Prediction modelling for trauma using comorbidity and 'true' 30-day outcome

Area of Science:

• Medical research
• Trauma care
• Epidemiology

Background:

• Trauma outcome prediction models routinely adjust for patient age.
• The role of comorbidity and its interaction with age in trauma outcome prediction remains uncertain.
• The Trauma Audit and Research Network (TARN) developed a revised risk adjustment model incorporating age and comorbidities, utilizing 30-day mortality data linked with the Office of National Statistics (ONS) database.

Purpose of the Study:

• To evaluate a revised Trauma Audit and Research Network (TARN) risk adjustment model for trauma outcomes.
• To assess the impact of incorporating comorbidity data into the TARN model.
• To determine if age and comorbidity interact in predicting trauma patient mortality.

Main Methods:

• Analysis of prospectively collected data from the TARN database (2010-2013) for 129,786 trauma admissions.
• Comparison of three models: current TARN, TARN with a modified Charlson comorbidity index, and TARN with ONS 30-day outcome data.
• Model performance evaluated using Area Under the Receiver Operating Curve (AuROC), Akaike information criteria, and goodness-of-fit tests.

Main Results:

• The revised model incorporating comorbidity (Model 2) showed a slightly improved predictive ability (AuROC 0.904) compared to the current TARN model (AuROC 0.896).
• The model using ONS 30-day outcome data (Model 3) had an AuROC of 0.897.
• No statistically significant interaction was detected between age and comorbidity in either of the revised models.

Conclusions:

• The inclusion of comorbidity in the TARN risk adjustment model enhances the prediction of trauma patient outcomes.
• Age and comorbidity appear to independently contribute to increased mortality risk following trauma.
• The findings support the use of comorbidity as an independent factor in trauma risk stratification.