Classification of traumatic brain injury severity using retrospective data

Abstract

Objective: Accurate classification of traumatic brain injury (TBI) severity is essential to brain injury research. TBI heterogeneity complicates classification of the injury; is a significant barrier in the design of therapeutic interventions; and results in retrospective data which is difficult to translate. The objective of this study is to describe the differences in two current tools used in the classification of TBI severity, the Glasgow Coma Scale (GCS) and the head Abbreviated Injury Score (AIS), using retrospective data to compare their performance.

Methods: Using correlational and descriptive statistics, this study examined two TBI severity classification methods across a large sample of TBI patients (N = 56,131), who were treated at level I and level II trauma centers in the United States and were included in the 2010 National Sample Program (NSP) of the National Trauma Data Bank (NTDB®).

Results: The study population was 67% male, 67% non-Hispanic white, treated most often in trauma centers in the South (38%), with blunt trauma (93%) and from non-motor vehicle collisions (MVC’s) (56%). Observation of the AIS classification system demonstrated that it tends to over-score TBI severity compared to the GCS classification. The methods (GCS & AIS) had a weak, inverse relationship with a correlation coefficient (Pearson’s r) of -0.3980, which was significant at p < .001.

Conclusions: The current study addressed the difficulties associated with categorizing TBI severity when analyzing retrospective data.  Although AIS is commonly used to classify severity in retrospective data when GCS is unavailable, the relationship between the two scales is relatively unknown. Results show that AIS and GCS are more closely related for severely brain injured patients but in cases of mild and moderate injury, AIS is less predictive of GCS. Since they are often used in conjunction in identifying brain injured severity in retrospective data, researchers cannot be certain that the tools are similarly classifying mild, moderate, and severe injuries. This study reinforces the need for additional TBI severity classification methods, such as neuroimaging techniques and biomarkers.