The retrospective application of a prediction model to patients who have had a decompressive craniectomy for trauma.

There is currently a resurgence of interest in the use of decompressive craniectomy. As the procedure is used more frequently, there may be an increasing number of patients surviving a severe traumatic brain injury with severe neurological impairment. The aim of this study was to determine if we could predict those cases that fall into this category. We used the web-based prediction model prepared by the CRASH collaborators and applied it to a cohort of patients who had a decompressive craniectomy in 2006 and 2007 at the two major trauma hospitals in Western Australia. All clinical and radiological data were reviewed and entered into the model, and predicted outcome and actual outcome were compared. Our analysis indicated that a significant cut-off point appeared at which the model predicted a 75% risk of an unfavorable outcome at 6 months; 19 of 27 patients with CRASH scores <75% returned to work, whereas none of the 14 patients with higher scores achieved this degree of rehabilitation at 18 months. Statistical analysis of the outcomes in our cohort confirmed that the CRASH model reliably predicted unfavorable outcome. This study demonstrated that our ability to predict poor outcome has improved.