Implementation of an Automated, Real-Time Machine Learning Algorithm Predicting Mortality in Trauma Patients: Survey Says Ready for Prime Time

Abstract

INTRODUCTION: Although artificial intelligence has been increasingly applied to patient care, many of these predictive models are retrospective and not readily available for real-time decision making. This survey-based study aims to evaluate implementation of a new, validated mortality risk calculator (MRC) embedded in our electronic medical record that calculates hourly predictions of mortality with high sensitivity and specificity. METHODS: This is a prospective, survey-based study at a Level I trauma center. An anonymous survey was sent to ICU providers about MRC implementation. The MRC score evaluates 23 variables including Glasgow Coma Score, vital signs, and laboratory data. RESULTS: Of the 36 completed surveys, 31 reported using MRC in decision-making. Before reviewing the MRC, providers proposed the top 3 predictors of mortality, including Glasgow Coma Score (22/33, 67%), age (16/33, 48%), and maximum heart rate (17/33, 51%). The actual MRC score also ranked Glasgow Coma Score (20/20, 100%), age (18/20, 90%), and maximum heart rate (14/20, 70%) as the top 3 predictors of mortality. Most providers reported MRC assisted their treatment decisions (26/33, 79%), timing of operative intervention (23/33, 70%), and ability to assess the patient’s stability for nonurgent intervention with a high MRC (greater than 50; 19/22, 86%). CONCLUSION: Artificial intelligence algorithms are mostly retrospective and lag in real-time prediction of mortality. To our knowledge, this is the first real-time, automated algorithm predicting mortality in trauma patients. This MRC assists in decision-making, timing of intervention, and improves accuracy in assessing mortality. Next steps include evaluating the short- and long-term effect on patient outcomes.