Newly proposed DSM-5 criteria reduce the need for “not otherwise specified” diagnoses

Research and Study of Eating Disorder Treatment and Etiology

Newly proposed DSM-5 criteria reduce the need for “not otherwise specified” diagnoses and can be reliability applied by clinicians in a residential eating disorder treatment setting

Jennifer J. Thomas, Ph.D.1,2,3; Kamryn T. Eddy, Ph.D.2,3;
Robert Hohe, B.S.1,2,4; Philip Levendusky, Ph.D. 1,3; Anne E. Becker, M.D., Ph.D., Sc.M.2,3,5

1Klarman Eating Disorders Center, McLean Hospital, Belmont, MA, USA
2Department of Psychiatry, Massachusetts General Hospital, Boston, MA, USA
3Department of Psychiatry, Harvard Medical School, Boston, MA, USA
4Technical University of Dresden, Dresden, Germany
5Department of Global Health & Social Medicine, Harvard Medical School, Boston, MA, USA

Objective: The DSM-5 Eating Disorders (ED) Work Group has proposed new diagnostic criteria that broaden anorexia nervosa (AN) and bulimia nervosa (BN); add new disorders (e.g., binge eating disorder, avoidant/restrictive food intake disorder); and create descriptive subtypes within ED not otherwise specified (EDNOS). However, the reliability and validity of these new criteria have not been prospectively tested. The purpose of this study was to evaluate the clinical utility of proposed DSM-5 criteria by assessing the prevalence of EDNOS under DSM-IV versus DSM-5; and the inter-rater reliability of research and clinician diagnoses in DSM-IV versus DSM-5.

Method: Consecutive female patients aged 13-23 (N = 42) admitted to a residential ED treatment facility received both DSM-IV and DSM-5 ED diagnoses from a research assessor (n = 3) via the Eating Disorder Examination; and a treating clinician (n = 13) via an unstructured interview in routine care. We tested our hypothesis that significantly fewer patients would receive EDNOS diagnoses under DSM-5 versus DSM-IV using Fisher’s exact test, and evaluated inter-rater reliability between research and clinician diagnoses using Cohen’s kappa.

Results: As hypothesized, significantly fewer patients received EDNOS diagnoses from researchers under DSM-5 (n = 14; 33%) versus DSM-IV (n = 30; 71%), p = .003. Clinicians also conferred fewer EDNOS diagnoses under DSM-5 (n = 7; 17%) than DSM-IV (n = 14; 33%), p = .003. While research and clinical diagnoses showed moderate inter-rater reliability under DSM-IV (κ = .41), they showed substantial reliability under DSM-5 (κ = .62).

Conclusion: Our preliminary data suggest that the proposed DSM-5 revisions will successfully reduce the predominance of EDNOS in this naturalistic clinical setting. Importantly, the observed improvement in clinical utility is not associated with a reduction in inter-rater reliability, suggesting that clinicians can reliably apply the new criteria in routine care.