In order to compare the predictive value of the Apache II, the Glasgow Coma Scale and Revised Trauma Scoring systems in relation to outcome in a neurosurgical department high dependency unit (HDU), all 109 patients entering the unit under the care of one consultant were studied. All patients in the HDU were self-ventilating, so that motor responses were not suppressed by muscle relaxants or sedation. Initial Minimum and Maximum Glasgow Coma Scale Scores, Revised Trauma Score, Apache II and the Apache minus neurological weighting (Apache-NW) scores, were compared as predictors of outcome (as assessed by the Glasgow Outcome Score at 6 months). Twenty-eight patients had a bad outcome, i.e. Glasgow Outcome Scores 1-3, and 72 individuals a good outcome, i.e. scores 4 or 5. Statistical analysis was by the Spearman Ranked Correlation Test, and comparison of Receiver Operational Characteristics Curves. Data were complete on 100 patients (91.7%) and show the Maximum Glasgow Coma Score, followed by the Apache II score, as the best predictors of outcome analysed. This was also true if all patients except those with head injury were analysed as a group. All scoring systems were significantly better predictors of outcome in the head injured patient. For this group, Apache II had an outcome predictive value of 97% compared with 93% for initial and 95% for minimum GCS. Removing the neurological weighting from Apache II weakened its predictive ability in all patients, emphasizing that it is the neurological status of the patient which best predicts overall functional outcome. Apache II data are also much more time-consuming to collect than GCS data.(ABSTRACT TRUNCATED AT 250 WORDS)