Abstract:
Objective: To develop and internally validate a prediction model examining the use of neonatal risk factors to predict neonatal mortality in infants with neonatal encephalopathy requiring therapeutic hypothermia using national-level data.
Methods: We used maternal and neonatal data from medical records of infants who received therapeutic hypothermia in 19 maternity hospitals over a six-year period (2016 to 2021), collected through the National Neonatal Therapeutic Hypothermia Development Project. Multivariable logistic regression with backward stepwise selection was used to develop the model. Potential predictors included variables on maternal characteristics, labour and delivery outcomes, and resuscitation and biochemical characteristics. An initial prediction model used all variables 2 significant in the univariable analysis, followed by a simplified model which included less candidate variables in the model while maintaining predictive performance. Discrimination was assessed using the area under the receiver operating characteristic curve (ROC) C- statistic. Internal validation was performed using bootstrapping to evaluate overfitting, and the THERM (Therapeutic Hypothermia Early Risk Model for Mortality) Excel-based tool was created to enable clinicians to calculate individualised mortality risk.
Results: Among 385 infant-mother pairs, 46 (11.9%) infants died within the first 28 days of life. Using univariable and multivariable logistic models, four variables emerged as the best predictors of neonatal mortality; mode of delivery, use of adrenaline, base excess ≤ - 22mmol/L, and seizures during the first day of life. The original apparent C-statistic in the final model was 0.903 (0.860, 0.948) suggesting excellent discriminative performance. There was minimal optimism adjustment to the C-statistic after bootstrapping, indicating good internal performance (optimism adjusted C-statistic: 0.885 [0.827, 0.936]).
Conclusions: Four routinely collected clinical and biochemical variables were identified as predictors of neonatal mortality in infants undergoing therapeutic hypothermia. The THERM tool offers a practical resource for clinicians, enabling personalised risk assessment and aiding parental counselling during the first day of life.
