**Household Food Insecurity and the Association with Cumulative Biological Risk among Lower-Income Adults: Results from the National Health and Nutrition Examination Surveys 2007–2010**

#### **Cindy W. Leung \* and Megan S. Zhou**

Department of Nutritional Sciences, School of Public Health, University of Michigan, Ann Arbor, MI 48109, USA; zhoumeg@umich.edu

**\*** Correspondence: cindyleung@post.harvard.edu; Tel.: +1-734-647-9087

Received: 10 April 2020; Accepted: 20 May 2020; Published: 23 May 2020

**Abstract:** Household food insecurity has been associated with adverse health outcomes; however, the mechanisms underlying these associations are not well-defined. Using data from 5005 adults from the 2007–2010 National Health and Nutrition Examination Surveys (NHANES), we examined associations between household food insecurity and cumulative biological risk, a measure of the body's physiological response to chronic stress. Household food security was assessed using the 18-item Household Food Security Survey Module. Marginal food security refers to 1–2 positive responses, and food insecurity refers to ≥3 positive responses. The cumulative biological risk scores were calculated based on the distributions of ten biomarkers from the cardiovascular, metabolic, and immune systems. Elevated biological risk was defined as a risk score of ≥3. Multivariable regression models were used to examine associations between food security and cumulative biological risk scores, adjusting for sociodemographic characteristics. After multivariable adjustment, food insecurity was associated with a 0.14-unit higher cumulative biological risk score (95% CI 0.05–0.22, *p*-trend = 0.003) and higher odds of elevated biological risk (OR 1.20, 95% CI 1.05–1.37, *p*-trend = 0.003). These associations differed by gender. Among women, food insecurity was associated with 0.30-unit higher cumulative biological risk score (95% CI 0.14–0.45, *p*-trend = 0.0004) and higher odds of elevated biological risk (OR 1.61, 95% CI 1.29–2.00, *p*-trend < 0.0001). These associations were not observed in men. Women experiencing food insecurity demonstrated elevated levels of biological risk. These findings support the hypothesis that food insecurity may be associated with women's chronic health outcomes through the pathway of chronic stress. Further research is needed to understand why these associations were not observed in men.

**Keywords:** food insecurity; allostatic load; biological risk; chronic stress; National Health and Nutrition Examination Surveys

#### **1. Introduction**

Food insecurity, defined as inadequate access and availability of food due to a lack of monetary resources, has persisted in the United States since its routine measurement in national population surveys in the 1990s [1]. In 2018, it was estimated that 14.3 million households or 11.1% of U.S. households experienced food insecurity during the year [1]. Household food insecurity has been associated with numerous physical and mental health outcomes among low-income adults, including higher levels of obesity [2–4], hypertension [5], diabetes [6,7] and poorer diabetes management [8,9], metabolic syndrome [10], lower cognitive function [11,12], depression [13,14], and poorer overall health [15,16]. The mechanisms underlying these associations have not been fully elucidated; however,

a number of studies have demonstrated associations between household food insecurity and modifiable health behaviors, such as poorer diet quality [17–19], lower levels of physical activity [20], higher rates of smoking [21–23], and poorer sleep outcomes [24,25]. Several researchers have also alluded to the role of chronic stress in explaining the observed associations [14,26–30], though none of these studies have been able to test the association between food insecurity and chronic stress directly.

Chronic stress refers to the repeated activation of major body systems in response to external stimuli perceived to be threatening [31]. In response to a stressor, the hypothalamic–pituitary–adrenal (HPA) axis and the sympathetic–adrenal–medullary (SAM) systems are activated, triggering a release of hormones and cytokines that act on multiple organ systems [31,32]. The concept of allostatic load has been used to describe how repeated exposure and prolonged response to stress can dysregulate these organ systems, resulting in "wear and tear" on the body over time [32]. Operationally, there is no single, gold-standard approach for measuring allostatic load, though allostatic load scores often include a combination of biomarkers from neuroendocrine, immune, metabolic, and cardiovascular systems—all known to be affected by the secretion of stress hormones and inflammatory markers [33]. Thus, allostatic load is theorized to represent the body's cumulative physiological response to chronic stress and to account for individual variability in the appraisal of different stress exposures, which may be more relevant to predicting subsequent health outcomes than the perception of stress [34].

Prior research has explored the role of the allostatic load as a framework through which to understand health disparities. Greater socioeconomic adversity [35–38] and minority race/ethnicity [39–41] have both been associated with elevated allostatic load in numerous studies. Household food insecurity is both a form of socioeconomic adversity and an issue with relatively high prevalence in low-income and minority racial/ethnic households [1]. However, its relation to allostatic load has only been explored in one study [42]. In an analysis of 733 Puerto Rican adults residing in Boston, food insecurity was associated with greater dysregulation of the neuroendocrine and inflammatory systems, but not total allostatic load [42]. In order to understand the pathways connecting food insecurity and adverse health outcomes, more research is needed to examine this association in a larger and more representative sample of adults.

