Balancing Early Indicators With Patient-Centered Care for HIE

Source: Glass HC, Wood TR, Comstock BA, et al. Predictors of death or severe impairment in neonates with hypoxic-ischemic encephalopathy. JAMA Netw Open. 2024;7(12):e2449188; doi: 10.1001/jamanetworkopen.2024.49188.Investigators from multiple institutions conducted a study to identify predictors of death or severe neurodevelopmental impairment (NDI) in neonates with hypoxic-ischemic encephalopathy (HIE) who were treated with hypothermia. For the study, they evaluated data originally collected for the High-dose Erythropoietin for Asphyxia and Encephalopathy (HEAL) trial. For this study, neonates born at ≥36 weeks’ gestation with HIE who were treated with hypothermia were enrolled. Data on multiple clinical findings were collected. In addition, participants had an electroencephalogram (EEG) obtained at approximately 24 hours of age and a brain MRI after cooling (about 5 days of life). Neurodevelopment was assessed in survivors at approximately 2 years of age. For the current study, the primary outcome was a composite of death or severe NDI at 2 years, with a classification of severe NDI based on neurodevelopmental assessment scores. Data on clinical characteristics, EEG, and MRI results were included in a recursive partitioning model to identify a set of risk factors for the primary outcome. Both a model at 24 hours and 5 days of life were developed. The model was trained on 70% of study participants and internally validated on 30%. In addition, the model was externally validated in a group of neonates with HIE, born in the United Kingdom (UK). The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the models in predicting death or severe NDI in each of the samples were calculated.Data on 424 infants enrolled in HEAL were analyzed; 59 (13.9%) died, and 46 (10.8%) had severe NDI. At 24 hours, 3 factors were highly predictive of the outcome: severely abnormal EEG, lowest cord blood gas pH ≤7.11, and 5-minute APGAR of 0. The presence of these 3 factors had a sensitivity of 27.4%, specificity of 99.6%, PPV of 95.2%, and NPV of 80.6% for predicting death or severe NDI in the training set and 21.9%, 97.9%, 77.8%, and 79.0%, respectively, in the internal validation set. At 5 days, the presence of abnormalities on MRI in 2 of 3 deep gray regions (thalamus, caudate, and putamen or globus pallidus) and a severely abnormal EEG at 24 hours had a sensitivity of 45.8%, specificity of 99.1%, PPV of 91.7%, and NPV of 89.4% in the training sample for predicting the outcome. Corresponding values in the internal validation sample were 56.5%, 98.9%, 92.9%, and 90.3%. These prediction models were assessed in 365 UK infants with HIE. The sensitivity, specificity, PPV, and NPV of the 24-hour model in predicting death or severe NDI in these infants were 6.6%, 97.6%, 46.2%, and 77.1%, respectively, and 38.5%, 98.6%, 83.3%, and 89.8%, respectively, for the 5-day model.The authors conclude that a limited number of factors had a high specificity and PPV in predicting death or severe NDI in neonates with HIE.Dr Beam has disclosed no financial relationship relevant to this commentary. This commentary does not contain a discussion of an unapproved/investigative use of a commercial product/device.Outcomes of HIE can be difficult to predict, as evidenced by multiple published predictive models.1 In the current study, the authors developed a simple predictive model utilizing key findings both in the first 24 hours as well as at 5 days of life to help predict overall outcomes of infants affected by HIE. The ability to utilize standardized elements that are common in all infants undergoing therapeutic hypothermia make this predictive model more broadly useful than prior single-center studies of smaller sample sizes.Predictive models have to be interpreted carefully, particularly in terms of a potential “self-fulfilling prophecy.”2 For example, the fact that a 5-minute Apgar of 0 or a 24-hour EEG predicts death or severe NDI may be confounded by the subjective assignment of the low Apgar as well as EEG findings that are concerning in the first 24 hours. The combination of these 2 things may affect how a provider counsels a family. This in turn can affect how a model’s predictive ability is determined. In the current study, the MRI results incorporated into the model at 5 days help to refine the prediction of the initial model.Predictive models for HIE are useful when discussing potential outcomes with families who have an infant affected by this condition.3 Simpler models are more easily incorporated into clinical care, and the ability to develop a model based on common factors is important for clinical use.4 Understanding how information is incorporated into models is key to implementation for clinical use.Simple predictive models can help to outline discussions with families around possible outcomes for infants with severe HIE, but understanding family goals and beliefs is key to providing appropriate family-centered care.In interpreting the results of this study, remember the “sensitivity-rule out, specificity-rule in” (SNOUT/SPIN) adage. The high specificity combined with low sensitivity of the prediction rules suggest that the presence of the risk factors was useful in “ruling in” a poor outcome, but the absence of the risk factors did not necessarily suggest a more positive outcome.