Objective To evaluate the predictive value of eight scoring systems (Acute Physiology and Chronic Health Evaluation Ⅱ [APACHE Ⅱ] and APACHE Ⅲ,Improved Edinburgh-Scandinavian Scale [CSS],U.S.National Institutes of Health Stroke Scale [NIHSS],Activity of Daily Living [ADL] scale,Glasgow Coma Scale [GCS],Previous History of Disease Scale and Concomitant Disease Scale) in severity and outcomes of patients with acute ischemic stroke by using discriminant analysis,and to establish their mathematical models to predict the status of early death of stroke patients.Methods Three hundred and ninety-nine patients with acute ischemic stroke,admitted to our hospital from January 2008 to December 2012,were chosen in our study; these patients were tested with APACHE Ⅱ,APACHE Ⅲ,CSS,NIHSS,ADL,GCS,Previous History of Disease Scale and Concomitant Disease Scale within 24 h of admission.All of them were divided into two groups according to groups of survival (n=278) or death (n=121) one month after disease onset.Discriminant analysis was performed on all the data and the predictive values of these eight scales in the prognosis were analyzed.Results Patients from group of death had significantly higher scores of APACHE Ⅱ,APACHE Ⅲ,CSS,NIHSS and Concomitant Disease Scale,and statistically lower scores of ADL and GCS scores than those from group of survival (P＜0.05).Cluster analysis showed that CSS and NIHSS,and APACHE Ⅱ and APACHE Ⅲ,respectively,belong to clusters,which enjoyed higher predictive values than other scales.The areas under receiver operator characteristic (ROC) curves were 0.808,0.818,0.796 and 0.794 of APACHE Ⅱ,APACHE Ⅲ,CSS and NIHSS scores,respectively,enjoying good definition; Discriminant analysis was used to analyze the eight scoring systems and mathematical models were established to predict the outcomes of stroke patients,enjoying more than 80％ of coincidence rate.Conclusion APACHE Ⅱ,APACHE Ⅲ,CSS and NIHSS are superior to the other four score systems in evaluating severity of stroke patients,whose mathematical models,having more than 80％ of accuracy rate.