Clinical Question

What is the long-term prognosis for patients with acute stroke?

Evidence Summary

A number of clinical decision models and scoring systems have been developed and validated to assist physicians in assessing the prognosis of patients with acute stroke. This assessment can be helpful to patients, families, and physicians as they plan for long-term care and prepare advance directives. The National Institutes of Health Stroke Scale (NIHSS) predicts seven-day and three-month prognoses and is widely used in the research setting. The NIHSS is available at http://www.ninds.nih.gov/disorders/stroke/strokescales.htm.

In a prospective study of 1,281 patients with acute stroke, a good or excellent outcome at three months was observed in 95 percent of patients with an NIHSS score of 0 to 3 points, 87 percent with 4 to 6 points, 78 percent with 7 to 10 points, 56 percent with 11 to 15 points, 42 percent with 16 to 22 points, and only 18 percent with 23 or more points.¹ However, the NIHSS has 13 items, is fairly complex, and requires training for accurate use and good reproducibility.

Other models are simpler and potentially more useful at the point of care. Guy's prognostic score (G-score), a simplification of the Guy's Hospital score, includes patient age and five clinical signs. It has been prospectively validated in several populations, but most of the validation studies are at least 10 years old.² Because of changes in the care of patients with acute stroke, this article presents only models validated since 2000.

The six simple variable model has been prospectively validated in a study of dietary management in 2,955 patients with stroke at 112 hospitals in 16 countries.³ The mean age of patients was 73 years, and about one half were men; more than 90 percent of patients were independent in daily activities before the stroke. Separate six simple variable models were created to predict survival free of dependency at six months (Table 1), overall survival at six months, and probability of being alive and at home within six months.³ The models had good accuracy, with an area under the receiver operating characteristic curve of 0.79. The models tended to be somewhat pessimistic in patients with severe strokes and optimistic in patients with milder strokes. The models require calculations but can be put into a spreadsheet fairly easily.

A simpler model was developed in 223 patients with acute stroke in an Australian teaching hospital and was validated in 217 patients at the same hospital (Table 2).⁴ The mean age of patients in the validation group was 69 years, and 58 percent were men. Although easier to use at the point of care, the model has not been prospectively validated in other populations and should be used with caution.