Food Security Impacts of the COVID-19 Pandemic: Longitudinal Evidence from a Cohort of Adults in Vermont during the First Year

This study assessed changes in household food insecurity throughout the first year of the COVID-19 pandemic in a cohort of adults in the state of Vermont, USA, and examined the socio-demographic characteristics associated with increased odds of experiencing food insecurity during the pandemic. We conducted three online surveys between March 2020 and March 2021 to collect longitudinal data on food security, use of food assistance programs, and job disruptions during the COVID-19 pandemic. Food security was measured using the USDA six-item module. Among the 441 respondents, food insecurity rates increased significantly during the pandemic and remained above pre-pandemic levels a year after the start of the pandemic. Nearly a third (31.6%) of respondents experienced food insecurity at some point during the first year of the pandemic, with 53.1% of food-insecure households being classified as newly food-insecure. The odds of experiencing food insecurity during the pandemic varied based on socio-demographic factors. Households with children (OR 5.5, 95% CI 1.782–16.936, p < 0.01), women (OR 8.1, 95% CI 1.777–36.647, p < 0.05), BIPOC/Hispanic respondents (OR 11.8, 95% CI 1.615–85.805, p < 0.05), and households experiencing a job disruption (OR 5.0, 95% CI 1.583–16.005, p <0.01) had significantly higher odds of experiencing food insecurity during the first year of the COVID-19 pandemic, while respondents with a college degree (OR 0.08; 95% CI 0.025–0.246; p < 0.001) and household income of ≥USD 50,000 (OR 0.01; 95% CI 0.003–0.038; p < 0.001) had lower odds of experiencing food insecurity. These findings indicate that food insecurity continued to be a significant challenge one year after the start of the pandemic, which is important, given the adverse health impacts associated with food insecurity and health disparities among certain socio-demographic groups.


Introduction
A year after the start of the COVID-19 pandemic in the United States, the pandemic itself and the policies and restrictions put in place to reduce the spread of the virus continued to disrupt economies and labor markets, with serious implications for food insecurity. Food insecurity, defined as the lack of consistent physical and economic access to sufficient, safe, and nutritious food for an active and healthy lifestyle [1], is closely aligned with national and household economic conditions. Trends in food insecurity rates typically parallel those of unemployment, poverty, and food prices [2][3][4], though food insecurity can also result from non-economic drivers including limited physical access to food retailers and lack of transportation [5][6][7]. The COVID-19 pandemic likely exacerbated these challenges to food access due to safety concerns about shopping in stores, limited hours at food retailers, and changes in public transit access.

1.
What was the trajectory of food insecurity during the first year of the COVID-19 pandemic among a cohort of respondents? 2.
What socio-demographic factors and life experiences were associated with increased odds of experiencing food insecurity during the first year of the COVID-19 pandemic? 3.
What factors, if any, contributed to the recovery from food insecurity during the first year of the COVID-19 pandemic?