In the present study, neuroendocrine markers are important components of allostatic load measurement but were unavailable in the National Health and Nutrition Examination Surveys (NHANES). Thus, we created a score to represent a cumulative biological risk, guided by the allostatic load framework. The objective of the present study was to examine the association between household food insecurity and cumulative biological risk in a national sample of adults. We hypothesized that adults who experienced more severe household food insecurity would demonstrate greater cumulative biological risk.

#### **2. Materials and Methods**

#### *2.1. Study Population*

Administered by the National Center for Health Statistics, NHANES is an ongoing, multistage survey designed to be representative of the civilian, noninstitutionalized U.S. population. NHANES collects information on demographic indicators and health outcomes through interviews, in-person examinations, and laboratory testing. Data from 2007 to 2008 and from 2009 to 2010 were combined in the present study, representing the most recent years in which all markers of cumulative biological risk were routinely collected (e.g., C-reactive protein was not assessed in NHANES 2011–2012 or 2013–2014). The analytic population was further restricted to adults between the ages of 20–65 and with household incomes ≤300% of the federal poverty level (FPL). An income threshold of 300% FPL was chosen because household food insecurity is relatively uncommon among households with incomes >300% FPL, and inclusion of a lower-income sample reduces the potential for confounding by socioeconomic status, as has been done in prior studies [6,17,43]. Pregnant women were also excluded, as body mass index (BMI) was one of the markers of biological risk. The analytic sample included 5005 adults.

#### *2.2. Food Security Status*

The primary exposure of interest was household food security, measured using the 18-item U.S. Household Food Security Survey Module [44]. Questions were asked in three stages and attribute related experiences or behaviors to insufficient resources to buy food over the past 12 months. A score was created by summing the affirmative responses of the 18 questions, with higher scores indicating more severe food insecurity. Food security was defined as 0 affirmative responses, meaning the household had no indicators of insufficient food access. Marginal food security was defined as 1–2 affirmative responses and refers to mild indicators of insufficient food access such as anxiety over the food supply. Food insecurity was defined as three or more affirmative responses and refers to multiple indicators of insufficient food access, including reducing the quality and the variety of the amount of food consumed by at least one member of the household. Food insecurity refers to the combined categories of low food security (i.e., 3–5 affirmative responses in households without children or 3–7 affirmative responses in households with children) and very low food security (i.e., 6–10 affirmative responses in households without children or 8–18 affirmative responses in households with children). Food insecurity categorizations and definitions are in accordance with the U.S. Department of Agriculture [44].

#### *2.3. Cumulative Biological Risk*

To assess cumulative biological risk, we included the following ten biomarkers: (1) systolic blood pressure, (2) diastolic blood pressure, (3) body mass index (BMI), (4) glycohemoglobin, (5) total cholesterol, (6) high-density lipoprotein (HDL) cholesterol, (7) total/HDL cholesterol ratio, (8) C-reactive protein, (9) albumin, and (10) estimated glomerular filtration rate (eGFR). The selection of biomarkers for the present study was guided by the availability of data within NHANES and on the basis of previous research [45–48]. Systolic blood pressure, diastolic blood pressure, total cholesterol, HDL cholesterol, and pulse represented the cardiovascular system. BMI, glycohemoglobin, albumin, and eGFR represented the metabolic system. C-reactive protein represented the immune system.

Each biomarker was categorized using clinically-relevant guidelines for low-risk, moderate-risk, and high-risk categories [47]. The cut-points used were: (1) systolic blood pressure: <120 mmHg, 120–<150 mmHg, and ≥150 mmHg; (2) diastolic blood pressure: <80 mmHg, 80–<90 mmHg, and <sup>≥</sup>90 mmHg; (3) BMI: <25 kg/m2, 25–<30 kg/m2, <sup>≥</sup>30 kg/m2; (4) glycohemoglobin: <5.7%, 5.7–<6.5%, and ≥6.5%; (5) total cholesterol: <200 mg/dL, 200–<240 mg/dL, ≥240 mg/dL; (6) HDL cholesterol: ≥60 mg/dL, 40–<60 mg/dL, and <40 mg/dL; (7) total/HDL ratio: <5; 5–<6, ≥6; (8) C-reactive protein: <1 mg/L, 1–<3 mg/L, and ≥3 mg/L; (9) albumin: ≥3.8, 3.0–<3.8, and <3.0; and (10) eGFR: <sup>≥</sup>60 mL/min/1.73 m2, 30–<60 mL/min/1.73 m2, and <30 mL/min/1.73 m2. Each biomarker was scored as zero points for low risk, 0.5 points for moderate risk, and one point for high risk. Adults who reported taking medication for high blood pressure, high cholesterol, or diabetes were also assigned to the high-risk groups for systolic and diastolic blood pressure, total cholesterol, and glycohemoglobin, respectively. A cumulative biological risk score was then created as the sum of the risk scores across the ten components, ranging from 0 (lowest) to 10 (highest). Similar to prior studies [45,46], elevated biological risk was defined as a score ≥3. Both continuous and dichotomous cumulative biological risk scores were examined as primary outcomes.

