Experiences participating in federal nutrition assistance programs during the early months of the COVID-19 pandemic: A mixed methods study in Vermont

Federal nutrition assistance programs serve as safety nets for many American households, and participation has been linked to increased food security and, in some instances, improved diet quality and mental health outcomes. The COVID-19 pandemic brought new and increased economic, social, and psychological challenges, necessitating inquiry into how nutrition assistance programs are functioning and associated public health outcomes. Using data from a representative statewide survey administered in Vermont (n = 600) during the early months of the COVID-19 pandemic, we examine participant experiences with the 3 major federal nutrition assistance programs: the Supplemental Nutrition Assistance Program (SNAP), the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC), and school meal programs. We explore quantitative and qualitative responses regarding perceptions of program utility, and use nearest neighbors matching analyses in combination with bivariate statistical tests to assess associations between program participation and food insecurity, perceived stress, and fruit and vegetable intake as indicators of dietary quality. modi�cations occurring at each stage. This study incorporates data collected online between July 29, 2020 and September 17, 2020 from a sample of Vermont residents recruited by the survey research �rm Qualtrics. The sample of 600 Vermonters (age 18 and older) re�ects the state’s population pro�le with respect to race, ethnicity and income (64).

Both WIC and School Meal participation have likewise been associated with signi cant food security bene ts for children (11)(12)(13)(14)(15). Metallinos-Katsaras et al. (12) analyzed longitudinal data collected from 21,863 WIC households and found that reduction in food insecurity was mediated by earlier program entry and longer duration of participation. In a nationally representative study, public-school children from food insecure households obtained a larger proportion of their total daily calories and nutrients from school meals as compared to children from highly food secure households (13). Cullen and Chen (14) found that among 448 children ages 5-18 who consumed both school breakfast and lunches, the two meals provided 47% of daily energy intake. Based on higher reported rates of food insecurity during the summer months among 15,241 households with children receiving free and reduced-price lunch, Huang and Barnidge (15) suggest that National School Lunch Program participation may be associated with a reduction in food insecurity of roughly 14%.
The ndings of studies into associations between federal nutrition assistance programs and dietary quality may depend on the program. In a systematic review of 25 studies examining diet, Andreyeva,Tripp and Schwartz (16) found that overall caloric, macro and micronutrient intakes were not signi cantly different between SNAP participants and income eligible nonparticipants. Results of speci c studies on dietary quality are mixed, with some nding that SNAP participants had poorer overall diet quality than both income-eligible and higher income nonparticipants (16)(17)(18)(19), while others have found improvements in dietary quality among SNAP eligible respondents who used the program (20). Focusing on fruit and vegetable intake, Saxe-Custack et al. (21) recently found that although SNAP participation did not increase the probability of participants meeting USDA intake recommendations, it did signi cantly increase the mean daily consumption of both fruits and vegetables for a cohort of child participants. Others have shown that trends in fruit and vegetable purchasing among SNAP households vary signi cantly according to the bene t cycle, although they are similar on average to non-participant households (22). Evidence suggests that speci c incentive programs for SNAP participants, including Double-Up Food Bucks and other targeted fruit and vegetable purchasing incentives, may increase intake more than SNAP alone (23,24).
Studies suggest that WIC participation is signi cantly associated with improved diet quality in children (25,26). In an analysis of 1250 children enrolled in WIC, Wein eld et al. (26) also found that longer duration of program participation was associated with signi cantly higher Healthy Eating Index (HEI) scores as compared to eligible candidates who discontinued participation after infancy. A systematic review by Zhang et al. (27) also shows consistent, although not universal, positive correlations between fruit and vegetable purchasing and/or consumption by WIC participants since the 2009 program revision. Participation in daily school breakfast and lunch was associated with modestly healthier dietary intakes among 5,106 US school children, ages 4-15 (28). In a study of 3944 fourth and fth graders, school lunch eaters had higher average HEI scores than those who ate lunch brought from home, although this trend did not hold for school breakfast participants (29).
Food insecurity has been associated with multiple indicators of poor mental health (30)(31)(32). In a systematic review of 12 studies, Breuning et al. (30) identify a bidirectional relationship between food insecurity and negative emotional health in US-based populations. Myers et al. (32) likewise report signi cant positive associations between food insecurity and multiple measures of psychological distress based on an assortment of cross-sectional, longitudinal and secondary data studies in numerous countries. Focusing on high-income countries, Maynard et al. (31) nd associations between food insecurity and mental health metrics, including symptoms of depression, anxiety and stress, among women in a review of 39 studies. Even more recently, using cross-sectional data from the 2020 U.S. Census Household Pulse Survey (N = 63,674), Nagata et al. (33) report independent associations between food insu ciency and all measured indicators of poor mental health, controlling for sociodemographic covariates. Interestingly, they nd that this association was mitigated by receipt of free groceries and meals (33). Additionally, a separate analysis of the dataset used in the present study shows a complex relationship between stress and home food procurement activities, with gardening associated with stress reduction while shing, hunting and canning are associated with higher stress (34).
