We hypothesized that ejection fraction (EF) best predicts cardiovascular death but only measures of ischemia predict relative survival benefit from revascularization compared with medical therapy. We followed up 5366 consecutive patients without prior revascularization who underwent stress electrocardiography-gated myocardial perfusion single photon emission computed tomography (MPS) for 2.8 +/- 1.2 years, during which 146 cardiac deaths occurred (2.7%, 1.0%/y). The treatment received within 60 days after MPS was used to define the subgroups (revascularization in 402 patients, with cardiac death occurring in 6.2%, vs medical therapy in 4964 patients, with cardiac death occurring in 2.4%; P < .0001, chi2 = 18.7). Adjustment for nonrandomized treatment assignment used a propensity score based on logistic regression modeling of referral to revascularization. The percent of myocardium that was ischemic was the most important predictor of revascularization. The overall model (multivariate chi2 = 728, c index = 0.89, P < 10(-5)) was used as a propensity score. Cox proportional hazards analysis, assessing the relationship between MPS results, non-MPS covariates, and cardiac death, revealed that EF was superior to percent ischemic myocardium in the prediction of cardiac death after adjustment for pre-MPS data and the propensity score. However, an interaction between percent ischemic myocardium and revascularization was present such that, irrespective of EF, patients with little or no ischemia had an improved survival rate with medical therapy, whereas with increasing ischemia, progressive improvements in survival rate were noted with revascularization. Although EF predicts cardiac death, only inducible ischemia identifies which patients have a short-term benefit from revascularization.