A new hybrid classification system for traumatic brain injury which helps predict long-term consciousness: a...

To develop and validate a refined traumatic brain injury (TBI) classification system to supplement the existing systems which have limited accuracy for predicting long-term consciousness recovery.

The refined classification system was developed using medical records of 527 patients according to clinical presentations within 12-24 hrs after injury. Multiple linear regression was applied to identify protective and risk factors for Glasgow Coma Scale (GCS) and Glasgow Outcome Scale (GOS) score at 12-month follow-up. The TBI severity was moved to a less or more severe level when more than half of the protective or risk factors were present. The capability and reliability of each system for predicting 12 month GCS and GOS scores, and mortality were assessed using ROC curve analysis and Cronbach's Alpha reliability coefficient.

One protective factor and four risk factors were identified for predicting long-term outcomes. The refined system had higher sensitivity and specificity in predicting 12-month GCS and GOS scores, and mortality than the other two systems. The refined system had lower reliability than the GCS system and higher reliability than the Chinese system.

The refined system incorporates the advantages of both GCS and Chinese systems and provides a better prediction of long-term consciousness outcome.