Poor accuracy of the Siriraj and Guy's hospital stroke scores in distinguishing haemorrhagic from ischaemic stroke in a rural, tertiary care hospital.

ABSTRACT

BACKGROUND: In India stroke is associated with a high morbidity and mortality. Bedside clinical examination does not always help in distinguishing cerebral infarction from intracranial haemorrhage. We evaluated the accuracy of the Guy's hospital and Siriraj stroke scores in distinguishing haemorrhagic from ischaemic stroke in a rural setting. METHODS: Patients with suspected stroke admitted to a rural teaching hospital were prospectively enrolled. Two investigators collected data for computing the Guy's hospital and Siriraj scores. Cut-off points, as described by the authors of the original scores, were used to predict haemorrhage and infarction. The scores were compared in a blind and independent manner with the computed tomography (CT) scan. The sensitivity, specificity, positive and negative likelihood ratios and agreement between the two scores were calculated. RESULTS: Of the 259 patients admitted for suspected stroke, 134 patients (73 men) underwent both clinical evaluation and a CT scan. The Siriraj score discriminated haemorrhage from infarction with a sensitivity of 78.5% (95% CI 66.5, 87.7) and specificity of 71% (95% CI 52, 85.8). The likelihood ratio of a positive test was 2.7 (95% CI 1.54, 4.75) and that of a negative test was 0.3 (95% CI 0.17, 0.53). For the Guy's hospital score the sensitivity was 81% (95% CI 68.6, 90.1), specificity 76.2% (95% CI 52.8, 91.8), likelihood ratio of a positive test 3.4 (95% CI 1.57, 7.39) and that of a negative test 0.25 (95% CI 0.11, 0.54). Both scores, when combined, were 80% sensitive (95% CI 66.3, 90) and 80% specific (95% CI 51.9, 95.7). The agreement between the two scores was modest (kappa = 0.51), but very good (kappa = 0.93) after exclusion of equivocal score results. CONCLUSION: Our study suggests that neither of the stroke scores is sufficiently accurate for distinguishing the type of stroke. CT scan, and not history and clinical signs, can accurately identify haemorrhage from infarction in acute stroke.