Several studies have attempted to examine how nutrition assistance programs may mediate relationships between food insecurity and various measures of mental health. Pulling data from the 2011-12 longitudinal SNAP Food Security Survey, Oddo and Mabli (35) nd that, among 3,146 U.S. households, 7.9% fewer household heads reported symptoms of psychological distress after 6 months of SNAP participation, and adjusted models show an associated decrease in psychological distress. Leung et al. (36) also examine the association between food insecurity and depression, as evaluated in the 2005-2010 NHANES dataset, restricted to adults earning no more than 130% of the federal poverty level. Controlling for sociodemographic and health covariates, they nd a signi cant positive association between food insecurity and depression, but SNAP participation lessened the strength of this relationship. However, Adynski et al. (37) found that, controlling for demographic variables, SNAP and WIC participation did not reduce the risk of depressive symptoms in a nationally representative sample from the NHANES 2013-14 and 2015-16 cohorts, while elevated levels of food insecurity were associated with higher risks of depressive symptoms.
Program Participation and Value Assignation SNAP Despite established bene ts, not all eligible participants take advantage of federal nutrition assistance programs. Demographic trends in participation vary between programs, although there is signi cant of overlap between programs. In nationally representative studies, SNAP participants are more likely to be female (38), and tend to be younger than eligible nonparticipants (39). As of 2018, a signi cant majority (81%) of SNAP households contained at least one child, elderly individual, or individual with a disability, and 61% of those with children were single-adult households (38). Additionally, 81% of SNAP households lived in poverty (38). Monthly SNAP participation has uctuated signi cantly since the program's origin, peaking at 15.2% of residents nationwide in 2013 after the Great Recession and subsequently declining (40). Notably, early ndings suggest that nationally, among households with children, participation in SNAP declined in the early months of the pandemic, whereas participation in WIC increased slightly and school meal participation remained consistent (41) Program administration challenges, including failures in customer service and di culties navigating administrative bureaucracy are common complaints, speci cally during the application and renewal processes (42). Some SNAP participants have expressed concerns over bene t adequacy (42)(43)(44). In a series of mixed methods studies by Leung et al. (42,44), SNAP bene ciaries expressed strong support for increasing total SNAP bene t allotment (84.2-89%), as WIC In contrast to uctuations in SNAP participation, WIC enrollment has been relatively stable, with rates roughly re ecting broader sociodemographic trends (45).
However, there is signi cant variability in the duration of WIC enrollment, with characteristics such as lower household income increasing the probability of sustained enrollment (46) and factors such as breastfeeding and home ownership associated with lower intent to maintain enrollment (47). Overall estimates of WIC-eligible nonparticipation rates range as high as 50% (48). In a population-based randomized survey of 1,634 postpartum women in New York City, WICeligible mothers facing structural barriers such as lack of transportation, unplanned pregnancy, and limited social supports were less likely to participate in WIC (49).
Insights from various qualitative studies indicate multiple other factors that in uence program value assignation. Themes include: (1) logistical challenges in meeting time and transportation demands; (2) program administration challenges, including failures in communication and organization, as well as variable exibility and amicability from program staff; and (3) challenges in the retail environment, including inconsistency and/or di culty identifying eligible foods, lack of choice, lack of training of store employees, and perceptions of stigma (43,48,(50)(51)(52)(53).
Perceptions of and value placed on federal nutrition assistance programs by participants in uence not only initial program participation, but intent to maintain enrollment despite sometimes stringent requirements. In a mixed-methods study of 150 WIC caregivers in Illinois (51), participants assessed perceived value of program packages at each stage of the program (pregnancy, infancy, childhood) against the time and effort required to maintain eligibility. During pregnancy and infancy, most participants believed bene ts to be worth the time and effort (70%; 91%), hypothesized to be a function of the high cost of formula. However, only 36% believed the child program package value to be worth the effort once past reliance on formula. Additionally, the more restrictive selection options under WIC may make WIC more di cult and stigmatizing to use than SNAP but are viewed by some as a useful incentive to eat more healthfully (50,51).

