The purpose of this work is the estimation of the Glasgow outcome scale (GOS) from a single continuous electroencephalogram (c-EEG) routinely recorded to monitor comatose patients in the neurosurgical intensive care unit. c-EEG was recorded from 13 patients in the acute phase: five with GOS = 5, four with GOS = 3 and four with GOS = 1. Different indexes were extracted from epochs of c-EEG (classical: amplitude and spectral estimators; non classical: from recurrence quantification analysis-RQA-and approximate entropy). Descriptors of different indexes (temporal variation and mean, standard deviation, skewness of the distribution across epochs) were used to train support vector machines to identify the correct GOS. We found classifiers allowing correct classification of the patients. Spectral indexes allowed to get optimal performances in classifying GOS 1 and 3. Nonlinear indexes (especially determinism from RQA) were optimal for identifying GOS = 5. Thus, the integration of information from classical/linear and nonlinear c-EEG descriptors in a multi-index classifier is important for GOS estimation.