The "DeyeCOM Sign": Predictive Value in Acute Stroke Code Evaluations.

BACKGROUND

Rapid diagnosis in stroke is critical. Computed tomography is often performed initially, even before a neurologic examination. Gaze deviation has been correlated with stroke diagnosis in some cohorts. Conjugate gaze deviation on stroke code imaging, the "DeyeCOM sign," may have emergency stroke care implications.

METHODS

We evaluated stroke code imaging from the University of California, San Diego database (2007-2013) for "DeyeCOM sign" diagnostic and predictive utility. Patients were grouped as DeyeCOM+ if conjugate gaze deviation was noted. The differences were assessed using the Fisher exact test for categorical and the Wilcoxon rank-sum test for continuous variables.

RESULTS

We evaluated 342 patients; 106 (31%) were DeyeCOM+. Mean age was 63. The most common diagnoses in the DeyeCOM+ group were ischemic stroke (50.94%), transient ischemic attack (8.49%), other (8.49%), somatization (6.6%), and hemorrhage (5.66%). The National Institutes of Health Stroke Scale was greater in stroke patients than that in nonstroke (8.2 versus 3.8; P < .0001), and in DeyeCOM+ compared with DeyeCOM- (6.8 versus 5.6; P = .03). DeyeCOM+ patients were more likely to have a +gaze score (26.4% versus 9.8%; P < .0001), and +gaze patients were more likely to have final stroke diagnosis (26.0% versus 3.6%; P < .0001). There was no overall difference between groups in final stroke diagnosis; however, patients with deviation of 15° or more were more likely to have final diagnosis stroke (63.9% versus 47.9%; P = .03).

CONCLUSIONS

DeyeCOM+ patients scored higher and were more likely to have +gaze on the stroke scale, and deviation of 15° or more was correlated with final diagnosis stroke. In current environments, there is pressure to complete stroke evaluations rapidly. Reliable imaging information obtained early (such as gaze deviation on scan correlating with scale score and final stroke diagnosis) could augment decision making even with negative imaging.