School Meals
School meal participation is highest among children eligible for free meals, and especially where universal free meals are offered (54,55). Rates of school breakfast participation are signi cantly lower than for lunch (56). Nontraditional breakfast formats, including meals served in classrooms after the start of the day and grab and go options, have been shown to promote participation (54,55). The implementation of the Healthy Hunger Free Kids Act in 2010 has not been associated with signi cantly altered rates of school lunch participation (57,58).
Multiple studies suggest that parental and student perceptions of school meals are signi cantly associated with program participation (59)(60)(61). Perceived healthfulness of meals by parents was associated with approximately 3.8 times greater meal participation rates in a survey of New Jersey households with children aged 7-18, and a signi cant majority viewed school lunch as healthy (59). In contrast, in a mixed methods survey of 488 school parents in Utah perceptions of healthfulness were mixed, with only 40% expressing the belief that the school breakfasts were healthy, and 45% unsure (62). Likewise, student perceptions of school meal healthfulness were associated with higher rates of participation in a cohort of 5,106 students from 130 US communities, although these perceptions did not correlate with objective measures of healthfulness (61). Sociocultural preferences and logistical challenges have also been shown to affect school meal participation. Examples include valuing of family mealtime, as well as concerns over school meal quality and food choice (56, 62, 63).

Data Collection
This study uses survey data collected by the National Food Access and COVID Research Team (NFACT), a multistate collaborative effort (64). Survey questions examine various aspects of Vermonters' experiences with food access and food security and related worries during the pandemic, in addition to a broad set of demographic characteristics (64). Multiple iterations of the survey, beginning in March 2020, have been administered, both within the state of Vermont and nationally, with modi cations occurring at each stage. This study incorporates data collected online between July 29, 2020 and September 17, 2020 from a sample of Vermont residents recruited by the survey research rm Qualtrics. The sample of 600 Vermonters (age 18 and older) re ects the state's population pro le with respect to race, ethnicity and income (64).

Relevant Variables
Independent variables for this study included select demographic characteristics and binary variables re ecting participation in three federal nutrition assistance programs: SNAP, WIC and school meals. Of note, we classify recipients of the special Pandemic Electronic Bene ts Transfer (P-EBT) program, offered to families of children who would have received free or reduced school meals prior to shutdowns, as school meal participants. To compensate for relatively small sizes within each category, most demographic variables (e.g., gender, race) are condensed or analyzed as binary (Supplementary Table S1).
Variables were also created to re ect participation in multiple or any federal nutrition assistance program. As one aim of this study was to distinguish between low-income and other program participants, we created a variable to re ect participants that fell above or below 200% of the federal poverty level (FPL) based on household size. To do so, rst, average household income was calculated based on reported categories. To calculate, we averaged each income category (i.e., if the respondent reported a household income between $10,000 and $14,999, this was averaged to $12499.50). These averages were compared to 200% of the FPL based on reported household size (65). If a respondent's average household income fell below the 200% threshold, they were classi ed as low income.
We also evaluate four dependent variables based on self-reported data: food security, fruit intake, vegetable intake, and perceived stress. Food security status is evaluated using the validated USDA 6-item short-form food security module, modi ed to re ect experiences since the start of the pandemic (March 2020) (66). Following established scoring procedures, respondents who answer 2 or more out of the 6 survey questions positively are classi ed as food insecure (66).
Survey respondents reported perceived fruit and vegetable intake based on binned categories (none, 1/2 cup or less, 1-2 cups… 4 or more cups). For analyses, these are condensed to re ect whether perceived intake did or did not meet USDA guidelines for fruit and vegetable intake (67). Given that established thresholds for fruit (2 cups) and vegetable (2.5 cups) intake do not neatly correspond with survey categories, any respondents reporting fruit intake of 1-2 cups or more and vegetable intake of 2-3 cups or more of vegetables are classi ed as meeting intake recommendations. Accordingly, our recategorization may slightly overestimate the proportion of respondents who meet fruit and vegetable intake recommendations.
Stress is measured using the validated 4-item perceived stress scale (68), which poses a series of scenarios to which respondents indicate that they occur never (0) to very often (4). The scale is corrected for all questions so that higher scores re ect higher stress, which requires reverse scoring on two of the four questions. Results are then summed for an overall perceived stress score of 0-16.
Finally, respondents who participated in federal nutrition assistance programs were asked to respond on a 5-point Likert scale (strongly disagree to strongly agree) to several statements regarding their experiences with the programs. Participants were given the option to make further optional comments on their experiences.