Survey Development and Recruitment
We surveyed a cohort of adults in Vermont three times during the first year of the COVID-19 pandemic. The original survey [27] was developed with feedback from key statelevel agencies and hunger relief organizations, as well as reviews of relevant literature [2,10], to measure food insecurity, food access challenges, and related concerns and experiences. When possible, this survey utilized existing validated questions and was updated after each round of data collection to add new relevant questions as the pandemic evolved. Using LimeSurvey [28], the instrument was piloted with 25 adults (18 years and older) from the target population of adults in Vermont.
The first survey ran online from 29 March to 12 April 2020, with a total of 3219 respondents. We used four methods for convenience sample recruitment: (1) paid advertisements via Front Porch Forum, a community-level listserv, which reaches approximately 2/3 of Vermont households [29]; (2) paid digital ads via Facebook to reach populations underrepresented in Front Porch Forum (e.g., males, lower-income households); (3) listservs of community partners; (4) a University of Vermont press release and subsequent newspaper, radio, and television media. Respondents could opt-in to be contacted for future surveys at the first survey. We conducted two follow-up surveys with the same respondents in May/June 2020 and March/April 2021. The second survey ran online from 21 May to 4 June 2020, with a total of 1236 respondents. The third survey ran from 29 March to 21 April 2021. The first two surveys were conducted using LimeSurvey, and the third was conducted through Qualtrics [30]. Respondents with ZIP Codes outside Vermont and empty responses (i.e., people who consented but did not fill in any responses) were removed, leaving 441 eligible individuals who responded to all three surveys ( Figure A1).
To measure household food security status, we adapted the U.S. Department of Agriculture's (USDA) Household Food Security Survey Module: Six-Item Short Form [31] to ask about different time periods. The six questions are related to having enough food or enough money for food, being able to afford a balanced diet, and disrupted eating patterns (i.e., cutting the size of meals, skipping meals, going hungry). For example, one of the questions reads "Did you ever eat less than you felt you should because there wasn't enough money for food?" with response options of "yes", "no", or "I don't know". We adapted the ends of the questions to ask about specific time periods. Across the three surveys, we assessed five timepoints. In the first survey, respondents were asked about food security both "in the year before the coronavirus outbreak" and "since the coronavirus outbreak". The responses to the pre-pandemic questions about food security were retrospective and were answered at the same time as the questions about current food security. The start of the coronavirus outbreak was set as 8 March 2020, based on the first positive COVID-19 test result in Vermont. In the second survey, respondents were asked about food security "in the last 30 days" to reflect the time since the first survey. In the third survey, respondents were asked about food security both "since June 2020" (to reflect the time since the second survey) and "in the last 30 days". We followed the procedures outlined in the USDA Six-Item Food Security Module [31] to code the responses to the six food security questions and calculate a food security status score for each respondent during each time period. We used a binary food security status score where two or more affirmative responses to the six questions is scored as food-insecure and fewer than two affirmative responses is scored as food-secure.
To determine respondents' urban/rural classification, we used their ZIP Codes and the Rural-Urban Commuting Area (RUCA) codes [32] and the four-category classification scheme (urban, large rural, small rural, and isolated) created by the Rural Health Research Center [33]. We condensed this into a binary variable of urban and rural (which includes the categories large rural, small rural, and isolated from the four-category classification).
In addition to measuring food security status, the survey included questions related to food access challenges, use of food assistance programs, food purchasing behaviors, concerns about food access and availability, COVID-19 perceptions, and behaviors and demographics. Table A1 lists the specific questions used in this analysis, which primarily focused on understanding food security status, use of food assistance programs, and job disruptions. Future analyses will explore other questions in the survey.

Statistical Analysis
To examine differences in household food security during the first year of the pandemic, we created two core categories of respondents: (1) households with food security (n = 307, including households that were always food-secure during the first year of the COVID-19 pandemic as measured for all five time points) and (2) households with food insecurity (n = 134, including households that were food-insecure at any time during the first year of the COVID-19 pandemic). We further categorized food-insecure households into two groups: (1) households with consistent food insecurity (n = 61, including households that were food-insecure both in the year before the COVID-19 pandemic and anytime during the first year of the pandemic) and (2) households with new food insecurity (n = 69, including households that were food-secure before the pandemic, but food-insecure at some point during the first year of the pandemic). There were four food-insecure households that did not respond to the questions about food insecurity in the year prior to the COVID-19 pandemic, and therefore, we could not categorize them as consistently or newly food-insecure. We also categorized food-insecure households into two groups based on their food security status in March 2021, as measured with the 30-day instrument: (1) households that recovered by March 2021, meaning they were food-insecure at any point since the start of the pandemic but were food-secure in March 2021 (n = 48) and (2)  that were food-insecure at any point since the start of the COVID-19 pandemic and were still food-insecure in March 2021 (n = 78). There were eight food-insecure households that did not respond to the food insecurity questions in March 2021, and therefore, we could not categorize them as recovered or still food-insecure.
We used t-tests and chi-square tests to determine statistically significant differences between the groups described above. To determine the factors correlated with food insecurity during the first year of the COVID-19 pandemic, we used a random effects logistic regression model with panel data and calculated cluster robust standard errors to account for intra-subject correlation. We included a variable indicating which survey the observation was from to capture time factors that might have influenced food security given the pandemic-related changes that occurred over time during the study period. The coefficients are reported as odds ratios. To examine differences between groups (newly vs. consistently food-insecure; recovered vs. still food-insecure), we used multivariate logistic regression models, with coefficients reported in odds ratios. We used all available data and assumed any missing data were missing at random. A p-value ≤ 0.05 was considered statistically significant for all tests. All statistical analyses were conducted using Stata v17.1 [34].

Demographic Characteristics of Respondents
The demographics of our respondents were comparable with Vermont state demographics on ethnicity and income distributions, but our respondents were more likely to have a college degree and to identify as female (Table 1; for full demographic results, see Table A2).

