Prognostication of traumatic brain injury outcomes in older trauma patients: A novel risk assessment tool based on initial cranial CT findings

INTRODUCTION: Advanced age has been traditionally associated with worse traumatic brain injury (TBI) outcomes. Although prompt neurosurgical intervention (NSI, craniotomy or craniectomy) may be life-saving in the older trauma patient, it does not guarantee survival and/or return to preinjury functional status. The aim of this study was to determine whether a simple score, based entirely on the initial cranial computed tomography (CCT) is predictive of the need for NSI and key outcome measures (e.g., morbidity and mortality) in the older (age 45+ years) TBI patient subset. We hypothesized that increasing number of categorical CCT findings is independently associated with NSI, morbidity, and mortality in older patients with severe TBI. METHODS: After IRB approval, a retrospective study of patients 45 years and older was performed using our Regional Level 1 Trauma Center registry data between June 2003 and December 2013. Collected variables included patient demographics, Injury Severity Score (ISS), Abbreviated Injury Scale Head (AISh), brain injury characteristics on CCT, Glasgow Coma Scale (GCS), Intensive Care Unit (ICU) and hospital length of stay (LOS), all-cause morbidity and mortality, functional independence scores, as well as discharge disposition. A novel CCT scoring tool (CCTST, scored from 1 to 8+) was devised, with one point given for each of the following findings: subdural hematoma, epidural hematoma, subarachnoid blood, intraventricular blood, cerebral contusion/intraparenchymal blood, skull fracture, pneumocephalus, brain edema/herniation, midline shift, and external (skin/face) trauma. Descriptive statistics and univariate analyses were conducted with 30-day mortality, in-hospital morbidity, and need for NSI as primary end-points. Secondary end-points included the length of stay in the ICU (ICULOS), step-down unit (SDLOS), and the hospital (HLOS) as well as patient functional outcomes, and postdischarge destination. Factors associated with the need for NSI were determined using matched NSI (n = 310) and non-NSI (n = 310) groups. All other analyses examined the combined patient sample (n = 620). Variables achieving a significance level of P < 0.20 were included in the logistic regression. Receiver operating characteristic curves, with corresponding area under the curve (AUC) determinations, were also analyzed. Statistical significance was set at α = 0.05. Data are presented as percentages, mean ± standard deviation, or adjusted odds ratios (AORs) with 95% confidence intervals (95% CIs). RESULTS: A total of 620 patients were analyzed, including 310 patients who underwent NSI and 310 age- and ISS-matched non-NSI controls. Average patient age was 72.8 ± 13.4 years (64.1% male, 99% blunt trauma, mean ISS 25.1 ± 8.68, and mean AISh/GCS of 4.63/10.9). CCTST was the only variable independently associated with NSI (AOR 1.23, 95% CI 1.06–1.42) and was inversely proportional to initial GCS and functional outcome scores on discharge. Increasing CCTST was associated with greater mortality, morbidity, HLOS, SDLOS, ICULOS, and ventilator days. On multivariate analysis, factors independently associated with mortality included AISh (AOR 2.70, 95% CI 1.21–6.00), initial GCS (AOR 1.14, 1.07–1.22), and CCTST (AOR 1.31, 1.09–1.58). Variables independently associated with in-hospital morbidity included CCTST (AOR 1.16, 1.02–1.34), GCS (AOR 1.05, 1.01–1.09), and NSI (AOR 2.62, 1.69–4.06). Multivariate models incorporating factors independently associated with each respective outcome displayed good overall predictive characteristics for mortality (AUC 0.787) and in-hospital morbidity (AUC 0.651). Finally, modified CCTST demonstrated good overall predictive ability for NSI (AUC 0.755). CONCLUSION: This study found that the number of discrete findings on CCT is independently associated with major TBI outcome measures, including 30-day mortality, in-hospital morbidity, and NSI. Of note, multivariate models with best predictive characteristics incorporate both CCTST and GCS. CCTST is easy to calculate, and this preliminary investigation of its predictive utility in older patients with TBI warrants further validation, focusing on exploring prognostic synergies between CCTST, GCS, and AISh. If independently confirmed to be predictive of clinical outcomes and the need for NSI, the approach described herein could lead to a shift in both operative and nonoperative management of patients with TBI.