Predicting functional outcome after stroke by modelling baseline clinical and CT variables
we aimed to assess whether the performance of strokeoutcome models comprising simple clinical variables could beimproved by the addition of more complex clinical variablesand information from the first computed tomography (CT) scan.
Methods: 538 consecutive acute ischaemic and haemorrhagic strokepatients were enrolled in a Stroke Outcome Study between 2001and 2002. Independent survival (modified Rankin scale 2) wasassessed at 6 months. Models based on clinical and radiologicalvariables from the first assessment were developed using multivariatelogistic regression analysis.
Results: three models were developed (I–III). Model Iincluded age, pre-stroke independence, arm power and a strokeseverity score (area under a receiver operating characteristiccurve, AUC = 0.882) but performed no better than Model II, whichcomprised age, pre-stroke independence, normal verbal componentof the Glasgow coma score, arm power and being able to walkwithout assistance (AUC 0.876). Model III, including two radiologicalvariables and clinical variables, was not statistically superiorto model II (AUC 0.901, P = 0.12). Model II was externally validatedin two independent datasets (AUCs of 0.773 and 0.787).
Conclusion: this study demonstrates an externally validatedstroke outcome prediction model using simple clinical variables.Outcome prediction was not significantly improved with CT-derivedradiological variables or more complex clinical variables.