Eating Disorder Pathology Among Individuals Living With Food Insecurity: A Replication Study

Abstract

Eating disorders (EDs) are stereotypically associated with thin, White, affluent women and girls. One result of the ED stereotype has been a relative dearth of ED research with marginalized communities. The aim of this study was to replicate recent findings showing an association between the severity of food insecurity (FI) and increased ED pathology. Participants included 891 clients of an urban food bank. Results were consistent with the findings of previous research; participants in the most severe FI group reported significantly higher levels of ED pathology, dietary restraint, anxiety, and depression. Findings provide further evidence that the thin, White, affluent, female ED stereotype offers a flawed portrait and highlight the need for additional psychological research that focuses on marginalized populations to address disparities in access to care. Both scholars and clinicians need to move away from the stereotypical portrait of who is and is not at risk for developing an ED.

Keywords

food insecurity eating disorders binge eating marginalized populations race and ethnicity

Kazdin and Blase (2011) argued that psychotherapy research needed to be rebooted if the field ever hoped to have a significant impact on the global burden of mental illness. They highlighted marked ethnic and racial disparities in who received mental-health treatment and proposed that the field needed to develop a portfolio of interventions, including lower-cost interventions that could reach marginalized communities. Other researchers have also noted the need to develop innovative intervention models that specifically address both the psychological and social-justice needs of impoverished communities (Goodman, Pugach, Skolnik, & Smith, 2013; Smith, 2005, 2010). However, innovative interventions cannot be developed if such needs go unrecognized secondary to stereotyped notions about who will and will not develop a particular form of mental illness. Thus, one key first step in addressing the calls to action by Kazdin and Blase and others is to document the types of problems encountered by marginalized communities.

Eating disorders (EDs) represent a significant public-health problem. EDs are associated with substantial psychological comorbidity (Swanson, Crow, Le Grange, Swendsen, & Merikangas, 2011; Chamay-Weber, Narring, & Michaud, 2005), significant medical complications (Mehler, 2017a, 2017b), elevated mortality rates (Arcelus, Mitchell, Wales, & Nielsen, 2011), and role impairment (Hudson, Hiripi, Pope, & Kessler, 2007). EDs are also associated with quality-of-life impairments that are similar to those associated with other disorders, such as depression, bipolar disorder, and Wilson’s disease (Carta et al., 2014). Moreover, research indicates that ED pathology assessed in patients who are in their mid-20s predicts increased psychological distress 10 years later, even after controlling for body mass index (BMI), self-rated health, psychological distress, and education (Kärkkäinen, Mustelin, Raevuori, Kaprio, & Keski-Rahkonen, 2018). Despite the burden of EDs, however, research suggests that only one quarter of individuals with an ED ever seek treatment and even fewer receive it (Hart, Granillo, Jorm, & Paxton, 2011).

The ED community increasingly has recognized that antiquated notions, held by both clinicians and the general public, about who is and is not at risk for EDs represent one barrier to addressing ED treatment needs (Mitchison, Basten, Griffiths, & Murray, 2017; Sonneville & Lipson, 2018). Specifically, EDs have long been stereotypically associated with thin, White, affluent girls and young women (Bruch, 1973; Sonneville, & Lipson, 2018). Some researchers have recently disputed the accuracy of this perception (e.g., Mulders-Jones, Mitchison, Girosi, & Hay, 2017); however, others have found that clinicians may be less likely to detect EDs in those who defy traditional stereotypes (Gordon, Brattole, Wingate, & Joiner, 2006). Further anecdotal reports suggest that those who differ from the ED stereotype are less likely to think they have an ED or perceive a need for treatment (P. Patel, n.d.; Sonneville & Lipson, 2018).

Although researchers are beginning to produce data challenging ED stereotypes, much of the ED literature focuses on girls and young women, most of whom are White. As a result, several potentially at-risk populations have been understudied, if not ignored, such as those living with food insecurity (FI). FI refers to a state in which households have limited access to sufficient and nutritious food as a result of inadequate resources, including money, to support active and healthy living (Coleman-Jensen, Rabbitt, Gregory, & Singh, 2017). In contrast, food security occurs when households have reliable access to sufficient food, both in terms of quality and quantity, to support active and healthy living (Coleman-Jensen et al., 2017; Lang & Barling, 2012). FI is often perceived to be a problem for low- and middle-income countries, yet almost 16 million households in the United States qualified as food-insecure in 2016 (Coleman-Jensen et al., 2017). Although some children in the United States do experience FI, in many cases they are protected from insufficient food intake and markedly disrupted eating patterns by adult caregivers and school-meal programs (Bove & Olson, 2006; Coleman-Jensen et al., 2017; McIntyre et al., 2003). As a result, in roughly half of food-insecure U.S. households with children, only adults experience FI (Coleman-Jensen, Rabbitt, Gregory, & Singh, 2016).

There are two primary mechanisms that may explain why adults living with FI might experience elevated levels of ED pathology. First, Keys et al. (1950) and Tucker (2006) documented the development of ED symptoms (e.g., binge eating, sense of loss of control over eating) in adult male participants in the Minnesota Starvation Study. Keys et al. (1950) has long been viewed by the ED community as providing robust evidence about the role of dietary restriction in the onset of ED pathology. Second, food-insecure households in urban environments often simultaneously have limited access to affordable nutritious food (i.e., conditions associated with food deserts; Dutko, 2012; Kato & Irvin, 2013) and abundant access to food retailers (e.g., corner stores, fast-food restaurants) selling inexpensive and highly palatable food (HPF; i.e., processed food high in salt, sugar, and fat); these latter conditions have been referred to as food swamps (Kato & Irvin, 2013). Animal-model research implicates HPF in the development of binge eating. In particular, some rats that consume normal amounts of food when on a diet of rat chow initiate binge eating when HPF is introduced (Boggiano et al., 2007).

On the basis of these two lines of reasoning, we previously tested a series of hypotheses that associated increased levels of FI with increased levels of ED pathology in a sample of 503 participants who were clients of food pantries affiliated with the San Antonio Food Bank (SAFB; Becker, Middlemass, Taylor, Johnson, & Gomez, 2017). Most participants belonged to ethnic and racial minority groups and were poor. Results revealed that adults reporting the most severe level of FI reported significantly higher levels of ED pathology relative to those with less severe FI or those living on the margins between FI and food security. In the most severe food-insecure group (participants with hungry children in the home), 17% met the cutoff for a current clinically significant ED. One finding that ran counter to our hypotheses but supported the high-risk nature of severe FI was that 20% of participants in this group reported self-induced vomiting to counteract the effects of eating or to prevent weight gain (Becker et al., 2017).

Becker et al. (2017) also found that more severe FI was associated with increased intentional dietary restraint. According to restraint theory (Herman & Polivy, 1975), intentionally restricting caloric intake increases the risk for binge eating. Restraint scales typically link intentional dietary restriction with weight and shape concerns. For instance, restraint items on the Eating Disorders Examination–Questionnaire (EDE-Q; Fairburn, 2008) ask participants whether they restrict to influence weight and shape. We hypothesized that adults living with FI might attempt to restrict for other reasons and that intentional restriction still might be associated with EDs regardless of the reason for which it was undertaken. Thus, we included three questions from the EDE-Q restraint scale but removed the weight and shape specifier (Becker et al., 2017). Instead, participants had the opportunity to explain why they restricted their intake in an open-ended format. A secondary qualitative analysis (Middlemass et al., 2018) indicated that those living with FI rarely reported restricting for weight and shape reasons (< 3%); instead, they reported restricting intake to stretch food to make it last, to save food for children, or because they did not have sufficient resources to buy sufficient food. Nonetheless, dietary restraint still correlated with ED pathology.

This finding is important because if ED researchers use standard, unadjusted dietary-restraint measures in food-insecure populations they may fail to detect important restricting behaviors secondary to the field’s historic focus on weight and shape concerns as a driver of EDs. Indeed, in a recent study of FI and objectively measured overeating (Stinson et al., 2018), food-insecure participants reported elevated disinhibition and hunger, but not restraint, relative to food-secure participants. However, restraint was assessed with a traditional measure (i.e., the cognitive restraint subscale from the Three-Factor Eating Questionnaire; Stunkard & Messick, 1985) that assesses restraint only for weight and shape concerns. Thus, it is plausible that food-insecure participants in Stinson et al. (2018) did restrict to a greater degree than food-secure individuals, but this was not detected secondary to the ED field’s selective attention to restraint as a result of weight and shape concerns.

