A comparison of methods for calculating expected body mass indices among youth with eating disorders

Percentage of median body mass index (mBMI) is an established method to calculate malnutrition severity for youth with eating disorders (ED) but does not account for individual variability in premorbid growth trajectories. We aimed to develop an individualized expected BMI (inBMI) measure and compare estimates of malnutrition severity derived from inBMI versus mBMI.

Using growth chart data for participants in a longitudinal ED registry (RECOVERY), we calculated percent of expected BMI at initial treatment presentation using two methods: (1) mBMI, the 50th percentile for age-and-sex, and (2) inBMI, defined as the 10% trimmed mean of each participant's premorbid BMI percentiles. We compared estimates of malnutrition severity between methods overall and by premorbid growth category.

Among 97 participants (mean age 15.7 years; 85% female), mean BMI percentile at presentation was 31.8 (SD = 26.6). A mean of 12.8 premorbid BMI measurements per participant were available to estimate inBMI. Mean %mBMI was 94.2 (SD = 17.7) vs. 86.8 (SD = 14.0) %inBMI, with modest correlation between measures (r = 0.29; p = 0.004), and mean difference (mBMI-inBMI) of 7.3 (SD = 19.2). Concordance between methods was high for those with premorbid growth trajectories in the 'normal' range (5th-84th). However, %mBMI was consistently lower than %inBMI for those with < 5th percentile trajectories (i.e., overestimating malnutrition) and higher for ≥ 85th trajectories (i.e., underestimating malnutrition).

Estimates of malnutrition severity differed meaningfully when using population median vs. individual growth trajectories. Individualized expected BMIs may offer a patient-centered alternative for assessing malnutrition severity, particularly among youth with premorbid growth in under or overweight ranges.