An Analysis of EMS and ED Detection of Stroke

Studies have shown a reduction in time-to-CT and improved process measures when EMS personnel notify the ED of a "stroke alert" from the field. However, there are few data on the accuracy of these EMS stroke alerts. The goal of this study was to examine diagnostic test performance of EMS and ED stroke alerts and related process measures.

The EMS and ED records of all stroke alerts in a large tertiary ED from August 2013-January 2014 were examined and data abstracted by one trained investigator, with data accuracy confirmed by a second investigator for 15% of cases. Stroke alerts called by EMS prior to ED arrival were compared to stroke alerts called by ED physicians and nurses (for walk-in patients, and patients transported by EMS without EMS stroke alerts). Means ± SD, medians, unpaired t-tests (for continuous data), and two-tailed Fisher's exact tests (for categorical data) were used.

Of 260 consecutive stroke alerts, 129 were EMS stroke alerts, and 131 were ED stroke alerts (70 called by physicians, 61 by nurses). The mean NIH Stroke Scale was higher in the EMS group (8.1 ± 7.6 vs. 3.0 ± 5.0, p < 0.0001). The positive predictive value of EMS stroke alerts was 0.60 (78/129), alerts by ED nurses was 0.25 (15/61), and alerts by ED physicians was 0.31 (22/70). The PPV for EMS was better than for nurses or physicians (both p < 0.001), and more patients in the EMS group had final diagnoses of stroke (62/129 vs. 24/131, p < 0.001). The positive likelihood ratio was 1.53 for EMS personnel, 0.45 for physicians, and 0.77 for nurses. The mean time to order the CT (8.5 ± 7.1 min vs. 23.1 ± 18.2 min, p < 0.0001) and the mean ED length of stay (248 ± 116 min vs. 283 ± 128 min, p = 0.022) were shorter for the EMS stroke alert group. More EMS stroke alert patients received tPA (16/129 vs. 6/131, p = 0.027).

EMS stroke alerts have better diagnostic test performance than stroke alerts by ED staff, likely due to higher NIH Stroke Scale scores (more obvious presentations) and are associated with better process measures. The fairly low PPV suggests room for improvement in prehospital stroke protocols.