Changes in Food Security, Healthfulness, and Access During the Coronavirus Disease 2019 Pandemic: Results From a National United States Survey

Background Coronavirus disease 2019 (COVID-19) disrupted access to food and adequate nutrition and the types of foods consumed. However, little empiric data exists on the changes in American’s food and nutrition habits 2 y into the pandemic. Objectives To assess current and altered food choices ∼2 y into the COVID-19 pandemic in the months after historic public pandemic relief. Methods A national sample of 1878 United States adults balanced by age, sex, race/ethnicity, and income completed a one-time, online, semi-quantitative, 44-item questionnaire in Fall 2021 asking about the demographics, COVID-19 food choice changes (including free-text), and consumer priorities. This analysis investigates COVID-19 impacts on food security, healthfulness, and access. Results More than 35% of respondents reported improved food security and >45% reported improved food healthfulness compared with prepandemic status. Improvement was reported in more than 30% of Black/African-American and Hispanic/Latinx adults, adults with lower annual income, and female sex, despite over 75% reporting reduced choice of where to eat or buy food. The pandemic offered occasion for many to improve diet, but a similar number expressed that the pandemic destabilized healthy habits. Conclusions Our novel findings suggest that by late 2021, most Americans had improved food security and food choice healthfulness, despite reduced access to food service and retail, although with worsening among a meaningful proportion of Americans as well as heterogeneity in these changes. Vigorous federal, state, city, and community responses to the pandemic may have played a role in improving the food security and food choice healthfulness during the COVID-19 pandemic. Health crises differently impact health behaviors, but when accompanied by vigorous civic and community response, food security, and food healthfulness can be fortified.


Introduction
Coronavirus disease 2019 (COVID-19) disrupted the United States food environment in multiple ways. After March 2020, access to amounts and types of food, and food settings, shifted due to supply disruptions, lock-downs, business closures, lost wages and unemployment, and the avoidance of public contacts and spaces [1][2][3][4][5]. Many news outlets and nonprofit organizations reported an increased demand in food pantries, food waste from farming outputs, and restaurant closures.
However, little empiric data exists on the ultimate impact on the changes in American's food and nutrition habits. Some surveys by June 2020 reported~35% food insecurity [6][7][8][9]10,11]. In response to the pandemic, the federal government launched major financial

Methods Population and design
We recruited a national sample of 1878 adults aged !18 y to complete an online, self-administered survey using the Qualtrics Online Panel Services (QOPS; Provo, UT), a network of national web-based market research panels with experience in marketing, academic, and health-based research. The study design, survey, and sampling goals were designed by the academic investigators. The recruitment was designed to recruit approximately balanced samples by age group (18-34 y, 35-49 y, 50-64 y, and !65 y), race/ethnicity (non-Hispanic White, Hispanic/Latinx, non-Hispanic Black/African-American, and Asian/Other), and sex (male and female), and annual household income levels [<25k/ y, 25k to <50k, 50 to <100k, and >100k United States dollars (USD). Sample size was determined for each demographic group with a confidence level of 0.95, a sample proportion of 0.50, and a margin of error of 0.05 which determined n ¼ 385 respondents in each demographic group. The power for this sample size using a 2-tailed t-test for α ¼ 0.05, and a large standard effect size 0.80 at n ¼ 385 resulted in a power of 1 (critical values: À1.96, 1.96) calculated using Stats Kingdom calculators [29]. Participants were recruited from the existing pools of respondents from previous surveys within the QOPS network, wherein participants received small incentives for completing voluntary surveys, and which prevented multiple participation [30]. Incentives for Qualtrics panels are proprietary and may vary based on the research panel to which participants belong, and the specifics of incentives were unknown to the academic investigators. This investigation was reviewed and approved by the Tufts University Institutional Review Board, and all participants provided informed consent by reading the consent materials and electing to complete the survey. A pilot study was first conducted in September 2021 on 100 participants, to define completion time, response quality, and understand the patterns in missing responses. The full study and all data collection were conducted between September and November 2021 to enroll 1878 participants.

