Eating disorder screening: EAT-26 for first-line screening, EDE-Q for in-depth assessment
What EDs are and why diagnosis gets missed
EDs are a heterogeneous group of disorders unified by disturbed eating patterns and associated psychological dysregulation. DSM-5 distinguishes six main diagnoses: Anorexia Nervosa (AN), Bulimia Nervosa (BN), Binge Eating Disorder (BED), ARFID, OSFED (other specified), and USFED (unspecified). The ICD-11 structure is close, with minor differences.
The central clinical underdiagnosis pattern is the stereotype "ED = a young thin girl". Reality, per Galmiche et al. (2019, *Am J Clin Nutr*, meta-analysis of 94 studies): lifetime prevalence of any ED is 8.4% in women and 2.2% in men; point prevalence rose from 3.5% (2000–2006) to 7.8% (2013–2018). Binge Eating Disorder at 3.5% lifetime is the most common of all EDs and equally prevalent in men and women. Atypical anorexia (anorexic psychopathology at normal or elevated weight) — a common OSFED form especially easy to miss.
EDs are not defined by weight. Anorexia nervosa can occur at normal weight (atypical AN — a formal DSM-5 category). Binge eating disorder is often accompanied by obesity, but it isn't "overeating" — it is a structural psychiatric disorder marked by compulsivity and emotional dysregulation. Hudson et al. (2007) in the NCS-R showed: only a minority of people with EDs ever received treatment specifically for the ED. Structured screening is the only reliable way to identify cases that don't fit the stereotype.
EAT-26: first-line screening
EAT-26 (Eating Attitudes Test-26, Garner, Olmsted, Bohr & Garfinkel 1982, *Psychological Medicine*) — 26 items, mostly with a 0–3 Likert scale. Total range 0–78. A cut-off of ≥ 20 indicates risk warranting further evaluation. The scale is freely distributed by the author for clinical and research use.
Psychometrics. Cronbach α across international samples typically runs 0.86–0.94. Sensitivity and specificity at the ≥ 20 cut-off vary widely by population: in clinical samples, Constaín et al. (2014, 2016) showed sensitivity up to 100% and specificity 98% in Colombian male and female clinical groups; in community samples, Nunes et al. (2005) on Brazilian women found sensitivity of only 40%. Kang et al. (2017) on a Chinese clinical sample reported sensitivity 0.66–0.68. The conclusion: EAT-26 is a strong detection tool in clinical contexts, but in universal screening (schools, primary care) it produces substantial false negatives and benefits from a two-stage approach.
Three empirically derived subscales: Dieting (dietary patterns), Bulimia and Food Preoccupation (bulimic and food-related thoughts), and Oral Control (control over eating). Subscale scores give a profile, but clinical decisions always rest on total score plus structured diagnostic assessment.
- 26 items, completion time 5–10 minutes
- Total score 0–78, cut-off ≥ 20 — risk threshold
- 3 subscales: Dieting, Bulimia/Preoccupation, Oral Control
- Sensitivity varies 40–100% across populations — clinical context matters
- Freely distributed by the author (eat-26.com)
How to read EAT-26: the ≥20 cut-off and its limits
Standard cut-off interpretation:
- 0–9: normal / non-clinical range
- 10–19: mild concern with eating patterns; monitor in the presence of other risk factors
- ≥ 20: clinically significant ED risk; requires further structured assessment (EDE-Q + clinical interview)
A core clinical caveat: an EAT-26 cut-off does not diagnose the ED type. A high score can come from anorexia nervosa, bulimia nervosa, OSFED, or BED — and the instrument covers BED least well (binge episodes are not represented separately). So a positive EAT-26 is a signal to move on to more specific instruments, not a final answer. Conversely: a negative EAT-26 does not rule out an ED, especially atypical anorexia or BED — clinical screening should run in parallel.
EDE-Q: in-depth DSM-5 assessment
EDE-Q (Eating Disorder Examination Questionnaire, Fairburn & Beglin 1994) — 28 items, assessing behavioral patterns over the past 28 days. Unlike EAT-26, EDE-Q covers all key DSM-5 diagnostic criteria: restraint over eating, compensatory behaviors, objective binge-eating episodes, and body image disturbance.
Four subscales: Restraint, Eating Concern, Shape Concern, Weight Concern. Each scale runs 0–6 (mean of items). The Global score is the mean across all 28 items. Clinical cut-offs vary 2.3–4.0 depending on population. Psychometrics are stable: Mond et al. (2004) on 5,255 Australian women yielded Cronbach α = 0.93 for the Global score and test-retest correlations of 0.66–0.94 over 2 weeks. Berg et al. (2012) reviewed EDE/EDE-Q and confirmed a stable factor structure across most samples.
