Introduction
Approximately one in five children in the United States experiences food insecurity, meaning that their access to adequate food is limited by a lack of money or other resources in the household [
1]. The COVID-19 pandemic further exacerbated household food insecurity, as significantly more families experienced periods of unemployment or underemployment, with associated financial hardships and an inability to meet basic needs [
2]. It is well-established that food insecurity can have deleterious short- and long-term effects on children’s health and development, including increased odds of being hospitalized, increased risk of asthma and other chronic health conditions, and lower test scores and behavioral outcomes in school [
3‐
7].
The Supplemental Nutrition Assistance Program (SNAP) is an important federal food assistance program that assists many low-income families in becoming and remaining food-secure [
8‐
12]. Approximately one-quarter of all US children receive SNAP supports, representing 17 million households of SNAP recipients [
13,
14]. Prior studies have demonstrated the positive effects of SNAP for both adults and children through subjective measures, such as associations with better self-reported health status in recipients, and objective measures, including increased birth weight and decreased likelihood of underweight status among children in recipient households [
12,
15,
16]. Reduced healthcare expenditures in SNAP recipients for adults are likely a result of improved overall health related to better nutrition. For example, in those with diabetes, proposed mechanisms linking SNAP receipt and improved health include increased fruit and vegetable intake and better glycemic control [
17]. While not the original intent of SNAP, the program’s measurable effects on the health of its recipients, which go beyond just alleviating hunger, raise the question of whether a corresponding effect on health care expenditures also occurs. A prior study demonstrated that receipt of SNAP in low-income adults is associated with lower health care expenditures over the course of 24 months when compared with eligible non-recipients [
18]. However, to date, it remains unknown whether this association holds true among children. Investigating this association is important because it carries significant policy and clinical practice implications. For example, findings could provide evidence supporting cost-saving, innovative partnerships between social services and medical homes, and for novel pediatric care delivery models that address food insecurity [
19]. Some insurers and state Medicaid agencies have already begun to implement direct nutrition supports under value based care arrangements, with the explicit goal of reducing costs; the relationship between SNAP, as a robust pre-existing food insecurity intervention, and health care costs outside the adult population is deserving of further investigation [
20,
21]. Thus, in this study we used data from a nationally representative survey to compare longitudinal health care expenditures among children in low-income households who received SNAP to those of children who did not receive SNAP, while accounting for sociodemographic and clinical differences in program enrollment.
Discussion
In this nationally representative sample of U.S. children, we found no statistically significant association between receipt of SNAP and lower health care expenditures (total, emergency department, inpatient, outpatient, prescription costs) when accounting for child-, parent-, and household-level sociodemographic and medical covariates. Consistent with previous studies, we found that SNAP participation is concentrated among families with high levels of unmet medical and social needs, limited parental educational attainment, and high rates of public insurance; the degree of vulnerability of SNAP-recipient families underlines the importance of this program, regardless of cost-saving considerations.
Our results differ from those of recent studies in adult SNAP recipients, which have found that SNAP was associated with lower health care expenditures [
8,
18]. One potential reason for this discrepancy between adult and child health care expenditures may be that the specific health conditions most impacted by receipt of SNAP in children might exert their effects on health care expenditures over a longer time period than was studied here. Whereas for adults, high prevalence conditions, such diabetes, may be more immediately impacted by food security and manifest in a more acute need for health care services [
17]. We studied cumulative expenditures over 24 months (the longest interval available in this dataset) and did note a trend towards significance (P=0.06), which supports this suggestion. Further, many studies of SNAP’s effects on child health focus on nutrition-related conditions such as low birth weight and childhood obesity, both of which have been found to be associated with significant lifetime costs [
32,
33]. The results of this study suggest that longer time horizons, likely well beyond 24 months, are needed to detect the effects that programs which moderate food insecurity have on health care expenditures. It is also worth noting that per-person child health care expenditures are lower than expenditures for adults ($2,479 per child in 2018, $5,644 per adult in 2018 based on MEPS data), so at baseline there may be less potential for short-term cost saving at baseline in children compared to adults.
Although we note that on an unadjusted basis, SNAP participants experience lower health care costs, this difference is not statistically significant when accounting for sociodemographic and clinical differences between the groups in the adjusted model. This finding suggests that unadjusted differences in health care expenditures likely relate, in large part, to underlying sociodemographic and clinical differences between continuous SNAP enrollees and non-SNAP participants, and not to participation in the program itself. We attempted to include as comprehensive a set of covariates as was possible with variables included in the survey, and found that SNAP recipients had significantly more social vulnerabilities than non-recipients, indicating that there may be other unmeasured covariates present. Recognizing that we could not comprehensively account for all co-factors, we also employed two statistical approaches aimed at accounting, to some degree, for unmeasured selection effects. Both approaches showed similar, non-significant effects of program enrollment on outcomes.
The results of this study carry implications for public policy pertaining to SNAP and other nutrition assistance programs. In recent years there has been vigorous political debate regarding the future of federal and state SNAP supports. On an unadjusted basis, policymakers may note lower (unadjusted) health care costs among SNAP participants, but our analysis suggests this may be driven primarily by the fact that SNAP participants often experience worse health and greater unmet social needs, increasing their risk for higher health care costs. When accounting for these differences, we note that over 24 months, unlike adults, children do not appear to have lower health care costs when enrolled in SNAP. It is critically important to note that the intended interpretation of this study is not to suggest a lack of program efficacy. The health and development benefits of SNAP participation for low-income children have been well-established, including fewer days of missed school, improved overall reported child health, fewer asthma-related ED visits, and improved birth outcomes, and these findings do not refute those points [
8,
16,
34‐
36]. Instead, we apply a health care cost framework onto SNAP participation, as a reflection of the growing movement towards the integration of social care into health care delivery. Better recognizing food insecurity and facilitating referrals to SNAP are immensely valuable functions for child health. Our results further reinforce the vulnerability of SNAP participants across other social domains, and suggest that expectations for moderating food insecurity to translate into immediate reduction in health care expenditures may be misplaced. We thus caution against misapplication of a health care financial model on a public health program which was not envisioned as a source of health care cost-savings, but instead as a societal and public health investment towards improving children’s nutrition, health, and well-being. It is our hope that these results might allow policymakers to manage their expectations on the financial returns of the program, and to encourage further research into the unique mechanisms through which SNAP affects health in children compared to adults.
Limitations of this study include the largely self-reported nature of the data, as well as the use of imputed data to replace missing values in MEPS. The accuracy of self-reported data in MEPS may vary; for example, a validation study found that MEPS households were accurate in their reports of inpatient admissions, but underreported ED visits by one-third [
37]. Restricting our analysis to children limited this study’s sample size relative to the entire population of MEPS participants, most of whom are adults. Our definition of receipt of SNAP benefits (24 consecutive months of benefits) further limited the sample size by excluding children who received SNAP benefits over a shorter time period from the analysis; however, because we detected no statistically significant difference in health care expenditures when using our definition of SNAP receipt, there is little reason to hypothesize that shorter-term exposure to SNAP would have a different effect. The version of the MEPS datasets used in this study did not include state-level identifiers, precluding us from using this information when adjusting for covariates. Finally, the longest longitudinal period used in this study was two years, and it is possible that the effects of SNAP on overall health and ensuing health expenditures in children could take longer than this relatively short time period to become apparent.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.