#### *2.4. Study Covariates*

Sociodemographic covariates were selected as variables hypothesized to be joint predictors of the association between food insecurity and cumulative biological risk, guided by the prior literature. These included age (continuous), gender, race/ethnicity (Non-Hispanic White, Non-Hispanic Black, Hispanic, Other), educational attainment (<12 years, high school graduate or equivalent, some college, or college graduate), household income relative to the federal poverty line (continuous), and marital status (married or living with a partner, never married, or separated, widowed, or divorced).

#### *2.5. Statistical Analysis*

Complex sampling weights for the mobile examination center were recalculated and used to account for different sampling probabilities and participation rates across the four-year period. Sampling weights were used in all subsequent analyses. Differences in sociodemographic characteristics by household food security were compared using univariate regression for continuous variables and χ<sup>2</sup> tests for categorical variables. Next, the distributions in individual biomarkers by household food security were compared using univariate regression and χ<sup>2</sup> tests. Multivariable linear and logistic regression models were used to examine associations between food security and cumulative biological risk. Differences in these associations by gender were evaluated by testing the significance of the interaction terms between household food security and gender on the outcomes. Models were first adjusted for age and gender, and second for all other sociodemographic covariates (race/ethnicity, educational attainment, household income, and marital status). In all models, age and household income were modeled as linear and quadratic terms to allow for a curvilinear relationship with cumulative biological risk. In a sensitivity analysis, we further examined the associations between the household food security and cumulative biological risk using a four-category household food security variable. Statistical tests were two-sided, and significance was considered at *p* < 0.05. Statistical analyses were performed with SAS 9.3 (SAS Institute Inc., Cary, NC, USA).

#### **3. Results**

In the analytic population of 5005 adults, 59.3% were food-secure, 14.4% were marginally food secure, and 26.2% were food-insecure. Table 1 shows the differences in the sociodemographic characteristics by household food security. Compared to food-secure adults, marginally food-secure and food-insecure adults were more likely to be of younger age, of minority race/ethnicity background, have lower educational attainment, have lower household income, and were more likely to be never married, or separated, divorced, or widowed.

Bivariate comparisons of biomarkers comprising cumulative biological risk and household food security are shown in Table 2. Compared to food-secure adults, marginally food-secure and food-insecure adults were more likely to have higher mean glycohemoglobin (*p* = 0.01), C-reactive protein (*p* = 0.005), and albumin (*p* = 0.0002). Significant differences were also evident for some risk categories. Marginally food-secure adults were more likely to be at moderate- or high-risk for systolic blood pressure (*p* = 0.04), and both marginally food-secure and food-insecure adults were more likely to be at moderate- or high-risk for HDL cholesterol (*p* = 0.05). There were no other significant bivariate associations between individual biomarkers and food security status.

Table 3 shows the associations between household food security and cumulative biological risk. Food insecurity was associated with a 0.22-point greater cumulative biological risk score (95% CI 0.11–0.32, *p*-trend = 0.0002), which remained significant after multivariate adjustment (β = 0.14, 95% CI 0.05–0.22, *p*-trend = 0.003). Although the associations between marginal food security and cumulative biological risk and between food insecurity and cumulative biological risk both appeared stronger in women than in men, the interaction was not statistically significant (*p*-interaction = 0.09).

When examining elevated biological risk (score ≥3), food insecurity was associated with higher odds of elevated biological risk (OR 1.20, 95% CI 1.05–1.37, *p*-trend = 0.003), after adjusting for sociodemographic characteristics. This association was significantly modified by gender (*p*-interaction = 0.03). Among women, food insecurity was associated with elevated biological risk (OR 1.61, 95% CI 1.29–2.00, *p*-trend < 0.0001). Among men, no association was observed between food insecurity and elevated biological risk (OR 0.93, 95% CI 0.72–1.20, *p*-trend = 0.70). A sensitivity analysis using a four-category household food security variable showed similar results with cumulative biological risk scores and elevated biological risk (Supplemental Table S1).


**Table 1.** Household food security and sociodemographic characteristics of 5005 adults (20–65 year) with household incomes ≤300% of the federal poverty level:National Health and NutritionExaminationSurveys2007–2010.

**Table 2.** Household food security and biomarkers of cumulative biological risk: National Health and Nutrition Examination Surveys 2007–2010.



**Table2.***Cont*.


**Table2.***Cont*.


#### **4. Discussion**

In this national sample of lower-income adults, food insecurity was significantly associated with elevated biological risk in women. Significant associations were also observed between marginal food security and food insecurity and higher cumulative biological risk scores, supporting the notion that marginal food security is similar to food insecurity with respect to adverse health risks [30,49]. These findings suggest that food insecurity, and potentially even experiences of marginal food security are associated with the dysregulation of the major body systems in women [31,50].

To date, only one other study has examined the association between food insecurity and allostatic load. Among 733 Puerto Rican adults in the Boston Puerto Rican Health Study (BPHC), McClain and colleagues found that food insecurity was associated with greater dysregulation of the neuroendocrine and immune systems over a five-year follow-up period, but not with the total allostatic load [42]. The differences in these findings may be due to BPHC study participants being older and already having higher burden of chronic disease at baseline than the general NHANES population, and the lack of neuroendocrine markers within NHANES to investigate this specific association.

