Multivariable analysis in the prediction of death in hospital after spontaneous supratentorial intracerebral hemorrhage

Abstract

The present research has the objective of proposing a prognostic model to predict death in hospital after spontaneous supratentorial intracerebral hemorrhage. Research design: Prospective descriptive study. Method The patients who were diagnosed supratentorial intracerebral hemorrhage by CT scan or MRI brain age between 41-87 years old were admitted during april to december 1994. Baseline characteristics, risk factors and predictive variables were recorded within 24 hours after admission. Result: The mortality rate of patient with spontaneous supratentorial intracerebral hemorrhage in this study was 42.4% and 91% of the diseased cases died within first week due to brain herniation. Univariate analysis comparing 11 baseline characteristics and 4 risk actors showed statistical significant difference in duration from onset to diagnosis (P = 0.033), Glasgow Coma Scale (P = 0.00004) volume of hematoma (P = 0.047), site of bleeding (P = 0.00412) and intraventricular hemorrhage (p = 0.00004). After stepwise logistic regression analysis adjusted among all variables, only 4 variables, Glasgow Coma scale (β = -0.2908) (P = 0.0039), intraventricular hemorrhage (β = -1.3922) (P = 0.0286). sex (β = -1.7214) (P = 0.0201), and volume more than 60 cm3 (β = 1.2209) (P = 0.005) were found to be potentially predictors of death in hospital after spontaneous supratentorial intracerebral hemorrhage. Hosmer-Lemeshow Chi-square for goodness of fit was 7.38 (with a degree of freedom of 8) (P = 0.4963), so the model fit the data quite well. Model sensitivity, specificity and accuracy were 75, 85.71 and 81.18k respectively. The area under the Receiver Operating Characteristic (ROC) curve was 0.8642.