Abstract 13467: Using Data Analytics Toward Precision Warfarin Dosing in Children With Heart Disease

Abstract

Introduction: Warfarin is the most common anticoagulant used in children with heart disease. Multiple factors make initial warfarin dose selection and management subjective and variable in children. This variability increases hospital length of stay and impacts efficacy and safety. We sought to develop a data-driven tool to optimize time to warfarin goal dose in children with heart disease. Methods: We collected all demographic and clinical data for all patients <18 years who received warfarin at our institution. We developed a warfarin sensitivity metric (Δ International Normalized Ratio (INR)/mg warfarin) to assess individual medication processing and characterize influential factors. Each inpatient interaction with warfarin was analyzed to assess the impact of concomitant medications, hepatic impairment, and procedure type on dosing, sensitivity, and resultant INR. A linear regression model was formulated to predict ending warfarin dose based on these influential factors. Results: A total of 714 patients were identified; 334 patients met inclusion criteria, and 447 unique encounters with warfarin were analyzed. Warfarin sensitivity was 94.3% higher in patients receiving major anticoagulation enhancing drugs and 13.6% higher in patients with minor liver injury. Patients undergoing a Fontan procedure had increased warfarin sensitivity compared to other procedures; 60% and 66% increase from ventricular assist device and valve replacement, respectively. A linear regression model was used to predict a patient’s warfarin ed dose based on age, baseline INR, other medications, liver injury, and Fontan procedure status. The Fontan binary indicator and age were the most impactful in predicting end dose with an adjusted R 2 of 0.1792. Conclusions: We have identified multiple factors that could affect individualized warfarin. Further analysis will refine our warfarin sensitivity metric and finalize and validate our tool to allow accurate and efficient warfarin dose prediction.