Students with Food Insecurity Are More Likely to Screen Positive for an Eating Disorder at a Large, Public University in the Midwest

Abstract

Background

College students experience a disproportionately high prevalence of both food insecurity and eating disorders. Food insecurity is associated with stress, irregular eating patterns, weight change, depression, and body dissatisfaction, making it a possible risk factor for the onset of eating disorders. However, the association between food insecurity and eating disorders among college students is not well understood.

Objective

This study explored the relation between food insecurity and screening positive for an eating disorder among students attending a large, public Midwestern university.

Design

Cross-sectional data were collected using an online survey administered from March through June 2018.

Participants/setting

Participants were recruited from a random sample of 2,000 students, with oversampling from the following groups: racial/ethnic minorities, first-generation students, and students from lower-income households. Of those sampled, 851 students (43%) responded. The final analytic sample comprised 804 students after excluding those with missing data.

Main outcome measure

The validated 5-item Sick, Control, One stone, Fat, Food (SCOFF) questionnaire was used to screen for the presence of an eating disorder.

Statistical analyses performed

Poisson regression was used to model prevalence ratios for positive SCOFF screens (≥ 2 affirmative responses) by levels of food security (ie, high, marginal, low, or very low). Models were adjusted for sex, age, race/ethnicity, degree type, financial aid, and first-generation student status.

Results

Compared to students with high food security, a higher prevalence of positive SCOFF screens was found among students with marginal food security (prevalence ratio [PR], 1.83, 95% CI 1.26 to 2.65; P = 0.001), low food security (PR 1.72, 95% CI 1.16 to 2.54; P = 0.007), and very low food security (PR 2.83, 95% CI 2.01 to 3.97; P < .0001).

Conclusions

Students with food insecurity at any level were more likely to screen positive for an eating disorder via the SCOFF questionnaire. Prospective studies are needed to determine whether food insecurity is a risk factor for the onset of eating disorders among college students.

The prevalence of food insecurity is alarmingly high among college students.¹ Food insecurity is defined by lack of reliable access to safe and nutritious food or the inability to acquire safe and nutritious food through socially acceptable means.² Current estimates suggest that approximately 35% to 42% of college students face food insecurity.^1,3 In comparison, 11.1% of households in the United States were food insecure in 2018.⁴

In addition to experiencing high levels of food insecurity, traditional college-aged students are at a peak age for eating disorder onset.⁵ A previous national study of college students found that 17.0% of female and 5.5% of male college students are likely to have an eating disorder,⁶ putting them at risk for myriad potential health consequences, including low bone mineral density,⁷ reduced oral health,⁸ reproductive concerns, ^9,10 poor psychosocial health,¹¹ gastrointestinal¹² and metabolic¹³ dysfunction, liver enzyme abnormalities,¹⁴ and cardiac complications.^15,16 People with eating disorders also have a higher risk of all-cause and suicide mortality.¹⁷ Subthreshold eating disorder symptoms, such as self-induced vomiting and loss of control eating, may act as precursors to full-syndrome eating disorders and are associated with poor long-term physical and mental health outcomes.¹⁸ In college students specifically, eating disorder symptoms have been associated with higher levels of suicidal ideation and more suicide attempts.¹⁹

Food insecurity during college may be a risk factor for the onset of eating disorder symptoms and eating disorders. Previous studies show that food insecurity is associated with a variety of behaviors and health outcomes that may increase risk of eating disorders: food insecure populations are more likely to experience psychological stress,^20–24 meal skipping and other irregular eating patterns,^24–27 weight change,²⁸ depression/depression symptoms,^{23,24,26,29,30} body dissatisfaction,^31,32 and drive for thinness,³² making food insecurity a potential risk factor for eating disorders. Body dissatisfaction and depression in particular are associated with the onset of eating disorder symptoms in adolescents and young adults.³³ It is also plausible that some characteristics of food insecurity, such as dietary restriction due to food scarcity, increase eating disorder risk.³⁴

