Few predictors of systemic embolism in patients with mitral stenosis have been identified by noninvasive methods. This study used the most powerful noninvasive diagnostic tool, transthoracic echocardiography, as well as other noninvasive clinical information to look for predictors. Five hundred consecutive patients with a mitral valve area of 2 cm² or less were studied. They were divided into two groups: group 1 consisted of 143 patients with a history of systemic embolism and group 2 consisted of 357 patients with no history of systemic embolism. Using a stepwise logistic regression on a random subsample of 400 patients, 4 independent predictors were found: the presence of atrial fibrillation (p=0.003, relative risk [RR]=2.3, 95% CI=1.3, 4.2), the absence of significant tricuspid regurgitation (p=0.008, RR=2.5, 95% CI=1.3, 4.9), the absence of aortic regurgitation (p=0.022, RR=2.2, 95% CI=1.1, 4.2), and the presence of left atrial smoky echoes (p=0.039, RR=1.7, 95% CI=1.1, 3.0). When the above model, together with significant interaction terms, was applied to the remaining 100 patients, both the Hosmer-Lemeshow and Brown goodness-of-fit statistics were not significant (p=0.888 and p=0.248, respectively), indicating that the fit was adequate and the model was validated. Thus, important noninvasive predictors of systemic embolism in patients with mitral stenosis can easily be obtained. Subgroups of patients with high risk of systemic embolism can be identified. This may refine our therapeutic strategies to prevent the catastrophe of systemic embolism. Copyright © 1994 The American College of Chest Physicians. Published by Elsevier Inc. All rights reserved.