Cross-sectional Analysis of Food Insecurity and Frequent Emergency Department Use

Introduction Emergency department (ED) patients have higher than average levels of food insecurity. We examined the association between multiple measures of food insecurity and frequent ED use in a random sample of ED patients. Methods We completed survey questionnaires with randomly sampled adult patients from an urban public hospital ED (n = 2,312). We assessed food insecurity using four questions from the United States Department of Agriculture Household Food Security Survey. The primary independent variable was any food insecurity, defined as an affirmative response to any of the four items. Frequent ED use was defined as self-report of ≥4 ED visits in the past year. We examined the relationship between patient food insecurity and frequent ED use using bivariate and multivariable analyses and examined possible mediation by anxiety/depression and overall health status. Results One-third (30.9%) of study participants reported frequent ED use, and half (50.8%) reported any food insecurity. Prevalence of food insecurity was higher among frequent vs. non-frequent ED users, 62.8% vs 45.4% (P <0.001). After controlling for potential confounders, food insecurity remained significantly associated with frequent ED use (adjusted odds ratio 1.48, 95% confidence interval, 1.20–1.83). This observed association was partially attenuated when anxiety/depression and overall health status were added to models. Conclusion The high observed prevalence of food insecurity suggests that efforts to improve care of ED patients should assess and address this need. Further research is needed to assess whether addressing food insecurity may play an important role in efforts to reduce frequent ED use for some patients.


INTRODUCTION
Even before the coronavirus disease (COVID-19) pandemic, food insecurity affected over 10% of United States households, including nearly 14% of households with children. 1 By May 2020, nearly 18% of US nonelderly adults and 22% of parents with children reported food insecurity in the prior month. 2 Food insecurity is associated with a wide range of negative health outcomes and with higher healthcare costs. [3][4][5][6][7][8][9][10][11] Large racial and ethnic inequities exist in food insecurity, as they do for health outcomes broadly; Black and Latinx households are disproportionately affected compared to White households. 2 A systematic review of the literature on social needs of emergency department (ED) patients found that prevalence of food insecurity is even higher among ED patients than among the general public. 12 Studies have also found associations between food insecurity and more frequent ED use among specific groups including people experiencing homelessness, 13,14 and people with diabetes, 15 and among lowincome Americans more generally. 16 Food insecurity may lead to increased ED use due to its association with poor physical and mental health, 11 worse control of chronic diseases, 11,17,18 and medication non-adherence, 19 which are in turn associated with ED use. [20][21][22] We build on past literature by examining the association of food insecurity and frequent ED use among a large, random sample of ED patients not restricted to any particular subpopulation. We aimed to increase understanding of potential pathways between food insecurity and frequent ED use by examining whether poor physical and mental health might be mediators of this relationship.

Study Design
We describe a cross-sectional, secondary analysis of baseline survey questionnaires conducted with randomly sampled ED patients as part of a larger study described previously. 23,24 The study was approved by the NYU School of Medicine Institutional Review Board.

Setting and Participants
Data collection occurred at a large, urban, public hospital in New York City from November 2016-September 2017. Adult (≥18 years) ED patients were eligible if they were medically/psychiatrically stable, not in prison/police custody, and spoke English or Spanish. Research assistant (RA) shift schedules rotated over time to cover all hours of the day and days of the week, with the number of shifts scheduled during a given time window over the course of the study approximately mirroring ED patient arrival volumes. The RAs approached patients following a random sampling scheme; they then read questions aloud and recorded responses using REDCap (Vanderbilt University, Nashville, TN) secure electronic data capture tools. 25 Participants provided written informed consent and received $15 compensation.