Although a few studies have found that food insecurity is associated with more eating disorder symptoms and/or full-syndrome eating disorders in populations such as adolescents,^27,35 parents of adolescents,²⁵ postpartum women,²⁰ prebariatric surgery patients,³⁶ food bank clients,^30,37 adult volunteers,³⁸ and US-based Amazon Mechanical Turk participants,^39,40 the association between food insecurity and eating disorders in college students has only been examined in a few studies. In one study, El Zein and colleagues²² demonstrated that current food insecurity was associated with a higher prevalence of self-reported eating disorder symptoms among college students from multiple universities. Darling and colleagues²³ also found that college students with prior food insecurity were more likely to report current eating disorder symptoms. Another study among male college athletes found that food insecurity, both prior and present, was associated with more food preoccupation, an eating disorder symptom.⁴¹ Although studies have found that women are more likely to experience household-level food insecurity⁴² and eating disorder symptoms are more prevalent among female college students,⁶ the possible modifying effect of sex on the relation between food insecurity and eating disorders in college students has not been previously explored. Potential effect modification by race/ethnicity has not been studied either, although race/ethnicity is associated with differential experiences of both food insecurity^3,22 and eating disorder symptoms⁶ in college students.

The aim of the present study was to examine the cross-sectional association between food insecurity and screening positive for an eating disorder, as well as to determine whether this association was modified by sex and/or race/ethnicity, among college students attending a large, public Midwestern university.

MATERIALS AND METHODS

Study Sample

This study used a cross-sectional sample of undergraduate, graduate, and professional degree students attending the University of Michigan during the Winter (January to April) 2018 semester. Details regarding survey design and data collection processes have been previously described.⁴³ Briefly, the Office of the Registrar generated a sample of 2,000 students, with oversampling of racial/ethnic minority students, students from households with annual incomes < $65,000, and first-generation students. These students were contacted via E-mail in multiple waves, until all students were contacted. They were invited to complete an online Qualtrics survey and were given 3 weeks to complete the survey. A total of 851 responses (response rate of 43%) were collected from March to June 2018. Students provided informed consent at the beginning of the survey, with 92% of respondents additionally consenting to link their responses to sociodemographic data from the Office of the Registrar. Those who did not consent to using data from the Office of the Registrar could self-report these sociodemographic variables. After excluding respondents with missing data for food security status, eating disorder screening, sex, race/ethnicity, residency status (in state vs out of state), or degree type (undergraduate vs graduate/professional), 804 students were included in the analytic sample. This study was approved by the Health Sciences and Behavioral Sciences Institutional Review Board at the University of Michigan.

Food Security Status

The 10-item US Adult Food Security Survey Module, developed by the United States Department of Agriculture (USDA), was used to measure food security over the past 12 months.⁴⁴ The survey assesses behaviors and experiences related to insufficient resources to acquire food and progressively increases in severity such that affirmative answers to the last questions suggest the most severe food insecurity. Using the USDA guidelines, students were classified as having high food security (score = 0), marginal food security (score = 1 or 2), low food security (score = 3 to 5), or very low food security (score = 6 to 10). Marginal food security suggests worry about acquiring food without a deficit of food quality or quantity; low food security suggests diminished food quality with sufficient quantity, and very low food security suggests that both food quality and quantity are diminished due to insufficient resources to acquire food. The term food security encompasses high and marginal food security, and the term food insecurity encompasses low and very low food security.

Eating Disorder Screener

The 5-item Sick, Control, One stone, Fat, Food (SCOFF) questionnaire was used to screen students for the presence of an eating disorder.⁴⁵ The SCOFF questionnaire has been validated as a sensitive and specific screener for eating disorders in college students.^46–49 A recent meta-analysis estimated a pooled sensitivity and specificity of 0.86 and 0.83, respectively, from SCOFF validation studies conducted in both college and noncollege populations.⁵⁰ The questionnaire is composed of five questions that can be represented by single words that together form the acronym SCOFF. These questions ask about thoughts and behaviors that can be risk factors or symptoms of eating disorders, with ≥2 affirmative responses defined as a positive screen for a possible eating disorder. The language of the original questionnaire was modified slightly for use in a Midwestern population by substituting the words 15 pounds for the British 1 stone in Question 3. Table 1 shows the five SCOFF questions as they were adapted for use in the present study.

Table 1.