The results of the present study also highlight differences in the associations between food insecurity and cumulative biological risk between men and women. Although it is unclear why the associations were significant among women and not significant among men, prior research on the connections between food insecurity and stress may provide insight into these differences. The development of the early Radimer/Cornell hunger scale drew primarily from women's experiences of household food insecurity and demonstrated that one of the earliest indicators was worrying about the household food resources being depleted [51]. This item, now in the U.S. Household Food Security Survey Module, is consistently endorsed by the vast majority of food-insecure households [1]. Qualitative research studies exploring the lived experiences of women have also expanded our understanding of the stressful experience of food insecurity. In a study of predominantly mothers from Quebec City, participants described the "psychological suffering" of food insecurity, as feelings of powerlessness, guilt, shame, feelings of inequity, and fears of being judged or labeled [29]. Mothers from another study in Philadelphia and Minneapolis discussed the continual trade-off between food and other basic necessities due to limited financial resources, characterizing their psychological response as sadness, frustration, resignation, worry and fear, and shame [52]. In quantitative studies of pregnant women, food insecurity has been related to higher reported levels of perceived stress, disordered eating behaviors, trait anxiety, and depressive symptoms, and lower self-esteem and mastery [53,54]. Furthermore, a systematic review on social position, stress, and obesity-related risk factors concluded that women not only perceive stress more strongly but also exhibit a greater physiological response to social stressors when compared to men [55]. At the present time, more qualitative and quantitative research is needed to better understand how men's psychological and physiological responses to food insecurity may differ from the responses of women and how those differences may translate into subsequent implications for health.

Although the present study did not include dietary intake as a mediator or outcome, the inverse association between food insecurity and diet quality has been well-established in prior studies. In a study of food pantry clients in Connecticut, food insecurity was associated with a lower likelihood of consuming fruits, vegetables, and fiber [56]. Another study in Texas found that urban and rural adults experiencing food-related hardship were more likely to consume sugar-sweetened beverages [57]. Within NHANES, a previous analysis showed inverse associations between household food insecurity and adult's dietary quality, as indicated by the lower scores on the Healthy Eating Index and the Alternate Healthy Eating Index [17]. The results from the present study on household food insecurity and some biomarkers of cumulative biological risk, e.g., glycohemoglobin, HDL cholesterol, and systolic blood pressure, may be driven, in part, by differences in dietary quality rather than chronic stress. However, the chronic stress and dietary pathways stemming from food insecurity are not mutually exclusive, and several studies have demonstrated how chronic stress could also alter eating behaviors to negatively impact dietary quality. Chronic stress activates the hypothalamus–pituitary–adrenal

(HPA) axis, which triggers a cascade of hormones, leading to the release of cortisol [32,58]. Cortisol stimulates food intake, particularly foods high in fat and sugar, and can lead to excessive caloric intake and cardiometabolic disease over time [59,60]. When food is available, food-insecure individuals may overeat not simply in response to the physical sensation of hunger, but as physiological and behavioral coping strategies to chronic stress. In a qualitative study by Tester and colleagues, food-insecure parents discussed disordered eating habits observed in their children, including binge eating, hiding food, and night-time eating behaviors not discussed by food-secure parents [61]. To date, there have been few studies on the associations between household food insecurity and disordered eating among adults, with most research limited to children [62,63] and pregnant women [54,63]. Further research, particularly using longitudinal study designs and robust measurement of food insecurity, chronic stress, diet quality and eating behaviors, and multiple systems comprising allostatic load, are needed to better understand the relationship between food insecurity and allostatic load and elucidate the pathways of chronic stress and dietary intake in the general population.

The results of this study have potential clinical and policy implications. The finding that even marginal food security was associated with higher cumulative biological risk scores among women is consistent with research showing that children in households with marginal food security exhibit poorer health and developmental outcomes than children in food-secure households [49]. Screening of food insecurity in health care settings using the validated Hunger Vital Sign measure can help identify adults with marginal food security for the referral to community food programs and social services [64]. Economic and nutrition programs and policies aimed at improving food security should also ensure that they are reaching populations with marginal food security to ameliorate any adverse health outcomes related to anxiety over household food resources and milder indicators of food insecurity that precede behavioral adaptations.

The primary limitation of this study is the cross-sectional design, which limits the ability to make causal inferences about the findings. Although we restricted the analytic sample to adults with household incomes ≤300% of the federal poverty level and further adjusted for household income in statistical analyses, we cannot rule out the potential for residual confounding by income or other proxies of socioeconomic status, which are known to have salient relationships with health behaviors, physical health, and mental health [65]. Relatedly, we cannot exclude the potential for reverse causation, where elevated biological risk might lead to increased health care costs, subsequently influencing household food security. Another limitation is the assessment of household food security in NHANES, which occurs over a 12-month period. By aggregating over the past year, our understanding is limited as to whether experiences of food insecurity were episodic or chronic. Prior research suggests food-insecure individuals may exhibit disordered eating behaviors corresponding to a monthly cycle of when food is plentiful or scarce [61,66]. How food insecurity-induced disordered eating is associated with cumulative biological risk is unknown given the long period over which food insecurity indicators are measured. Further, no information is available on the history of food insecurity in the family, as cumulative experiences of food insecurity experienced before the study window may have also contributed to biological risk.

