Evaluating EHR-based prediction of obesity using who weight velocity standards and peak BMI

Cincinnati Children's Hospital Medical Center received a $456,914 Project Grant from the National Institute of Child Health and Human Development (NICHD) under the Child Health and Human Development Extramural Research program (CFDA 93.865), effective September 1, 2025, through August 31, 2027. The research project, "Evaluating EHR-Based Prediction of Obesity Using WHO Weight Velocity Standards and Peak BMI," addresses the critical gap in infant growth assessment methodologies for identifying early obesity risk. The project will evaluate electronic health record (EHR)-based approaches to predict obesity risk using World Health Organization (WHO) weight velocity standards and peak body mass index (BMI) metrics, aiming to develop practical clinical tools that improve upon current imprecise infant growth screening methods used in pediatric practice.

The research deliverables include the development and validation of computationally-enhanced EHR systems capable of applying WHO weight velocity standards—which account for age-specific and sex-specific growth trajectories across varying measurement intervals—to enable practical obesity risk screening in clinical settings. Duke University's Office of Research Administration serves as a sub-grantee collaborating on the project. The expected outcomes will provide clinicians and researchers with more accurate, implementable metrics for identifying infants at risk of early-onset obesity, addressing limitations of currently used growth metrics such as rapid infant weight gain thresholds that suffer from inconsistent definitions and misinterpretation.

Abstract

Obesity is a common, pervasive, chronic disease often beginning in early childhood. Although infant growth characteristics are key signals of lifelong chronic disease risk, there are no current guidelines about how best to screen for infant growth patterns associated with future obesity risk. Several growth metrics have been used in research settings to evaluate patterns of infant growth using attained weight or BMI z-scores, changes in attained weight-for-age (WFA) z-scores, and modeled trajectories of weight or body mass index (BMI). However, each of these metrics suffers from imprecision or difficulty being implemented in pediatric clinical practice. In particular, the frequently-used metric of change in WFA >+0.67 z-score units, dubbed "rapid infant weight gain", makes incorrect assumptions about infant weight gain, is inconsistently defined, and is often misinterpreted. In 2009, the who published globally-representative infant weight velocity (wtvel) standard distributions of normal physiological growth by SEX, age (monthly to 24 months), and the specific time interval between measurements (1, 2, 3, 4 or 6 months). These wtvel standard distributions account for changes in expected growth rates as infants age yet are rarely used in either clinical practice or research, largely because they are impractical to apply (e.g., 194 different reference charts, one for each combination of interval length, SEX, and age). Inappropriate or poorly-specified infant growth metrics or misinterpretation of metrics by clinicians threaten our ability to identify infants at risk of developing early-onset obesity. Thus, we have an urgent need to rethink how infant growth is assessed and interpreted both in clinical practice and in research settings. With the advancement in computational capacity in electronic health records (EHR) systems, IT is now necessary to examine the clinical performance of more complex research-based infant growth velocity metrics. The present study aims to assess the feasibility and predictive validity of various metrics of infant growth velocity among 1.8 million children from a large national clinical EHR dataset. In aim 1, we will compare the practical application and demographic inclusivity of 1) who wtvel z-scores, 2) modifications to these who methods (wtvel-m), 3) infant peak weight velocity (PWV), 4) infant peak BMI, and 5) change in WFA z-scores using individual longitudinal clinical data. In aim 2, we will test the ability of each metric to predict obesity by age 5. We will also externally validate the predictive performance of each metric using an independent EHR dataset. By the completion of this study, we expect to have rigorously evaluated both existing research-based and novel metrics of infant growth, characterizing their generalizability and validity to predict childhood obesity. Our findings will be a critical step towards evidence-based tailored risk prediction of childhood obesity in the clinical setting, bridging the GAP in evidence between research-based growth standards and clinical practice.