Identifying Opportunities for Food-as-Medicine Programs: A Latent Class Analysis of Cardiometabolic Risk, Health Care Engagement, and Dietary Behavior Among US Adults

Abstract

Background: Food-as-medicine programs are expanding rapidly in US healthcare systems, but eligibility is generally determined solely by diagnosis and food insecurity. Whether this identifies the populations with the most room to benefit is unclear. Recruitment occurs through clinical channels, so opportunity depends on the intersection of cardiometabolic risk, healthcare engagement, and room for dietary improvement—a combination not yet characterized at the population level.

Methods: The 2021 Behavioral Risk Factor Surveillance System (n = 438,670 adults) and the 2024 National Health Interview Survey (NHIS; n = 32,629) were analyzed, the latter as a confirmatory sample. Survey-weighted latent class analysis was applied to 15 binary indicators encompassing cardiometabolic risk, healthcare engagement, dietary behavior, physical activity, psychological distress, and self-rated health. Dietary behavior was limited to fruit and vegetable intake frequency; sodium, saturated fat, and ultra-processed food intake were not measured. Models with one to eight classes were estimated using expectation maximization with full-information maximum likelihood
for item nonresponse. Class profiles were compared across the 13 indicators common to both surveys.

Results: A six-class solution was identified. One segment (16.3% of adults, 40.1 million) combined elevated cardiometabolic risk (37.1% obesity, 23.3% high cholesterol) with substantial healthcare engagement (75.0% routine checkup, 98.1% insured) and the lowest fruit and vegetable intake of any class (72.6% and 47.5%, respectively, less than once daily). Two additional engaged segments carried greater disease burden yet reported higher intake. A sixth segment (5.6%, 13.7 million) displayed similar risk but disrupted access, with 66.0% reporting cost barriers and 36.9% food insecure in NHIS. Five of six classes had corresponding NHIS profiles (r ≥ 0.76); the diet-defined segment did not.

Conclusions: Population segments differ substantially in cardiometabolic need, opportunity for dietary improvement, and clinical reachability. A large, clinically reachable segment reporting low fruit and vegetable intake is underrecognized by current eligibility criteria, whereas the segment with the greatest need remains largely unreachable through clinical referral. These characteristics can identify populations in whom intervention may be feasible and potentially beneficial, but willingness to change, uptake, adherence, and subsequent dietary and clinical response require prospective assessment.