Comparison of Retrospective and Prospective Measurements of the National Institutes of Health Stroke Scale

For large-scale stroke studies a measure of clinical neurological severity is often needed but may not be available. This issue has recently been addressed by Goldstein and Chilukuri [1] who described a number of stroke outcome studies that were limited by the lack of such a clinical measure of clinical severity. The authors demonstrated that a modified version of the Canadian Neurological Scale was reliable when used to extract neurological scores from medical charts [1]. As the National Institutes of Health (NIH) Stroke Scale score developed by Brott et al. [2] has become one of the most widely used measures of neurological severity, we tested the reliability of a modified version of this scale for retrospective NIH Stroke Scale measurements. The admission neurological examinations of 23 patients admitted to the Neurology Service at the Beth Israel Deaconess Medical Center were reviewed by a neurologist to extract a retrospective NIH Stroke Scale score. The retrospective scores were compared with scores that had been prospectively measured in the same patients by nurses and doctors trained in the use of the NIH Stroke Scale score. The NIH Stroke Scale items were modified to correlate with features of the neurological examination as it is typically recorded. The retrospective NIH Stroke Scale scores correlated significantly with the NIH Stroke Scale score that had been measured prospectively (r = 0.945, p = 0.001, Î = 0.596, p ! 0.001, table 1). Moderate to substantial agreement was present in 12/15 items (Î 1 0.4), the highest agreement values being for orientation, commands and dysphasia (Î 1 0.8). The items with the lowest Î values indicating fair agreement were visual fields, ataxia and sensory examination (Î = 0.2–0.4). A high intrarater reliability for the retrospective NIH Stroke Scale measurements was found (r = 0.946, p ! 0.001). The results compare favorably with values reported by Goldstein et al. [3] and Lyden et al. [4] (table 1). NIH Stroke Scale scores extracted from the charts of a neurology service are reliable and may provide a method for grading neurological deficits where prospective measurements are not available.