Background and purpose: The burden of stroke in sub-Saharan Africa is already high and likely to increase, but few patients with stroke have access to brain imaging. Distinguishing pathologic stroke types is relevant both for clinical management and epidemiologic studies. We assessed the accuracy of two stroke scores in distinguishing stroke types in a population known to have a high prevalence of intracranial hemorrhage but low prevalence of atherosclerosis and compared them with the clinicians' assessment of stroke type with computed tomography brain scanning as the "gold standard."
Methods: We assessed the stroke scores and the clinicians' blind assessment of pathologic stroke type in consecutive black patients with stroke included in the Johannesburg Hospital Stroke Register over 23 months. We calculated the accuracy of the scores and clinicians compared with computed tomography brain scan (sensitivity, specificity, positive predictive value, likelihood ratio, kappa statistic).
Results: Two hundred twenty-two patients were scanned and assessed within 15 days. Sixty-two (28%) had cerebral hemorrhage and nine (4%) subarachnoid hemorrhage. Neither the Siriraj (sub-Saharan Africa) nor Guy's Hospital score was accurate or offered much advantage over clinician assessment (sensitivity 0.60 and 0.34, specificity 0.88 and 0.95 for intracranial hemorrhage in the Siriraj Stroke Score and Guy's Hospital Stroke Score, respectively; sensitivity 0.70 and 0.71, specificity 0.84 and 0.74, respectively, for ischemic stroke). Although the scores were more accurate when we used new cutoff points, they then failed to diagnose over 80% of stroke types.
Conclusions: The Siriraj Stroke Score and Guy's Hospital Stroke Score are not sufficiently accurate for use in either epidemiologic studies or to guide clinical management in sub-Saharan Africa at present.