Comparison of two prognostic models in trauma outcome

Area of Science:

• Trauma care research
• Medical statistics
• Injury severity scoring

Background:

• The UK's Trauma Audit and Research Network (TARN) uses a specific model for hospital performance reporting.
• TARN's risk adjustment model relies on a fractional polynomial transformation of the Injury Severity Score (ISS).
• The Trauma Mortality Prediction Model (TMPM) presents an alternative injury severity measure.

Purpose of the Study:

• To compare the anatomical injury components of the TARN model with the TMPM.
• To evaluate the predictive performance of TMPM against TARN's ISS-based model.

Main Methods:

• Analysis of 2011-2015 National Trauma Data Bank data (438,058 patient records).
• Estimation of probability of death using TARN's ISS transformation and TMPM.
• Model comparison via calibration curves, AUROC, AIC, and Hosmer-Lemeshow test on 80% training and 20% validation data.

Main Results:

• TMPM showed superior discrimination with an AUROC of 0.882 compared to TARN's 0.845.
• TMPM demonstrated better proximity to the true model (AIC: 18,204 vs. 21,163) and goodness-of-fit (32.4 vs. 153.0).

Conclusions:

• TMPM exhibits greater discrimination, model proximity, and goodness-of-fit than TARN's anatomical injury component.
• TMPM is recommended as a preferred injury severity measure for comparative trauma center assessments.