Objective: Our goal was to generate a preoperative risk-of-death prediction model in selected neonates with congenital heart disease undergoing surgery with deep hypothermic circulatory arrest.
Methods: We completed a single-center, prospective, randomized, double-blind, placebo- controlled neuroprotection trial in selected neonates with congenital heart disease requiring operations for which deep hypothermic circulatory arrest was used. An extensive database was generated that included preoperative, intraoperative, and postoperative variables. Variables (delivery, maternal, and infant related) were evaluated to produce a preoperative risk-of-death prediction model by means of logistic regression. An operative risk-of-death prediction model including duration of deep hypothermic circulatory arrest was also generated.
Results: Between July 1992 and September 1997, 350 (74%) of 473 eligible infants were enrolled with 318 undergoing deep hypothermic circulatory arrest. The mortality was 52 of 318 (16.4%), unaffected by investigational drug. The resulting preoperative risk model contained 4 variables: (1) cardiac anatomy (two-ventricle vs single ventricle surgery, with/without arch obstruction), (2) 1-minute Apgar score (</=5 vs >5), (3) presence of genetic syndrome, and (4) age at hospital admission for surgery (</=5 or >5 days). Mortality for two-ventricle repair was 3.2% (4/130). Mortality for single ventricle palliation was 25.5% (48/188) and was significantly influenced by Apgar score, genetic diagnosis, and admission age. The preoperative model had a prediction accuracy of 80%. The operative risk model included duration of deep hypothermic circulatory arrest, which significantly (P =.03) increased risk of death, with a prediction accuracy of 82%.
Conclusions: In this selected population, postoperative mortality risk is significantly affected by preoperative conditions. Identification of infants with varying mortality risks may affect family counseling, therapeutic intervention, and risk stratification for future study designs.