Another important limitation of the NHANES dataset is that it lacks neuroendocrine markers to better measure allostatic load. Further research is needed to understand how food insecurity is associated with neuroendocrine dysregulation and the primary system of allostatic load to better understand its relationship with chronic stress.

#### **5. Conclusions**

Understanding the mechanisms underlying food insecurity and adverse health outcomes is critical to designing effective interventions to reduce socioeconomic and health disparities. The findings of the present study show higher cumulative biological risk scores among marginally food-secure and food-insecure women, providing additional evidence to suggest even mild experiences of food insecurity may affect physical and mental health outcomes through chronic stress. Further research is needed to understand why these associations were not observed in men and to better elucidate the role of food insecurity in promoting chronic stress independent of other forms of socioeconomic adversity in women.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6643/12/5/1517/s1, Table S1: Associations between household food security (using the four-category variable) and cumulative biological risk: National Health and Nutrition Examination Surveys 2007–2010.

**Author Contributions:** Conceptualization, C.W.L. and M.S.Z.; Formal analysis, C.W.L.; Writing—original draft, C.W.L. and M.S.Z.; Writing—reviewing & editing, C.W.L. and M.S.Z. All authors have read and agreed to the published version of the manuscript.

**Funding:** C.W.L. was supported by grant 4R00 HD084758 from NIH.

**Conflicts of Interest:** The authors have no conflict of interest to disclose.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Food Insecurity and COVID-19: Disparities in Early E**ff**ects for US Adults**

**Julia A. Wolfson 1,2,\* and Cindy W. Leung <sup>2</sup>**


Received: 15 May 2020; Accepted: 29 May 2020; Published: 2 June 2020

**Abstract:** The COVID-19 pandemic has dramatically increased food insecurity in the United States (US). The objective of this study was to understand the early effects of the COVID-19 pandemic among low-income adults in the US as social distancing measures began to be implemented. On 19–24 March 2020 we fielded a national, web-based survey (53% response rate) among adults with <250% of the federal poverty line in the US (*N* = 1478). Measures included household food security status and COVID-19-related basic needs challenges. Overall, 36% of low-income adults in the US were food secure, 20% had marginal food security, and 44% were food insecure. Less than one in five (18.8%) of adults with very low food security reported being able to comply with public health recommendations to purchase two weeks of food at a time. For every basic needs challenge, food-insecure adults were significantly more likely to report facing that challenge, with a clear gradient effect based on severity of food security. The short-term effects of the COVID-19 pandemic are magnifying existing disparities and disproportionately affecting low-income, food-insecure households that already struggle to meet basic needs. A robust, comprehensive policy response is needed to mitigate food insecurity as the pandemic progresses.

**Keywords:** covid-19; food insecurity; low-income adults; disparities; survey

#### **1. Introduction**

Food insecurity, a condition defined by limited or uncertain access to sufficient, nutritious food for an active, healthy life, disproportionately affects low-income communities and communities of color [1]. Food is a core social determinant of health [2] and food insecurity is associated with numerous poor health outcomes in both the short and long term [3–9]. The unprecedented COVID-19 pandemic, and the associated social and economic response [10] (e.g., school closures, stay at home orders, business closures, and job losses) have the potential to dramatically increase food insecurity and its related health disparities among already at-risk populations. Early evidence suggests that food insecurity is indeed rapidly rising above pre-epidemic levels [11–13]. Household food insecurity has risen from 11% in 2018 to 38% in March 2020; in April 2020, 35% of households with a child aged 18 and under were food insecure [12,13]. Households already struggling with food insecurity may find their current situations exacerbated by COVID-19 with fewer resources to comply with social distancing recommendations. Food insecure individuals also may have less flexibility in their jobs to allow them to earn income while staying home, or may be at higher risk of losing their jobs completely, thereby decreasing (or eliminating) their incomes. These factors may put food insecure households both at higher risk of contracting COVID-19 and of greater food insecurity due to the economic effects of COVID-19 mitigation efforts.

In addition to the long-term health and economic effects of the COVID-19 pandemic, it is important to understand the immediate impact of social distancing measures to fight COVID-19 on vulnerable populations who already struggle to meet their basic needs. To do so, we fielded a national survey of low-income adults in the US on 19–24 March 2020 to understand the immediate effects of how COVID-19 was impacting low-income Americans and any disparities in its effects based on food security status.

