Abstract 19029: Cardio-Oncology Incidence, Cardiac Arrest, and Mortality for Adolescent versus Pediatric Patients: Machine Learning Propensity Score National Analysis

Introduction: The incidence of cardio-oncology disease, arrest, and inpatient mortality among adolescent and pediatric versus adult patients is largely unknown, which undermines optimal care of these high-risk patients.

Methods: BAyesian Machine learning-augmented Propensity Score (BAM-PS) translational statistics with multivariable regression was performed on the National Inpatient Sample.

Results: Of the 142,651,868 hospitalizations from 2017-2020, 13.69% were pediatric (0-13 years old), 1.13% were adolescent (14-17 years old), and 85.18% were adult patients (over 17 years old). Adults were significantly more likely than adolescent and pediatric patients to have cancer and cardiac disease (99.37% versus 0.15% versus 0.47%, p<0.001). The most common primary malignancies among adolescent or pediatric patients with cardio-oncology disease were leukemia (39.66% versus 52.60%), bone/connective (27.00% versus 12.43%), brain/nervous (8.02% versus 13.80%), Hodgkin lymphoma (7.17% versus 1.09%), and renal (6.01% versus 1.27%). In active cardio-oncology disease by BAM-PS multivariable regression controlling for clinical confounders and severity, adolescent and pediatric patients had comparable mortality odds to adults. Outside of active cardio-oncology disease, pediatric patients (OR 4.95, 95%CI 4.70-5.20, p<0.001) had significantly greater mortality odds than adults, as did adolescent patients but with less magnitude of increased odds (OR 2.03, 95%CI 1.78-2.31, p<0.001). In active cardio-oncology disease, pediatric patients (OR 0.17, 95%CI 0.04-0.68, p=0.013) had significantly lower odds of arrest compared to adults, with adolescents having comparable arrest odds versus adults. Among adolescent and pediatric patients with active cardio-oncology disease, African American versus Caucasians (OR 4.67, 95%CI 1.18-18.42, p=0.028) had significantly greater mortality odds.

Conclusions: This machine learning multi-year national study suggests the existence of increased mortality in African American adolescent and pediatric patients with concomitant cancer and cardiovascular disease. Overall, adolescent and pediatric patients with cardio-oncology have better outcomes with more effective long-term management.