Abstract Using Exploratory Data Mining to Identify Diagnostic Severity PRESENTATIONS for Other Specified Feeding or Eating Disorder (OSFED)
Shelby N. Ortiz, Lauren N. Forrest, April R. Smith
Miami University, Oxford, OH, United States

1: Unlike other eating disorders (EDs), no severity specifiers are noted for other specified feeding or eating disorder (OSFED). However, shape/weight overvaluation has been proposed as a transdiagnostic ED severity specifier. We used structural equation modeling trees, an exploratory data mining approach, to identify OSFED severity “splits” based on shape/weight overvaluation. 2: Participants were 595 females with OSFED in residential ED treatment. Using the semtree R package, clinically-relevant variables (all Eating Disorder Examination-Questionnaire [EDE-Q] subscales, body mass index, ED onset) determined meaningful severity splits indexed by shape/weight overvaluation, which was assessed with two EDE-Q items. After identifying splits, analysis of variance tests compared severity categories on demographic and comorbid characteristics. 3: Severity splits occurred at shape/weight overvaluation scores <1.75 (n=47), 1.75-4.75 (n=109), 4.75-5.75 (n=119), and >5.75 (n=320). Individuals within categories significantly differed on severity of depressive and anxiety symptoms, trauma history, and length of stay but not on ED duration. Nearly all symptoms increased with each severity category. 4: Shape/weight overvaluation can meaningfully differentiate OSFED presentations. Validated and meaningful OSFED severity specifiers may inform treatment planning regarding level of care and core treatment targets.