Multivariable Predictive Models of Death or Neurodevelopmental Impairment Among Extremely Low Birth Weight Infants Using Heart Rate Characteristics

Objective: We hypothesized that a cumulative heart rate characteristics (HRC) index in real-time throughout the neonatal intensive care unit (NICU) hospitalization, alone or combined with birth demographics and clinical characteristics, can predict a composite outcome of death or neurodevelopmental impairment (NDI).

Study design: We performed a retrospective analysis using data from extremely low birth weight infants who were monitored for HRC during neonatal intensive care. Surviving infants were assessed for NDI at 18-22 months of age. Multivariable predictive modeling of subsequent death or NDI using logistic regression, cross-validation with repeats, and step-wise feature elimination was performed each postnatal day through day 60.

Results: Among the 598 study participants, infants with the composite outcome of death or moderate-to-severe NDI had higher mean HRC scores during their stay in the NICU (3.1 ± 1.8 vs 1.3 ± 0.8; P < .001). Predictive models for subsequent death or NDI were consistently higher when the cumulative mean HRC score was included as a predictor variable. A parsimonious model including birth weight, sex, ventilatory status, and cumulative mean HRC score had a cross-validated receiver-operator characteristic curve as high as 0.84 on days 4, 5, 6, and 8 and as low as 0.78 on days 50-52 and 56-58 to predict subsequent death or NDI.

Conclusions: In extremely low birth weight infants, higher mean HRC scores throughout their stay in the NICU were associated with a higher risk of the composite outcome of death or NDI.

Trial registration: ClinicalTrials.gov: NCT00307333.