Survey
The 44-question online survey contained items on demographics, including age, sex, income, race/ethnicity, education, and geographic location, as well as 5-items on food and beverage changes during the COVID-19 pandemic. Consumers responded to questions using a 5-point Likert scale from "worsened" to "improved." A separate, free-text item asked participants to share additional information regarding pandemic impacts on food and beverage choices. (Supplemental Table 1). The USDA defines food security as "access by all people at all times to enough food for an active, healthy life" [15,17], and the questions in the COVID-19 section of the survey were developed to address the constructs of food security (access to adequate amounts and types of food), food access, and food choice healthfulness in the context of the pandemic crisis. For this section, healthfulness refers to responses to the question "the healthfulness of the products I choose now."." This term was used to prompt participants to consider food choices specifically in support of improving or maintaining health. Elsewhere in the questionnaire, when not referring to choices, the term "health" or "healthy" was used and specifically defined as "having impact on personal or household health." After the pilot, 3 attention verification items were added to the questionnaire, and additional quality controls were determined, including a minimum completion time, and limits on response straight lining, and other unit nonresponse. These quality controls were then used in the final sample to identify and remove surveys with low-quality responses (Supplemental Figure 1). Other survey questions assessed consumers' ethical priorities for food and beverage businesses and products, unrelated to the pandemic, and will be analyzed in future reports.

Statistical analysis
Descriptive analyses were performed to assess improvement/ worsening of Likert scaled items overall, as well as stratified by age, race/ethnicity, sex, income, education level, and the United States region (West, Midwest Northeast, and South) [31].
Logistic regression was used to evaluate the magnitude and significance of the association between the ability to choose where to eat or buy food with food security or food choice healthfulness, adjusted for age, race/ethnicity, sex, and income.
Three attention verification items were added after the pilot, finalized and prespecified, and then used in the final sample to identify and remove surveys with low-quality responses. (Supplemental Figure 1).
Analyses were conducted using StataSE17 (64-bit) and SAS9.4, with 2-tailed α ¼ 0.05. We did not correct for multiple comparisons, and P values should be considered as a guide.

Qualitative analysis
Free-text responses were analyzed with content and thematic analysis, first using a modified bag-of-words natural language processing (NLP) semantic classification and content analysis [32][33][34], with further NLP conducted with ResearchAI for thematic analysis, sentiment analysis, and outlier analysis [35,36]. Word-based content analysis was performed in 2 parts 1) lemmatization, English translation, and spell-checking; and 2) frequency sorting. After content analysis, a 6-step inductive thematic analysis was performed using the cloud-based software ResearchAI (version February-July 2022). ResearchAI uses 2 unsupervised learning-based-topic-modeling techniques (Latent Dirichlet Allocation) [37,38] and Latent Semantic Indexing [39] in addition to AI-trained keyword matching from publicly available custom topic models via python packages [40] that contribute to thematic detection and sentiment detection (For thematic analysis steps see Supplemental Table 2) [41][42][43].
Compared with before the pandemic, >80% of respondents reported having the same or improved food security, food choice healthfulness, and ability to choose where to eat and buy food in Fall 2021 ( Figure 1). Forty percent of the participants reported improved food security, and 45% reported improved food choice healthfulness. In contrast, only 12% reported worsened food security or healthfulness. Findings for the ability to choose where to eat and buy food were similar, with~40% reporting improvement in each and 12% reporting worsening in each.