#### **2. Materials and Methods**

We designed a web-based (Qualtrics) survey to measure the initial effects of COVID-19 on low-income adults in the United States (US) in mid-March 2020, just as some states were beginning to implement school closures and "stay at home" orders. The web-based survey was formatted to be accessible when access both via smart phones and on a personal computer or laptop. The survey was fielded using TurkPrime, an online crowdsourcing platform that is designed to be used for academic research [14]. TurkPrime allows researchers to use quotas to recruit a sample that matches their specific needs and has been used in numerous academic studies from a variety of disciplines published in the peer-reviewed literature [15–19]. In the present study, we used a census matched panel of US adults (matched on age, gender, and race/ethnicity to the overall population) and limited the sample to low-income adults with household incomes <250% of the federal poverty line (FPL). The FPL is calculated based on both household size and annual household income. For example, 100% of the FPL for a four-person household is \$26,200, and 250% of the FPL for a four-person household is \$655,000 per year. The annual income for a two-person household at 250% FPL is \$43,100.

The survey was open to participants on 19–24 March 2020. We invited 2840 eligible panel members to participate and 1497 participants completed the survey (53% completion rate). Additional exclusions included participants who completed the survey in <4 min (*n* = 7), indicated they did not live in the US (*n* = 3), and were missing food insecurity data (*n* = 9) resulting in a final analytic sample size of 1478. Forty-four percent of participants took the survey on a personal computer or laptop and 56% took the survey on a smart phone or mobile device. This study was determined to be exempt by the Institutional Review Board at the University of Michigan.

#### *2.1. Measures*

*Food security:* Food security status over the past 30 days was measured using the 18-item US Household Food Security Module [20]. Questions are ordered by severity and include three levels of screening for adults, and an additional level of questions only for households with children. Affirmative responses to questions were summed to create a total food security score (out of 10 for adults and out of 18 for households with children). Food security categories (high, marginal, low, very low) were assigned according to US Department of Agriculture scoring guidelines [21]. The term food insecurity refers to the combined categories of low and very low food security.

*COVID-19-related basic needs challenges:* We inquired about challenges related to meeting basic needs people may have faced in the early weeks of the US COVID-19 epidemic and response. First, we asked about participants' ability to comply with recommendations to purchase two weeks of food (which was recommended by public health efforts to limit grocery shopping trip and facilitate social distancing). We also asked participants whether they had encountered any of the following challenges due to the coronavirus: the ability to feed their family, availability of household items such as toilet paper, access to healthcare, access to medications, the ability to pay bills, ability to rent or pay mortgages, whether they had been unable to work due to lack of childcare, and whether they had been unable to work due to illness.

*COVID-19-related workplace reactions:* At the time of data collection, some, but not all, states had begun issuing stay at home orders and mandatory business closures. Even in states without stay at home mandates, some businesses were making adjustments to operations due to COVID-19. We asked working adults (i.e., those with full- or part-time work outside the home) what their employer was doing to adjust to the pandemic. Specifically, we asked "Workplaces in the US are adjusting to the coronavirus situation in different ways. What is your workplace doing to adjust?" Participants were given the following response options: nothing, my workplace is proceeding as normal; all employees are encouraged to work at home; all employees must work at home; essential employees must come in to work but others can work from home; hours are being reduced for hourly employees; my place of employment has temporarily closed due to the coronavirus; my place of employment has closed and I have been laid off; work is busier and employees need to work longer hours; other.

*Expected impact of COVID-19 on employment and income:* We asked working adults what they expected would happen at their job if they or someone in their family became ill with COVID-19. Response options focused on whether they would be able to stay home, whether they had vacation or sick days they could use, and what they expected would happen if they missed work due to illness.

#### *2.2. Analysis*

All analyses were conducted in 2020 with Stata, Version 15 (StataCorp LP, College Station TX, USA). First, we describe the socio-demographic characteristics of the study sample overall and by food security status using cross tabulations and chi-squared tests of significant differences. Next, we examine differences in COVID-19-related basic needs and workplace challenges (among participants working full or part time), by food security status using cross tabulations. Missing data was treated using listwise deletion. Significant differences by food security status were assessed using chi-squared tests. All tests were two tailed and significance was considered at *p* < 0.05.

#### **3. Results**

The characteristics of the sample of low-income adults are presented in Table 1. Overall, 36% of this sample was food secure, 20% had marginal food security, and 44% were food insecure (17% low food security; 27% very low food security). Individuals with low or very low food security were more likely to be non-Hispanic Black or Hispanic, to have children in the home, and have less than a college education. Individuals with very low food insecurity were also more likely to rent their homes, not have health insurance or have Medicaid, and were more likely to be receiving SNAP benefits. The distribution of the sample by state of residence is shown in Appendix A.


**Table 1.** Demographic characteristics of the study sample overall and by food security status (*n* = 1478).


**Table 1.** *Cont.*

<sup>a</sup> Column percentage; <sup>b</sup> Row percentage.

Figure 1 shows the ability of low-income US adults to comply with public health recommendations to stock up on two weeks of food to avoid excess grocery store trips and facilitate social distancing. Nearly 2/3 (60%) of food-secure, low-income adults reported being able to comply with that recommendation, compared to less than one in five (18.8%) of low-income adults with very low food security. Adults with very low food security were more likely to report their local stores were sold out of products, and not being able to afford to purchase an extra two weeks of food at one time.

