Endovascular thrombectomy is a highly efficacious treatment for large vessel occlusion (LVO). LVO prediction instruments, based on stroke signs and symptoms, have been proposed to identify stroke patients with LVO for rapid transport to endovascular thrombectomy-capable hospitals. This evidence review committee was commissioned by the American Heart Association/American Stroke Association to systematically review evidence for the accuracy of LVO prediction instruments. Medline, Embase, and Cochrane databases were searched on October 27, 2016. Study quality was assessed with the Quality Assessment of Diagnostic Accuracy-2 tool. Thirty-six relevant studies were identified. Most studies (21 of 36) recruited patients with ischemic stroke, with few studies in the prehospital setting (4 of 36) and in populations that included hemorrhagic stroke or stroke mimics (12 of 36). The most frequently studied prediction instrument was the National Institutes of Health Stroke Scale. Most studies had either some risk of bias or unclear risk of bias. Reported discrimination of LVO mostly ranged from 0.70 to 0.85, as measured by the C statistic. In meta-analysis, sensitivity was as high as 87% and specificity was as high as 90%, but no threshold on any instruments predicted LVO with both high sensitivity and specificity. With a positive LVO prediction test, the probability of LVO could be 50% to 60% (depending on the LVO prevalence in the population), but the probability of LVO with a negative test could still be ≥10%. No scale predicted LVO with both high sensitivity and high specificity. Systems that use LVO prediction instruments for triage will miss some patients with LVO and milder stroke. More prospective studies are needed to assess the accuracy of LVO prediction instruments in the prehospital setting in all patients with suspected stroke, including patients with hemorrhagic stroke and stroke mimics.