Food Insecurity, Job Disruptions, and Food Assistance Program Use
Food insecurity rates increased during the pandemic and remained above pre-COVID levels a year after the start of the pandemic ( Table 2). It is important to note that the pre-COVID levels are based on retrospective responses to the questions about food insecurity. Among this cohort, 24.1% of respondents were classified as food-insecure in March 2020 (95% CI 20.0-28.2%) compared with 14.8% in the year before the COVID-19 pandemic (95% CI 11.5-18.2%), representing a 62.8% increase (p < 0.001) (Table A3). Compared with March 2020, food insecurity prevalence decreased in May/June 2020 to 17.4% (95% CI 13.9-21.0%) (p < 0.05) (Table A4) but then fluctuated throughout the first year of the pandemic. By March 2021, the food insecurity prevalence was 18.2% (95% CI 14.6-21.9%), representing a 24.5% decrease as compared with March 2020 (p < 0.05) but was 22.9% higher than pre-COVID levels (p = 0.18) (Tables A5 and A6). Nearly a third (31.6%) of respondents experienced food insecurity at some point during the first year of the pandemic. Among those experiencing food insecurity during the first year of the pandemic, 46.9% also experienced food insecurity at some point in the year prior, while 53.1% were newly food-insecure. Of the respondents who experienced food insecurity at any point since the start of the pandemic, 61.9% were still classified as food-insecure in March 2021.
More than half of respondents (54.2%) reported suffering a job disruption (i.e., job loss, reduction in work hours or income, furlough) during the COVID-19 pandemic, and 18.7% were still reporting a job disruption in March 2021 ( Table 3). The most common type of job disruption was a loss of hours or income (70.7%), followed by job loss (45.6%). The duration of these job disruptions varied, with 35.6% of respondents experiencing a job disruption lasting more than 6 months. More than one in five respondents (23.3%) received unemployment at some point since March 2020. Among those who reported job disruptions during the pandemic, 40.5% received unemployment.
Among our respondents, participation in food assistance programs increased during the first year of the COVID-19 pandemic compared with the year before, except for the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC), though these differences were not statistically significant (Figure 1). However, by March 2021, participation in all programs had declined compared with time points measured earlier in the pandemic, with an 18.2% decrease in Supplemental Nutrition Assistance Program (SNAP/3SquaresVT), 49.3% decrease in Pandemic-EBT (P-EBT), 8.0% decrease in WIC, 19.1% decrease in school meal programs, and 34.7% decrease in the use of food pantries. Only the decline in P-EBT use was statistically significant, dropping from 7.3% at any time during the first year of the COVID-19 pandemic (95% CI 4.9-9.8%) to 3.7% in March 2021 (95% CI 1.9-5.4%) (p < 0.05) (Table A10). Food-insecure households (67.4%; 95% CI 59.3-75.5%) used food assistance programs at over three times the rate of foodsecure households (16.3%; 95% CI 12.2-20.5%) (p < 0.001) (Tables 3 and A11).

Factors Correlated with Food Insecurity
A random effects logistic regression model predicted the factors contributing to higher odds of experiencing food insecurity at any time during the first year of the COVID-19 pandemic ( Table 4). Households that experienced any type of job disruption during the first year of the pandemic had greater odds of experiencing food insecurity (OR 5.03; 95% CI 1.583-16.005; p < 0.01). The odds of experiencing food insecurity were also higher among households with children (OR 5.49; 95% CI 1.782-16.936; p < 0.01), respondents who identified as BIPOC and/or Hispanic (OR 11.77; 95% CI 1.615-85.805; p < 0.05), and women (OR 8.07; 95% CI 1.777-36.647; p < 0.05). Older respondents (OR 0.05; 95% CI 0.012-0.180; p < 0.001) had lower odds of experiencing food insecurity compared with respondents under 63. Having a college degree (OR 0.08; 95% CI 0.025-0.246; p < 0.001) or a household income of ≥USD 50,000 (OR 0.01; 95% CI 0.003-0.038; p < 0.001) were also associated with reduced odds of household food insecurity. Table 4. Random effects logistic regression predicting odds of food insecurity compared with food security during the first year of COVID-19.

Variable
Odds We found no statistically significant differences (p < 0.05) using multivariate logistic regression models between newly and consistently food-insecure respondents (Table 5) and between respondents who were still food-insecure in March 2021 and those who had recovered (Table 6).