The adapted Sick, Control, One stone, Fat, Food (SCOFF) questionnaire used to screen for the presence of an eating disorder among students attending the University of Michigan, along with the specific eating disorder characteristic assessed by each SCOFF question

Question	Eating disorder characteristic
1) Do you ever make yourself sick because you feel uncomfortably full? (S = “Sick”)	Self-induced vomiting
2) Do you worry you have lost control over how much you eat? (C = “Control”)	Loss of control eating
3) Have you recently lost more than 15 pounds in a 3month period? (O = “One stone”)	≥15 lb weight loss
4) Do you believe yourself to be fat when others say you are too thin? (F = “Fat”)	Feel fat despite thinness
5) Would you say that food dominates your life? (F = “Food”)	Food dominates life

Covariates

Sex, age, race/ethnicity, residency status (in state vs out of state), and degree type (undergraduate vs graduate/professional) were collected from the Office of the Registrar or self-reported when students did not consent to have survey responses linked to data from the registrar. Receipt of financial aid and parental education level were self-reported. If both parents had a high school education or less, students were considered first-generation students.

Statistical Analysis

To account for student nonresponse and oversampling of certain demographic groups, we used poststratification weights to make analyses representative of the entire student body enrolled during the Winter 2018 academic term. Weights were created based on all possible combinations of sex (male vs female), race/ethnicity (White, Black, Asian, Hispanic, or other), residency status (in state vs out of state), and degree type (undergraduate vs graduate/professional) using student enrollment data from the Office of the Registrar. These weights were scaled to reflect the analytic sample size (n 804) and were used in all analyses to produce results representative of all students enrolled at the University of Michigan, with respect to sex, race/ethnicity, residency status, and degree type.

Poisson regression was used to model prevalence ratios for positive SCOFF screens by sociodemographic characteristics and by food security status. Wald tests and likelihood ratio tests were used to examine the statistical significance of comparisons between two categories and multiple categories, respectively. When effect estimates suggested a linear trend for multiple categories, a Wald test for trend was used. An interaction term for food security status and sex was used to examine whether sex modified the association between food security and positive SCOFF screens. Separately, an interaction term for food security status and race/ethnicity was used to examine whether race/ethnicity modified the association between food security and positive SCOFF screens. In a secondary analysis, Poisson regression was used to model prevalence ratios for affirmative responses to each SCOFF question by levels of food security. Multivariable analyses were adjusted for potential confounding variables, including sex (male vs female), age (continuous), race/ethnicity (White, Black, Asian, Hispanic, or other), degree type (undergraduate vs graduate/professional), receipt of financial aid (yes vs no), and first-generation status (first-generation vs not first-generation). All statistical analyses were performed using SAS version 9.4.⁵¹

RESULTS

Accounting for poststratification weights, the analytic sample was 49.6% male and 50.4% female, with a median age of 21.6 years. Based on the USDA’s categorization of food security status, 47.8% of students had high food security, 17.1% had marginal food security, 16.3% had low food security, and 18.8% had very low food security. Males had a higher prevalence of food insecurity, defined as low or very low food security, than females (37.5% and 32.7%, respectively). Among all students, 29.0% screened positive for an eating disorder using the SCOFF questionnaire, with the prevalence of positive SCOFF screens higher among females (33.4%) than males (24.4%). Unadjusted associations between sociodemographic characteristics and having a positive SCOFF screen can be seen in Table 2. For the individual SCOFF questions, among the total sample, 20.3% endorsed self-induced vomiting, 30.0% endorsed loss of control eating, 9.0% endorsed recent weight loss of ≥15 lb over a 3-month period, 18.3% endorsed feeling fat despite being thin, and 15.0% said that food dominates their life.

Table 2.

Prevalence of positive Sick, Control, One stone, Fat, Food^a (SCOFF+) screens by sociodemographic characteristics among students attending the University of Michigan

Characteristic	nb	Weighted % of total sample	Weighted % SCOFF+	Unadjusted prevalence ratio (95% CI)c
Sex d
Male	291	49.6	24.4	1.00
Female	513	50.4	33.4	1.37 (1.06–1.78)
P valuee				0.02
Age group (y)
< 21	419	41.3	26.2	1.00
21 to < 24	287	30.9	26.9	1.03 (0.75–1.42)
≥24	97	27.8	34.2	1.31 (0.96–1.78)
P valuef				0.20
Race/ethnicity
White/Caucasian	219	65.0	28.1	1.00
Black/African American	100	5.0	28.5	1.02 (0.56–1.86)
Hispanic	146	6.9	20.6	0.73 (0.40–1.34)
Asian	277	14.5	34.5	1.23 (0.87–1.74)
Multirace/ethnicity or other	62	8.7	33.0	1.17 (0.76–1.82)
P valuef				0.53
Degree type
Undergraduate	731	72.1	28.5	1.00
Graduate/professional	73	27.9	30.2	1.06 (0.80–1.41)
P valuee				0.69
Financial aid
No	121	15.6	29.7	1.00
Yes	683	84.4	28.8	0.97 (0.68–1.38)
P valuee				0.86
First-generation student
No	476	76.6	27.8	1.00
Yes	328	23.4	32.6	1.17 (0.88–1.57)
P valuee				0.29

^aFor a US population, concepts in the SCOFF questionnaire roughly translate as self-induced vomiting, loss of control eating, ≥15 lb weight loss in the past 3 months, feeling fat despite being thin, and feeling that food dominates life.

