Enhancing Detection of Cerebral Palsy: Multimodal Developmental Assessments in High-Risk Infants

OBJECTIVE To test the predictive performance of CHIP, an assessment that integrates the Capute Scales, Hammersmith Infant Neurological Examination, and Prechtl’s General Movement Assessment, in relation to future Cerebral Palsy (CP) outcomes. METHODS We designed CHIP, an assessment that combines the Capute Scales (CAT/CLAMS), Hammersmith Infant Neurological Examination, and Prechtl’s General Movement Assessment (GMA) to evaluate cognitive, language, and motor development in high-risk 3-4-month-old infants. To test our hypothesis that CHIP predicts later CP outcomes, we conducted a prospective cohort study in former neonatal intensive care unit infants, using CP status at 2-3 years of age as the primary outcome. We performed univariate and multivariable logistic regression analyses with leave-one-out cross-validation to create Receiver Operator Characteristic curves and assess the accuracy, sensitivity, and specificity of CHIP for classifying infants with and without CP. Best Subset Selection was used to assess the performance of reduced models incorporating every subset of CHIP assessments. For 10 children lost to follow-up, the prediction equations were used to predict their outcomes. RESULTS Prior to the 2-year mark, 10 children were lost to follow-up from our initial cohort of 108 infants. Of the remaining 98 children, 17 had CP, and 81 did not have CP. The CHIP screening model predicted CP well (accuracy: 0.898; sensitivity: 1.000; specificity: 0.877). Best Subset Selection identified some reduced models that performed similarly well (e.g., GMA+CAT+CLAMS). CONCLUSIONS CHIP can provide an early indicator for the prognostication of CP in the clinical setting.