To our knowledge, Becker et al. (2017) is the first study to use a validated ED measure to assess the full range of ED pathology, including compensatory behaviors, in those living with FI. Two other studies investigated binge eating and found that food-insecure participants reported more binge eating than food-secure participants (Bruening, MacLehose, Loth, Story, & Neumark-Sztainer, 2012; Rasmusson, Lydecker, Coffino, White, & Grilo, 2019).

Although there are scattered references to disordered and emotional eating in studies of FI (e.g., Bove & Olson, 2006), these references are often side notes in studies that are predominantly focused on the association between FI and obesity (e.g., Adams, Grummer-Strawn, & Chavez, 2003). To date, Becker et al. (2017) is still the only study conducted with the primary aim of documenting the association between FI and ED pathology, including compensatory behaviors.

The aim of the present study was, first and foremost, to determine whether we could replicate the results from Becker et al. (2017) in a larger sample that included more participants in the household level of FI, as this group was underrepresented in Becker et al., increasing the chance of spurious findings. In addition, we hoped to increase the number of participants who reported being on the margins of food security and FI compared with the 2017 study. Given the so-called replication crisis in psychology (Lindsay, 2015) and the fact that Becker et al. involves one aspect of the “troubling trio” identified by Lindsay (i.e., a surprising result, as indicated by numerous personal communications from those in the ED field), an attempt at replication seemed warranted.

We extended Becker et al. by including measures of depression and anxiety that could be used to identify participants who are likely to meet criteria for clinically severe depression and anxiety. To address critiques of the 2017 study, we collected self-reported weight and height data so that BMI could be calculated. On the basis of the findings in Becker et al., we tested six hypotheses. First, we hypothesized that the highest level of FI (hungry children in the home) would be associated with higher levels of ED pathology, dietary restraint, depression, and anxiety. Second, we hypothesized that those living on the margins between FI and food security would report significantly lower scores on those same dependent variables. This was not based on a significant finding in Becker et al. (2017) but rather on (a) observed patterns of scores (i.e., this group typically scored the lowest on the aforementioned variables) and (b) a recognition that the number of participants in this group in Becker et al. was relatively low (n = 41; 8.2%).

Third, we hypothesized that dietary restraint would significantly correlate with ED pathology per Middlemass et al. (2018). Fourth, we hypothesized that a greater percentage of those participants with the most severe FI would meet criteria for clinically significant ED pathology, depression, and anxiety. Fifth, we hypothesized, on the basis of Becker et al. (2017), that frequencies would not differ by gender, ethnicity, or race for ED pathology (for depression and anxiety, these analyses were exploratory). Finally, we hypothesized that rates of reported ED symptoms, including compensatory behaviors such as vomiting, would show a linear trend; the lowest levels would occur in those experiencing marginal FI and the highest levels would occur in those with the most severe FI. We continue to report descriptive data on major ED symptoms because of the unexpected finding in Becker et al. (2017) that participants with the most severe level of FI reported not only increased binge eating but also compensatory behaviors. We submit that it is important to determine whether this finding was also replicable.

Method

Participants

Participants included 891 adult clients of the SAFB. Participants received a $7 gift card to the largest grocery store in the area in exchange for completing a self-report survey. The mean age of the sample was 42.07 years (SD = 14.36), and the mean BMI, calculated from self-reported weight and height, was 30.73 (SD = 7.71). Table 1 provides demographic information of the participants, including gender, highest level of education, ethnicity and race, and annual household income. Approximately 75% of participants self-identified their ethnicity as Latinx/Hispanic; the city of San Antonio is a majority-minority city; approximately 64% of the population identifies as Latinx/Hispanic (U.S. Census Bureau, 2016).

Table 1.

Participant Gender, Education, Ethnicity, Race, Marital Status, and Household Income

Characteristics	Total sample(N = 891)		Not food-insecure(n = 86)		Household food insecurity(n = 364)		Individual food insecurity(n = 246)		CHH food insecurity(n = 192)
Characteristics	n	%	n	%	n	%	n	%	n	%
Gender
Male	287	32.4	31	36.0	125	34.4	71	29.0	60	31.3
Female	600	67.3	55	64.0	238	65.6	174	71.0	132	68.6
Education
No/some grade school	74	8.4	12	14.0	31	8.6	15	6.2	15	7.9
Finished grade school	59	6.7	9	10.5	24	6.6	10	4.1	16	8.4
Some high school	189	21.4	15	17.4	74	20.4	53	21.9	47	24.7
High school/GED	278	31.5	20	23.3	125	34.4	77	31.8	56	29.5
Some college/technical	234	26.5	20	23.3	91	25.1	77	31.8	46	24.2
College graduate/postgraduate	47	5.4	20	11.7	17	4.7	10	4.1	10	5.3
Ethnicity/race
Latinx/Hispanic	673	76.2	64	75.3	267	74.0	186	75.6	155	80.7
Black/African American	52	5.9	7	8.2	20	5.5	15	6.1	10	5.2
White	89	10.1	8	9.4	46	12.7	24	9.8	11	5.8
Other	69	7.8	6	7.2	28	7.8	20	8.1	15	7.8
Marital status
Single	351	39.5	32	37.2	148	40.7	105	42.7	65	34.0
Married or living with a partner	287	32.3	29	33.7	99	27.2	83	33.7	76	39.8
Separated	88	9.9	4	4.7	36	9.9	30	12.2	18	9.4
Divorced	113	12.7	12	14.0	56	15.4	23	9.3	22	11.5
Widowed	49	5.5	9	10.5	25	6.9	5	2.0	10	5.2
Annual household income
≤ $10,000	477	56.2	34	41.0	215	61.8	125	54.1	102	54.8
> $10,000–$20,000	210	24.7	21	25.2	81	23.3	60	26.0	48	25.8
> $20,000–$30,000	85	10	12	14.4	25	7.1	24	10.4	24	13
> $30,000	71	8.3	15	18.0	24	7.0	22	9.5	10	5.3

Note: Three participants did not provide complete food-insecurity data and could not be classified into a food-insecure group. Percentages are based on the number of participants responding to a given item. CHH = child-hunger household; GED = General Educational Development.

Procedure

This study was approved by the Trinity University Institutional Review Board and the SAFB, which served as an organizational partner. The SAFB service area consists of 16 counties in southwest Texas, and research by Feeding America ranks Texas as the second-most food-insecure state in the United States (Feeding Texas, 2015). Moreover, San Antonio ranks among the nation’s poorest cities; San Antonio proper had a poverty rate of 17.3% in 2017 (Royall, 2018). In collaboration with more than 500 partner agencies, the SAFB distributes food to approximately 58,000 individuals per week and provides several client services at its headquarters, including help in applying for federal and state programs (e.g., Supplemental Nutrition Assistance Program and Children’s Health Insurance Program).

After a series of planning meetings with SAFB staff, we collectively agreed that data collection should take place in the waiting area of client services at SAFB’s headquarters. To avoid disrupting those important services while collecting data, the research team made several trips to SAFB’s headquarters to observe day-to-day operations (e.g., the ways in which SAFB staff interacted with clients, length of waiting and meeting times, client flow at various times during the day). These observations were then used to develop the research protocol. Data collection began in July 2017 and finished in February 2018.

After clients signed the client-services waiting list, research assistants (RAs) approached individual clients using a standardized script, either in English or Spanish depending on the client’s preference, to determine whether the client wanted to participate in a short research study while he or she waited for an appointment. For clients who chose to participate, RAs used a second script to provide additional details about the study and to obtain consent. At least one RA who spoke Spanish was available at all times. After providing consent, clients received the survey in their preferred language and completed it. Surveys were precoded with identification numbers so that all responses were anonymous. RAs remained nearby in case clients had questions. If clients encountered difficulty reading and/or understanding the survey, a RA would offer to read the survey and assist in selecting the appropriate response. To try to maximize confidentially and privacy if the room was crowded, RAs used soft voices when communicating with clients, which clients tended to mimic. If the room was not crowded, RAs encouraged clients to move to a less populated area of the waiting room to enhance privacy and confidentiality. After completing the survey, clients received a list of low-cost/free mental-health resources, a behavioral activation handout, and their gift card. Clients were also informed of the opportunity to participate in a supplemental qualitative interview; those data are not reported here.