Measures
Measures were self-reported. We defined frequent ED use as self-report of ≥4 ED visits in the past 12 months, to any ED, including the current visit. While there is no standard definition of "frequent ED use," ≥4 visits within one year is commonly used. 26 Past research has found patients self-report ED visits with good accuracy. 27 Participants answered four questions about food insecurity in the past 12 months from the widely used US Department of Agriculture (USDA) Household Food Security Survey. 28 Questions were as follows: 1) I/we worried whether my/our food would run out before I/we got money to buy more; 2) The food that (I/we) bought just didn't last, and (I/we) didn't have money to get more; 3) (I/we) couldn't afford to eat balanced meals; and 4) Did you ever eat less than you felt you should because there wasn't enough money for food?
For the first three questions, participants responded "never true," "sometimes true," or "often true." For the last question-a measure that identifies very low food security 1 (sometimes called food insufficiency)-participants responded "yes" or "no," with yes being considered an affirmative response. The primary independent variable was any food insecurity, defined as participants giving an affirmative response (both "sometimes true" or "often true" were included as affirmative) to any of the four items. We separately examined association of frequent ED use with only Estrella et al.
Cross-sectional Analysis of Food Insecurity and Frequent ED Use the more severe form of food insecurity, food insufficiency.
In bivariate analyses we also examined each food insecurity question separately and the number (0-4) of questions answered affirmatively. Covariates included age, gender, race and ethnicity, insurance status, difficulty meeting essential expenses (past year), 29 homelessness (living in a shelter, unsheltered, or doubled up; past year), unhealthy alcohol use, 30 and moderate or greater problems with drug use as measured by the Drug Abuse Screening Test (DAST-10). We decided a priori to examine physical health, anxiety, and depression as potential mediators based on prior literature and theory positing these factors as sensitive to food insecurity and as strong drivers of ED use. 11,22,31,32 Self-reported overall health was measured using a single item from the US Centers for Disease Control and Prevention Health-Related Quality of Life score, asking "Would you say that in general your health is"; possible answers were excellent, very good, good, fair, and poor. 33 Anxiety was measured using the GAD-2 (general anxiety disorder) and depression using the PHQ-2 (patient health questionnaire); both are previously validated two-item screeners. 34,35 Analysis We examined bivariate associations using chi-squared tests of independence for categorical variables. We conducted multivariable logistic regression to examine the independent association of any food insecurity and food insufficiency with frequent ED use while adjusting for potential confounders. As described above, we examined for mediation by anxiety or depression and overall self-rated health. Anxiety and depression were combined into a single binary variable due to significant collinearity; no other variables demonstrated significant collinearity (Spearman correlation coefficients < 0.4). We examined mediation by including the hypothesized mediators in adjusted regression models and determining whether effect estimates were attenuated compared to effect estimates in models including confounders but without hypothesized mediators. 36 Complete case deletion was used in regression models; the amount of missing data was small (3.7%).

RESULTS
Research assistants approached 6097 patients, of whom 2924 (48%) were eligible. The most common reasons for ineligibility were being medically unstable, intoxicated, not speaking English/Spanish, or in police/prison custody. Of eligible patients, 2396 (82%) agreed to participate. After removing duplicate records for patients who participated more than once (n = 84) there were 2312 participants. Three did not answer the question about past ED use (n = 3) and were excluded from bivariate and multivariable analyses. Participants were diverse in gender, race and ethnicity, and age (Table 1). Half (50.8%) reported any food insecurity. Many also reported difficulty meeting basic expenses and past year homelessness. Cross-sectional Analysis of Food Insecurity and Frequent ED Use Estrella et al.

DISCUSSION
We found a robust association between food insecurity and frequent ED use, including in multivariable analyses adjusting for potential confounders. This relationship was partially attenuated by controlling for anxiety/depression and overall health status, suggesting the possibility of mediation. Notably, while prevalence of food insecurity was highest among participants who reported frequent ED use, even participants without frequent ED use had a high prevalence of food insecurity.
Our findings are consistent with past research showing ED patients have a high prevalence of social needs, including food security. A systematic review 12 showed that while studies varied, food insecurity prevalence of ED patients was generally above 20%, with several studies finding prevalence of one-third or even higher. 12 A few studies examining the association of food insecurity and frequent ED use, among specific subgroups [13][14][15] and more generally, 16 have uniformly found a significant association. Other studies have found foodinsecure ED patients are more likely to have chronic pain, mental health concerns, substance use, and homelessness, all of which are known to be associated with frequent ED use. 3,37 Our study was unique in randomly sampling a large number of ED patients and including multiple measures of food insecurity, as well as examining the independent association of food insecurity with frequent ED use while controlling for possible confounders and exploring mediation. Although we cannot prove causality in this cross-sectional study, one potential hypothesis is that food insecurity contributes to anxiety/depression and poor overall health, which in turn contributes to frequent ED use. This hypothesis could be examined in future longitudinal research.
The strong association observed between food insecurity and frequent ED use in this study has implications for programs aiming to reduce frequent ED use. Frequent ED use has been the subject of persistent programmatic and policy attention in the US, although programs to address it have had variable success, particularly when examined using robust study designs. 38 Our study adds to evidence suggesting the importance of assessing and addressing the social and structural conditions of people's lives as an integral part of programs developed to reduce frequent ED use.
There has been increased interest nationally in screening for patient social needs in healthcare settings. For food insecurity, a two-item screening tool called the Hunger Vital Sign has been well tested and validated in healthcare settings. [39][40][41][42] The items are based on two of the USDA Food Security Survey questions used in our study, on worry about food running out and food not lasting. Research indicates patients generally feel positively about being asked such questions in healthcare settings, including EDs. 43,44 Some studies have suggested patients may prefer and more readily disclose food insecurity when electronic tablet-based screening is used, 45  Nearly one-third (30.9%) were frequent ED users. Participants who reported frequent ED use had significantly higher prevalence of food insecurity than other ED patients (Table 2). This finding held true across each individual food insecurity question and for food insecurity overall, with 62.8% of participants with frequent ED use endorsing any food insecurity vs 45.4% of participants who did not report frequent ED use (P <0.001).
In multivariable analyses (Table 3), both any food insecurity (adjusted odds ratio [aOR] 1.48, 95% confidence interval [CI], 1.20-1.83) and food insufficiency (aOR 1.45, 95% CI, 1.16-1.83) were associated with frequent ED use. These relationships were partially attenuated in models adding depression/anxiety and overall health status, with persistently significant yet reduced aORs for the associations of food insecurity/insufficiency and frequent ED in mediation models. a Percentages shown are among those who answered a given question; denominators for some questions are <2,312 due to a small amount of missing data for some questions (never exceeding 1.6%).  a Percentages shown are among those who answered a given question; denominators for some questions are lower due to a small amount of missing data for some questions (never exceeding 1.7%). b P-values for bivariate associations tested using chi-squared tests of independence for categorical variables.  To date, few studies have rigorously examined how to best assist ED patients who screen positive for food insecurity. A systematic review by De Marchis et al found 23 studies-most of which were of low quality-that examined interventions addressing food insecurity in healthcare settings. 47 One study found having an electronic health record order for referral to a local food bank partner (with patient contact information sent to the food bank and the food bank proactively contacting patients) resulted in more ED patients receiving referrals, 63% of whom ultimately received assistance. 48