Discussion
This longitudinal study in Vermont documented a statistically significant increase in food insecurity compared with the year prior to the COVID-19 pandemic among a cohort of respondents. Furthermore, we show that while the food insecurity prevalence in March 2021 had decreased compared with the early months of the pandemic, rates remained higher than the year before the pandemic. This trend aligns with evidence from most other studies. For example, a longitudinal study conducted nationally found a reduced risk of food insecurity in November 2020 compared with March/April 2020 [23]. Adams et al. (2021) conducted surveys of a cohort of U.S. households with children in May and September 2020, finding that food insecurity increased in May 2020 compared with before the pandemic and then decreased in September 2020, but food insecurity remained above pre-pandemic levels [20]. However, a recent government report found no change in overall food insecurity prevalence in 2020 compared with 2019 [35], counter to most existing research.
We also demonstrated that certain demographic groups were at higher odds of experiencing food insecurity during the first year of the COVID-19 pandemic, including women, younger people (under 63), BIPOC/Hispanic respondents, people without a college degree, lower income households (<USD 50,000), households with children, and people who experienced a job disruption during the pandemic. These findings are consistent with other research on food insecurity during COVID-19 [22,[35][36][37][38] and with research on food insecurity before the pandemic [2][3][4].
The increase in food insecurity documented here could have serious implications both in the short-and longer-term for physical and mental health [8][9][10][11]. Furthermore, health disparities among racial and ethnic minorities and people from lower socio-economic status are well-documented [39][40][41]. Given that our findings show that BIPOC/Hispanic respondents and lower income households were at greater risk for food insecurity during the first year of the COVID-19 pandemic, these groups may be particularly vulnerable to experiencing the associated health impacts in both the short and long term. Further research is needed to better understand how food insecurity during the COVID-19 pandemic has impacted diet quality and health, especially among socio-demographic groups that are at greater risk of food insecurity and adverse health outcomes.
Importantly, our results also show that a significant number of food-insecure households (32.6%) were not using federal food assistance programs or food pantries during the first year of the COVID-19 pandemic. Additionally, we found that only 40.5% of respondents who reported a job disruption during the pandemic received unemployment insurance. These findings demonstrate why we may have seen such a significant increase in food insecurity at the onset of the pandemic and the continued higher prevalence as compared with pre-pandemic periods. Particularly since more than half of our respondents were newly food-insecure, these households may have faced new barriers to receiving available assistance or may have gone without it altogether. Previous research has identified potential barriers to using food assistance programs, including stigma and administrative burden [42,43]. Lack of assistance may explain why food insecurity prevalence remained high, as many households in the United States live paycheck to paycheck and do not have the financial means to adapt to an economic shock [44]. Additional research is needed to continue understanding the barriers to using federal programs and to develop targeted solutions to ensure that households facing economic shocks can find necessary assistance.
Perhaps most surprising, our findings suggest that using federal programs designed to reduce the impact of the economic recession did not necessarily alleviate food insecurity during the first year of the pandemic when controlling for other demographic factors. Use of unemployment and/or use of food assistance programs did not predict recovery from food insecurity among our cohort. However, selection bias is a known issue when studying the impact of SNAP participation on food security [45,46], and it may have impacted our results. Additionally, two limitations of our work are the small sample size and the treatment of food insecurity as a binary outcome. Though the use of food assistance programs and/or unemployment did not necessarily move households out of food insecurity, it may have reduced the severity of the food insecurity they were experiencing and/or allowed households to reallocate money they would normally spend on food towards other essentials such as housing and healthcare. Future research using this longitudinal dataset will examine in more depth how various interventions, including federal and community food assistance programs and unemployment benefits, affected food insecurity outcomes in the first year of the COVID-19 pandemic, while treating food insecurity as a continuum.
Overall, our findings indicate that food insecurity continued to be a significant challenge one year after the start of the pandemic, despite loosened restrictions and new policies that aimed to provide economic relief. Trends from previous economic recessions show that it can take years after economic recovery begins for food insecurity rates to return to prerecession levels [4,47] and that the financial hardships experienced during the pandemic will linger for some households even after they return to work, as they catch up on past due bills and replenish depleted savings. As some of the support programs come to an end (e.g., enhanced unemployment insurance, mortgage relief, eviction moratorium, student loan forbearance, etc.), it is imperative to continue monitoring the impact on food insecurity rates and continue providing assistance as the economy recovers, especially given the recent rise in food prices [48]. The recent update to SNAP benefits will increase the average SNAP benefit and could help alleviate some of the ongoing food insecurity challenges [49], though our findings show there is a gap in participation among food-insecure households.
This study provides important insights into the impact of COVID-19 on food insecurity throughout the first year of the pandemic. The study's strengths include its longitudinal design, early administration, population-based assessment, and survey instrument addressing the multiple dimensions of food security. The limitations are partly rooted in the need to rapidly administer this survey in the early days of the pandemic in order to provide data that could be tracked over time. Though our respondent population matches statewide census statistics closely on many metrics, this was a convenience sample; further research is expanding these results using similar questions with representative samples across states and populations. It is worth noting that our observed overall rate of food insecurity prior to COVID-19 (14.8%) is above the most recently available state figure (9.6%) in 2019 [2]. There are multiple possible reasons for this. First, this is likely to be due, in part, to a higher than average number of female respondents and respondents in households with children; both groups have been documented, in Vermont and elsewhere, to have elevated rates of food insecurity [50]. Second, our measurement instrument for documenting food security, the USDA Six-Item Food Security Module, includes a subjective experience domain that measures concern about household food supplies. According to the local media [51], anxiety about household food supplies preceded the Stay Home/Stay Safe order and may explain the higher than expected level of food insecurity prior to COVID-19. Further, we used an internet-based survey, given the necessity of social distancing during COVID-19 and the need for a rapid response, which may limit the capacity of some people to participate, although 84.1% of adults in Vermont do have internet plans [52]. Another limitation is the wide confidence intervals for some variables (e.g., race/ethnicity, gender) in the logistic regression model predicting odds of food insecurity during the first year of the COVID-19 pandemic, which reduces the reliability of these estimates.

