Accuracy of paramedic identification of stroke and transient ischemic attack in the field.

PURPOSE: To determine the accuracy of acute stroke identification by paramedics in an urban emergency medical services system.
METHODS: Retrospective chart review of all patient encounters by paramedics resulting in transport to two university hospitals during a six-month period. Subjects were identified by paramedic coding of stroke/transient ischemic attach (TIA) or final hospital discharge ICD-9 diagnosis of acute stroke and TIA. The sensitivity and positive predictive value for paramedic identification of stroke were calculated, and the time intervals from symptom onset to various points along the patients' prehospital and hospital courses were identified.
RESULTS: Ninety-six patients were identified, of whom 81 met the diagnosis of acute stroke or TIA. Paramedics identified 49 of these 81 patients (sensitivity 61%). Fifteen patients were identified by paramedics as having a stroke when the patient ultimately had a different diagnosis (positive predictive value 77%) Patients or their families waited on average 2.5 +/- 3.6 (SD) hours before accessing 911, and a mean of 5.1 +/- 4.0 (SD) hours elapsed from symptom onset until head imaging studies were obtained.
CONCLUSION: Paramedics in San Francisco County were correct three-fourths of the time when their documentation listed patients as having stroke/TIA. However, they did not identify 39% of stroke victims, a patient population who may benefit from urgent therapy. A substantial period elapses before stroke victims access 911. This highlights the need to develop an educational program for the community at risk for stroke, and another for paramedics directed toward more accurate identification of acute stroke victims.