The key difference from EAT-26 — EDE-Q gives the frequency of specific behaviors over 28 days: objective binge episodes, compensatory strategies, excessive exercise. This makes it possible to apply the DSM-5 algorithm to differentiate anorexia / bulimia / BED / OSFED. EAT-26 doesn't offer this.
EDE-Q is available in Soveria. Completion takes 10–15 minutes. The platform automatically computes the four subscales and Global score, and reports behavioral-episode frequencies separately for clinical interpretation.
Screening vs diagnosis: where the boundary lies
The boundary between screening and diagnosis is a central methodological anchor in ED work. Screening (EAT-26 or EDE-Q as self-report) provides a signal: "there is probably a problem". The diagnosis is made through a structured clinical interview verifying DSM-5/ICD-11 criteria, duration, frequency, functional impact, and medical status.
The gold-standard diagnostic interview is the EDE (Eating Disorder Examination, Cooper & Fairburn 1987, current version 17.0D) — the interviewer-administered counterpart of EDE-Q. It takes 60–75 minutes and requires training. The alternative is SCID-5-ED (eating disorder module of SCID-5).
Why specifically a structured interview? Because EDs include behaviors patients frequently conceal: compensatory episodes, dietary restriction, body image disturbance. Self-report scales can systematically underestimate these. The EDE adds probe questions, clarifications, and calibration to standardized definitions — something no questionnaire alone provides.
Comorbidity and why AN has the highest mortality of any mental disorder
EDs almost never exist in isolation. Per Hudson et al. (2007, NCS-R, n = 9,282), comorbidity with other psychiatric disorders for anorexia nervosa, bulimia nervosa, and BED runs 56–94% across categories: depression, anxiety disorders, impulse-control disorders, substance use. The implication: a positive EAT-26 or EDE-Q should always be accompanied by comorbidity screening (PHQ-9, GAD-7 at minimum; PCL-5/ITQ when a traumatic background is suspected).
A separate clinical reality is medical status. Anorexia nervosa has the highest mortality of any psychiatric disorder: a standardized mortality ratio of 5.86 (Arcelus et al. 2011, meta-analysis). That means people with AN have a death rate roughly six times higher than the general population of the same age and sex. About one-fifth of AN deaths are suicides; the rest are medical complications of prolonged starvation and compensatory behaviors.
A positive ED screen does not mean "book 8 sessions of CBT". It is a signal for a multidisciplinary assessment: psychotherapist + physician (physical status, ECG when AN is suspected, lab markers when purging is involved) + dietitian where appropriate. AN with significant weight loss often requires medical care as the first priority; psychotherapy works effectively only after physical stabilization.
How to use in practice
Bottom-line guidance — five typical workflows for private practice.
- Initial appointment with suspected ED: EDE-Q (available in Soveria) + comorbidity screening (PHQ-9, GAD-7). When positive → referral for structured EDE interview or an ED specialist with medical evaluation
- Universal screening in primary care or university clinics: EAT-26 as a first-line filter; positives → EDE-Q or interview. Keep in mind sensitivity ranges from 40% to 100% by population — a two-stage approach is needed
- Tracking treatment progress in ED therapy: EDE-Q every 4–8 weeks — following change across the four subscales and behavioral-episode frequencies
- Client self-seeking an "ED test": EAT-26 as online screening — but with a mandatory recommendation to consult a specialist on a positive result. Self-screening without subsequent clinical interpretation is methodologically incorrect
- Does not replace: medical evaluation (physical status, ECG, lab markers) — that is the physician's domain, not the psychologist's
A brief note on Russian-language adaptations. EAT-26 and EDE-Q are used in Russian clinical practice via working translations; formally published psychometric validations on Russian-speaking samples are not currently indexed in PubMed. For practice this means: use the instruments anchored to international psychometric data, always confirm interpretation in clinical context, and avoid claims of "officially RF-validated version" — that fact is not in the published literature.
The core principle — self-report tells you "there is something to discuss"; a structured interview plus medical evaluation make the diagnosis. EAT-26 and EDE-Q are the first and second steps in differential assessment, not the third and not the last. In Soveria, EDE-Q runs as a self-report with automatic four-subscale breakdown and behavioral-episode frequencies — this turns "let's talk about how you've been eating this month" into specific numbers that can anchor clinical decisions.