Conclusions
The prevalence of food insecurity increased during the first year of the COVID-19 pandemic and remained higher than pre-pandemic levels a year after the pandemic began.
Our findings indicate that food insecurity continued to be a significant challenge for many people one year after the start of the pandemic, despite loosened restrictions and new policies that aimed to provide economic relief. Furthermore, the odds of experiencing food insecurity during the pandemic varied based on socio-demographic factors. Further research is needed to better understand how food insecurity during the COVID-19 pandemic has impacted diet quality and health, especially among socio-demographic groups that are at greater risk of food insecurity and adverse health outcomes.
This work also highlights a significant number of food-insecure households that were not using federal food assistance programs or food pantries during the first year of the COVID-19 pandemic, as well as fewer than half of respondents who reported a job disruption during the pandemic receiving unemployment insurance. Future research using this longitudinal dataset will further examine the characteristics of individuals and households that did and did not utilize these available services and will assess their relationship to food security and other outcomes.  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The survey instrument used in this study is publicly available and can be accessed at: https://dataverse.harvard.edu/dataverse/foodaccessandcoronavirus (accessed on 2 February 2022). tion, and analysis. We also thank those who assisted in the dissemination or review of the survey including Bytes Co, Community College of Vermont, Farm to Institute New England, Front Porch Forum, Hunger Free Vermont, Representative Welch's staff, Rural Vermont, Salvation Farms, Senator Sander's staff, Senator Leahy's staff, Support and Services at Home, University of Vermont, University of Vermont Extension, Vermont Academy of Nutrition and Dietetics, Vermont Department of Agriculture, Vermont Department of Children and Families, Vermont Department of Health, Vermont Farm to Plate Network, Vermont Foodbank, Vermont Retail and Grocers Association, and the Vermont Sustainable Jobs Fund.

Conflicts of Interest:
The authors declare no conflict of interest.

Appendix A
Nutrients 2022, 14, x FOR PEER REVIEW 12 of 22 Institutional Review Board Statement: The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of University of Vermont (STUDY00000873 approved on 24 March 2020).

Informed Consent Statement:
Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The survey instrument used in this study is publicly available and can be accessed at: https://dataverse.harvard.edu/dataverse/foodaccessandcoronavirus (Accessed on 2 February 2022).

Conflicts of Interest:
The authors declare no conflict of interest.
Appendix A Figure A1. Breakdown of response to each survey. Figure A1. Breakdown of response to each survey. Table A1. Complete list of variables, questions, and scales used in the analysis.

Variable Question Scale
Food-secure       Table A7. Results of a two-sided t-test for differences in reporting job disruptions during the COVID-19 pandemic between food-secure and food-insecure households.  Table A11. Results of a two-sided t-test for differences in use of food assistance programs during the COVID-19 pandemic between food-secure and food-insecure households.