Components of traumatic brain injury severity indices

The purpose of this study was to determine whether there are underlying dimensions common among traditional traumatic brain injury (TBI) severity indices and, if so, the extent to which they are interchangeable when predicting short-term outcomes. This study had an observational design, and took place in United States trauma centers reporting to the National Trauma Data Bank (NTDB). The sample consisted of 77,470 unweighted adult cases reported to the NTDB from 2007 to 2010, with International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) TBI codes. There were no interventions. Severity indices used were the Emergency Department Glasgow Coma Scale (GCS) Total score and each of the subscales for eye opening (four levels), verbal response (five levels), and motor response (six levels); the worst Abbreviated Injury Scale (AIS) severity score for the head (six levels); and the worst Barell index type (three categories). Prediction models were computed for acute care length of stay (days), intensive care unit length of stay (days), hospital discharge status (alive or dead), and, if alive, discharge disposition (home versus institutional). Multiple correspondence analysis (MCA) indicated a two dimensional relationship among items of severity indexes. The primary dimension reflected overall injury severity. The second dimension seemed to capture volitional behavior without the capability for cogent responding. Together, they defined two vectors around which most of the items clustered. A scale that took advantage of the order of items along these vectors proved to be the most consistent index for predicting short-term health outcomes. MCA provided useful insight into the relationships among components of traditional TBI severity indices. The two vector pattern may reflect the impact of injury on different cortical and subcortical networks. Results are discussed in terms of score substitution and the ability to impute missing values.