Measures

Assessment of FI

In addition to collecting demographic information, we assessed FI. There are multiple ways to measure the continuum of food security to extreme FI. The U.S. Department of Agriculture uses a classification system that first categorizes households as either food-secure or food-insecure; food-insecure households are then divided into low food security and very low food security, and there are additional notations regarding the experience of children (Coleman-Jensen et al., 2016). In contrast, the Radimer/Cornell food-insecurity measure (RCFIM; Kendall, Olson, & Frongillo, Jr., 1995; Radimer, Olson, Greene, Campbell, & Habicht, 1992) is designed to classify people according to the Radimer continuum of FI, which consists of four groups that include a food-secure group and three levels of FI; the most severe level including households with children who do not have enough to eat (child-hunger household, or CHH). We used the RCFIM because it is shorter, which reduced the response burden of participants. Sample statements, which are rated on a Likert scale (0 = not true, 1 = sometimes true, 2 = always true), include “The food that I bought didn’t last and I didn’t have money to buy more” and “I know my child(ren) are hungry sometimes, but I can’t afford more food.” If a participant chooses not true for all items, they are considered food-secure according to the Radimer continuum, and if they identify any item as true then they are considered food-insecure. To determine the degree of FI, we divided RCFIM questions into three clusters that match the three conceptual levels of FI, as described in the standard scoring instructions. Participants who select sometimes true or always true for any statement in the lowest-severity cluster of questions (e.g., repeatedly eating the same thing secondary to lack of resources) and not true on all of the statements in the highest clusters (e.g., reporting going hungry because of food scarcity) are classified in the lowest level of FI: household FI. Participants in this category report being anxious about having sufficient food or are eating the same thing repeatedly because they cannot afford to eat a more diverse diet.

Those who endorse some level of true on any higher-severity questions but not true for child-hunger questions (e.g., reporting children going hungry) are designated food-insecure at the middle level: individual FI. Participants in this food-insecure group report that they themselves are not eating sufficiently because of a lack of resources. If participants endorse any level of true for questions regarding child hunger, they are categorized as CHH FI. The rationale in designating CHH as the most severe level of FI is based on the presumption that most adults will do their best to shield children from hunger, yet adult participants reported that they could not feed their children sufficient food. Thus, if there are hungry children at home, presumably the adults are hungrier.

To be consistent with our previous work, we labeled the least severe group as “not food-insecure” (NFI), as opposed to food-secure, because these individuals sought services from the SAFB, which suggests that they are living on the margins of food security and FI. One possible reason for this is because they collectively have slightly more income compared with the other groups in our sample; 41% of the NFI group reported an annual income of less than $10,000, which is lower than the other groups (54.1% to 61.8%). Note, however, that two thirds of the NFI group earned $20,000 or less annually, and approximately 60% of the respondents had one or more children living at home; the federal poverty level in 2017 for a three-person household was $20,420. Per the RCFIM, the sample was distributed as follows: NFI (n = 86), household FI (n = 364), individual FI (n = 246), and CHH FI (n = 192).

Further, the NFI group (M = 1.88, SD = 1.50) reported a similar number of monthly visits to food pantries to obtain food compared with both the household FI (M = 1.93, SD = 1.42) and individual FI (M = 1.93, SD = 1.28) groups. The CHH FI group reported a somewhat more elevated rate of food-pantry visits (M = 2.34, SD = 1.52). Both the internal consistency and the construct and criterion-related validity of the RCFIM have been supported (Kendall et al., 1995). We found excellent internal consistency within our sample (Cronbach’s α = .941). Table 1 provides demographics for each subgroup.

Assessment of psychopathology

Participants completed the self-report Eating Disorder Diagnostic Scale 5 (EDDS–5), which is an updated version of the EDDS that assesses core ED diagnoses in the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM–5; American Psychiatric Association, 2013). Research with the EDDS-4, which was designed to assess the core ED diagnoses in the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM–IV) supports its internal consistency, convergent validity with ED risk factors, criterion validity with interview-based diagnoses, sensitivity to change, and predictive validity (Stice, Fisher, & Martinez, 2004). However, we used the updated EDDS-5 because it assesses night eating, which is a form of overeating, and because K. M. Middlemass et al. (2018), who has extensive experience with marginalized, low-income populations, thought the EDDS-5 was somewhat easier to understand. Because the night-eating question was complicated and the education levels reported in Becker et al. (2017) were low, we divided this item into two simple questions to help participants with comprehension. To score EDDS, we computed a standardized summed composite score per Stice et al. (2004) and Krabbenborg et al. (2012). In the present sample, EDDS items had excellent internal consistency (α = .906).

We used three items from the dietary-restraint subscale of the EDE-Q (Fairburn, 2008). The three items assessed deliberately trying to limit the amount of food eaten, going long periods (8 or more waking hr) without eating, and trying to exclude liked foods from the diet. To assess dietary restraint, the EDE-Q uses a 7-point Likert scale (0 = no days to 6 = every day) to measure dietary restraint over the past 28 days. The three items included in this study demonstrated acceptable internal consistency (α = .725). As described earlier, we eliminated the statement specifying that dietary restraint was for weight and shape concerns.

To assess depression, we used the eight-item Patient Health Questionnaire (PHQ-8; Razykov, Ziegelstein, Whooley, & Thombs, 2012). The PHQ-8 is identical to the nine-item Patient Health Questionnaire (PHQ-9), which maps onto the nine core symptoms of depression in DSM-5, except that the PHQ-8 excludes the question about passive thoughts of death or hurting oneself. The PHQ-9 was designed as a brief screening instrument for medical settings; research supports both criterion validity and construct validity (Kroenke, Spitzer, & Williams, 2001). Research also indicates that the PHQ-8 and PHQ-9 are highly correlated (r = .997) and have similar sensitivity and specificity (Razykov et al., 2012). Internal consistency in this sample was excellent (α = .939).

For anxiety, participants completed the seven-item Generalized Anxiety Disorder questionnaire (GAD-7), which was developed as a brief screening measure for generalized anxiety disorder, one of the most common types of anxiety (Spitzer, Kroenke, Williams, & Löwe, 2006). Research indicates that the GAD-7 has good criterion and construct validity and good reliability (Spitzer et al., 2006). Research also supports its validity and reliability in the general population (Löwe et al., 2008). In the present sample, internal consistency was excellent (α = .947).

Modification of measures

Because individuals with FI often have lower education levels, we reviewed all items using the Microsoft Word proofing tool, which provides a Flesch-Kincaid Grade Level. We then altered each question on the basis of the results and as needed to meet a sixth-grade reading level. As we previously considered this issue in choosing the questions for the survey, many questions were unchanged. Because of the prevalence of Spanish speakers in San Antonio, all measures needed to be translated into Spanish. During the original Becker et al. (2017) study, we translated the full packet of measures into Spanish and then had them back-translated by an independent bilingual consultant to ensure the meaning of each question remained the same in English and Spanish. The Spanish version of the overall survey was then reviewed by a second bilingual consultant who grew up in San Antonio. During this process, we made a few minor changes to reflect San Antonio–based Spanish dialect, which was described to us by local speakers as a derivative of Mexican Spanish. This consultant provided feedback regarding any confusing constructs and made suggestions. The questionnaire was then piloted with both local English and Spanish speakers; during the pilot phase, participants were asked to highlight any concepts or questions they found confusing. Both versions were further refined on the basis of the feedback that we received. As a final check, one of the authors of Becker et al. (2017) subsequently walked through both the Spanish and English questionnaire with a bilingual speaker from San Antonio whose first language was Spanish and whose highest level of education was first grade. A few minor changes were made in response to this additional assessment to ensure accessibility. Copies of the adjusted measures are available upon request.

Approximately 90% of the Latinx population in San Antonio is of Mexican descent (Rentiera, 2011). The present study included two measures, the PHQ-9 and GAD-7, that were not included in Becker et al. (2017). Both measures have Mexican Spanish versions (available at https://www.phqscreeners.com/select-screener/36); these measures were used in the present study. There is a Spanish version of the EDDS-4, but it was developed by Chilean researchers and piloted in Chile. Because Chilean Spanish differs from Mexican Spanish, we used our initial translation, which had been reviewed by multiple local Spanish speakers of Mexican descent and because no problems arose when it was originally implemented for our 2017 study.

