A simple risk score for early ischemic stroke mortality derived from National Institutes of Health Stroke Scale: a discriminant analysis

Abstract

Objectives: The aim of the current study was to design a new simpler form of National Institutes of Health Stroke Scale (NIHSS) for use in emergency settings, and compare its predictive ability with original NIHSS score for mortality.

Methods: A total of 152 consecutive patients with first ever ischemic stroke admitted to a university affiliated hospital were recruited. NIHSS score on admission was estimated and the predictive ability of NIHSS items for mortality at 28 days was evaluated by logistic regression. Stepwise discriminant analysis was performed on NIHSS items to obtain a discriminant function with the best discriminative ability for mortality. Further, receiver operating characteristics (ROC) curves were depicted to compare the new determined discriminant function with the original NIHSS score.

Results: Cumulative rate of mortality was 11.8% for 28-day follow-up period. Among NIHSS items, scores of visual field, limb ataxia and extinction neglect were not associated with mortality (P>0.05). On the contrary, level of consciousness-commands, language and gaze were determined as independent indicators of mortality (P<0.05), and their coefficients on discriminant function were equal to 0.65, 0.44 and 0.30, respectively. In addition, area under the ROC curve of the calculated discriminant function was not statistically different from NIHSS score (P>0.05).

Conclusions: The suggested discriminant function, comprising NIHSS items of level of consciousness-commands, language and gaze, can predict 28-day mortality after ischemic stroke in a similar way to the original NIHSS score and can provide a baseline for stroke severity in emergency settings.