Health effects of reductions in food assistance benefits in the USA

In the USA, the Supplemental Nutrition Assistance Program (SNAP)—the nation’s largest anti-hunger programme—serves more than 40 million individuals and is undergoing unprecedented retrenchment. In July, 2025, policy makers enacted the One Big Beautiful Bill Act (H.R. 1), which reduces SNAP funding by US$186 billion over the next decade—the steepest cuts in the programme’s history.^1,2 Approximately 4 million people are expected to lose SNAP benefits or experience reductions in monthly support. Although SNAP benefit reductions following the expiration of COVID-19-era emergency allotments led to a sharp rise in food insecurity,^3,4 the clinical implications of large-scale SNAP cuts are unknown. Understanding these consequences is increasingly important as the USA implements substantial reductions in nutritional support through H.R. 1, and as countries globally reassess the scope of social protection programmes that provide nutritional assistance.

To evaluate these potential consequences, we leveraged the time-staggered termination of SNAP emergency allotments across US states between 2021 and 2022, which created a natural experiment for assessing the effects of abrupt reductions in nutrition assistance on all-cause hospitalisations and all-cause mortality. We used 2021–22 Medicaid claims data from the Transformed Medicaid Statistical Information System. The study population included SNAP-eligible adults aged 20 years and older, defined as individuals with low incomes, based on continuous enrolment in Medicaid during the study period (January, 2021 to December, 2022). Although claims do not identify individuals enrolled in SNAP, the majority of SNAP participants are Medicaid enrollees⁵ and categorical eligibility for SNAP closely overlaps with income eligibility for Medicaid.⁶ The intervention group included adults residing in six states where emergency allotments were terminated early, resulting in immediate reductions in SNAP benefits. The control group included those in eight region-matched states where emergency allotments remained in place throughout the study period (appendix p 2). Only states with high-quality claims data were included, as determined by the Medicaid Data Quality Atlas.⁷

A quasi-experimental difference-in-differences design with an event study framework was used to estimate changes in outcomes (per 100 000 Medicaid enrollees) in the intervention group relative to the control group following the time-staggered termination of emergency allotments. The calendar quarter in which emergency allotments ended was set as time zero, with pre-policy and post-policy periods of up to five quarters determined by the timing of emergency allotments termination during the study period. We estimated dynamic treatment effects using the Callaway-Sant’Anna estimator⁸ for each quarter relative to the termination of emergency allotments, with quarter −1 serving as the reference period. This approach accounted for state-specific and time-specific factors,⁸ and models further adjusted for age, state Medicaid expansion status, and state-level quarterly SNAP enrolment as a proportion of Medicaid enrollees. 95% CIs were derived using wild cluster bootstrap procedures to account for the small number of state-level clusters.⁹ Pre-period trends were visually similar across all outcomes in the event study plots. We also did a sensitivity analysis that did not require continuous Medicaid enrolment, instead requiring enrolment only during the first 3 months of the study period. In addition, because SNAP enrolment could itself be influenced by the termination of emergency allotments, we repeated the analyses after excluding adjustment for state-level SNAP enrolment in the difference-in-differences models. All analyses were conducted in R, version 4.3.2, with statistical significance defined as a two-sided p value of less than 0·05.

The study population included 5 591 669 adults (mean age 48·2 years [SD 0·01]; 63·7% female [n=3 563 464], 36·3% male [n=2 028 205]) residing in states where SNAP emergency allotments were terminated early, of whom 942 376 (16·9%) were hospitalised at least once during the study period, and 9 146 854 adults (mean age 45·8 years [SD 0·01]; 62·5% female [n=5 712 264], 37·5% male [n=3 434 428]) in control states, of whom 1 473 254 (16·1%) were hospitalised at least once. Overall, 5·5% (n=804 793) of the study population had more than one admission during the study period.

The figure shows changes in outcomes among adults residing in states that terminated emergency allotments early (intervention group) compared with those in control states. Although there was initially no differential change in adjusted all-cause hospitalisations in the first three quarters after emergency allotments ended, hospitalisations increased significantly in intervention states relative to control states beginning in the fourth quarter (852·5 per 100 000 Medicaid enrollees, 95% CI 144·8–1560·1) and continued to rise in the fifth quarter (1310·1, 272·2–2347·9; figure; appendix p 3). In contrast, there was no significant change in all-cause mortality between these groups (figure; appendix p 3). Findings were also highly consistent across sensitivity analyses using less restrictive Medicaid enrolment criteria (appendix p 4) and in models estimated without adjustment for state-level SNAP enrolment (appendix p 5).

Figure: Differential changes in all-cause hospitalisations and all-cause mortality following SNAP emergency allotment termination.

Quarterly changes in all-cause hospitalisations (A) and all-cause mortality (B) per 100 000 Medicaid enrollees in intervention states (Florida, Georgia, Iowa, Indiana, Kentucky, and Nebraska) where SNAP emergency allotments were terminated early, compared with those in control states (Kansas, Maryland, Michigan, North Carolina, Ohio, Texas, Virginia, and Wisconsin), where emergency allotments remained in place during the study period. Each point represents the estimated treatment effect relative emergency allotment termination (vertical dashed line), using quarter −1 as the reference period, after adjustment for age, state Medicaid expansion status during the study period, and state-level quarterly SNAP enrolment. Estimates are shown with 95% CIs. SNAP=Supplemental Nutrition Assistance Program.

All-cause hospitalisations increased in states that terminated SNAP emergency allotments early compared with control states that maintained them, whereas all-cause mortality remained unchanged. Our findings indicate that reductions in nutritional assistance can have measurable clinical impacts. The increase in all-cause hospitalisations observed in our study might be driven by multiple factors, including declines in diet quality leading to the exacerbation of diet-sensitive conditions and heightened financial strain forcing individuals to make difficult trade-offs between food and health-care needs.¹⁰ A study limitation is the potential for residual confounding due to unmeasured factors, despite the quasi-experimental design.

These findings underscore that reductions in SNAP benefits under H.R. 1 could lead to a substantial rise in preventable hospitalisations among US adults with low incomes. More broadly, they illustrate how the scaling back of nutritional support programmes—even in high-income settings—can rapidly lead to adverse health outcomes.