Findings in demographics
By age, the frequency of improvements in all 4 metrics were generally larger for respondents younger in age, compared with older in age and were generally the largest in middle-aged adults (35-49 y) (Figure 2). At the same time, worsening of food healthfulness and ability to choose where to buy food were also the largest in younger compared with older age groups.
Among different racial/ethnic subgroups, the frequency of improvements was largest in Hispanic/Latinx adults, followed by Black adults then White adults, and the lowest for Asian/Other adults-who also most frequently reported worsening of food security (15%). Largest overall improvements by race/ethnicity were seen in food choice healthfulness among Hispanic/Latinx adults and Black adults. However, the largest increase in the frequency of reduced food choice healthfulness was also reported by Hispanic/Latinx adults (15%).
Comparing males to females, results were generally similar by sex, with slightly more frequent improvement, and slightly less frequent worsening, across these domains reported by males.
As household income increased, reported improvements increased and worsening decreased. Similarly, the reported frequency of worsening decreased as education level increased. However, the frequency of improvements was also smaller for college degree holders compared with the less educated groups.
By region, Americans in the West and Midwest tended to have the most frequent improvements, and the least frequent worsening, in food security, but these differences were smaller than Association of food security, healthfulness, and where to eat and buy food In unadjusted analyses, an improved ability to choose where to eat was highly associated with improved food security (OR: 3.49, P < 0.0001) and improved food choice healthfulness (OR: 2.66, P < 0.0001). Findings were similar after adjusting for age, race/ethnicity, sex, and income, (OR: 3.35, P < 0.0001 and OR: 2.60, P < 0.0001; respectively).
Similar strong relationships were seen between the choice of where to buy food and food security (OR: 3.76, P < 0.0001) and food healthfulness (OR: 3.75, P < 0.0001), with little change after adjustment for age, race/ethnicity, sex, and income (OR: 3.63, P < 0.0001 and OR: 3.68, P < 0.0001, respectively).

Free-text responses
Free-text was entered by 1423 respondents (76%). Six unintelligible responses were removed (n ¼ 1417). Demographics of free-text respondents did not substantially differ from the overall sample (Table 1). After lemmatization, 8968 unique words were identified and coded into 1381 unique codes ( Table 2, Supplemental Figure 2). Overall, the sentiment of the free-text responses was 22% positive, 38% neutral, and 37% negative. Two hundred ninety five respondents (n ¼ 20%) wrote in that they experienced little or no changes in their food and beverage eating or purchasing habits due to the pandemic.
As a result of the thematic analysis, 9 overall themes emerged from our analysis: 1) Behavioral characterizations; 2) food settings; 3) intensity changes; 4) food descriptions; 5) economics; 6) time; 7) nutrition; 8) relationships; and 9) health ( Table 2). Details of the text responses for each of these themes are described below.

Behavior
Most responses described behavior (59.8%). Of these, 20% had negative sentiment, 53% were neutral, and 27% were positive. Reported behaviors were often fear-motivated and conveyed urgency to change, seek protection, exercise, and deliberate choices and that safety and caution were needed (Table 3, Quotes 4, 6, and 17).

Setting
Thirty three percent of responses referenced food settings, with 17% having a negative sentiment; 58%, neutral; and 25%, positive. Responses often compared different settings, such as in-store grocery shopping compared with online/delivery (Table 3, Quotes 9 and 10), cooking at home compared with fast-food (Table 3, Quote 8), and eating indoors compared with outdoors (Table 3, Quote 10). Less in-store shopping was offered as a reason fresh food purchases declined, for example owing to respondent preference to select fresh food in person (Table 3, Quote 9). Additionally, respondents offered subjective experiences of reduced safety in shopping due to COVID-19, such as discomfort around people not wearing masks and fear of infection while shopping or handling foods. A small, but passionate, number of respondents criticized COVID-19 safety procedures, stating that they boycotted the food sector business with safety requirements. These respondents also typically indicated that they had no other changes in their food choices, and no change in food healthfulness during the pandemic.

Intensity
Changes in the intensity of their experience were described in 22.2% of the responses, such as "much more able" to perform a task, or "much worse than" a previous experience. Twenty six percent had negative, 30% had neutral, and 45% had positive sentiments. The most common reports described access to food, food locations, and other resources being more plentiful now than in the earlier stages of the pandemic. In contrast, intensity was also often used to describe increases in fear regarding the safety or longevity of the pandemic despite increasing access and returning the availability of foods of preference.   Table 3.

