Predictors of Death or Severe Impairment in Neonates With Hypoxic-Ischemic Encephalopathy
Author Information
Author(s): Glass Hannah C., Wood Thomas R., Comstock Bryan A., Numis Adam L., Bonifacio Sonia L., Cornet Marie-Coralie, Gonzalez Fernando F., Morell Adriana, Kolnik Sarah E., Li Yi, Mathur Amit, Mietzsch Ulrike, Wu Tai-Wei, Wusthoff Courtney J., Thoresen Marianne, Heagerty Patrick J., Juul Sandra E., Wu Yvonne W.
Primary Institution: University of California, San Francisco
Hypothesis
Readily available data from clinical, neurophysiology, and magnetic resonance imaging (MRI) reports could be used to develop a decision tree to predict neonates with at least 85% likelihood of death or severe neurodevelopmental impairment.
Conclusion
The study found that clinical, EEG, and MRI results can accurately predict death or severe neurodevelopmental impairment in infants with hypoxic-ischemic encephalopathy.
Supporting Evidence
- Among 424 neonates, 59 (13.9%) died and 46 (10.8%) had severe neurodevelopmental impairment.
- The presence of severely abnormal EEG, low pH, and a 5-minute Apgar score of 0 had a specificity of 99.6% for predicting death or severe NDI.
- External validation showed high specificity and positive predictive value for the models developed.
Takeaway
Doctors can use certain tests done in the first day of life to guess if a baby will have serious problems later on.
Methodology
The study used recursive binary partitioning to analyze clinical, EEG, and MRI data from neonates with hypoxic-ischemic encephalopathy.
Potential Biases
The study may have inherent biases due to the exclusion criteria of the HEAL Trial and the potential for a self-fulfilling prophecy in predicting death.
Limitations
The models were designed to optimize specificity rather than balance sensitivity and specificity, which may limit their use in predicting outcomes for children without identified risk factors.
Participant Demographics
Mean gestational age was 39.1 weeks; 45.3% were female, 6.6% Asian, 11.8% Black, and 73.3% White.
Statistical Information
P-Value
p<0.001
Confidence Interval
95% CI, 97.5%-100%
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website