The College Risk Behavior Survey (CRBS) is an annual survey of undergraduate students at a large public university in the mid-Atlantic region of the United States that has been conducted since the mid-1990s. Each spring semester, researchers send an email invitation to participate in the survey to a sample of 3,000 undergraduate students currently enrolled on the main campus. A new sample is generated each year. The survey is administered via Qualtrics, takes about 15 min for students to complete, and students are offered a $5 gift card as an incentive for participating [24]. The CRBS measures demographics, substance use, mental health, as well as other health knowledge, health risk behaviors, and social practices of students.

In this study, we examined DE measures over four years of survey data, from 2019 to 2022. In 2019, 1,138 students completed the survey (38% response rate); in 2020, 764 students completed the survey (25% response rate). In 2021, 765 students completed the survey (25% response rate); in 2022, 640 students completed the survey (21% response rate). For this study, data from years 2019 and 2020 are combined and classified as “pre-pandemic”, while data from years 2021 and 2022 were combined and classified as “after the onset of the pandemic”. The CRBS survey is administered each year during the spring semester, typically running from early to middle March to the end of April. The 2020 survey administration was a unique cycle; while the survey was launched in March, that same week the campus enacted its initial shutdown due to the COVID-19 pandemic, at which point students vacated the residence halls and the majority of classes shifted to a virtual format. While the 2020 survey technically remained open to students after the campus shutdown, student engagement with the survey virtually halted. The only survey responses from 2020 that were analyzed for this paper were collected prior to the campus shutdown when student life was still relatively normal. Lower survey response rates in 2020, 2021, and 2022 compared to 2019 are likely attributable to pandemic-related disruptions in that influenced student engagement.

Demographics

The study included the following key demographic variables: genderFootnote 1 (woman, man); race/ethnicity (Non-Hispanic Black, Hispanic, Non-Hispanic Asian, Non-Hispanic Another Race/Multiple Races (this includes Middle Eastern, Native Hawaiian or Pacific Islander, Alaska Native or American Indian, or other or biracial/multiracial, Non-Hispanic White), year in college (first year, second year, third year, fourth year and beyond); and place of residence (home with family, off campus, on campus).

Social characteristics and behavioral health

The CRBS survey measured binary variables related to student life and behavioral health. These variables included sorority or fraternity membership (yes/no); participation in intercollegiate or intramural athletic teams (yes/no); past month marijuana use (yes/no); past month alcohol use (yes/no); and whether students seriously considered suicide in the past year (yes/no).

Since 2019, the CRBS has also included two brief screeners for EDs, described below. We selected these two screeners after consulting the Academy for Eating Disorders’ Epidemiology and Public Health Practice Special Interest Group.

SCOFF

The SCOFF was designed to be a brief screening tool with a simplistic scoring algorithm [13]. The SCOFF asks the respondent to answer yes/no to five items, each associated with a letter in the acronym SCOFF (Sick, Control, One, Fat, Food).

1.

Do you make yourself Sick because you feel uncomfortably full?

2.

Do you worry that you have lost Control over how much you eat?

3.

Have you recently lost more than One stone (14 lb.) in a three-month period?

4.

Do you believe yourself to be Fat when others say you are thin?

5.

Would you say that Food dominates your life?

Per the original publication, answering “yes” to two or more items provided 100% sensitivity and 87.5% specificity for anorexia nervosa and bulimia nervosa. A 2020 systematic review and meta-analysis of 25 studies reported a pooled sensitivity of 86% and specificity of 83% [25].

ESP

Another brief screener, the Eating Disorder Screener for Primary Care [14], was created for both use in primary care and college samples. It contains four questions:

1.

Are you satisfied with your eating patterns?

2.

Do you ever eat in secret?

3.

Does your weight affect the way you feel about yourself?

4.

Do you currently suffer with or have you ever suffered in the past with an eating disorder?

The original publication reported 100% sensitivity and 71% specificity for two or more abnormal responses (an abnormal response is considered a “no” to question 1, “yes” to questions 2–4) as a threshold for having an ED.

Analysis

Chi-squared analyses were used to compare the prevalence of DE before versus after the onset of the pandemic. Two logistic regression models were used to test our hypotheses regarding correlates of DE, one regression model using the SCOFF score to define the dependent variable and one regression model using the ESP score. The dependent variable was DE, defined by the original publications of the SCOFF and the ESP as ≥ 2 abnormal responses. An abnormal response is considered “yes” to any item on either questionnaire, except for item 1 for the ESP, in which case “no” is an abnormal response. The number of abnormal responses per individual was tallied separately for each questionnaire, and the proportion of individuals with at least two abnormal responses was calculated.

Substance use, biological sex, athlete, and fraternity/sorority membership were coded using dummy variable binary responses; whereas, race, living situation, and class year were categorical variables. Analyses were conducted using Stata v14. The logistic command in Stata was used, which produces maximum-likelihood dichotomous logistic models. Alpha was set as p ≤ 0.05. For each respective model (SCOFF and ESP), all independent variables were entered at the same step.