Prediction of intracerebral hemorrhage survival

The Pilot Stroke Data Bank obtained information on 94 patients with intracerebral hemorrhage. These data were used to identify factors predictive of 30-day outcome from among 85 demographic, historical, clinical, and laboratory variables generally available to clinicians on the day of admission. The 9 univariate factors statistically associated with outcome were Glasgow Coma Scale score, systolic blood pressure, pulse pressure, horizontal and vertical gaze palsies, severity of weakness, presence of brainstem-cerebellar deficits, interval stroke course, and parenchymal hemorrhage size. Beginning with these factors, a step-down variable selection procedure was used to derive a logistic regression model, containing only Glasgow Coma Scale score, pulse pressure, and hemorrhage size, that could be used to categorize correctly 92% of the patients as alive or dead at 30 days after onset.