Data analyses

Preliminary analyses revealed a significant difference in age between the four food-insecure groups, F(3, 871) = 9.523, p = .0001, η_p² = .032. Participants in the individual FI group (M = 38.53, SD = 12.70) and CHH FI group (M = 41.16, SD = 11.44) were younger than those in the NFI FI group (M = 44.86, SD = 17.59) and household FI group (M = 44.34, SD = 15.40). Thus, age was covaried in subsequent analyses. No differences between groups emerged on the basis of the self-reported BMI.

For the main analyses, we ran four planned one-way analyses of covariance, with age as the covariate, to test hypotheses regarding the degree to which ED pathology, any-reason dietary restraint, depressed mood, and anxiety worsened as FI increased. These analyses tested clear a priori hypotheses; thus, we did not adjust for multiple tests. Post hoc tests were conducted only if we found a significant omnibus test. Further, we investigated whether any-reason dietary restraint correlated with ED pathology in this sample so as to potentially replicate findings from Middlemass et al. (2018).

Those living with FI are from marginalized populations and have been understudied by ED researchers specifically and mental-health researchers more generally. To assess and help FI individuals in culturally appropriate ways, a critical role for researchers is to facilitate partnerships with community stakeholders, such as nonprofit organizations and policymakers; it is important to communicate in a way that is understandable to everyone involved. Frequencies are easier to comprehend than means, and this is particularly true in the case of communicating the degree to which pathological behaviors are sufficiently common to be a public-health problem (Becker, 2017; Fiske, Fallon, Blissmer, & Redding, 2014). Thus, we include descriptive data (i.e., frequencies) for those who met the clinically significant cutoff for ED pathology, depression, and anxiety and to test our hypotheses regarding lack of gender or ethnicity and race differences using χ2 tests. We also provide frequencies of specific ED behaviors/symptoms given the unexpected finding in Becker et al. (2017) that FI severity was associated with increased compensatory behaviors and elevated weight/shape concerns in addition to increased binge eating. Note that we did not conduct statistical analyses on symptom-frequency data secondary to concerns about multiple tests but rather observed whether the data were consistent with the hypothesized linear trend.

Results

Group differences in mean scores and ED pathology/dietary restraint correlation

As predicted in our hypotheses, ED pathology significantly differed between the four food-insecure groups, F(3, 791) = 6.22, p = .0001, η_p² = .023. Post hoc analyses indicated that the CHH FI group reported increased ED pathology relative to the other three FI groups (ps < .02; η_p² range = .015 to .067; see Table 2 for means and standard deviations). In contrast, NFI participants reported significantly lower ED pathology relative to the other three groups (ps < .02; η_p² range = .014 to .067). No significant difference in ED pathology emerged between individual and household FI groups.

Table 2.

Means and Standard Deviations for Age, Eating Disorder Pathology, Dietary Restraint, Depressive Mood, and Anxiety

Measures	NFI	Household food insecurity	Individual food insecurity	CHH food insecurity
Age	44.86 (17.59)_a	44.34 (15.40)_a	38.53 (12.70)_b	41.16 (11.44)_c
EDDS	−3.87 (9.99)_a	−0.54 (11.63)_b	0.03 (10.56)_b	2.68 (12.45)_c
EDE-Q DR	0.61 (1.23)_a	1.32 (1.56)_b	1.23 (1.53)_b	1.87 (1.58)_c
PHQ-8	3.38 (5.39)_a	6.38 (6.74)_{b, c}	5.99 (6.11)_b	7.32 (6.59)_c
GAD-7	3.72 (5.30)_a	8.05 (7.06)_b	7.58 (6.25)_b	9.49 (6.75)_c

Note: Within each row, values with different subscripts are significantly different. Some participants did not complete enough items to score the full Eating Disorder Diagnostic Scale (EDDS), eight-item Patient Health Questionnaire (PHQ-8), and/or seven-item Generalized Anxiety Disorder questionnaire (GAD-7) or did not complete the three Eating Disorder Examination Questionnaire (EDE-Q) dietary restraint items. These participants were excluded from this table and corresponding analyses for each measure for which this occurred. Scores on the EDDS are standardized. Significance levels for each item are as follows—age: child-hunger household (CHH) versus all other groups, p < .05, and individual versus household and not food-insecure (NFI), p < .0001; EDDS: CHH versus all other groups, p < .02, and NFI versus all other groups, p < .02; EDEQ-DR: CHH versus all other groups, p < .0001, and NFI versus all other groups, p < .001; PHQ-8: CHH versus individual and NFI, p < .05, and NFI versus all other groups, p < .0001; and GAD-7: CHH versus all other groups, p < .02, and NFI versus all other groups, p < .0001.

A similar pattern of overall significant differences between the four groups also emerged for any-reason dietary restraint, F(3, 801) = 14.21, p = .0001, η_p² = .051; depression, F(3, 848) = 7.54, p = .0001, η_p² = .026; and anxiety, F(3, 825) = 15.31, p = .0001, η_p² = .053. Post hoc tests largely supported the same pattern as was found for ED pathology. For dietary restraint, the CHH FI group scored significantly higher relative to the other three groups (ps < .0001; η_p² range = .029 to .138), whereas the NFI group scored significantly lower (ps < .001; η_p² range = .037 to .138). Once again, individual and household FI groups did not significantly differ. With regard to depression, the CHH FI group differed from the individual FI and NFI groups (p values < .05; η_p² range = .010 to .086) but not the household FI group. The NFI group differed from the other three groups (ps < .0001; η_p² range = .033 to .086) on depression, consistent with ED pathology and dietary restraint, but no difference emerged between the individual and household FI groups. Finally, for anxiety, post hoc tests showed that the CHH FI group reported significantly increased anxiety relative to the other three groups (ps < .02; η_p² range = .011 to .165), whereas the NFI group had significantly lower anxiety relative to the other three groups (ps < .0001; η_p² range = .064 to .165). Household and individual FI groups again did not differ from one another on the level of anxiety.

According to the RCFIM, the most severe level of FI for adults can occur only if there are children in the household. Thus, it could be argued that results would differ if analyses included only participants who had children in the household. However, when the sample was limited to the 493 participants with children and the main analyses rerun, results were fundamentally unchanged. All omnibus tests remained highly significant. Further, of the 24 post hoc comparisons, only 3 changed: One comparison became nonsignificant and two became significant. In all three cases, effect sizes remained within the ranges reported above. As a result, we report results for the full sample.

We hypothesized that dietary restraint would be significantly correlated with ED pathology in this food-insecure sample. As with Becker et al. (2017), we assessed dietary restraint for any reason as opposed to solely for weight and shape concerns. Dietary restraint was correlated with ED pathology, even when controlling for the level of FI (R_s = .24, n = 739, p = .0001), which was consistent with our hypothesis.

Frequency of clinically significant ED pathology, depression, and anxiety

To be consistent with Becker et al. (2017), we used a cutoff of > 16.5 on the EDDS composite score (following Krabbenborg et al., 2012) to identify those with clinically significant EDs. The use of this cutoff in Becker et al. (2017) produced two groups with a mean difference (29.30) similar to that found in Krabbenborg et al. (28.29); the same was true with this sample (28.76). For depression and anxiety, a cutoff of greater than 10 identified those at a probable clinically significant level (Razykov et al., 2012; Spitzer et al., 2006). Overall, 10.6% of the total sample met criteria for clinically significant ED pathology, 24.8% met criteria for clinically significant depressed mood, and 34.3% met criteria for clinically significant anxiety. Table 3 provides the percentages of each food-insecure group that met criteria for clinically significant ED pathology, depression, and anxiety. Given the stereotype that EDs are a problem predominantly for White women and girls, and given the marginalized nature of this sample and the limited information generally on the mental health of those living with FI, the data are broken out by gender, ethnicity, and race for each food group. Information for depression and anxiety also is included.

Table 3.

