A Longitudinal Comparison of Alternatives to CDC BMI Z-Scores for Children With Very High BMIs.

Abstract Objective The current CDC BMI z-scores are inaccurate for BMIs ≥ 97th percentile. We, therefore, considered 5 alternatives that can be used across the entire BMI distribution: modified BMIz, %CDC95th percentile, extended BMIz, %median, and %median adjusted for the dispersion of BMIs. Study design We illustrate the behavior of the metrics among children of different ages and BMIs. We then compared the longitudinal tracking of the BMI metrics in electronic health record (EHR) data from 1.17 million children in PEDSnet using the intraclass correlation coefficient (ICC) to determine if one metric was superior. Results Our examples show that using CDC BMIz for high BMIs can result in nonsensical results. All alternative metrics showed higher tracking than CDC BMIz among children with obesity. Of the alternatives, modified BMIz performed poorly among children with severe obesity, and %median performed poorly among children who did not have obesity at their first visit. The highest ICCs were generally seen for extended BMIz, adjusted %median, and %CDC95th percentile. Conclusions Based on the examples of differences in the BMI metrics, the longitudinal tracking results, and current familiarity BMI z-scores and percentiles, extended BMIz and extended BMI percentile may be suitable replacements for the current z-scores and percentiles. These metrics are identical to those in the CDC growth charts for BMIs