Ann Emerg Med. 2026 Mar 23:S0196-0644(26)00093-4. doi: 10.1016/j.annemergmed.2026.02.010. Online ahead of print.
ABSTRACT
STUDY OBJECTIVE: Health care researchers, policymakers, and managers have long been concerned with heavy emergency department (ED) use for low-acuity conditions that can be addressed in office-based settings. Clinically, urgent care clinics are a viable substitute. Yet, we know little about how ED use and urgent care use interact, and how their combined or separate use varies across population groups. People may cluster in groups with distinct patterns of combined health care utilization, which we consider a meaningful way to study health care use. We aimed to identify latent classes of adult health care utilization based on observed characteristics.
METHODS: We conducted a latent class analysis to identify distinct classes of health care utilization among adults (18+ years old) using publicly available, de-identified data from the 2022-2023 National Health Interview Survey (N=56,181). The latent class model included 4 indicators and 2 ordinal variables: having the last visit being a wellness visit, having a usual place of care, delaying or foregoing care due to cost, having a hospitalization, urgent care use (0, 1-2 visits, and 3+ visits), and ED use (0, 1, and 2+ visits) in the past year. We compared the fit of 2, 3, and 4-class models using Akaike’s Information Criterion and Bayesian Information Criterion statistics. We then estimated regression models of class probabilities on sociodemographic and health-related characteristics.
RESULTS: A 4-class model had the best model fit. Two classes were distinguished by low health care use: one with barriers to care (“nonusers with access barriers,” the smallest class, 6.8%) and one without (“nonusers without access barriers”; the largest class, 59.7%). The class of “heavy health care users” (15.7%) is characterized by the highest probability of ED use (mean probability of having 1 visit 0.471 and 2+ visits 0.351) and the highest probability of hospitalization (0.515) of all classes, alongside moderate urgent care use (probability of 1+ visits 0.430). The class of “urgent care users” (17.7%) is marked by the highest probability of urgent care use (zero probability of no visit, 0.786 of 1-2 visits and 0.214 of 3+ visits), alongside low probability of ED use and the lowest probability of hospitalizations (<0.01). In adjusted regression analyses, the probability of being in the “nonusers with access barriers” class was substantially higher for the uninsured and the probability of being in the “heavy health care users” class was substantially higher for Medicaid enrollees. The probability of being an “urgent care user” was higher for those with higher educational attainment and private insurance.
CONCLUSION: Our findings suggest that urgent care is either complementary to the ED (in the “heavy health care users” class) or is used alongside low to no use hospital-based care and low to no barriers to care (“urgent care users” class). At the same time, our analysis did not identify a distinct class of ED users, with low to no urgent care use. Our findings can inform health system decisionmaking, especially in areas of health care delivery and improving access.
PMID:41874493 | DOI:10.1016/j.annemergmed.2026.02.010
