Severe head injuries: An outcome prediction and survival analysis

Objective: To identify the predictors determined early after admission and associated with unfavorable outcome or early (within 48 h) death after severe head injury. Design: Prospective cohort study. Setting: A neurosurgical intensive care unit in a university hospital. Patients: 198 consecutive comatose patients hospitalized from 1989 to 1992. Results: Logistic regression showed that a combination of age, best motor response score from the Glasgow Coma Scale, and hypoxia provided a good prediction model of unfavorable outcome (sensitivity = 0.93). The length of participation of survivors was 6 to 61 months (median 27.1). The Cox model demonstrated age, motor score less than 3, mydriasis, and hypoxia as poor prognosis factors. Conclusions: Clinicians can determine the odds of a good outcome from the combination of three easily measurable factors using a simple diagram constructed from logistic regression. Survival analysis showed that motor score adjusted values greater than 3 had the same prognosis.