An early 'prognosis' based on initial findings can influence clinical decisions. To evaluate the quality of first-day outcome prediction based on either clinical or neuroradiological information, we prospectively examined 100 consecutive severely head-injured patients from the surgical intensive care unit. The prognoses were always made by the same experienced neurosurgeon and neuroradiologist according to a contracted Glasgow Outcome Scale (GOS). Every patient's outcome was predicted according to a three-class GOS within 24 h after injury. Correct 'first-day' prognoses were made in 59 and 56% of the cases by the neuroradiologist and neurosurgeon, respectively. In those instances where the prognoses made by the clinician and the radiologist coincided, the prognoses were correct for 73% of the patients. Based on accepted criteria, the neuroradiologist tended to overpredict favorable outcomes and missed many of the unfavorable outcomes (dead or vegetative) whereas the clinician overestimated unfavorable outcomes. Our study showed that outcome prognosis of patients with severe head injury has limited accuracy when made within 24 h after the injury, although an improvement in accuracy occurred when there was agreement between clinical and radiological predictions. Even with sophisticated clinical and radiological technologies, it is not possible to predict outcome on the first day after the accident with sufficient accuracy to guide early management.