Frequencies of Clinically Significant Eating Disorder Pathology, Depression, and Anxiety

Characteristics	Not food-insecure	Household food insecurity	Individual food insecurity	CHH food insecurity
Clinically significant eating disorder^a	Availablen = 78	Availablen = 327	Availablen = 223	Availablen = 174
Total sample, N (%)	5 (6.4)	37 (11.3)	14 (6.3)	29 (16.7)
Male	3 (10.3)	12 (10.7)	2 (3.0)	7 (12.7)
Female	2 (4.1)	25 (11.6)	12 (7.7)	22 (18.5)
Latinx/Hispanic	3 (5.2)	27 (11.3)	12 (7.0)	24 (16.9)
White	0 (0)	5 (12.5)	1 (5.6)	2 (20.0)
Black	0 (0)	3 (15.0)	0 (0.0)	2 (22.2)
Other	2 (33.3)	2 (8.3)	1 (5.6)	1 (7.7)
Clinically significant depression^b	Availablen = 82	Availablen = 349	Availablen = 241	Availablen = 187
Total sample, N (%)	7 (8.5)	92 (26.4)	54 (22.4)	60 (32.1)
Male	6 (20.0)	37 (30.6)	13 (18.8)	14 (23.3)
Female	1 (1.8)	55 (24.1)	41 (24.0)	46 (36.2)
Latinx/Hispanic	5 (8.2)	63 (24.7)	40 (21.9)	49 (32.7)
White	0 (0)	15 (34.9)	5 (20.8)	3 (27.3)
Black	0 (0)	5 (25.0)	3 (21.4)	5 (50.0)
Other	2 (33.3)	9 (32.1)	6 (30.0)	3 (20.0)
Clinically significant anxiety^c	Availablen = 85	Availablen = 343	Availablen = 232	Availablen = 178
Total sample, N (%)	10 (11.8)	124 (36.2)	73 (31.5)^d	81 (45.5)
Male	6 (19.4)	49 (41.2)	25 (37.3)	21 (36.8)
Female	4 (7.4)	75 (33.6)	47 (28.7)	60 (49.6)
Latino/Hispanic	7 (11.1)	82 (32.0)	53 (29.8)	69 (47.6)
White	1 (12.5)	18 (43.9)	8 (36.4)	4 (36.4)
Black	0 (0)	10 (52.6)	6 (46.2)	3 (42.9)
Other	2 (33.3)	14 (58.3)	6 (33.3)	5 (35.7)

Note: The available n values are total/composite scores for each measure within each food-insecurity category for the total sample; valid percentages (in parentheses) were calculated on the basis of the total number of scores for the main dependent variable within each food-insecurity category for the total sample; the gender and ethnicity/race sample percentages were calculated as valid percentages of scores within the food-insecure group and gender or ethnicity/race. CHH = child-hunger household.

Participants whose standardized cutoffs was > 16.5 on the Eating Disorder Diagnostic Scale. ^bParticipants who scored ≥ 10 on the eight-item Patient Health Questionnaire. ^cParticipants who scored ≥ 10 on the seven-item Generalized Anxiety Disorder questionnaire. ^dOne participant who met the criteria for generalized anxiety disorder did not list a gender; thus, male and female combined = 72.

Consistent with our hypotheses, the percentage of those scoring above 16.5 on the EDDS composite score in the total sample generally increased as FI increased; 16.7% of those in the CHH FI group met criteria for clinically significant ED pathology. Also consistent with our hypotheses, rates of those scoring above the ED cutoff did not significantly differ between male and female participants in either the total sample, χ²(1, N = 882) = .458, p = .498, or the CHH FI group, χ²(1, N = 174) = .899, p = .343. We found a similar pattern for depression and anxiety; the highest levels of both emerged in the CHH FI group. To better understand the relationship between gender and depression and anxiety, we conducted exploratory analyses in the total sample only. No gender frequency differences emerged for either depression, χ²(1, N = 859) = .009, p = .923, or anxiety, χ²(1, N = 174) = 1.20, p = .274.

With regard to race and ethnicity, as hypothesized, there were no differences in the frequency of clinically significant ED pathology in either the total sample, χ²(3, N = 713) = .061, p = .996, or the CHH FI group, χ²(3, N = 174) = 1.04, p = .792. On an exploratory basis, we tested for differences in levels of clinically significant depression, χ²(3, N = 855) = .998, p = .802, and anxiety by ethnic and racial group, χ²(3, N = 833) = 4.39, p = .223, within the total sample; once again there were no significant differences.

Frequencies of ED behaviors and weight/shape concerns

Table 4 presents frequencies of objective overeating (i.e., eating a large amount of food without a sense of loss of control), objective binge eating (i.e., eating a large amount of food with a sense of loss of control), and night eating (i.e., eating after waking from sleep) with distress. In addition to overeating and binge-eating behaviors, the EDDS specifically asks about compensatory behaviors that aim to “prevent weight gain or counteract the effects of eating” and includes a question assessing shape/weight concerns in language that is similar to the EDE-Q (Fairburn, 2008). A score of 4 or greater on the EDE is considered clinically significant for weight/shape concerns, and both the EDDS and the EDE-Q use a 7-point measure that is scaled similarly. Table 4 includes the following compensatory behaviors: vomiting, laxative/diuretic use, skipping two meals in a row, and exercising harder than usual because of overeating, as well as the frequency of participants endorsing 4 or greater on weight and shape importance.

Table 4.

Frequencies of Overeating and Eating Disorder Symptoms

Symptoms	Not food-insecure	Household food insecurity	Individual food insecurity	CHH food insecurity
Objective overeating	Availablen = 85	Available n = 352	Available n = 243	Available n = 186
	10 (11.8%)	68 (19.3%)	40 (16.3%)	41 (22.0%)
Objective binge eating	Available n = 84	Available n = 350	Available n = 240	Available n = 188
	10 (11.9%)	47 (13.4%)	39 (16.3%)	50 (26.6%)
Night eating with distress	Available n = 82	Available n = 350	Available n = 238	Available n = 187
	8 (9.8%)	72 (20.6%)	62 (26.1%)	857 (30.5%)
Vomiting^a	Available n = 84	Available n = 352	Available n = 242	Available n = 186
	8 (9.5%)	29 (8.2%)	18 (7.4%)	32 (17.2%)
Laxatives/water pills^a	Available n = 84	Available n = 348	Available n = 242	Available n = 187
	7 (8.3%)	31 (8.9%)	18 (7.4%)	23 (12.3%)
Skipped at least two consecutive meals^a	Available n = 82	Available n = 352	Available n = 242	Available n = 185
	13 (15.9%)	127 (36.1%)	106 (43.8%)	102 (55.1%)
Exercised harder than usual because of excessive eating^a	Available n = 83	Available n = 350	Available n = 240	Available n = 187
	72 (20.6%)	72 (20.6%)	65 (27.1%)	66 (35.3%)
Weight and shape concerns score ≥ 4	Available n = 84	Available n = 354	Available n = 243	Available n = 186
	10 (11.9%)	74 (20.9%)	63 (25.9%)	52 (28.0%)

Note: The available n values are the available responses for each food-insecurity group; valid percentages were calculated on the basis of the total number of responses for each dependent variable in each food insecurity-category. CHH = child-hunger household.

To prevent weight gain or counteract the effects of eating.

As hypothesized, we observed a linear relationship between frequencies of reported overeating behaviors; for instance, the NFI group reported the lowest levels of objective overeating, objective binge eating, and night eating with distress, whereas the CHH FI group reported the highest levels. The other two groups fell in between these extremes. Consistent with our hypotheses, the highest level of all compensatory behaviors was reported by the CHH FI group. Contrary to our hypotheses, the other FI groups and the NFI group reported similar levels of compensatory vomiting, laxative/diuretic use, and exercising harder. With regard to weight and shape concerns, consistent with our hypotheses, the percentage of those scoring 4 or greater was lowest in the NFI group and most elevated in the CHH FI group.

Discussion

The primary aim of the present study was to determine whether we could replicate findings from a previous study investigating the association between FI and EDs. Second, we sought to address the call to action by Kazdin and Blase (2011) and others to document the types of problems faced by marginalized communities, particularly those who largely have been understudied (and largely ignored) by mental-health researchers. Both aims are critical.