**Figure 1.** Ability to comply with recommendation to stock up on two weeks of food among low-income US adults, by food security status (*n* = 1478). Question text: "Experts have recommended stocking up on two weeks of food for your household to prepare for the coronavirus. Have you been able to do this? [Please check all that apply. One respondent was missing data for this question and was excluded from analysis. Differences within each response option by food security status are significant at *p* < 0.001 based on chi-squared tests.

Potential basic needs challenges related to COVID-19 are displayed in Figure 2. For every challenge asked about, food-insecure adults were significantly more likely to report dealing with that challenge, with a clear gradient effect based on severity of food security status. Strikingly, 41.3% of adults with very low food security reported not having enough food to feed themselves or their family compared to 10.7% of adults with low food security, 3.1% of adults with marginal food security and 1.6% of adults with high food security. Half (49.9%) of adults with very low food security did not have enough money to pay their bills compared to 36.9% of those with low food security, 23.1% of those with marginal food security and 8.8% of food secure adults.

**Figure 2.** Challenges encountered by low-income US adults as a result of COVID-19, as of 19–24 March, by food security status (*n* = 1478). Question text: "Have you experienced any of the following challenges due to the coronavirus (COVID-19) so far?" [Please check all that apply] Percentages reflect the percent of respondents in each food security category that said they encountered that problem. Ten percent of respondents (*n* = 161) did not indicate any of the response options were challenges for them and are counted as 'missing'.

Food-secure, low-income adults working full or part time (44.3% of the overall sample) were more likely than their food-insecure counterparts to work in jobs that were either proceeding as normal, were busier than usual, or had closed and laid off employees. In contrast, working adults with food insecurity were more likely to have their hours reduced (Table 2). When asked what they thought would happen if they or someone in their family got sick with COVID-19, working adults with very low food security were less likely than their food-secure counterparts to have sick days or vacation

days they could use, and were more likely to say they would lose their job if they missed too many days of work (52% very low food security vs. 18% high food security, *p* < 0.001).

**Table 2.** COVID-19 effects on workplaces among low-income adults working full or part time in the US overall and by food security status as of 19–24 March 2020 (*n* = 655).


<sup>a</sup> Column percentage; <sup>b</sup> Row percentage; <sup>c</sup> Asked among respondents who are working full or part time.

#### **4. Discussion**

This study presents results from a national survey of low-income adults in the US in the days immediately following the first major policy steps to enforce COVID-19-related social distancing measures on a wide scale in the US. Though large-scale school and business closures were only beginning to be implemented [22], we find that the effects of the COVID-19 pandemic were already impacting low-income adults, with disproportionately negative effects for low-income adults experiencing food insecurity. This initial evidence from a time period before even greater economic effects and job losses took place demonstrate that the COVID-19 pandemic threatens to greatly exacerbate existing health disparities related to food security status. Indeed, evidence from later surveys show that food insecurity in the US has dramatically increased well beyond levels seen during the Great Recession [12,13].

Results from this study illuminate the extent to which, very early in the COVID-19 trajectory in the US, individuals with food insecurity were disproportionately vulnerable to the severe economic and health consequences of the crisis. Our findings show that as early as mid-March 2020, food-insecure adults currently working outside the home were at greater risk of losing their income or their jobs if they got sick from COVID-19. Regardless of whether they get sick or their employment status, food-insecure individuals were also more likely to report expecting that they will lose income during the pandemic. For already low-income households, loss of income puts them at high risk of severe food insecurity and an inability to meet other basic needs, both of which can lead to future physical and mental health problems [23–26]. Compared to low-income, food-secure adults, food-insecure adults were more likely to report that they had already been laid off, and that their income would go down substantially. Fifty-four percent of food secure adults reported they expected their income would remain the same compared to 23% of adults with very low food security (results not shown, but available upon request). Subsequent massive job losses [27] and more extensive social distancing measures [10] after data collection ended have likely exacerbated the trends we document in our results.

Across the lifespan, food insecurity is associated with a range of negative health outcomes over the short and long term, including poor mental health outcomes such as depression, stress, and anxiety [4,9,28], poor diet quality [7,29], high rates of chronic diseases such as diabetes and obesity [6,30,31], and lower overall health status [3,5,32]. Food insecurity is also associated with higher healthcare expenditures, in part due to the higher burden of chronic health conditions among food-insecure patients and the known tradeoffs between food and medicine [24,26]. As the COVID-19 pandemic and the associated economic fallout progress, it will be critical for policymakers, health systems, and the public health community to proactively and comprehensively address access to food and other basic needs, particularly for populations at risk of, and already experiencing, food insecurity. Failure to do so will have long-term implications for population health, health disparities, and the health care system as a whole.

Food-insecure adults are more likely to be people of color, with lower social standing, who have less flexible and secure jobs, and are more vulnerable to chronic stress and other basic needs insecurities [1]. In 2018, 11.1% of adults in the US were food insecure; among low-income adults (<185% FPL), 29.1% were food insecure [1]. We find that, as of mid-March, 2020, 44% of adults with an income <250% of the FPL were food insecure in the past 30 days, and these individuals were more likely to be non-Hispanic Black or Hispanic. This disparity in food security status based on race/ethnicity is an additional way in which COVID-19 is disproportionately impacting communities of color in the US. Since mid-March, adult and child food insecurity rates in the US have dramatically risen [12,13]. In our study, adults currently experiencing food insecurity were not able to buy food in bulk quantities and therefore are at greater risk of exposure to the virus (due to the need for more frequent food shopping trips) as well as being at greater risk of an acute hunger crisis (due to lack of financial resources to purchase sufficient food). In addition, as individuals already at risk for food insecurity are more vulnerable to losing their jobs, rates of food insecurity will climb higher as the pandemic progresses.

