The Westmead Head Injury Project outcome in severe head injury. A comparative analysis of pre-hospital, clinical and CT variables

TLDR

The study aimed to identify pre‑hospital and hospital factors that predict mortality and morbidity in patients with severe head injury. A prospective cohort of 315 severe head‑injury patients (GCS ≤ 8) was followed for death or 6‑month outcome, with logistic regression used to assess pre‑hospital, clinical, and CT predictors. Increasing age, hypotension, low GCS, abnormal motor responses, pupillary non‑reactivity, raised ICP, and specific CT findings (cerebral oedema, intraventricular blood, midline shift, subarachnoid blood, intracerebral haematoma/contusion) were the strongest mortality and poor‑outcome predictors, with an 84.4 % accurate mortality model and a 72.5 % accurate poor‑outcome model.

Abstract

A prospective study of 315 consecutive patients with a severe head injury was undertaken to study factors contributing to mortality and morbidity, both in the pre-hospital and hospital phases. Entry criteria were a Glasgow Coma Scale (GCS) score of 8 or less after non-surgical resuscitation within 6 h of the injury, or a deterioration to that level within 48 h. Patients with gunshot wounds or who were dead on arrival were excluded. End points of the study were either death or at 6 months after the injury. Predictors of mortality were increasing age, the presence of hypotension, a low GCS, abnormal motor responses and pupillary non-reactivity. In the 167 patients in whom intracranial pressure (ICP) was measured, raised ICP and failure to respond to treatment for raised ICP also predicted mortality. Three CT predictors of mortality were the presence of cerebral oedema, intraventricular blood and the degree of midline shift. When analysed using logistic regression, the most accurate model (accuracy 84.4%) included increasing age, abnormal motor responses and the three CT indicators. Analysis of the data for 'good' (Glasgow Outcome Score (GOS) 1 and 2) vs 'poor' (GOS 3 and 4) survival at 6 months was also performed using logistic regression. The model which provided the most accurate prediction of poor outcome included age, hypotension and three different CT characteristics, subarachnoid blood, intracerebral haematoma or intracerebral contusion (accuracy 72.5%).