Statistical analysis

We used SAS (SAS Studio) and R (version 4.0.8) for statistical analyses. We described the demographic characteristics of the study population and calculated the median and interquartile range (IQR) of age at first height or weight data, length of follow-up, and counts of height and weight measurements. Next, we determined the number of exclusions by growthcleanr for both pediatric and adult data. For pediatric data, we compared the results of growthcleanr with the other 2 cleaning methods (cp and peanof) using a Sankey diagram (R package, networkD3²¹), a flow diagram in which the width of the arrows depicts the proportion to the flagged values across each cleaning method. Based on the 2000 CDC Growth Chart,²² we calculated age- and sex-specific BMI percentile and defined severe obesity as >120% above the 95th percentile,²³ using same-day height and weight measurements below 20 years of age. In the case of missing same-day height, we imputed height using the closest height measurement ±60 days (for measurements <18 years) or mean of all height measurements above age 18 (for measurements ≥18 years).