Data Analysis
We use descriptive statistics to assess individual demographic characteristics of federal nutrition assistance program participants and their experiences with these programs during the early months of the COVID-19 pandemic, and bivariate tests (chi-2 or t-tests, based on data type) to assess demographic differences between low-income program participants and nonparticipants with an alpha level of .05 indicating signi cant differences. Where sample size allows, statistical tests are conducted on SNAP and WIC participants separately.
We also summarize open-ended comments provided by nutrition assistance program participants about the programs. Notably, open-ended responses were optional, and many respondents elected not to provide substantive comments, such that broad trends are di cult to identify, particularly among the small WIC subsample. Given the small sample of qualitative data, comments were divided by relevant program and coded into three broad themes using NVIVO version 20 (69): program challenges, program bene ts, or both.
In order to estimate the effects of federal nutrition assistance program participation on food security, fruit and vegetable intake, and perceived stress, we use chi-square tests, t-tests, and nearest neighbors matching techniques. Nearest neighbors matching is useful to approximate causal treatment effects where only observational data is available (70). In simple terms, matching techniques attempt to compensate for selection bias by selecting those untreated individuals who are most similar to a sample of treated individuals based on a set of prede ned relevant characteristics. In the context of this study, federal nutrition assistance program participation serves as the treatment. However, given the signi cant variation in programming between school meal programs, SNAP and WIC, we run matching analyses on participants of any nutrition assistance program and with school meal participants broken out into their own treatment group to compare results. Additionally, school meal eligibility during this time was universal rather than income-based, further distinguishing the program from SNAP and WIC. SNAP and WIC participants are analyzed together as a single treatment group due to the small sample size of WIC participants.
All analyses are strati ed by income (low/high) to assess the differential impacts of nutrition assistance program participation on these groups. In each of these analyses, we match program participants to nonparticipants based on a set of six demographic covariates that are likely to be associated with program participation or relevant outcomes: age (under 35), children in household, negative job change, education (at least a bachelor's degree), household size (4 or more individuals), and rurality.
Nearest neighbors matching techniques are based on a measure of distance, called a caliper, that evaluates the relative similarity of demographic covariates between individuals. Matches are selected based on the shortest "distance" that can be found, but to ensure quality matches we set a maximum caliper for each analysis at the smallest caliper (shortest distance) that allowed at least 2 matches to be found. Calipers vary slightly between analyses, but all were set at a maximum of between .15 and .25, with higher calipers employed for analyses in which matches were more di cult to nd. Using these matches, we report average treatment effect on the treated, which assesses the difference between expected outcomes (food security, fruit intake, vegetable intake, and perceived stress) with and without treatment (nutrition assistance programs) for those who participate in treatment (70).

Section 1: Descriptive Statistics
Sample demographics  Note. Sample size is adjusted for several variables based on missing data. Within the full sample, n = 593 for children in household and household size; n = 594 for job disruptions; n = 599 for rural/urban residence. Within the low-income non-participants, n = 85 for rural/urban residence.

Federal nutrition assistance program participation
Of all respondents, 27.3% (n = 164) reported that their household used at least one federal nutrition assistance program and 5.67% (n = 34) reported that their household used two or more programs. Divided by program, 68 respondents participated in a school meal program, 114 participated in SNAP and 25 participated in WIC.
Over half (57.4%, n = 116) of low-income respondents participated in at least one federal nutrition assistance program. Among these respondents, there is a signi cant association between age and program participation, with 47.6% of 18-34-year-olds participating, compared to 71% of 35-54-year-olds and 56.1% of those 55 and over (Table 2). We also nd signi cant associations between program participation and gender (x 2 (1) = 4.778, p = .029), and presence of children in the household (x 2 (1) = 6.075, p = .014) with higher participation among those who do not identify as female and those living in a household with children. Finally, we nd a signi cant association between program participation and education, with the highest rates of participation among those who have some college or an associate degree (compared to those with more or less education), and the lowest rates of participation among those with a college or advanced degree.
We also nd signi cant associations between program participation and gender (x2(1) = 4.778, p = .029), presence of children in the household (x2(1) = 6.075, p = .014) and education, with higher participation among those who do not identify as female, have some college or an associate's degree (compared to those with more or less education), and those living in a household with children Table 2 Crosstab of federal nutrition assistance program participation by select demographic characteristics among low-income respondents (n = 202).

