Predicting Underweight Status in Infants With Cleft Lip and/or Palate in 10 US Children's Hospitals

Abstract

ObjectiveTo develop and compare models predicting which infants with cleft lip and/or palate (CL/P) will be underweight at 4 months, using multicenter electronic health record data.DesignUsing data from birth to 3 months, we evaluated 63 anthropometric, clinical, healthcare-utilization, and demographic variables across 6 regression and machine-learning models to identify the strongest predictors of underweight status, overall and by cleft type.SettingTen US children's hospitals in PEDSnet, a national pediatric learning health network, 2009 to 2024.PatientsInfants with CL/P aged 0 to 4.5 months.Main Outcome MeasureUnderweight status at 4 months (range 3.5-4.5), defined as weight-for-age z-score (WAZ) < -2.ResultsAmong 5717 infants with CL/P (19% cleft lip, 35% cleft lip and palate, 46% cleft palate), 25.9% were underweight at 4 months. The strongest predictors were early (0-3 months) growth measures: mean WAZ (58% of predictive value), WAZ velocity, mean length-for-age z-score (LAZ), and LAZ velocity, plus the Child Opportunity Index (a neighborhood socioeconomic measure; ∼2%). These 5 predictors accounted for >85% of predictive value in every cleft type. All 6 models discriminated similarly (area under the receiver operating characteristic curve 0.87-0.90); GPBoost, which also accounts for hospital differences, produced risk estimates closest to the actual proportion of underweight infants.ConclusionsRoutine early-life growth measures (WAZ, LAZ, and their trajectories), together with neighborhood socioeconomic status, can identify infants with cleft at elevated risk of being underweight before 4 months, helping craniofacial and primary care teams provide timelier feeding and nutrition support and referrals to the highest-risk infants.