Predicting the potential for recovery within 24 hours of a head injury

A linear discriminant model is applied to clinical and radiological signs from 492 consecutive head injuries transferred to a neurosurgical unit over a period of four years. Unstable discriminators may be eliminated and those remaining transformed to interval scales in order to release their full prognostic potential. Patient selection criteria included being in coma, judged in the accident and emergency (A & E) department to be susceptible to developing a complicating secondary event, having a depressed fracture and presenting evidence of leakage of cerebrospinal fluid. These criteria are considerably less restrictive than in most other head injury series. The study is also ambitious in attempting prognosis after only 24 hours. The virtues of early and accurate prognosis include counselling of relatives, patient selection for new forms of treatment and to provide an aid towards medical audit. Neither the relaxed attitude towards patient selection nor the speed of prognosis is found to jeopardize the overall efficiency of allocation which is as high as 91.5%. In fact misallocation is only slightly increased if prognosis is attempted within 60 minutes of admission to the A & E department.