A simple scoring system for outcome prediction of ischemic stroke

Abstract

Objectives: According to most existing models, a computer is usually needed for predicting stroke outcome. Our purpose was to construct a simple and reliable prognostic scale not requiring the use of a calculating machine.

Materials and methods: The scale [the Bologna Outcome Algorithm for Stroke (BOAS)] was obtained in 221 patients with ischemic stroke not undergoing thrombolysis and was then validated in a test group of 100 different patients. Outcome was assessed at 9 months as the number of dependent or dead patients (modified Rankin scale - mRS > 2).

Results: By a preliminary systematic univariate analysis, 25 of 415 baseline variables were found to be associated with a mRS > 2 independently of stroke severity and age. Subsequent multivariable analyses led to a final model based on five dichotomous risk factors (RF): National Institutes of Health Stroke Scale score ≥10, age ≥78, need of urinary catheter, oxygen administration, and persistence of upper limb paralysis at discharge from stroke unit. The patients with two or more RF (53%) had a mRS > 2 in 91% of cases and were dead in 42% of cases. With 0-1 RF, the two percentages were 24% and 2%, respectively (overall accuracy of prediction 83.9%, area under ROC curve [AUC] 0.891). In the test group, the accuracy was 79.0% and the AUC was 0.839.

Conclusions: The need of urinary catheter, oxygen administration, and persistence of upper limb paralysis, together with stroke severity and advanced age, allow a simple and accurate prediction of dependency or death after ischemic stroke.