Figure 3

Graphical depiction of the Net Reclassification Improvement (NRI) where univariable Cox regression models with cardiovascular magnetic resonance (CMR) ejection fraction (EF) predicting all cause mortality (panel A) or death or cardiac transplant (panel B) are compared to Cox regression models containing echocardiography (echo) EF. Since the reclassification improved using CMR EF relative to echocardiography EF in both those with events and those without events, the NRI for all cause mortality and death/cardiac transplant are 0.51 (includes rounding error; panel A) and 0.61 (panel B), respectively, after summing these net improvements for events and nonevents.