Variable
Participating in any program (n = 115), n (%) Not participating in any program (n = 86), n

Program Experiences
When asked to express their level of agreement with a series of position statements, most of both SNAP and WIC participants, 78% and 80% respectively, agreed or strongly agreed that the bene ts are easy to use (Fig. 1). Only 14% of SNAP participants agreed with the statement "we are not able to use our full months' worth of SNAP bene ts," while, 60% of WIC participants agreed or strongly agreed that they could not use a full month's worth of WIC bene ts.
However, nearly half (47%) of participants disagreed with the statement "SNAP bene ts are enough to meet our household's needs," suggesting that bene ts alone did not fully compensate for household food security needs. Just over half (54%) of SNAP participants neither agreed nor disagreed that they were unable to use their bene ts to order groceries online, suggesting that these respondents may not have attempted to do so. However, of WIC participants, 72% agreed or strongly agreed that they would be interested in online shopping for WIC foods with delivery or curbside pickup options. Seventy-two percent also agreed with the observation that there is a limited selection of foods that can be purchased with WIC bene ts (Fig. 2).
Most school meal participants (78%) and P-EBT recipients (71%) agreed that these programs had been helpful to their families. When asked to report their level of agreement with speci c challenges related to school meals during the pandemic, the most common complaints were that school meal sites were not consistently open (28%), home delivery was not available or was di cult (27%), and that participants were unable to pick up at the sites (27%) and times (23%) offered. Fewer than 20% of participants reported running out of meal provisions before the next delivery dates (19%) or limitations related to inadequate kitchen equipment needed to store and reheat meals (11%).
Among all program participants, 35% agreed that they did not want to rely on nutrition assistance programs because they valued personal independence and 22% expressed worry that others would nd out they used programs. Others expressed pragmatic concerns with qualifying and recertifying for programs, including possessing too many personal assets to be eligible (27%), di culties travelling to program sites for appointments (23%), and worries over the paperwork needed to enroll (17%).
In open-ended comments, participants responded with a mix of gratitude for the programs along with discussion of challenges and limitations. About a quarter of WIC participants provided further comments on their experiences, of which a couple commented that the selection of foods offered was limited, and not always available during the pandemic, e.g., "its been harder to get certain WIC items since COVID". Over a dozen SNAP participants, or roughly a third of those who provided qualitative data, commented on the helpfulness of the program during the pandemic, with particular emphasis on the necessity of the temporary increase in bene ts provided: "the increase was very much appreciated and needed"; "the extra money is necessary for both before and after the pandemic".
However, echoing responses to closed-ended questions, numerous SNAP participants elaborated on challenges they faced with their bene ts. Some argued that, even with temporary increases, bene ts were inadequate to meet their needs, whether due to rising costs or supply shortages: "It's not enough given the rising costs of everything," said one participant, while another stated, "I feel like the bene ts didn't go as far because I had to buy name brand items due to [the] store brand [being] sold out". Another respondent observed cyclical challenges associated with bene ts noting that "everyone shops on the rst of the month, if the store is out, some people go without. I get SNAP & SSI [social security insurance], and my money is all gone by the 10th of every month". Other observed challenges included bureaucratic issues in qualifying, limited bene t eligibility due to age and seemingly arbitrary changes to bene ts, as well as limited opportunities to shop online.
Participants were also given the opportunity to comment on the P-EBT and school meal programs, resulting in substantive comments from about 20 participants. Most responses suggested that the programs had been helpful: "I don't know what we would have done with[out] the school meals. We appreciate them more than many people can imagine." However, a small subset reiterated that the programs were still "Not enough to feed the kids," or wished for their continuation, i.e., "P-EBT was a blessing and I wish we had more". Although few speci c challenges were discussed, one respondent did note that their family did not prefer the taste of school meals, and another that delivery options were important to the value of school meals during the COVID-19 pandemic: "When meals were being delivered by bus, they were very helpful. Grocery stores didn't have items we needed in stock, and buying groceries was more expensive than school lunches had been. Getting food deliveries was very helpful to my family. When they stopped delivering, we were unable to pick up meals at the allotted time."

Outcome variables
Among all respondents, 71.0% were consistently food secure since the start of the COVID-19 pandemic. Most did not meet USDA recommendations for either fruit intake (58.5%) or vegetable intake (72.3%). The average perceived stress score (out of 16) was calculated to be 6.85 for the full sample. Low-income respondents were signi cantly less likely than higher income respondents to meet fruit and vegetable recommendations (p < .001) and were signi cantly more likely to have experienced food insecurity and higher perceived stress since the start of the COVID-19 pandemic (p < .001). Note. Sample size is adjusted for several variables based on missing data. For the food security variable, n = 583 for the full sample; n = 193 for the lowincome sub-sample; n = 81 for program non-participants; n = 65 for school meal participants; n = 160 for all program participants; n = 122 for SNAP/WIC participants. For the perceived stress variable, n = 597 for the full sample; n = 201 for the low-income subsample; n = 163 for all program participants; n = 123 for SNAP/WIC participants.

