Predictive factors of in-hospital mortality in 986 consecutive patients with first-ever stroke

The aim of this study was to determine predictors of in-hospital mortality in 986 consecutive first-ever stroke patients by means of a multivariate analysis. In a first predictive model based on 13 demographic, anamnestic and clinical variables, age, previous or concomitant pathologic conditions (atrial dysrhythmia, chronic obstructive pulmonary disease, congestive heart failure, cirrhosis or chronic liver disease), deteriorated level of consciousness, vomiting, cranial nerve palsy, seizures and limb weakness appeared to be independent prognostic factors of in-hospital mortality. In addition to these 7 variables, intraventricular haemorrhage and middle cerebral topography were selected in a second predictive model based on clinical and neuroimaging variables. In both models, setting a cutoff point of 0.50 for predicting vital status at hospital discharge resulted in a sensitivity of 52 %, specificity of 96% and total correct classification of 88 %. Clinical factors indicative of the severity of stroke available at the time of initial diagnosis have a predominant influence upon in-hospital mortality irrespective of the stroke subtype, and may help clinicians to establish prognosis more accurately.