Results supported the overwhelming majority of our ED hypotheses, and we were able to replicate key results from Becker et al. (2017). More specifically, as hypothesized, those in the CHH FI group reported significantly higher levels of ED pathology compared with the other FI groups, whereas those living on the margins of FI and food security (the NFI group) endorsed the lowest level of ED pathology (Table 1). Further, 16.7% of the CHH FI group crossed the cutoff for a current and clinically significant ED. This closely compares to the 17% previously found in Becker et al. (2017). Thus, two studies now support the argument that severe FI may be associated with elevated levels of ED pathology. Further, rates of clinically significant EDs in the CHH FI group markedly exceeded those found in previous community-based samples in the United States (e.g., 7% 12-month prevalence in Hudson et al., 2007). In addition, as found in Becker et al. (2017), rates of clinically significant ED pathology did not differ by gender or ethnicity and race.

It should be noted that the use of self-report measures, which are often required when conducting unfunded research in a novel area, may lead to inflated reporting of symptoms relative to interviewer-based assessments, such as those used in Hudson et al. (2007). Thus, results of this study need to be confirmed with interviewer-based measures. However, results do support the importance of conducting such follow-up studies with food-insecure individuals, a population that has largely been ignored by the ED field.

Results indicated that elevated ED pathology scores were not exclusively driven by binge-eating behaviors, consistent with Becker et al. (2017). Indeed, 17.2% of the CHH FI group in the present study reported engaging in self-induced vomiting to control weight or offset the effects of eating; this is relatively similar to the 20.4% rate found in Becker et al. (2017). With the exception of laxative use (12.3% in the current study vs. 22.8% in Becker et al., 2017), rates of both compensatory behaviors and elevated weight/shape concerns were remarkably consistent in the CHH FI group across both studies. Thus, we were also able to replicate a key unexpected finding of Becker et al. (i.e., elevated rates of compensatory behaviors and weight and shape concerns in CHH FI group).

In conversations with researchers who specialize in FI (e.g., H. K. Seligman, personal communication, May 9, 2018), one question that has emerged is whether behaviors that seem to be consistent with ED pathology are simply normative coping responses within a U.S.-based urban food-insecure population. This does not seem to be the case given that almost 90% of the total sample did not meet criteria for a clinically significant ED, and most participants within each food group did not endorse ED behaviors, such as binge eating, night eating, and vomiting. Thus, FI cannot be equated with an ED; the overwhelming majority of participants in both studies who are living with FI do not seem to report experiencing elevated ED pathology. Rather, findings indicate that a distinct subgroup of individuals living with FI may develop an ED and that severe FI may be a potent stressor that triggers the onset of EDs in vulnerable individuals, leading to elevated rates compared with community norms.

It is important to note that because the present study was cross-sectional it is impossible to determine whether FI is, in fact, a prospective ED risk factor; longitudinal research is needed to tease apart the nature of the association. Longitudinal and qualitative studies are also needed to determine the degree to which existing explanatory models of the development and maintenance of EDs apply to or need to be altered to address the experiences of those living with FI (e.g., transdiagnostic model of EDs; Fairburn, Cooper, & Shafran, 2003). For instance, results from Middlemass et al. (2018), as noted above, suggest that those living with FI restrict their food intake to stretch food to make it last, to save food for children, or because they do not have sufficient resources to buy sufficient food. Thus, models may need to be expanded to account for intentional dietary restriction, which once again correlated with ED pathology in this study, when it is undertaken for economic reasons versus weight and shape concerns.

The present study supports the contention that FI severity is correlated with ED pathology. Further research is needed to determine whether FI serves as a prospective risk factor for the development of an ED, given that there are multiple plausible pathways (e.g., increased intentional dietary restraint, easy access to HPF, magnified economic stress secondary to FI) that might drive such a relationship. The present study provides additional evidence that the thin, White, affluent, female ED stereotype offers a flawed portrait of who is and is not at risk for EDs. Future research needs to determine the degree to which findings generated from research with those who largely match the stereotype also apply to those who differ from the stereotype in significant ways.

Results of the present study also supported our hypotheses that FI severity would be associated with a worsening of both depression and anxiety; indeed, 32.1% and 45.5% of the CHH FI group seem to be contending with clinically significant depression and anxiety, respectively. In contrast to EDs, a small but significant body of research does exist on the association of FI with depression and anxiety. Consistent with this study, previous research supports an association between FI and depression and anxiety (Kleinman et al., 1998; Palar et al., 2015; Sorsdahl et al., 2010; Weaver & Hadley, 2009). For instance, Alaimo, Olson, and Frongillo (2002) found that food-sufficient adolescents reported a lower frequency of major depression (5.9%) compared with food-insecure adolescents (12.2%). Likewise, in two related studies, Hadley and Patil (2006, 2008) identified that FI was associated with both depression and anxiety symptomatology in Tanzanian participants.

It is important to note that, as with EDs, the etiology underlying the association between FI and anxiety and depression remains unknown. Much of the existing data, including results from the present study, are correlational, which limits our ability to draw conclusions about causation. Some limited research, however, does support the argument that FI may be a risk factor for depression and anxiety (Palar et al., 2015). For instance, despite a lack of prestarvation depression, volunteers in the Keys et al. (1950) study experienced significant increases in depression (Tucker, 2006), whereas men who were not deprived of food yet were living in the same study conditions did not develop depression. Further, Hadley and Patil (2008) found that changes in FI, dependent on season, predicted changes in anxiety and depression symptoms. Nonetheless, as with EDs, not everyone who is food-insecure develops anxiety and depression. Thus, it is important for future research to identify additional risk factors that interact with FI to trigger clinically significant levels of anxiety and depression.

A critical feature of the present study is the fact that participants were impoverished; 56.2% of the total sample earned less than $10,000 per year and an additional 24.7% earned less than $20,000 per year. Poverty and mental illness are interconnected; poverty increases the risk for mental illness, and mental illness increases the odds that someone will stay or become impoverished (Lund et al., 2011). Studies conducted in low- and middle-income countries suggest that mental-health interventions positively affect economic status (Lund et al., 2011). Given that more than 4 decades of research support the contention that impoverished people are both interested in psychotherapy and can benefit from it (Smith, 2005; Goodman et al., 2013), it is imperative that increased attention is focused on fully understanding the psychological impact of FI so that scalable interventions can be developed to help this understudied and marginalized population.

People living with FI want help. Indeed, during both the collection of data for this study and for Becker et al. (2017), RAs were repeatedly surprised by expressions of gratitude from participants who reported being shocked that anyone cared enough to ask about their experiences; we must ask about their experiences. We cannot presume that interventions developed for food-secure populations can be imported wholesale for food-insecure populations. Indeed, a common component of ED treatment is to establish a regular pattern of eating; this obviously becomes more challenging when people simply cannot afford to eat regularly. At the same time, an untreated ED may worsen the experience of FI because out-of-control binge eating may hamper a person’s ability to ration food throughout the month. Untreated anxiety and depression may also negatively affect a person’s ability to cope with the chronic stress of FI.

Despite the importance of addressing mental-health concerns in marginalized populations, research indicates that poor communities are less likely to seek treatment for mental illness and are more likely to stop treatment early. However, positive outcomes improve when psychological interventions are tailored to attend to contextual stressors associated with poverty (Goodman et al., 2013). Research by V. Patel et al. (2010) indicates that empirically supported behavioral interventions can be both tailored for impoverished communities and task-shifted to lay providers to reduce the cost of scaling, but tailoring is critical (Smith, 2010). Further, it will require clinical psychology to both better understand the lived experience of those who are forced to contend with FI and to adopt a social-justice perspective that recognizes that poverty is not limited to material deprivation but also includes social-capital deprivation (Smith, 2010).

This study has several limitations. First, although we back-translated Spanish translations to ensure that the English and Spanish versions were comparable, we did not have the resources to conduct a validity study of the measures in Spanish and did not use a specific cultural framework to structure translations of the measures (e.g., Alegria et al., 2004). Second, to address the typical low educational levels of this population, we modified some questions in our measures to make them understandable. Although this is both common and necessary when conducting research with poorly educated, marginalized populations, it is less than optimal from a psychometric perspective. Third, this study is cross-sectional and cannot determine causation. As noted above, both longitudinal and qualitative studies are needed to elucidate the relationships identified and replicated in this study. Fourth, although we collected self-reported weight and height data to calculate BMI, there are good reasons to think that participants may not have an accurate perception of their weight, given that both visiting a doctor and scales are luxuries for low-income populations who do not have enough food. For this reason, we did not conduct analyses on BMI beyond checking to see whether significant differences existed across groups. Fifth, this study recruited only participants who were food-insecure or living on the margins of FI. Thus, participants did not represent the full continuum of food security to FI. Not surprisingly, the number of participants in the lowest level of FI represented only a small percentage of the total sample. Although we more than doubled the number in the NFI group compared with our first study, they still remain a distinct minority of the sample, which is a limitation.