Food
Furthermore, 32% of responses referenced personal choices around food selection and 18% had negative, 45% had neutral, and 37% had positive sentiments. Topics included aiming to increase fruit and vegetable intake for health, including often for perceptions of immune boosting; stocking up on nonperishable foods, especially canned food, and seeking foods for comfort to manage stress (

Economic
Economic factors were referenced by 30.7% of the responses, with 23% having negative, 48% had neutral, and 29% had positive sentiments. Topics included unemployment, depletion of savings, grateful reliance on financial and food assistance programs (Table 3, Quotes 2 and 3), and the inability to afford food preferences. Respondents often reported deliberating the personal value of spending money on healthy food, including some respondents who increased, and some who decreased, their budgets for buying healthy foods (Table 3, Quote 17). Food shortages and rising prices were both commonly reported as altering shopping habits, from seeking food in places other than grocery stores, especially from fast-food and delivery (Table 3, Quote 1), to reducing the consumption of meat and processed food for both affordability and health.

Time
Time was referenced by 23.4% of responses, and 27% of the responses had negative, 38% had neutral, and 34% had positive sentiments. Present-tense and time-of-day characterizations were the most common, followed by references to earlier pandemic periods. Additionally, few mentioned the future-those that did mostly describe the hypothetical end of COVID-19.

Nutrition
Only 11.5% of responses referenced nutrition, and 22% had negative, 34% had neutral, and 45% had positive sentiments. Responses around nutrition were generally categorized into 2 opposing camps: 1) those able to cook more at home, learn about nutrition, increase fruits and vegetables, and decrease sugar and alcohol intake in response to the pandemic (Table 3, Quote 5); and 2) those with heightened stress and/or economic difficulties, such as increased isolation, decreased food access, and diminished community resources, departed loved ones, and being separated from families, resulting in increased fast-food and junk food consumption (Table 3, Quotes 1, 8, and 15). Responses about eating at home more were similarly divided, with some respondents indicating that it promoted unhealthier, and others healthier, food choices. Respondents who endorsed the impacts of the pandemic on increasing unhealthy choices often referenced fast-food, processed food, and sugar-sweetened beverages.

Relationships
Eleven percent of responses referenced the self, others, and emotions, and 23% had negative, 27% had neutral, and 51% had positive sentiments. Many used internally-focused descriptions, including feeling isolated, as well as described decisions and experiences from an individual perspective (Table 3, Quote 9, 18). However, some respondents reported as "we" and described making decisions on behalf of their family (Table 3, Quotes 7 and  16). Additionally, some respondents characterized pro-social concepts, such as "we are in this together."."

Health
Only 10% of respondents discussed health, and 28% of responses had negative, 37% had neutral, and 35% had positive sentiments. Fear of getting sick from COVID-19 was frequently reported, including from community transmission and contact with food packages (Table 3, Quotes 4, 10, 11, 12, and 13). Some reported beliefs that ingesting specific foods could result in COVID-19 infection; others reported changes in social interactions (Table 3, Quote 18) and hygiene to reduce risk. Mental health concerns were expressed by many, including around wellness, stress, and diagnoses such as depression. Reported health improvements were typically in comparison to health early in the pandemic, but a few people reported initial health behavior improvements that backslide into worse habits and worse health.

