Logistic regression model with 26 predictive features has area under receiver operating curve of 0.77
By Elana Gotkine HealthDay Reporter
WEDNESDAY, Dec. 27, 2023 (HealthDay News) — Predictive modeling in neonatal care can identify very preterm infants at risk for developing cognitive impairment at 2 years of age, according to a study published online Dec. 26 in JAMA Network Open.
Andrea K. Bowe, M.P.H., from the University College Cork in Ireland, and colleagues conducted a prognostic study to predict cognitive delay at 2 years of age using routinely available clinical and sociodemographic data for very preterm infants (gestational age younger than 32 weeks). To predict cognitive delay, various machine learning models were trained and tested.
The modeling process included 1,062 children, of whom 231 (21.8 percent) had cognitive delay. The researchers found that the area under the receiver operating characteristic curve was 0.77 for a logistic regression model that included 26 predictive features. Non-Scandinavian family language, prolonged duration of hospitalization, low birth weight, discharge to destination other than home, and not receiving breastmilk on discharge were the five most important features for cognitive delay. The full model could correctly identify 605 of 650 infants at discharge from the neonatal unit who would have cognitive delay at 24 months and 1,081 of 2,350 who would not (sensitivity and specificity, 0.93 and 0.46, respectively).
“Other research groups using neurophysiological, microbiome, or imaging data should consider inclusion of the 26 clinical features identified herein in coming predictive models,” the authors write.
One author disclosed ties to Liltoda Ltd., a company that focuses on technology for early cognitive assessment in children.
Copyright © 2023 HealthDay. All rights reserved.
