Prediction of recovery of continuous memory after traumatic brain injury

To evaluate the ability of measures of initial severity, tests of attention, and demographic characteristics to predict recovery of continuous memory for words over a 24-hour period in patients with acute traumatic brain injury. Recovery of continuous memory was assessed prospectively in 94 patients with nonpenetrating traumatic brain injury. A classification and regression tree analysis identified a hierarchical subset of variables that may be used as a simple guideline for predicting recovery of continuous memory. Weibull regression models evaluated and compared the predictive ability of multiple variables. Four groups of patients were identified based on measures of severity of injury and demographic characteristics. These four groups had recovery profiles that were more precise than could be obtained by using the Glasgow Coma Scale alone: mild, about 1 week to recovery of continuous memory; moderate, 1 to 4 weeks; severe, 2 to 6 weeks; and extremely severe, 4 to 8 weeks. Regression analysis confirmed that measures of capacity (inherent resources such as indicated by age) and compromise (general functional brain state measured neuropsychologically) improved prediction over models based only on injury severity measures, such as the Glasgow Coma Scale. Approaches to predicting recovery of continuous memory in the acute period after traumatic brain injury that take into account multiple measures provide a more sensitive predictive index.