LIMITATIONS
Our study results should be interpreted in light of a few limitations. First, measures were self-reported. We used validated questions when available and chose measures for which we expected self-report to be accurate. Second, this study was conducted in a single public hospital ED serving a patient population with high levels of social needs. However, multiple other studies conducted in geographically diverse EDs have found that ED patients have a high prevalence of food insecurity. 12 Even if prevalence of food insecurity and other participant characteristics in our study differed from those of patients at other EDs, we do not expect the relationship between food insecurity and frequent ED use would be unique to the ED patients we studied.
Finally, we cannot suggest causality for relationships observed in this cross-sectional study. Although we controlled for multiple potential confounders, including other measures a Adjusted models include: gender, race and ethnicity, age category, insurance status, homelessness in past 12 months, difficulty meeting essential expenses in past 12 months, unhealthy alcohol use, and moderate or greater drug use problems. b Mediation models additionally adjust for anxiety/depression (positive GAD-2 or PHQ-2) and self-rated overall health. Independent effect estimates for the association between the mediators and outcomes in the mediation models were all positive and statistically significant (not shown). c Any food insecurity defined as affirmative response (often true or sometimes true coded as affirmative) to ≥1 of the 4 US Department of Agriculture (USDA) food insecurity questions asked (over the past 12 months). d Food insufficiency defined as a "yes" response to USDA question of whether participant had eaten less than they felt they should because there was not enough money for food in the past 12 months. e Parameter estimates for any food insecurity and food insufficiency reduced in mediation models by 21.2% and 31.5%, respectively. OR, odds ratio; CI, confidence interval. Table 3. Association of emergency department (ED) patient food insecurity with frequent ED use.

Estrella et al.
Cross-sectional Analysis of Food Insecurity and Frequent ED Use of socioeconomic status, there is a possibility of unmeasured confounders. Additionally, although we postulate one hypothetical causal pathway in our mediation analyses, the data remain cross-sectional; we were unable to prove causal associations, and mental health and health status could potentially be confounders as well as mediators. Additional longitudinal and qualitative research could further elucidate the relationship of food insecurity and frequent ED use. We also suggest implementing and studying programs to assist ED patients with food insecurity.

CONCLUSION
We found a high prevalence of food insecurity among ED patients in our study population. Food insecurity was significantly associated with frequent ED use. Efforts to improve care of patients who frequently visit the ED should assess and address social needs including food insecurity; even apart from any potential effects on reducing future ED use, having adequate food is a critical human need that such efforts could be well-positioned to help address. More generally, EDs have long been described as "social welfare institutions," 49 and there has been a recent resurgence of interest within emergency medicine in patient social needs. 50 This study adds to the body of evidence supporting the potentially important role of EDs in assisting patients with food insecurity.