Related Experiment Video
Updated: Oct 16, 2025
07:21
Systems Analysis of the Neuroinflammatory and Hemodynamic Response to Traumatic Brain Injury
Published on: May 27, 2022
3.3K
Parkland Trauma Index of Mortality: Real-Time Predictive Model for Trauma Patients
Adam J Starr1, Manjula Julka2, Arun Nethi2
1Department of Orthopaedic Surgery, UT Southwestern Medical Center, Dallas, TX.
Journal of Orthopaedic Trauma
|October 15, 2021
Summary
The Parkland Trauma Index of Mortality (PTIM) is a dynamic machine learning model that accurately predicts trauma patient mortality within 48 hours. This tool uses electronic health record data to improve clinical decision-making for critically injured patients.
Area of Science:
- Trauma care
- Machine learning in medicine
- Prognostic modeling
Background:
- Traditional methods for guiding trauma care decisions rely on static vital signs and lab values.
- Existing mortality prediction models have limited utility due to their inability to dynamically capture evolving patient physiology.
- There is a critical need for dynamic prognostic tools that adapt to real-time physiological changes in trauma patients.
Purpose of the Study:
- To develop and validate a dynamic machine learning model, the Parkland Trauma Index of Mortality (PTIM), for predicting in-hospital mortality in trauma patients.
- To assess the accuracy and clinical utility of PTIM in forecasting mortality within 48 hours during the initial days of hospitalization.
- To provide a tool that aids early clinical decision-making in trauma care.
Main Methods:
- The Parkland Trauma Index of Mortality (PTIM) was developed as a machine learning algorithm utilizing electronic medical record data.
- The model was trained on 1935 trauma patient encounters (2009-2014) and validated on 516 encounters (2015-2016).
- PTIM updates hourly, recalculating mortality predictions based on continuously changing physiological data.
Main Results:
- In validation data, PTIM demonstrated high accuracy with 82.5% sensitivity and 93.6% specificity for predicting mortality within 48 hours.
- The negative predictive value was 99.3%, indicating strong reliability in identifying patients likely to survive.
- During its first year of clinical use, PTIM achieved 86.9% sensitivity and 94.7% specificity, with a 99.6% negative predictive value.
Conclusions:
- The PTIM model dynamically adapts to patient physiology using only electronic medical record data, overcoming limitations of prior static models.
- PTIM shows significant potential to enhance early clinical decision-making for trauma patients.
- This dynamic prognostic tool offers improved risk assessment for critically injured individuals.
