Accuracy of two scores in the diagnosis of stroke subtype in a multicenter cohort study

Ann Emerg Med. 2009 Mar;53(3):373-8. doi: 10.1016/j.annemergmed.2008.06.005. Epub 2008 Aug 16.

Abstract

Study objective: The distinction between hemorrhagic and ischemic stroke has critical implications for management. For that purpose, clinical scores have been proposed to be used in areas with limited health care resources where brain computed tomographic (CT) scan is not readily available. We conducted this study to evaluate the predictive value of the Allen and Siriraj scores in the differential diagnosis of stroke subtypes.

Methods: We prospectively collected data for 4 years on the clinical characteristics of patients with stroke in a multicenter study. For all patients, we calculated the Allen and the Siriraj scores and we assessed their accuracy in predicting stroke subtypes with receiver operating characteristics (ROC) curves.

Results: We assessed 1,023 patients. Of these, 82.7% (n=846) had ischemic stroke. The area under the ROC curve was higher for Siriraj score compared with the Allen score (0.780 versus 0.702; P=.04). Using the original cutoff points, Siriraj score has a sensitivity for the diagnosis of hemorrhage of 60% and a specificity of 95%; the corresponding values for the Allen score are 55% and 70%, respectively. The negative predictive value was higher for Siriraj score compared to the Allen score (90% versus 80%). The diagnosis of stroke subtype was best predicted at Siriraj score less than -4.

Conclusion: Siriraj score is a valid and useful tool for predicting stroke subtype in a clinical setting in which financial constraints make systematic brain CT scan unfeasible.

Publication types

  • Comparative Study
  • Multicenter Study
  • Validation Study

MeSH terms

  • Brain Ischemia / diagnosis*
  • Diagnosis, Differential
  • Female
  • Humans
  • Intracranial Hemorrhages / diagnosis*
  • Male
  • Middle Aged
  • Prospective Studies
  • ROC Curve
  • Sensitivity and Specificity
  • Stroke / diagnosis*