2020 American College of Cardiology/American Heart Association (ACC/AHA) Guidelines state that the ideal time for mitral valve surgery in primary mitral regurgitation (PMR) is when the LV approaches but has not yet reached echocardiographic LV ejection fraction (EF) < 60% or LV end-systolic dimension (ESD) > 40 mm. However, it is difficult to know the imminent risk of crossing this threshold when the surgical outcome is less optimal. Using machine learning and statistical models, we have shown that cardiac magnetic resonance (CMR) LV sphericity index (SI) and LV mid circumferential strain rate (SRcirc) added to LVEF and LVESD predict LVEF < 50% after mitral valve surgery. Here we test the hypothesis that these CMR features predict LVEF < 60% in asymptomatic PMR patients at 18 months. 33 asymptomatic PMR patients with moderate to severe mitral regurgitation had CMR with tissue tagging at baseline and every 6 months for 18 months. Two types of models were employed to predict LVEF < 60% at 18 months: a model using CMR features at a single time point (e.g., baseline) and a model utilizing repeated measurements over time. CMR LVEF decreased below 60% in 13 patients over 18 months. LVEF varied over time with an inverse relation to mean arterial pressure and mean end-systolic wall stress. Random Forest models utilizing LV SI, LV mid SRcirc, LVESD, and LVEF at a single time point (baseline) had a predictive accuracy of 64%. LV SI, LV mid SRcirc, LVESD and LVEF at baseline, 6, and 12 months achieved a higher predictive accuracy of 79%, improved sensitivity from 57% to 85% than baseline alone and identified a threshold of CMR LVEF 63%-64% signaling LVEF < 60%. The variability of LVEF due to blood pressure dependence may require a longitudinal study that incorporates LVEF, LVESD, SRcirc at multiple time points to identify the threshold at which LVEF is at risk for decline to less than 60%.