Predicting outcomes in severe traumatic brain injury: A prospective analysis of clinical, radiological, and laboratory indicators

Background: Traumatic brain injury (TBI) is a primary cause of mortality and disability globally, with severe cases often resulting in long-term impairment. This study investigated key admission characteristics that predict outcomes in severe TBI patients, enhancing early mortality prediction through a comprehensive prognostic tool. Methods: This prospective study involved 240 patients with severe TBI (Glasgow Coma Scale [GCS] ≤8) admitted to our university hospitals over 2.5 years. Clinical, radiological, and laboratory data were collected, including GCS, Best motor response (BMR), pupil reactivity, hypotension, and the Marshall computed tomography (CT) score. Transcranial Doppler (TCD) pulsatility index (PI) and S100B protein levels were also assessed. Outcomes were assessed by the Glasgow outcome score (GOS). Results: Key findings demonstrated significant correlations between GCS, BMR, pupil reactivity, hypotension, and CT scores with GOS outcomes. Lower GCS scores and higher PI on Day 3 were associated with increased mortality. The madras head injury prognostic scale (MHIPS) effectively predicted outcomes with a strong correlation to GOS ( P < 0.001). For patients with severe TBI, predictive thresholds for PI were established. On Day 1, a PI cutoff of ≤1.27 showed a sensitivity of 75%, specificity of 72.7%, and accuracy of 72.9% (area under the receiver operating characteristic curve [AUROC] 0.760, P = 0.001). On Day 2, a cutoff of ≤1.30 had 100% sensitivity, 49.4% specificity, and 53% accuracy (AUROC 0.694, P = 0.018). On Day 3, a cutoff of ≤1.03 provided a sensitivity of 87.5%, specificity of 56.3%, and accuracy of 69.4% (AUROC 0.729, P = 0.024). For mortality prediction, a Day 3 PI cutoff of >0.97 yielded a sensitivity of 86.5%, specificity of 51.4%, and accuracy of 69.4% (AUROC 0.692, P = 0.002). S100B protein measured on Day 1 at >264 ng/L had a sensitivity of 76.9%, specificity of 65%, and accuracy of 69.7% (AUROC 0.723, P = 0.027). Conclusion: This study underscores the importance of integrating clinical, radiological, and laboratory parameters into prognostic models for severe TBI. The findings emphasize the predictive power of the TCD-measured PI and S100B protein levels as practical tools for estimating patient outcomes and mortality risk, facilitating improved clinical decision-making and resource allocation in intensive care settings.