Abstract 15657: Risk Based Scheduling for Pediatric Cardiac Catheterization

Abstract

Introduction: Machine learning techniques can predict operational capacity needs and improve understanding of procedural risk in heterogeneous patient populations. Our goal is to combine machine learning and computer simulation to design scheduling formats for pediatric cardiac catheterization maximizing patient safety and system efficiency. Methods: Data for 13,115 pediatric cardiac catheterizations from 2010-19 were used to develop a simulation model including 3 catheterization interventional laboratories. Data on case time, patient and procedural complexity, and adverse events were used to determine statistical distributions on patient, procedure, and system factors for key components of the catheterization lab process. The simulation model was used to evaluate scheduling plans implemented for historical and synthetic populations of patients, comparing both system efficiency and total lab time spent where the system risk of an adverse event is high. Results: A simple scheduling heuristic model was developed where one lab worked cases from lowest to highest risk and another lab worked from highest to lowest risk. This scheduling heuristic was applied to all cases in 2018 and compared against the original schedule. Using thresholds of 10% and 20% probability of an adverse event to separate low, medium- and high-risk times. Implementing the sorting schedule resulted in the reduction of minutes spent at high-risk equivalent to 11.57 days per year. Conclusions: This study demonstrates that predictive analytics and simulation using simple scheduling heuristics results in reduced time at increased risk. Future work includes improving the fidelity of the model to include specific providers, case cancellations, and other system complexities.