^bTotals may be less than 804 due to missing values.

^cPrevalence ratio estimates from Poisson regression models with positive SCOFF result as the dichotomous outcome and indicator variables for each characteristic as predictors. Models accounted for survey weights.

^dSex, age, race/ethnicity, and degree type were obtained from the Office of the Registrar with participant consent. Participants who did not consent could choose to self-report these characteristics. Financial aid and first-generation status were self-reported.

^eWald test.

^fLikelihood ratio test.

In multivariable-adjusted analysis, there was a higher prevalence of positive SCOFF screens in students with marginal food security (prevalence ratio [PR], 1.83, 95% CI 1.26 to 2.65), low food security (PR 1.72, 95% CI 1.16 to 2.54), and very low food security (PR 2.83, 95% CI 2.01 to 3.97), compared to students with high food security. This association was modified by sex (P for interaction = 0.02) but not by race/ethnicity (P for interaction = 0.95). Among male students only, there was a higher prevalence of positive SCOFF screens in students with marginal food security (PR 3.06, 95% CI 1.60 to 5.85), low food security (PR 2.34, 95% CI 1.19 to 4.59), and very low food security (PR 5.08, 95% CI 2.85 to 9.06) compared to students with high food security. Among female students only, there was also a higher prevalence of positive SCOFF screens in students with very low food security (PR 1.86, 95% CI 1.16 to 2.95), compared to students with high food security. Marginal and low food security were not statistically significantly related to positive SCOFF screens in females, but there was evidence of a statistically significant trend, where the prevalence of positive SCOFF screens increased monotonically with increasing food insecurity (P for trend = 0.007) (Table 3).

Table 3.

Prevalence of positive Sick, Control, One stone, Fat, Food^a (SCOFF+) screens by food security status among students attending the University of Michigan

	nb	Weighted % of sample	Weighted % SCOFF+	Unadjusted prevalence ratio (95% CI)c	Adjusted prevalence ratio (95% CI)d
Total sample
High food security	378	47.8	18.4	1.00	1.00
Marginal food security	141	17.1	34.4	1.88 (1.30–2.71)	1.83 (1.26–2.65)
Low food security	143	16.3	30.9	1.68 (1.14–2.48)	1.72 (1.16–2.54)
Very low food security	142	18.8	49.3	2.69 (1.94–3.72)	2.83 (2.01–3.97)
P valuee				< 0.0001	< 0.0001
Male (n = 291) f
High food security	151	47.7	9.8	1.00	1.00
Marginal food security	49	14.8	34.6	3.53 (1.88–6.60)	3.06 (1.60–5.85)
Low food security	47	17.4	24.6	2.50 (1.30–4.83)	2.34 (1.19–4.59)
Very low food security	44	20.1	51.3	5.23 (3.03–9.04)	5.08 (2.85–9.06)
P valuee				< 0.0001	< 0.0001
Female (n = 513)
High food security	227	47.9	26.7	1.00	1.00
Marginal food security	92	19.4	34.3	1.28 (0.81–2.04)	1.40 (0.87–2.24)
Low food security	96	15.2	38.0	1.42 (0.87–2.32)	1.53 (0.92–2.54)
Very low food security	98	17.5	47.0	1.76 (1.14–2.72)	1.86 (1.16–2.95)
P value for trendg				0.009	0.007

^aFor a US population, concepts in the SCOFF questionnaire include self-induced vomiting, loss of control eating, ≥15 lb weight loss in the past 3 months, feeling fat despite being thin, and feeling that food dominates life.

^bTotals may be less than 804 due to missing values.

^cPrevalence ratio estimates from Poisson regression models with positive SCOFF result as the dichotomous outcome and indicator variables for each characteristic as predictors. Models accounted for survey weights.

^dAdjusted for sex, age, race/ethnicity, degree type, financial aid, and first-generation student status.

