Implementation of an automated, real-time mortality prediction tool in trauma patients: Can it do more than just predict mortality?

Abstract

Introduction: The Parkland Trauma Index of Mortality (PTIM) is a validated machine learning tool for real-time prediction of 48-hour mortality in trauma patients. Methods: This retrospective study at a Level I Trauma center analyzed PTIM scores and their association with clinical outcomes in 171 patients from December 2020 to August 2021. A PTIM score of 1.0 predicts 100 % mortality. Results: We included a total of 171 patients, with a mean PTIM score of 0.26 [median = 0.1825, range =0.89, IQR=0.37]. For every 0.1 increase in PTIM score, ICU days increased by 0.966 days, ventilator days increased by 0.724 days, hospital stay increased by 1.668 days, and blood transfusion needs increased by 0.37 units. An increase in PTIM score was also associated with higher odds of shock, pneumonia, and mortality. PTIM did not significantly impact time to the operating room. Conclusion: These findings demonstrate PTIM's utility in anticipating resource utilization and patient outcomes. Future research aims to validate PTIM externally and explore additional applications for this real-time prediction model.