Summary of findings
In this large, diverse national sample, we identified several novel trends around changes in eating habits during the COVID-19 pandemic. These included notable improvements in the  frequency of food security, food choice healthfulness, and food access compared with prepandemic status. Improvements in the ability to choose where to eat and buy food were strongly associated with both food security and food choice healthfulness. Smaller numbers of respondents described worsening in these domains. Interestingly, both improvements and worsening of food healthfulness and choice about where to buy food were more frequent among younger adults (aged 18-34 y and 35-49 y), compared with older adults (aged 50-64 y and !65 y), suggesting less overall stability for younger respondents. Among different racial/ethnic subgroups, the largest frequency of improvements in the healthfulness of foods occurred among Hispanic/Latinx adults and Black adults; however, Hispanic/Latinx adults also most frequently reported worsening in healthfulness choice. Based on extensive free-text responses, the pandemic afforded many the opportunity to make improvements in health behaviors, but many others expressed, in the free-text responses, that healthy habits from before the pandemic were destabilized. The qualitative responses were illuminating as they showed that many participants became more cognizant of the healthfulness of their food choices, across different venues, including food service and others not typically associated with healthier choices. Text responses also highlighted common economic concerns, ranging from personal finances to food supply and access, as well as safety concerns around the additional effort needed to ensure safety while accessing food. Mentions of fear, constraint, changes in food settings, and limited socialization together characterize both the disruption and ongoing calibration processes experienced by Americans during the shifting landscape of the COVID-19 pandemic.
Our results differ from prior surveys conducted during early time points in the pandemic, when food insecurity was reported by survey participants at~35%. However, our findings are supported by later USDA reports, both by the conclusion of the first year of the pandemic, and during the timepoint in which our survey was launched, in September of the second year of the pandemic. The potential reasons for these differences may n ¼ 1878 participants were recruited to complete a 44-item survey about ethical priorities for food and beverage businesses and products including a section asking about changes in food and beverage habits during the pandemic compared to prepandemic conditions. In this section, a free-text item asked participants to describe changes in food and beverage consumption and purchase, n ¼ 1423 responses were received (Table 1). Quotes listed above represent examples of multi-theme constructs as determined by Natural Language Processing, selected to represent multiple participants' reports with several overlapping elements. Quotes were also selected to report approximately balanced perspectives by demographic group. COVID-19, coronavirus disease 2019.
include the robust financial assistance in the later stages of the pandemic from the United States pandemic response, which includes, for example, the FFCRA expansion of the SNAP; programs such as the ESAP that increased access for older adult, and the comprehensive ARP, which passed~6 mo before our survey. These landmark assistance programs included over $12 billion in nutrition programming expanding food and nutrition access, in response to the financial burdens caused by the pandemic. [12][13][14]44]. Additional differences in food choice could also be attributed to individual adaptation to pandemic conditions, and a return of willingness to shop in grocery stores in later stages of the pandemic, as well as the expanded access to online food ordering.
Although our results also document heterogeneity in response by demographics, on average, the plurality of respondents in this large national sample reported stability or improvement in these factors. Our results support the efficacy of federal, state, and local responses to the pandemic, many of which were implemented in 2021, at abating the emergent food and nutrition security challenges of the early pandemic. Our findings support the inference that these programs may have not only successfully arrested but may have also reversed the serious impacts of financial, social, and access hardships resulting from the pandemic for a large number of Americans.
The present analyses detailed text descriptions provide novel insights into personal sentiments around food, nutrition, and health among United States adults during the pandemic. The responses highlight lived experiences of stress and adaptation, relating to both economic concerns and elaborate behavioral steps needed to procure food and ensure personal safety. Overall, the responses conveyed that much additional effort was needed to function in daily life during the pandemic. At the same time, characterizations of "Nutrition" and "Relationships" were often strongly positive in sentiment, indicating that perhaps respondents saw these areas as means to improving their quality of life during a time of stress and uncertainty.