^eLikelihood ratio test.

^fThe interaction between sex and food security was statistically significant (P = 0.02, likelihood ratio test) in the multivariable-adjusted model.

^gWald test for a variable representing ordinal categories of food insecurity entered into the model as a continuous covariate.

In multivariable-adjusted analyses examining food security status and individual SCOFF questionnaire items, students with very low food security, the most severe level of food insecurity, were most likely to endorse self-induced vomiting, loss of control eating, ≥15 lb weight loss, and feeling fat despite thinness. Compared to students with high food security, students with very low food security had 3.04 times (95% CI 2.00 to 4.62) the prevalence of self-induced vomiting, 1.78 times (95% CI 1.27 to 2.48) the prevalence of loss of control eating, 4.32 times (95% CI 2.36 to 7.91) the prevalence of ≥15 lb weight loss, and 2.07 times (95% CI 1.36 to 3.15) the prevalence of feeling fat despite thinness. Students with very low food security were also more likely to feel that food dominates life (PR 2.09, 95% CI 1.21 to 3.58) compared to those with high food security. However, the prevalence of students affirming that food dominates life was highest among those with marginal (PR 2.83, 95% CI 1.74 to 4.62) and low food security (PR 2.81, 95% CI 1.70 to 4.64) (Table 4).

Table 4.

Prevalence of eating disorder characteristics (self-induced vomiting, loss of control eating, ≥15 lb weight loss in the past 3 months, feeling fat despite being thin, and feeling that food dominates life) by food security status among students attending the University of Michigan

Eating disorder characteristic	Weighted % affirmative response	Unadjusted prevalence ratio (95% CI)a	Adjusted prevalence ratio (95% CI)b
Self-induced vomiting
High food security	12.1	1.00	1.00
Marginal food security	27.4	2.26 (1.47–3.48)	2.35 (1.52–3.63)
Low food security	24.6	2.03 (1.30–3.19)	2.20 (1.39–3.47)
Very low food security	31.0	2.56 (1.71–3.84)	3.04 (2.00–4.62)
P valuec		< 0.0001	< 0.0001
Loss of control eating
High food security	25.2	1.00	1.00
Marginal food security	34.6	1.37 (0.97–1.94)	1.38 (0.97–1.95)
Low food security	25.0	0.99 (0.67–1.48)	1.06 (0.71–1.59)
Very low food security	42.2	1.68 (1.22–2.30)	1.78 (1.27–2.48)
P valuec		0.008	0.006
≥15 lb weight loss
High food security	4.6	1.00	1.00
Marginal food security	11.9	2.60 (1.33–5.10)	2.59 (1.30–5.17)
Low food security	6.5	1.43 (0.63–3.23)	1.43 (0.63–3.28)
Very low food security	19.9	4.36 (2.42–7.86)	4.32 (2.36–7.91)
P valuec		< 0.0001	< 0.0001
Feel fat despite thinness
High food security	13.2	1.00	1.00
Marginal food security	22.7	1.71 (1.10–2.67)	1.58 (1.01–2.48)
Low food security	15.4	1.16 (0.69–1.95)	1.06 (0.63–1.80)
Very low food security	29.6	2.24 (1.50–3.34)	2.07 (1.36–3.15)
P valuec		0.0008	0.005
Food dominates life
High food security	8.2	1.00	1.00
Marginal food security	24.5	2.99 (1.84–4.85)	2.83 (1.74–4.62)
Low food security	24.0	2.93 (1.79–4.80)	2.81 (1.70–4.64)
Very low food security	16.1	1.96 (1.15–3.32)	2.09 (1.21–3.58)
P valuec		< 0.0001	< 0.0001

^aPrevalence ratio estimates from Poisson regression models with affirmation of SCOFF item as the dichotomous outcome and indicator variables for each characteristic as predictors. Models accounted for survey weights.

^bAdjusted for sex, age race/ethnicity, degree type, financial aid, and first-generation student status.

^cLikelihood ratio test.

DISCUSSION

In this cross-sectional study of college students attending a large, public university in the Midwest, students with marginal, low, and very low food security were significantly more likely to screen positive for an eating disorder compared to their peers with high food security. After stratifying by sex, students with very low food security had the highest adjusted prevalence of screening positive for an eating disorder among both males and females, and the association between very low food security and positive SCOFF screens was stronger in males compared to females. Among all students, those with very low food security were most likely to endorse self-induced vomiting, loss of control eating, ≥15 lb weight loss over a 3 month period, and feeling fat despite being thin. Relative to high food security, very low food security was also associated with higher prevalence of feeling that food dominates life, but students with marginal or low food security were most likely to report this feeling.

