All 7 predictive scales analyzed were found to be highly accurate, with some more feasible than others.
A recent study analyzed the accuracy of 7 prehospital scales utilized in the assessment of patients with acute stroke codes, all of which were found to have good accuracy, high specificity, and low sensitivity. Particularly, the Los Angeles Motor Scale (LAMS) and the Rapid Arterial Occlusion Evaluation (RACE) led the group in accuracy.
In addition to LAMS and RACE, the study assessed the Cincinnati Stroke Triage Assessment Tool prediction (C-STAT), the Prehospital Acute Stroke Severity (PASS) scale, gaze-face-arm-speech-time (G-FAST), Field Assessment Stroke Triage for Emergency Destination (FAST-ED), and gaze, facial asymmetry, level of consciousness, extinction/inattention scales for symptomatic large anterior vessel occlusion (sLAVO).
Accuracy of the scales ranged from 0.79 to 0.89, with LAMS (0.89; 95% CI, 0.87–0.90) and RACE (0.88; 95% CI, 0.86–0.89) scales yielding the highest scores. Sensitivity ranged from 38% to 62%, and specificity from 80% to 93%. Scale feasibility rates, which should be taken into account before implementing a scale, ranged from 78% to 88%, with the highest rate for the PASS scale (87.9%, 95% CI, 86.5–89.4).
First author T. Truc My Nguyen, MD, Department of Neurology, Leiden University Medical Center, and colleagues wrote that “in this prospective cohort study assessing more than 2000 patients with acute stroke codes, we found that several established sLAVO prediction scales had good accuracy when used in the EMS setting...feasibility rates could increase by using all available information to reconstruct the scale’s cut point, thereby enabling additional inclusion of acute stroke codes with incomplete or untestable items.”
Nguyen and colleagues examined data from 2007 patients who received acute stroke codes from 3 comprehensive stroke centers and 4 primary stroke centers in an urban area of the Netherlands. These patients had a mean age of 71.1 years (standard deviation [SD], 14.9) and 1021 (50.9%) were male. The median NIH Stroke Scale (NIHSS) score was 4 (interquartile range [IQR], 2-8) and 158 (7.9%) had sLAVO.
The final diagnosis after 3 months was ischemic stroke in 842 patients (41.9%), intracerebral hemorrhage in 148 patients (7.4%), transient ischemic attack (TIA) in 264 patients (13.2%), and stroke mimic in 753 patients (37.5%). In addition, 158 patients (7.9%) with an acute stroke code were later diagnosed with sLAVO. Patients without sLAVO had a greater history of ischemic stroke or TIA (n = 638; 34.5%) when compared to patients with sLAVO (n = 33; 20.9%; P = .001). Patients with sLAVO had a median NIHSS score of 11 (IQR, 5-17) and patients without had a median NIHSS score of 3 (IQR, 2-6; P <.001). Similarly, 13 (8.2%) patients with sLAVO experienced de novo atrial fibrillation vs 66 (3.6%) patients without (P = .01).
Median symptom-onset-to door time was 115 minutes (IQR, 45–340) in patients with sLAVO as compared to 142 minutes (IQR, 62–446) in patients without sLAVO (P = .02). More patients with sLAVO (n = 61; 38.6%) received intravenous therapy (IVT) than those without (n = 253; 13.7%; P <.001), and endovascular therapy (EVT) was performed in 100 (63.3%) patients with sLAVO , with a median door-to-groin-puncture time of 72 (IQR, 54–105) minutes.
Nguyen and colleagues found that the LAMS and RACE scales significantly outperformed the others, with C-STAT receiving an accuracy score of 0.79 (95% CI, 0.77–0.81), PASS with 0.81 (0.79–0.83), G-FAST with 0.82 (0.81–0.84), and FAST-ED with 0.83 (0.81–0.85). Specificity was high for all scales (range, 80–93%), whereas sensitivity was low (range, 38–62%). The Youden index ranged from 0.30 to 0.47, and RACE had the highest index score of 0.47 (95% CI, 0.37-0.56).
In addition, negative predictive value was high for all scales (range, 95–96%), and positive predictive value was low (range, 21–32%). FAST-ED had the highest area under the curve (AUC) of 0.80 (95% CI, 0.74–0.85) but this score was not statistically significantly different from G-FAST (AUC, 0.77; 95% CI, 0.72–0.82; P = .46), LAMS (AUC, 0.76; 95% CI, 0.71–0.81; P = .10), or RACE (AUC, 0.75; 95% CI, 0.69–0.82; P = .53).
Nguyen and colleagues found that the scales’ cut point mean reconstruction rate was 84.1% (range, 78.1–87.9). The PASS scale had the highest reconstruction rate, demonstrating the highest feasibility of the scales. Calculating the rates for the cut point yielded 6.1% to 24.1% more acute stroke codes compared with reconstruction rates of the full scale.
“Our study is the first to our knowledge to provide external validation in the field as well as to offer head-to-head comparisons of several established sLAVO prediction scales... it is important to take feasibility into account before implementing a prediction scale in the field because focused training could substantially increase these rates,” Nguyen and colleagues concluded.