Nutrition Assistance Program Participation and Outcomes
Federal nutrition assistance program participation and food security We nd a signi cant association between federal nutrition assistance program participation and food insecurity through chi-square tests, with 18.1% of nonparticipants classi ed as food insecure, compared to 57.5% of participants (x 2 (1) = 87.436, p = < .001) within the full sample. This association held for SNAP, WIC and school meal participation (p = .001) for the full sample of respondents, but when low-income respondents were examined alone, there was no signi cant relationship between school meal participation and food insecurity, and the association was weaker but still signi cant for SNAP/WIC participants (p = .031). However, when low-income SNAP and WIC respondents were examined separately, the signi cance held for SNAP but not WIC respondents.
Using matching techniques to approximate the effects of federal nutrition assistance program participation on food security, we nd a signi cant association between participation in any program and increased food insecurity since the start of the COVID-19 pandemic, both for the higher income (p = .001) and the low-income group (p < .001). In other words, among similar higher-income households, those using any federal nutrition assistance program are more likely to be food insecure compared to those who are not using any program. The same is found when comparing among otherwise similar low-income households.
When school meal participation and SNAP/WIC participation are evaluated as separate treatments, we nd that this association holds true for SNAP/WIC participation among respondents that are not low-income (p < .001) ( Table 4). Note. Each program participation variable was used as a "treatment" in a separate matching analysis while using six demographic controls (gender, children in household, education negative job change, household size, rural/urban) to conduct the matching. Negative coe cients re ect an association with increased food security.

Federal nutrition assistance program participation and fruit and vegetable intake
We use matching analysis and chi-square tests to evaluate associations between program participation and fruit and vegetable intake. Using chi-square tests we nd a signi cant association between SNAP/WIC program participation and reduced probability of meeting fruit intake recommendations in both the full sample (p = .001) and low-income subgroup (p = .049), but not for school meals. This trend is only signi cant in the full sample when all programs are grouped (p = .008). When SNAP and WIC are examined separately, only SNAP participation within the full sample is signi cantly associated with reduced fruit intake (p = .001). When matching analysis is used to account for select demographic controls, we nd no signi cant associations for individual programs, but we do see a signi cant association between participation in any program and reduced probability of meeting fruit intake recommendations within the lowincome group (Table 5; p = .048). Meaning that for low-income households participating in a program, compared to other low-income households not participating in a program, there is lower likelihood of meeting fruit intake recommendations. Note. Each program participation variable was used as a "treatment" in a separate matching analysis while using six demographic controls (gender, children in household, education negative job change, household size, rural/urban) to conduct the matching. Negative coe cients re ect an association with a reduced probability of meeting recommended fruit intake levels.
Using chi-square tests we nd a signi cantly reduced probability of meeting vegetable recommendations for both SNAP/WIC participants (p < .001) and all participants grouped (p < .001). Within the full sample, this trend holds for SNAP but not WIC when each is examined alone, but sample size precludes conducting this same analysis within the low-income subsample. We nd a weaker association between SNAP/WIC participation and a reduced probability of meeting vegetable recommendations within the low-income group by matching analysis (Table 6; p = .035). Note. Each program participation variable was used as a "treatment" in a separate matching analysis while using six demographic controls (gender, children in household, education negative job change, household size, rural/urban) to conduct the matching. Negative coe cients re ect an association with a reduced probability of meeting recommended vegetable intake levels.

Federal nutrition assistance program participation and perceived stress
Within the full sample, we nd signi cantly higher rates of perceived stress among all grouped program participants (p < .001), SNAP and WIC together (p < .001) and SNAP participants alone (p < .001) by t-tests, with an average score of 8.17 for SNAP participants as compared to 6.53 for nonparticipants. However, school meal and WIC participants analyzed alone do not exhibit signi cantly higher rates of perceived stress, and no signi cant associations hold for any program when the sample is restricted to only low-income respondents.
We also use matching techniques to examine the effects of program participation on perceived stress. The only model for which we nd signi cant effects of program participation on stress scores is that for school meal participation among low-income respondents, wherein we nd a signi cant negative association indicating reduced stress (Table 7). Note. Each program participation variable was used as a "treatment" in a separate matching analysis while using six demographic controls (gender, children in household, education negative job change, household size, rural/urban) to conduct the matching. Negative coe cients re ect an association with a reduced perceived stress score.