Direct income support, expanded unemployment benefits, and additional support for federal food assistance programs included in the CARES act (passed on 27 March 2020) and the Families First Coronavirus Response Act (passed on 18 March 2020) are important first steps to supporting low-income families in the US [33,34]. However, longer-term support for individuals, as well as institutions and organizations that provide food, is needed. Some communities are already experiencing unprecedented demand in the emergency food system [35], and the federal government, states, and cities are scrambling to ensure that families with children who depend on free or reduced price meals at school do not go hungry [36]. Given the scale of the pandemic and the likely duration of social distancing measures and the associated economic impacts, more support is urgently needed to mitigate the toll of COVID-19 on the most vulnerable members of society. In particular, in addition to direct economic support to individuals, financial support for the emergency food system, greater flexibility for school systems to provide food to families, and long-term, expanded food assistance support via the Supplemental Nutrition Assistance Program (SNAP) are all urgently needed. Expanded SNAP benefits were critical for providing needed support for low-income families during the Great Recession, and were effective at reducing food insecurity [37]. Congress and the Trump administration should urgently increase SNAP benefits and expand eligibility for the program to help low-income families afford food during this extraordinary time.

#### *Limitations*

Results from this study should be considered in light of some limitations. First, the web-based survey panel does not use probability-based sampling and is not nationally representative. However, the TurkPrime panel is national in scope, and uses census-matched quotas to achieve a sample that closely aligns with the demographics of the population in the US which mitigates some of this concern. However, because we limited our sample to households <250% of the FPL (based on income and household size), and because the survey was only available in English, the demographics of our sample may be more similar to the US population overall and undercount some key demographic groups, particularly non-Hispanic Blacks and Hispanics, non-English speakers, and immigrants. Relatedly, this data was collected via a web-based survey which by definition required participants to have internet access via a computer or a smart phone. This method of data collection could also have undercounted some groups (e.g., those with very low income, without high school degrees, and those living in rural areas without broadband internet access) [38], likely those especially vulnerable to food insecurity. It is important to note, however, that any bias introduced from these factors would have biased results into the direction of undercounting, rather than overcounting, food insecurity and the other outcomes we document here. Second, respondents could choose whether or not to participate in the survey which may introduce some selection bias. Third, all measures in the study are self-reported and may be subject to a social-desirability bias. However, the fact that the survey was fielded online and was completely anonymous may mitigate this concern. Fourth, this survey is cross-sectional and we cannot make any statements about causal relationship between the coronavirus and our measure of food insecurity in the past 30 days. Finally, data were collected very quickly after social distancing measures and business and school closures began to be implemented in some (but not all) states. This is a strength as we were able to capture the immediate, real-time impacts on low-income adults. However, some measures also focused on anticipated effects. It is possible that the respondents did not accurately assess the likely effect of the coronavirus pandemic on their employment and income. However, initial evidence in the weeks after our data were collected show clearly that unemployment and rates of food insecurity have skyrocketed. The longer-term effects on low-income adults in the US, and associated disparities based on food insecurity, may be better or worse than those expected by participants in this survey. It will be imperative for future research to examine the long-term effects of the coronavirus pandemic and associated social distancing measures on food insecurity and associated health outcomes, particularly among vulnerable communities that were already struggling at the start of the pandemic.

The strengths of our study include the fact that we were able to collect these data so quickly after a national emergency was declared and states began implementing policies to slow the spread of COVID-19. Our large national sample of low-income adults is another key strength as is our use of the gold standard 18-question USDA food security screener module.

#### **5. Conclusions**

The social and economic upheaval caused by the COVID-19 pandemic is magnifying existing disparities and disproportionately affecting low-income, food-insecure households that already struggle to meet basic needs. The early effects documented in the present study are likely to continue to worsen as the pandemic continues unless extensive policy and economic supports are swiftly implemented.

**Author Contributions:** Conceptualization, J.A.W. and C.W.L.; Methodology, J.A.W.; Formal analysis, J.A.W.; Data curation, J.A.W.; Writing—Original draft preparation, J.A.W.; Writing—Review and editing, J.A.W. and C.W.L.; project administration, J.A.W.; funding acquisition, J.A.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** Funding for this study was provided by a Faculty Research Grant from the University of Michigan Poverty Solutions. J.A.W. was also supported by the National Institutes of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health (Award #K01DK119166) and C.W.L. was supported by the Eunice Kennedy Shriver National Institute for Child Health and Human Development (Award #4R00HD084758).

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **Appendix A**


**Table A1.** Distribution of the Sample by State.


**Table A1.** *Cont.*

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
