Background and purpose: The objective of this study was to determine the performance of the Recognition Of Stroke In the Emergency Room (ROSIER) scale in risk-stratifying Chinese patients with suspected stroke in Hong Kong.
Methods: This was a prospective cohort study in an urban academic emergency department (ED) over a 7-month period. Patients over 18 years of age with suspected stroke were recruited between June 2011 and December 2011. ROSIER scale assessment was performed in the ED triage area. Logistic regression analysis was used to estimate the impacts of diagnostic variables, including ROSIER scale, past history and ED characteristics.
Findings: 715 suspected stroke patients were recruited for assessment, of whom 371 (52%) had acute cerebrovascular disease (302 ischaemic strokes, 24 transient ischaemic attacks (TIA), 45 intracerebral haemorrhages), and 344 (48%) had other illnesses i.e. stroke mimics. Common stroke mimics were spinal neuropathy, dementia, labyrinthitis and sepsis. The suggested cut-off score of>0 for the ROSIER scale for stroke diagnosis gave a sensitivity of 87% (95%CI 83-90), a specificity of 41% (95%CI 36-47), a positive predictive value of 62% (95%CI 57-66), and a negative predictive value of 75% (95%CI 68-81), and the AUC was 0.723. The overall accuracy at cut off>0 was 65% i.e. (323+141)/715.
Interpretation: The ROSIER scale was not as effective at differentiating acute stroke from stroke mimics in Chinese patients in Hong Kong as it was in the original studies, primarily due to a much lower specificity. If the ROSIER scale is to be clinically useful in Chinese suspected stroke patients, it requires further refinement.