Discussion
This study builds on prior literature examining food insecurity during the ongoing COVID-19 pandemic by exploring the role of federal nutrition assistance programs within the context of Vermont, a predominantly rural state characterized by a relatively robust policy and social response to the pandemic (71). Rural populations experience higher rates of food insecurity and are more likely participate in nutrition assistance programs as compared to their urban counterparts (4,72). Vermont readily accepted and implemented numerous federal waivers made available in association with the Families First Coronavirus Response Act to increase the exibility of programs in response to COVID-19, from provision of P-EBT to suspension of certain face-to-face interview requirements to the temporary restructuring of school meal delivery (73). The state also boasted among the lowest COVID-19 caseloads at the time of survey administration (71).
We found that, despite shortcomings, participants generally perceived federal nutrition assistance programs as helpful or easy to use. We documented notable levels of food insecurity, suboptimal fruit and vegetable intake, and perceived stress among participants and non-participants alike. Understanding food and nutrition security and perceived stress outcomes under these conditions can provide insights regarding how nutrition assistance programs can provide for the most vulnerable even in such challenging times.
Our ndings regarding nutrition assistance program perception correspond closely to prior studies on several counts. In line with calls for increased total SNAP bene t allotment in other studies (44,74), we nd that 47% of participants felt that bene ts were not adequate to meet their household's needs. Likewise, restricted product options available through WIC are a continuing topic of debate (51), which is re ected in our results, although our respondents were not prompted make a value judgment on these limitations. Our results also suggest that online utilization options for SNAP and WIC participants could improve accessibility and e cacy of these programs, which echoes perceptions of convenience and time savings associated with online options in earlier studies (48, 53, 75, 76). However, in considering expanding online options for these programs, it is important to revisit reported concerns, including associated transaction costs and limited control over selection and quality (75,77,78). Logistical challenges associated with school meal delivery in other studies, such as scheduling con icts with school breakfasts (56, 63), are more di cult to evaluate against our results, given the altered format of meal delivery during the pandemic. However, new and additional logistical challenges, including accessibility of school meals related to site location, timing, and lack of delivery options continue to pose challenges for a substantial number of participants. However, at-home options for school meals may overcome other perceived challenges, such as infringement on family mealtimes (63) Despite a perception of utility among participants, we found that -aligning with prior literature (4, 42) -federal nutrition assistance program participation was signi cantly associated with food insecurity for both low-income and not low-income respondents, even when we explore this using matching techniques to compare similar households. We also found a reduced probability of meeting fruit intake recommendations for low-income program participants, and a reduced probability of meeting vegetable intake recommendations for low-income SNAP/WIC participants. There is a substantial body of literature supporting the e cacy of federal nutrition assistance programs in alleviating food insecurity and, to some degree, improving diet quality (7,8,79). Our ndings should be considered within the unique context of the time and place at which our data were collected, and of course, as subject to limitations in the analytical models employed.
It has been documented that the pandemic exacerbated food insecurity (80) and changed eating patterns. While some studies have found evidence of increased fruit and vegetable intake during the pandemic (81), these impacts were not universal. In a survey of Michigan adults, Litton and Beavers (82) found that food insecure respondents not only consumed fewer fruits and vegetables than their food secure counterparts but were more likely to report decreasing fruit and vegetable consumption in the early months of the pandemic, for reasons including quality, availability, price, desire to reduce store trips, and fears of Pandemic-speci c challenges to food access may be particularly potent in a relatively rural context. Rural populations experience food insecurity at elevated rates, which may be exacerbated by structural barriers affecting access, such as large distances to supermarkets (88, 89). It is possible that, under these conditions, online delivery or curbside pickup options were more challenging or prohibitively expensive for some rural residents. Additionally, social support systems and community networks may play a unique role in the mitigation of food insecurity in rural settings (90), and constraints associated with new social distancing regulation compliance may have impacted such avenues. Additionally, although roughly one-third of our sample is classi ed as rural using the Rural Urban Commuting Area 4 category (RUCA) designation, the population density is low across the state, with only approximately 45,000 people, or about 10% of the state's population residing in its largest city (91). In combination, exacerbated experiences of food and nutrition insecurity in concert with new structural barriers to nutrition assistance program utilization may have limited the capacity of programs to operate optimally or fully compensate for negative shifts in these outcomes for Vermont's rural population. These limitations leave room for new and expanded strategies to improve access to healthy foods among program participants.
Although few studies have focused speci cally on perceived stress during the pandemic, food insecurity is positively associated with various indicators of poor mental health (31,32,92). Some research indicates that this relationship may be attenuated by participation in SNAP and WIC (35,36), but -in line with our ndings -other studies have not identi ed a moderating effect (37). Interestingly, we do nd that perceived stress was signi cantly lower among lowincome respondents with a household member enrolled in a school food program. Few studies have explicitly evaluated the relationship between school meal participation and perceived stress of household members. Our results merit further investigation -Could school meal delivery or pickup have reduced the necessity of public ventures and associated risk of exposure for some families? Could school meal delivery and pick up options during the pandemic have in uenced the frequency of family meals or reduced the burden of meal planning? While Nagata et al. (33) did not examine school meal participation during the pandemic, they do report a mediating effect of free groceries and meals on the association between food insecurity and poor mental health. Similarly, in a qualitative analysis, commercial family meal kit use has been associated with perceived bene ts including reduced mental load for family meal providers (93). Additionally, evidence suggests that frequent family meals may be associated with lower rates of depressive symptoms and stress among parents (94).
While these results cannot be directly applied to the role of school meals during the pandemic, further research could illuminate the pathways through which pandemic school meal formats might have impacted perceived stress in the household. It is also relevant to consider that the reduced burden of applying for school meal programs was removed due to the universal delivery approach taken during the time period under study.
As with any study, our survey and methods are subject to reasonable limitations. Although our sample size is relatively modest, it has a margin of error of 4% and was intentionally designed to re ect Vermont's population on key demographic factors including income, race and ethnicity. Due to the limited sample of WIC participants, this population is only analyzed in conjunction with SNAP or both SNAP and school food participants, limiting the capacity of this study to distinguish between the impacts of these distinct programs. By design, our survey captures a breadth of data related to food insecurity during the COVID-19 pandemic but covering a substantial range of material naturally limits the depth of data that can practically be captured in any one area. Additionally, given the evolving nature of the pandemic context, subjects of importance to individual experiences of food and nutrition security continue to shift and emerge over time. For example, the administration of school meal programs was evolving at the time of data collection, making it challenging to clearly assess which speci c program components and iterations participants responded to. Additionally, the limited number of open-ended responses collected suggests that the full breadth of participant experiences may not have been captured, although the supplementary qualitative data nonetheless adds depth.
Nearest neighbors matching analysis seeks to address weaknesses associated with selection bias in non-experimental study designs, but in the absence of perfect knowledge, such tools cannot perform perfectly. Although our survey captured many demographic variables known to be associated with food and nutrition security, perceived stress and federal nutrition assistance program participation, these are complex constructs in uenced by a myriad of interrelated factors. It is likely that additional confounding variables exist that we were unable to fully evaluate. For example, our matching analysis was not able to account for differences in social support, which has been shown to in uence food insecurity and perceived stress. Additionally, we were unable to control for the role that disability, physical health and comprehensive mental or emotional health may have on these outcomes, although such factors are known to interact with the experience of food insecurity (92,95). Furthermore, our survey design could not meaningfully capture the speci c food environments of participants, which can signi cantly impact food and nutrition security (88). Integrating these variables into future analyses might better isolate the impacts of federal nutrition assistance programs on food and nutrition security and perceived stress.
Of note, our matching analysis used binary outcomes to maximize power, except for perceived stress. However, by examining outcomes as binary, our models do not account for changes in the intensity of an outcome. While federal nutrition assistance program participation did not reduce food insecurity in our analysis, it is possible that participants may have experienced a reduction in the degree of food insecurity experienced, which would not be re ected in our models.

