O2185: Higher Presenting BMI Associated with Lower Odds of Medical Admission for Anorexia Nervosa in Adolescents, Independent of Physiologic Parameters

Background: Atypical anorexia nervosa (AAN) is a diagnosis mirroring features of anorexia nervosa (AN) with the exception that, despite significant weight loss, the individual remains within or above normal weight range. A challenge for treatment of AAN is that providers often use normal weight as reassurance against illness severity. We sought to examine weight-related factors that predict illness severity and describe the impact of presenting BMI percentile (BMI%ile) on medical admission recommendations for AN and AAN.

Methods: Included patients were aged < 20 years, seen for a new eating disorder (ED) visit at a single center between 1/2015-2/2020, and diagnosed with AN or AAN. Patients with previous ED treatment or were weight restored were excluded. Medical admission vs. outpatient management treatment was compared across demographic and physiologic variables. Linear regression was used to assess whether BMI%ile and percentage of total weight loss (% weight loss) were independently associated with severity of bradycardia or whether % weight loss mediated a relationship between BMI%ile and bradycardia. Logistic regression was used to examine if BMI%ile significantly predicted admission recommendation while controlling for age and vitals.

Results: Of the overall cohort (N=292), 97 (33%) were recommended medical admission, most commonly for bradycardia (90%). Admission vs. outpatient groups did not differ by age (15.8 vs. 15.2 years), sex (87% vs. 84% female), gender identity (97% vs. 99% cis-gender), or race (88% vs. 94% white). The admission cohort had significantly lower BMI%ile (13.3% vs. 37.7%, p< 0.001), heart rate (HR; 44 vs. 70 BPM, p< 0.001), systolic (SBP; 101 vs. 112 mmHg, p< 0.001) and diastolic blood pressures (DBP; 61 vs. 65 mmHg, p< 0.001). Linear regressions suggested that lower BMI%ile (p < 0.001) and higher % weight loss (p=0.002) were independent predictors of lower HR when controlling for rate of weight loss and age. However, the relationship between BMI%ile and HR was fully mediated by % weight loss (p=0.02). Logistic regression found that higher BMI%ile was associated with lower odds of admission recommendation (aOR=0.98[95%CI=0.96-0.99], p=0.006) when controlling for age, HR, SBP, and DBP. Two sensitivity analyses were performed to assess clinical significance; when controlling for age and vitals, findings remained consistent with bradycardia as a binary variable (using admission cutoff of HR≤45 BPM) (aOR=0.97[95%CI=0.96-0.99], p=0.002) and with BMI as a binary variable (≤10th percentile=AN, >10th percentile=AAN). Patients with AAN had significantly lower odds of admission recommendation (aOR=0.30[95%CI=0.14-0.66], p=0.003).

Conclusion: Percentage of total weight loss appears to be a better predictor than presenting BMI percentile for medical severity in youth with AN/AAN. However, patients with higher presenting BMI percentiles are less likely to be recommended for medical admission even when controlling for physiologic parameters indicating illness severity. Practitioners should take care to avoid biases regarding presenting weight when evaluating illness severity.