Prediction of outcome in non-traumatic coma in childhood

The data in 104 comatose children were analysed, neonates and those with head injury being excluded. The median age was 30 months, with a range of one month to 17 years. 50% were normal, 32% died and the remainder had handicaps ranging from mild to severe. A number of individual clinical variables related with outcome. Stepwise multivariate discriminate analysis individually classified 75% of 102 cases correctly into one of 5 outcome groups, using clinical information obtained at the time of initial assessment within 12 hours of the onset of coma; 67% of 66 cases could be classified correctly on the basis of information available at 24 hours after the onset of coma. There were, respectively, 8% and 3% serious misclassification errors for the 2 time periods. Our data suggest that the stepwise procedure may be a useful method for the early prediction of outcome in comatose children. Classification errors may be minimized by enlarging the data base, by using alternate statistical methods and by taking into account information from investigative variables, the most promising of which seems to be multimodality evoked potentials.