The sample in the present study was largely Latinx/Hispanic. Given the historic focus on White individuals in the ED literature, we argue this is a strength of the present study. However, it also is a weakness because it is unclear to what degree findings from the present study will generalize to other ethnic and racial populations living with FI.

In conclusion, the present study replicates previous research showing that higher levels of FI are associated with increased levels of ED pathology, depression, and anxiety. We add our voices to previous calls for increased research with marginalized, understudied populations so that their lived experiences can be understood and used to tailor existing behavioral interventions to meet their needs. Only then will our field be able to fully respond to calls by Kazdin and Blase (2011) and others (Kazdin, 2017; Smith, 2010) to address the treatment gap and provide psychological services for those who are currently underserved.

Footnotes

Acknowledgements

We thank the San Antonio Food Bank for their assistance in this research; the H-E-B grocery company, McNairs Scholars Program, and Murchison Summer Undergraduate Research Fellowship Program for financial support; and Paola Gutierrez, Calista Struby, Autumn Sutherland, and Eden White for their contributions to this study.

Action Editor

Kelly L. Klump served as action editor for this article.

Author Contributions

F. Gomez and A. Martinez-Abrego collected, entered, and checked the data. C. B. Becker analyzed the data. All of the authors designed and conceived the study, wrote and revised the manuscript, and approved the final manuscript for submission.

Declaration of Conflicting Interests

The author(s) declared that there were no conflicts of interest with respect to the authorship or the publication of this article.

References

Adams

E. J.

Grummer-Strawn

Chavez

(2003). Food insecurity is associated with increased risk of obesity in California women. The Journal of Nutrition, 133, 1070–1074.

Alaimo

Olson

C. M.

Frongillo

E. A.

(2002). Family food insufficiency, but not low family income, is positively associated with dysthymia and suicide symptoms in adolescents. The Journal of Nutrition, 132, 719–725.

Alegria

Vila

Woo

Canino

Takeuchi

Vera

. . . Shrout

(2004). Cultural relevance and equivalence in the NLAAS instrument: Integrating etic and emic in the development of cross-cultural measures for a psychiatric epidemiology and services study of Latinos. International Journal of Methods in Psychiatric Research, 13, 270–288.

American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author.

American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Washington, DC: Author.

Arcelus

Mitchell

Wales

Nielsen

(2011). Mortality rates in patients with anorexia nervosa and other eating disorders: A meta-analysis of 36 studies. Archives of General Psychiatry, 68, 724–731. doi:10.1001/archgenpsychiatry.2011.74

Becker

C. B.

(2017). From efficacy to global impact: Lessons learned about what-not-to-do in translating our research to reach. Behavior Therapy, 48, 718–730.

Becker

C. B.

Middlemass

Taylor

Johnson

Gomez

(2017). Food insecurity and eating disorder pathology. International Journal of Eating Disorders, 50, 1031–1040.

Boggiano

M. M.

Artiga

A. I.

Pritchett

C. E.

Chandler-Laney

P. C.

Smith

M. L.

Eldridge

A. J.

(2007). High intake of palatable food predicts binge-eating independent of susceptibility to obesity: An animal model of lean vs obese binge-eating and obesity with and without binge-eating. International Journal of Obesity, 31, 1357–1367. doi:10.1038/sj.ijo.0803614

10.

Bove

C. F.

Olson

C. M.

(2006). Obesity in low-income rural women: Qualitative insights about physical activity and eating patterns. Women & Health, 44, 57–78.

11.

Bruch

(1973). Evolution of a psychotherapeutic approach to eating disorders: Obesity, anorexia nervosa, and the person within. New York, NY: Basic Books.

12.

Bruening

MacLehose

Loth

Story

Neumark-Sztainer

(2012). Feeding a family in a recession: Food insecurity among Minnesota parents. American Journal of Public Health, 102, 520–526. doi:10.2105/AJPH.2011.300390

13.

Carta

M. G.

Preti

Moro

M. F.

Aguglia

Balestrieri

Caraci

. . . Hardoy

M. C.

(2014). Eating disorders as a public health issue: Prevalence and attributable impairment of quality of life in an Italian community sample. International Review of Psychiatry, 26, 486–492.

14.

Chamay-Weber

Narring

Michaud

P. A.

(2005). Partial eating disorders among adolescents: A review. Journal of Adolescent Health, 37, 416–426.

15.

Coleman-Jensen

Rabbitt

Gregory

Singh

(2016). Household food security in the United States in 2015 (Research Report No. ERR-215). Washington, DC: U.S. Department of Agriculture. Retrieved from https://www.ers.usda.gov/webdocs/publications/79761/err-215.pdf

16.

Coleman-Jensen

Rabbitt

Gregory

Singh

(2017). Household food security in the United States in 2016 (Research Report No. ERR-237). Washington, DC: U.S. Department of Agriculture. Retrieved from https://www.ers.usda.gov/webdocs/publications/84973/err-237.pdf

17.

Dutko

(2012). Food deserts suffer persistent socioeconomic disadvantage. Choices, 27(3). Retrieved from http://www.jstor.org/stable/choices.27.3.02

18.

Fairburn

C. G.

(2008). Cognitive behavior therapy and eating disorders. New York, NY: Guilford Press.

19.

Fairburn

C. G.

Cooper

Shafran

(2003). Cognitive behaviour therapy for eating disorders: A “transdiagnostic” theory and treatment. Behaviour Research and Therapy, 41, 509–528.

20.

Feeding Texas. (2015, April 17). National study maps food insecurity in every Texas County. Retrieved from https://www.feedingtexas.org/press/2015/04/National-Study-Maps-Food-Insecurity-In-Every-Texas-County

21.

Fiske

Fallon

E. A.

Blissmer

Redding

C. A.

(2014). Prevalence of body dissatisfaction among United States adults: Review and recommendations for future research. Eating Behaviors, 15, 357–365.

22.

Goodman

L. A.

Pugach

Skolnik

Smith

(2013). Poverty and mental health practice: Within and beyond the 50-minute hour. Journal of Clinical Psychology, 69, 182–190.

23.

Gordon

K. H.

Brattole

M. M.

Wingate

L. R. R.

Joiner

T. E.

(2006). The impact of client race on clinician detection of eating disorders. Behavior Therapy, 37, 319–325.

24.

Hadley

Patil

C. L.

(2006). Food insecurity in rural Tanzania is associated with maternal anxiety and depression. American Journal of Human Biology, 18, 359–368.

25.

Hadley

Patil

C. L.

(2008). Seasonal changes in household food insecurity and symptoms of anxiety and depression. American Journal of Physical Anthropology, 135, 225–232.

26.

Hart

L. M.

Granillo

M. T.

Jorm

A. F.

Paxton

S. J.

(2011). Unmet need for treatment in the eating disorders: A systematic review of eating disorder specific treatment seeking among community cases. Clinical Psychology Review, 31, 727–735.

27.

Herman

C. P.

Polivy

(1975). Anxiety, restraint, and eating behavior. Journal of Abnormal Psychology, 84, 66–72.

28.

Hudson

J. I.

Hiripi

Pope

H. G.

Kessler

R.C.

(2007). The prevalence and correlates of eating disorders in the National Comorbidity Survey Replication. Biological Psychiatry, 61, 348–358.

29.

Kärkkäinen

Mustelin

Raevuori

Kaprio

Keski-Rahkonen

(2018). Do disordered eating behaviours have long-term health-related consequences? European Eating Disorders Review, 26, 22–28.

30.

Kato

Irvin

(2013). Flow of food and people across the city: An examination of local food access in a New Orleans food desert. Spaces and Flows, 3, 45–56. doi:10.18848/2154-676/CGP/v03i02/53693

31.

Kazdin

A. E.

Blase

S. L.

(2011). Rebooting psychotherapy research and practice to reduce the burden of mental illness. Perspectives on Psychological Science, 6, 21–37.

32.

Kazdin

A. E.

(2017). Addressing the treatment gap: A key challenge for extending evidence-based psychosocial interventions. Behaviour Research and Therapy, 88, 7–18.

33.

