A three-item scale for the early prediction of stroke recovery

Background Accurate assessment of prognosis in the first hours of stroke is desirable for best patient management. We aimed to assess whether the exte nt of ischaemic brain injury on magnetic resonance diffusion-weighted imagi ng (MR DWI) could provide additional prognostic information to clinical fac tors. Methods In a three-phase study we studied 66 patients from a North American teaching hospital who had: MR DWI within 36 hours of stroke onset; the Nat ional Institutes of Health Stroke Scale (NIHSS) score measured at the time of scanning; and the Barthel Index measured no later than 3 months after st roke. We used logistic regression to derive a predictive model for good rec overy. This logistic regression model was applied to an independent series of 63 patients from an Australian teaching hospital, and we then developed a three-item scale for the early prediction of stroke recovery. Findings Combined measurements of the NIHSS score (p=0.01), time in hours f rom stroke onset to MR DWI (p=0.02), and the volume of ischaemic brain tiss ue on MR DWI (p=0.04) gave the best prediction of stroke recovery. The mode l was externally validated on the Australian sample with 0.77 sensitivity a nd 0.88 specificity. Three likelihood levels for stroke recovery-low (0-2), medium (3-4), and high (5-7)-were identified on the three-item scale. Interpretation The combination of clinical and MR DWI factors provided bett er prediction of stroke recovery than any factor alone, shortly after admis sion to hospital. This information was incorporated into a three-item scale for clinical use.