Background Cardiac magnetic resonance (CMR) permits robust risk stratification of discharged ST-segment–elevation myocardial infarction patients, but its indiscriminate use in all cases is not feasible. We evaluated the utility of left ventricular ejection fraction (LVEF) by echocardiography for a selective use of CMR after ST-segment–elevation myocardial infarction. Methods Echocardiography and CMR were performed in 1119 patients discharged for ST-segment–elevation myocardial infarction included in a multicenter registry. The prognostic power of CMR beyond echocardiography-LVEF was assessed using adjusted C statistic, net reclassification improvement index, and integrated discrimination improvement index. Results During a 4.8-year median follow-up, 136 (12%) first major adverse cardiac events (MACE) occurred (47 cardiovascular deaths and 89 readmissions for acute heart failure). In the entire group, CMR-LVEF (but not echocardiography-LVEF) independently predicted MACE occurrence. The MACE rate significantly increased only in patients with CMR-LVEF<40% (≥50%: 7%, 40%–49%: 9%, <40%: 27%, P <0.001). Most patients displayed echocardiography-LVEF≥50% (629, 56%), and they had a low MACE rate (57/629, 9%). In patients with echocardiography-LVEF<50% (n=490, 44%), the MACE rate was also low in those with CMR-LVEF≥40% (24/278, 9%) but significantly increased in patients with CMR-LVEF<40% (55/212, 26%; P <0.001). Compared with echocardiography-LVEF, CMR-LVEF significantly improved MACE prediction in the group of patients with echocardiography-LVEF<50% (C statistic, 0.80 versus 0.72; net reclassification improvement index, 0.73; integrated discrimination improvement index, 0.10) but not in those with echocardiography-LVEF≥50% (C statistic 0.66 versus 0.66; net reclassification improvement index, 0.17; integrated discrimination improvement index, 0.01). Conclusions A straightforward strategy based on a selective use of CMR for risk prediction in ST-segment–elevation myocardial infarction patients with echocardiography-LVEF<50% can provide insights into patient care. The cost-effectiveness of this approach, as well as the direct implications in clinical management, should be further explored.