Comparison with previous reports
Our findings contrast with surveys from earlier periods in the pandemic that reported food insecurity surpassing a prevalence of 35% [6][7][8][9]10,11], and frequent negative changes to health behaviors, such as increased snacking and increased consumption of sugar-sweetened beverages [18,22,25]. However, our findings support later USDA national surveys on food security, performed in late 2020 through 2021, which found that food security had slightly improved during the pandemic period, compared with prepandemic (2019). In contrast with the USDA reports, we found that the food security was most frequently improved in the racial/ethnic groups typically disproportionately impacted by food insecurity, such as Black/African-American adults and Hispanic/Latinx adults. However, similar to the USDA report, we also found that adults with <$50,000 in annual income more frequently reported worsening of food security and less frequently reported improved food security than high income adults. Our analysis indicates that overall food choice healthfulness increased to a greater extent, and independently from, changes in food security. These results suggest that the pandemic may have motivated the United States adults to prioritize healthful foods in late 2021 independently of their food security status and support the differentiation of assessing food security compared with nutrition security [45,46]. Similar to USDA reports later in the pandemic, we observed that many Americans were resilient to the crisis, likely owing to robust federal and state-level assistance. However, our novel findings extend that observation to the reported changes in the healthfulness of food choices, and in the settings in which people eat. Although food security is often associated with improved healthfulness, participants in our study seemed to indicate increases in healthfulness more frequently than associated reports of access.
In general, high income and old age were most associated with resilience to changes in these domains as a result of the pandemic, perhaps the former owing to surplus financial resources and associated social support and the latter from more experience in navigating crises. In contrast, younger age groups and households with lower annual incomes reported more changes-both positive and negative-suggesting that these groups are both more vulnerable to crisis and more responsive to corresponding interventions. Notably, although lower age groups reported both more frequent improvements and more frequent worsening, lower income groups reported the most frequent worsening of any subgroup. Additionally, access to food increased for some, but worsened for others. The response to COVID-19 was uneven across United States regions and local neighborhoods, which may have contributed to these differences [47,48]. In addition, different people had different levels of comfort with going to grocery stores, online ordering, etc., which could have also contributed to disparities in efficacy. Not all people take advantage of the eligibility for food assistance programs, which could further compound these disparities.
Our findings build upon and extend previous studies that described the challenges to food security, access, and health behaviors in the context of the early pandemic and called for policy and programs to alleviate financial barriers to healthy eating [6,7,9,10,12,19,[49][50][51][52]. Although these early reports found trends toward unhealthy eating, our findings suggest that later in the pandemic, these trends were reversed.

Limitations
Although our food security questions were adapted from food security instruments and included similar items, the survey did not use a validated food security instrument, limiting direct comparison to federal measures of food security. This survey did not follow people longitudinally through multiple stages of the pandemic, nor did it seek to assess dietary intakes, or verify participant comprehension of complex constructs such as "food security" and, thus, presents a cross-section of selfreported changes in habits, experiences, and perspectives. Thus, although participants are asked to compare their precompared with postpandemic experiences, they are not necessarily linear trajectories, which may be more complex. Although we recruited a large, diverse national sample, like all surveys, the results are unlikely to be fully nationally representative, and likely undercount individuals with severe illness or social stressors. In addition, this survey was conducted online, using pre-existing pools of survey users from the Qualtrics panel service. This sample likely overrepresents participants who have more reliable access to and familiarity with online surveys.
We incorporated several important participants characteristics, such as geographic region and educational attainment, but these participant characteristics were not specifically used for balanced stratification of recruitment as was done for age, sex, race/ethnicity, and income. Although we recruited balanced groups demographically, the subgroup analyses were not adjusted for multiple comparisons. These analyses resulted in levels of statistical significance P <0.0001, nonetheless these analyses are associative and descriptive by design.
In conclusion, our novel findings suggest that by late 2021, most Americans had improved food security and food choice healthfulness, despite reduced access to preferred food settings, although with worsening among a meaningful proportion of Americans as well as heterogeneity in these changes. Results suggest a need to collect more robust, real-time data on food during future public health emergencies, including on both food security and nutrition security and on access to healthful foods to prevent or treat disease. Assessing not only access to sufficient calories but also the healthfulness of foods during moments of crisis can help future policy and disaster-relief programs to address food and nutrition security in the future. Our findings suggest that a public health crisis can increase the salience of health for many, increasing motivation toward better choices and well-being. Future responses to public health emergencies might better take advantage of these moments to promote not only greater access to food but also food healthfulness.

Declaration of interests
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Dariush Mozaffarian reports financial support was provided by Vail Innovative Global Research.