at hospitals and emergency rooms with conditions that clinic-based management might have controlled, had they been able to get there.

Housing-related problems formed the second significant signal, associated with a 4.1 percentage point increase in the probability of at least one emergency department visit. The authors suggest several plausible mechanisms: unstable or inadequate housing can impede chronic disease self-management, heighten exposure to environmental triggers such as mold or pests, elevate psychosocial stress, and disrupt care continuity when patients change addresses frequently or become difficult to reach. Notably, the other screened needs, including food insecurity, the most prevalent need of all, did not show statistically significant associations with acute care use in the primary models. The researchers are careful to stress that this should not be read as evidence that food insecurity is unimportant. Its effects may operate through slower pathways, such as worse glycemic control in diabetes, that do not translate into hospital visits within a single year.

The Bronx context may explain why transportation looms so large. Planning documents for the Cross Bronx corridor note that subway lines in the borough run largely north to south, making buses the primary option for east to west travel, with high ridership and slow speeds on key routes. Reported transportation needs, the authors argue, may reflect structural constraints in transit infrastructure, clinic location, and scheduling rather than individual deficits. The demographic patterns reinforce this structural reading: patients with at least one positive screen were more likely to be Hispanic, 50.6 percent versus 48.4 percent, and non-Hispanic Black, 35.3 percent versus 31.8 percent, and less likely to be non-Hispanic White, a distribution the study attributes to unequal exposure to structural racism and related inequities in housing and transportation services.

The study also documented a heavy burden of co-occurring behavioral health conditions among patients with unmet social needs. Depression was roughly twice as prevalent in the positive-screening group, 23.5 percent versus 12.0 percent, as were substance use disorders, 5.0 percent versus 2.1 percent, and psychoses, 8.3 percent versus 4.3 percent. Patients with positive screens were also younger on average, 52 versus 55 years, and carried more Elixhauser comorbidities, including higher rates of nutritional deficiency anemias and chronic lung disease. This clustering of social, physical, and behavioral health risks underscores why single-issue interventions often fall short and why the authors advocate for workflows that connect screening results directly to coordinated support.

Operationally, the findings suggest a concrete pathway: a positive transportation screen could trigger an electronic health record alert prompting a care navigator or community health worker to assess the urgency of the need, arrange transportation support through Medicaid non-emergency medical transportation or ride-coordination services, and document referral completion. Programs such as NYC Health + Hospitals’ AfterCare offer scalable models, and city-level infrastructure efforts like Reimagine the Cross Bronx could indirectly reduce the delayed outpatient care that feeds avoidable acute utilization. The authors caution, however, that evidence on whether social needs screening programs consistently reduce utilization remains mixed, with results likely shaped by heterogeneous implementation and variable referral uptake.

The study’s limitations deserve honest weight. As an observational analysis, it cannot establish causation, and residual confounding by factors such as health-seeking behavior or insurance details, individual payer information was unavailable, cannot be excluded. Screening was recommended but not universal, so patients who completed it may differ systematically from those who did not. Social needs were captured by self-report at a single point in time, and needs fluctuate, which could attenuate measured associations. Still, the core message survives these caveats: in one of America’s poorest urban communities, a single question about whether transportation has delayed a patient’s care identifies people at markedly elevated risk of landing in the hospital within the year. For health systems investing millions in social care programs, that question may be the highest-yield one they can ask.

Subject of Research: Associations between patient-reported social needs and subsequent acute care utilization in a safety-net health system

Article Title: Social Needs Screening and Subsequent Acute Care Utilization in a Large Safety-Net Health System: A Retrospective Cohort Study

Article References: Jia, M., Dong, H., Fiori, K., Adamson, E., Zhang, M., Tanner, D., Weeks, W. B., Speyer, P., Aerts, A., & Zhang, Y. (2026). Social Needs Screening and Subsequent Acute Care Utilization in a Large Safety-Net Health System: A Retrospective Cohort Study. Journal of General Internal Medicine. https://doi.org/10.1007/s11606-026-10908-3

Image Credits: AI Generated

DOI: 10.1007/s11606-026-10908-3

Keywords: social determinants of health, social needs screening, acute care utilization, transportation barriers, emergency department visits, preventable hospitalizations, safety-net health system, food insecurity, housing instability, electronic health records, health equity, Bronx