Kendall

Olson

C. M.

Frongillo

E. A.

Jr. (1995). Validation of the Radimer/Cornell measures of hunger and food insecurity. The Journal of Nutrition, 125, 2793–2801.

34.

Keys

Brozek

Henschel

Michelsen

Taylor

H. L.

Simonson

. . . Wells

(1950). The biology of human starvation. Volumes 1 and 2. Minneapolis, MN: University of Minnesota Press.

35.

Kleinman

R. E.

Murphy

J. M.

Little

Pagano

Wehler

C. A.

Regal

Jellinek

M. S.

(1998). Hunger in children in the United States: Potential behavioral and emotional correlates. Pediatrics, 101(1), Article E3. Retrieved from https://pediatrics.aappublications.org/content/101/1/e3.long

36.

Krabbenborg

M. A.

Danner

U. N.

Larsen

J. K.

Veer

Elburg

A. A.

Ridder

D. T.

. . . Engels

R. C.

(2012). The eating disorder diagnostic scale: Psychometric features within a clinical population and a cut-off point to differentiate clinical patients from healthy controls. European Eating Disorders Review, 20, 315–320. doi:10.1002/erv.1144

37.

Kroenke

Spitzer

R. L.

Williams

J. B.

(2001). The PHQ-9: Validity of a brief depression severity measure. Journal of General Internal Medicine, 16, 606–613.

38.

Lang

Barling

(2012). Food security and food sustainability: Reformulating the debate. The Geographical Journal, 178, 313–326. doi:10.1111/j.1475-4959.2012.00480.x

39.

Lindsay

D. S.

(2015). Replication in psychological science. Psychological Science, 26, 1827–1832.

40.

Löwe

Decker

Müller

Brähler

Schellberg

Herzog

Herzberg

P. Y.

(2008). Validation and standardization of the Generalized Anxiety Disorder Screener (GAD-7) in the general population. Medical Care, 46, 266–274.

41.

Lund

De Silva

Plagerson

Cooper

Chisholm

Das

. . . Patel

(2011). Poverty and mental disorders: Breaking the cycle in low-income and middle-income countries. The Lancet, 378, 1502–1514.

42.

McIntyre

Glanville

N. T.

Raine

K. D.

Dayle

J. B.

Anderson

Battaglia

(2003). Do low-income lone mothers compromise their nutrition to feed their children? Canadian Medical Association Journal, 168, 686–691.

43.

Mehler

P. S.

(2017a). Medical complications of anorexia nervosa. In Brownell

K. D.

Walsh

B. T.

(Eds.), Eating disorders and obesity: A comprehensive handbook (3rd ed., pp. 214–218). New York, NY: Guilford Press.

44.

Mehler

P. S.

(2017b). Medical complications of bulimia nervosa. In Brownell

K. D.

Walsh

B. T.

(Eds.), Eating disorders and obesity: A comprehensive handbook (3rd ed., pp. 219–222). New York, NY: Guilford Press.

45.

Mitchison

Basten

Griffiths

Murray

S. B.

(2017). Beneath the tip of the iceberg: Why so many people with eating disorders are not referred for treatment. Australian Family Physician, 46, 539–540.

46.

Middlemass

K. M.

Cruz

Gamboa

Johnson

Taylor

Gomez

Becker

C. B.

(2018). Moving beyond weight & shape: Dietary restraint and food insecurity in a marginalized population. Manuscript submitted for publication.

47.

Mulders-Jones

Mitchison

Girosi

Hay

(2017). Socioeconomic correlates of eating disorder symptoms in an Australian population-based sample. PLOS ONE, 12(1), Article e0170603. doi:10.1371/journal.pone.0170603.

48.

Palar

Kushel

Frongillo

E. A.

Riley

E. D.

Grede

Bangsberg

Weiser

S. D.

(2015). Food insecurity is longitudinally associated with depressive symptoms among homeless and marginally-housed individuals living with HIV. AIDS and Behavior, 19, 1527–1534. doi:10.1007/s10461-014-0922-9

49.

Patel

. (n.d.). Filmmaker Tachaiko Omawale is opening a conversation about Black women and eating disorders [Blog post]. Retrieved from https://www.nationaleatingdisorders.org/blog/filmmaker-tchaiko-omawale-opening-conversation-about-black-women-and-eating-disorders

50.

Patel

Weiss

H. A.

Chowdhary

Nalk

Pednekar

Chatterjee

. . . Kirkwood

B. R.

(2010). Effectiveness of an intervention led by lay health counsellors for depressive and anxiety disorders in primary care in Goa, India (MANAS): A cluster randomized controlled trial. Lancet, 18, 2086–2095.

51.

Radimer

K. L.

Olson

C. M.

Greene

J. C.

Campbell

C. C.

Habicht

J. P.

(1992). Understanding hunger and developing indicators to assess it in women and children. Journal of Nutrition Education, 24, 36S–44S.

52.

Rasmusson

Lydecker

J. A.

Coffino

J. A.

White

M. A.

Grilo

C. M.

(2019). Household food insecurity is associated with binge-eating disorder and obesity. International Journal of Eating Disorders, 52, 28–35.

53.

Razykov

Ziegelstein

R. C.

Whooley

M. A.

Thombs

B. D.

(2012). The PHQ-9 versus the PHQ-8: Is item 9 useful for assessing suicide risk in coronary artery disease patients? Data from the Heart and Soul Study. Journal of Psychosomatic Research, 73, 163–168.

54.

Rentiera

(2011, September 29). Hispanics in San Antonio. San Antonio Express News. Retrieved from https://www.mysanantonio.com/sacultura/conexion/article/Hispanics-in-San-Antonio-2192950.php

55.

Royall

(2018, September 13). San Antonio ranks among nation’s highest-poverty cities. The Rivard Report. Retrieved from https://therivardreport.com/san-antonio-ranked-among-nations-highest-poverty-cities

56.

Smith

(2005). Psychotherapy, classism, and the poor: Conspicuous by their absence. American Psychologist, 60, 687–696.

57.

Smith

(2010). Psychology, poverty, and the end of social exclusion: Putting our practice to work (Vol. 7). New York, NY: Teachers College Press.

58.

Sonneville

K. R.

Lipson

S. K.

(2018). Disparities in eating disorder diagnosis and treatment according to weight status, race/ethnicity, socioeconomic background, and sex among college students. International Journal of Eating Disorders, 51, 518–526. doi:10.1002/eat.22846

59.

Sorsdahl

Slopen

Siefert

Seedat

Stein

D. J.

Williams

D. R.

(2010). Household food insufficiency and mental health in South Africa. Journal of Epidemiology and Community Health, 65, 426–431. doi:10.1136/jech.2009.091462

60.

Spitzer

R. L.

Kroenke

Williams

J. B.

Löwe

(2006). A brief measure for assessing generalized anxiety disorder: The GAD-7. Archives of Internal Medicine, 166, 1092–1097.

61.

Stice

Fisher

Martinez

(2004). Eating disorder diagnostic scale: Additional evidence of reliability and validity. Psychological Assessment, 16, 60–71. doi:10.1037/1040-3590.16.1.60

62.

Stinson

E. J.

Votruba

S. B.

Venti

Perez

Krakoff

Gluck

M. E.

(2018). Food insecurity is associated with maladaptive eating behaviors and objectively measured overeating. Obesity, 26, 1841–1848.

63.

Stunkard

A. J.

Messick

(1985). The three-factor eating questionnaire to measure dietary restraint, disinhibition and hunger. Journal of Psychosomatic Research, 29, 71–83.

64.

Swanson

Crow

Le Grange

Swendsen

Merikangas

(2011). Prevalence and correlates of eating disorders in adolescents: Results from the national comorbidity survey replication adolescent supplement. Archives of General Psychiatry, 68, 714–723. doi:10.1001/archgenpsychiatry.2011.22

65.

Tucker

(2006). The great starvation experiment: The heroic men who starved so that millions could live. New York, NY: Simon and Schuster.

66.

U.S. Census Bureau. (2016). American Community Survey 5-year Estimates. Retrieved from https://censusreporter.org/profiles/16000US4865000-san-antonio-tx

67.

Weaver

L. J.

Hadley

(2009). Moving beyond hunger and nutrition: A systematic review of the evidence linking food insecurity and mental health in developing countries. Ecology of Food and Nutrition, 48, 263–284.