Conclusions
Within the context of a pandemic in the state of Vermont, federal nutrition assistance programs were broadly not adequate to address the experience of food insecurity and stress or increase fruit and vegetable intake. However, participants nonetheless perceived these programs as helpful and may have experienced other bene ts, including reduced stress among low-income school meal program participants. Continuing research on the delivery and impacts of nutrition assistance programs, particularly in other rural contexts, is needed to inform their implementation. If federal nutrition assistance programs are to function effectively as vital safety nets, they must continue to evolve as new challenges to food and nutrition security emerge. This study was determined to be exempt from review by The Institutional Review Board of the University of Vermont. Consent was implied by survey completion following review of the study's Information Sheet. The researchers complied with all human subject protection regulations.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
Funding for this study was provided by the University of Vermont College of Agriculture and Life Sciences and the O ce of the Vice President of Research, as well as a COVID-19 Rapid Research Fund grant from the Gund Institute for Environment. The funders had no role in the design of the study and collection, analysis, and interpretation of data.
Authors' contributions MTN, EHB, and FB designed and disseminated of the survey instrument. EHS and MTN analyzed the data, with input from EHB and FB. EHS drafted the manuscript and prepared all tables and gures. All authors read and provided edits and revisions on the full manuscript. Figure 1 Experiences of SNAP participants during the early months of the COVID-19 pandemic in Vermont.

Figure 2
Experiences of WIC participants during the early months of the COVID-19 pandemic in Vermont.

Figure 3
Experiences of School Meals recipients during the early months of the COVID-19 pandemic in Vermont.