Heart rate patterns predicting cerebral palsy in preterm infants.

Area of Science:

• Neonatal neurology
• Cardiovascular physiology
• Biomedical data science

Background:

• Heart rate (HR) patterns offer insights into central nervous system dysfunction.
• Previous research utilized highly comparative time series analysis (HCTSA) to predict mortality in neonatal intensive care unit (NICU) patients.
• This study applies HCTSA to identify HR patterns predicting cerebral palsy (CP) in preterm infants.

Purpose of the Study:

• To discover novel heart rate (HR) metrics for predicting cerebral palsy (CP) in preterm infants.
• To develop a parsimonious prediction model for CP using HR data from the NICU.
• To leverage highly comparative time series analysis (HCTSA) for identifying subtle HR patterns indicative of neurological conditions.

Main Methods:

• Studied NICU patients (<37 weeks' gestation) with archived every-2-second HR data.
• Performed HCTSA on over 2000 HR metrics from two 7-day periods: week 1 and 37 weeks' postmenstrual age.
• Utilized multivariate modeling to optimize a prediction model incorporating HR metrics and birthweight.

Main Results:

• Early HR metrics (week 1) showing low variability, such as "RobustSD" (AUC 0.826), were significant predictors of CP.
• At week 37, a novel metric "LongSD3" (indicating recurrent HR acceleration) combined with early HR metrics improved CP prediction.
• A combined model including birthweight, and early and late HR features achieved an AUC of 0.853 for CP prediction.

Conclusions:

• HCTSA successfully identified novel HR characteristics for CP prediction in preterm infants.
• A parsimonious model combining early (low variability) and late (recurrent acceleration) HR patterns, along with birthweight, effectively predicts CP.
• These findings offer a new non-invasive method for early identification of CP risk in high-risk neonatal populations.