Cohort included 2667 patients, median age 72, 51% female. LAMS scores: 19% 0; 14% 1; 13% 2; 17% 3; 16% 4; 21% 5.
We generated three ROC curves for the interventions of mechanical intervention, tPA, and CTA perfusion imaging, with C statistics of .706, .632, and .668 respectively. All three models were significant using the likelihood ratio test, p < .0001. In all three models, the cut point was a LAMS of 4. Accordingly, a “high LAMS” was defined as ≥4, a “low LAMS” was defined as <4.
We calculated relative risk and found that high LAMS results in the following risk ratios: 3.868x for mechanical intervention, 3.290x for in-hospital death, 1.979x for receiving advanced imaging, 2.089x for receiving tPA, 2.364x for hemorrhagic stroke, 1.515x for ischemic stroke, and 1.344x for not being discharged home. All of these associations were significant with p < .0001, using a χ2 test. In particular, we found 74.9% sensitivity and 61.4% specificity for mechanical intervention.
The median NIHSS at hospital arrival was 6 (IQR 2-13). For low LAMS it was 4 (IQR 1-8); for high LAMS it was 13 (IQR 7-21), p <.0001 using Wilcoxon’s rank-sum test. At hospital discharge, median NIHSS was 2 (IQR 0-7). For low LAMS it was 1 (IQR 0-5); for high LAMS it was 5 (IQR 1-14.75), p <.0001.