Clinical diagnosis of intracranial haemorrhage using Guy's Hospital score

We tested the Guy's Hospital stroke diagnostic score using the clinical data from two independent samples of patients with acute stroke. These were 228 patients from the Oxfordshire community stroke project and 130 referred to the National Hospital for Nervous Diseases in London. The diagnosis was confirmed by computed tomography or necropsy in each case. The optimum cut off point on the clinical score for the differentiation of intracranial haemorrhage from infarction was found to be the same for both the patients in our study and those from whose data the score was derived originally. Set at this level, the score achieved a sensitivity for the diagnosis of haemorrhage of 81% and 88% in the patients from Oxford and London, respectively. In those from Oxford infarction was diagnosed with a sensitivity of 78% with an overall predictive accuracy of 78% with an overall London the sensitivity for infarction was also 78% with an overall predictive accuracy of 82%. When it is essential to exclude intracerebral blood before starting treatment in the small proportion of patients with stroke who require anticoagulation the Guy's Hospital score is not sufficiently accurate to replace computed tomography. The score is, however, the most accurate clinical means of differentiating haemorrhage from infarction as the cause of stroke. It is suggested that it should be used as a screening test in epidemiological studies and in large scale trials of low risk treatment for the secondary prevention of stroke when computed tomography in all cases is impracticable.