The overall prevalence of food insecurity in this sample was 35.1%, which is near the low end of the range of food insecurity prevalence estimates of 35% to 42% in other college student literature.^1,3 The slightly lower prevalence of food insecurity in this sample might be due to underlying qualities of this particular student population. For example, students attending this university might be more likely to come from financially privileged backgrounds than students at other colleges. The prevalence of positive eating disorder screens in this sample was 29.0%, which is somewhat higher than the 21.2% prevalence found in a recent, national study that used the SCOFF questionnaire to screen for eating disorders in more than 70,000 students from 77 US college campuses.¹⁹ Again, this difference might be explained by qualities unique to the university and the student body in the present study. For example, this institution might have unusually high academic standards, recruiting a higher proportion of students with perfectionistic tendencies who are at increased risk of eating disorders.

The finding that students with food insecurity were more likely to screen positive for an eating disorder is supported by previous studies reporting that food insecurity is associated with more eating disorder symptoms among college students.^22,23,41 Calorie/nutrition deprivation or risk of calorie/nutrition deprivation for any reason, including insufficient resources to obtain food, could result in increased focus on food and/or body image and provoke eating disorder behaviors. This theory is supported by findings from the historic Minnesota Starvation Study, which demonstrated that food restriction imposed by study investigators caused otherwise healthy adult men to engage in a variety of eating disorder symptoms.⁵² In the present study, students with marginal food security had a higher prevalence of positive SCOFF screens compared to students with high food security, suggesting that even students who worry about having resources to obtain food, without experiencing a deficiency of food quality or quantity, might be at increased eating disorder risk. These results underscore the importance of promoting food security for all students, rather than developing interventions only for students with the most severe forms of food insecurity.

The overall association between food security status and screening positive for an eating disorder was statistically significantly modified by sex. In sex-stratified analyses, male and female students with very low food security were most likely to have a positive SCOFF screen compared to students with higher levels of food security. The effect estimate for males was stronger than the estimate for females (PR males = 5.08, PR females = 1.86). This finding suggests that food insecurity may be a salient risk factor for eating disorders in both males and females, but especially in males. The stronger association among male students is an unexpected and interesting finding. One possible explanation for a stronger association in males might be that factors besides college food security contribute to a more saturated eating disorder risk in female students, allowing food security to have a greater relative influence on eating disorder risk in male students. For example, females are substantially more likely to have experienced calorie deprivation from intentional dieting to lose weight before college,⁵³ which could contribute to eating disorder symptoms via the same mechanisms as food insecurity (ie, calorie restriction). Thus, food insecurity may be a less influential risk factor for eating disorder symptoms among participants with a history of these behaviors. Another possible explanation may be that males are more likely to experience calorie deprivation in the context of food insecurity. Previous research suggests that food insecurity is associated with elevated weight status in women but not in men,⁵⁴ whereas other studies have found that food insecurity is associated with lower weight among men but not women.^55,56 Hernandez and colleagues⁵⁴ hypothesize that gender differences in the association between food insecurity and weight status might be due in part to women having more knowledge about or experience using strategies to acquire food in the context of food insecurity. These coping strategies may help female college students mitigate some negative consequences of food insecurity, in contrast to male students.

In analyses examining individual eating disorder risk factors and symptoms, very low food security was associated with recently losing ≥ 15 lb over a 3-month period. Weight loss could be a symptom of food insecurity without the presence of eating disorder pathology. Very low food security results in reduced quantity of food,⁴ and the US Adult Food Security Survey Module includes a question about losing weight due to lack of money for food to assess food security status.⁴⁴ In 2018, 47% of adults/individuals in households with very low food security reported losing weight due to lack of money for food.⁵⁷ However, weight loss is also a common symptom of eating disorders. Qualitative research suggests that some adults see weight loss resulting from food insecurity as a benefit,⁵⁸ which may indicate underlying body image concerns. Dietary restriction for any reason, including reasons not related to concern about weight or shape, was correlated with more eating disorder symptoms in a previous study.³⁴

