A dynamic Norwood mortality estimation: Characterizing individual, updated, predicted mortality trajectories after...

Post-Norwood mortality remains high and unpredictable. Current models for mortality do not incorporate interstage events. We sought to determine the association of time-related interstage events, along with (pre)operative characteristics, with death post-Norwood and subsequently predict individual mortality.

From the Congenital Heart Surgeons' Society Critical Left Heart Obstruction cohort, 360 neonates underwent Norwood operations from 2005 to 2016. Risk of death post-Norwood was modeled using a novel application of parametric hazard analysis, in which baseline and operative characteristics and time-related adverse events, procedures, and repeated weight and arterial oxygen saturation measurements were considered. Individual predicted mortality trajectories that dynamically update (increase or decrease) over time were derived and plotted.

After the Norwood, 282 patients (78%) progressed to stage 2 palliation, 60 patients (17%) died, 5 patients (1%) underwent heart transplantation, and 13 patients (4%) were alive without transitioning to another end point. In total, 3052 postoperative events occurred and 963 measures of weight and oxygen saturation were obtained. Risk factors for death included resuscitated cardiac arrest, moderate or greater atrioventricular valve regurgitation, intracranial hemorrhage/stroke, sepsis, lower longitudinal oxygen saturation, readmission, smaller baseline aortic diameter, smaller baseline mitral valve z-score, and lower longitudinal weight. Each patient's predicted mortality trajectory varied as risk factors occurred over time. Groups with qualitatively similar mortality trajectories were noted.

Risk of death post-Norwood is dynamic and most frequently associated with time-related postoperative events and measures, rather than baseline characteristics. Dynamic predicted mortality trajectories for individuals and their visualization represent a paradigm shift from population-derived insights to precision medicine at the patient level.