Students with marginal, low, and very low food security were more likely to report that food dominates their life compared to students with high food security. However, associations were strongest for students with marginal and low food security. Like weight loss, the feeling that food dominates life could reasonably be a symptom of food insecurity, independent of eating disorder pathology. For example, worry about obtaining food is an early indicator of food insecurity. However, food preoccupation is a salient symptom of eating disorders, and it is plausible that food preoccupation due to a lack of financial resources could increase eating disorder risk. The weaker association between very low food security and the feeling that food dominates life might be explained by the possibility that students with very low food security, who lack sufficient resources to obtain an adequate quantity of food, repress thoughts about food as a coping mechanism to avoid being reminded of hunger. This type of coping mechanism has been previously reported in qualitative research among children.⁵⁹

It is noteworthy that very low food security among students was also associated with a higher prevalence of self-induced vomiting, loss of control eating, and feeling fat despite thinness. These behaviors/attributes are not characteristics of food insecurity itself, supporting the hypothesis that food insecurity could be associated with eating disorder risk factors and symptoms. Previous studies support these findings: Feeling fat despite thinness is an indicator of poor body image; food insecurity has also been associated with higher levels of body dissatisfaction in children³¹ and adolescents³² and weight self-stigma in adults.³⁷ Loss of control eating is a component of binge eating behavior, and previous studies in noncollege populations found that food insecurity was associated with more binge eating behavior.^{25,30,35–39} There are also studies that found a higher prevalence of bulimia nervosa and inappropriate compensatory behaviors (such as self-induced vomiting) among people with lower food security.^27,30,37,40

This study has many strengths. First, this study had a large sample size that was weighted to be sociodemographically and academically representative of the entire University of Michigan student body. Validated measurement tools were used to assess exposure and outcome variables. The USDA food security measurement tool is the most widely used instrument for assessing food security status in population-based studies and is recommended for use in college students.⁶⁰ The SCOFF questionnaire is also well-validated as a screening tool for eating disorders.⁵⁰ Much of the covariate data was directly provided by the registrar and thus less susceptible to measurement error due to self-report. This study also explored a novel research question of public health significance in a vulnerable population.

This study does have some limitations. Its cross-sectional design prevents establishment of temporality between food security status and eating disorder pathology. Students were asked to report food security status over the past year, but food security status can change substantially over the course of a year among college students.²⁴ Furthermore, the SCOFF questionnaire cannot provide definitive eating disorder diagnoses, although it is a well-validated and widely used eating disorder screener.⁵⁰ There could be measurement error because food insecurity and eating disorder screening items were self-reported. Eating disorders are stigmatized conditions that may be intentionally underreported. However, social desirability bias may have been mitigated by administering the survey online, which allowed students to report their responses privately. In addition, under-report of eating disorder risk factors and symptoms would very likely bias effect estimates toward the null, meaning results in this study are estimated conservatively. Nonresponse could lead to bias in the case that students who responded to the survey were systematically different from students who did not respond. Still, responses were weighted to reflect the sociodemographic and academic diversity of the student body. Given the consistency of our findings with other studies on college campuses that have also found a relation between food insecurity and more eating disorder symptoms,^22,23,41 the results of the present study may be generalizable to other college student populations. Future research should employ prospective cohort studies with more comprehensive eating disorder measures to better understand whether food insecurity is a risk factor for the onset of eating disorder symptoms and eating disorders.

CONCLUSIONS

This study found that food insecurity at any level of severity was associated with a higher prevalence of positive eating disorder screens among students attending a large public university. These findings underscore a need for prospective studies to better establish whether food insecurity increases risk of eating disorder onset in college students. Food insecurity is widespread on college campuses, and college students are at high risk for onset of eating disorders. If future research confirms a prospective association between food insecurity and eating disorders in college students, there might be important public health implications. For example, colleges might be able to reduce the onset or worsening of eating disorders and related symptoms among students by combating food insecurity on their campuses. Eating disorder prevention/intervention programs might also be improved by incorporating strategies to address food insecurity.

RESEARCH SNAPSHOT.

Research Question:

Is there an association between food security status and eating disorders among college students?

Key Findings:

In this cross-sectional study of 804 college students, students with marginal, low, and very low food security were more likely to screen positive for an eating disorder compared to their peers with high food security. Sex-stratified results reflected similar associations in both males and females, but the associations were stronger in males. Analyses were adjusted for sex, age, race/ethnicity, degree type, financial aid, and first-generation student status.