BACKGROUND: Retrospective studies of patients with ischemic mitral regurgitation(iMR) undergoing coronary artery bypass grafting(CABG) with concomitant mitral valve surgery frequently report improved survival with mitral valve annuloplasty(MVr) over replacement(MVR). However, the only randomized controlled trial found no survival difference. METHODS: Medicare claims data were queried to identify beneficiaries with iMR undergoing CABG+MVr or CABG+MVR. Kaplan-Meier estimates of survival after CABG+MVr and CABG+MVR were generated, and 20-year restricted mean survival times(RMST) were compared. Then, surgeons were stratified by their rate of CABG/MVr into groups with a demonstrated preference for MVr(PA) or MVR(PR). Outcomes were re-analyzed by surgeon preference. Overlap propensity score weighting was used for risk-adjustment in all analyses. RESULTS: Among 10,471 beneficiaries with iMR, 6,457(61.7%) underwent CABG+MVr and 4,014(38.3%) underwent CABG+MVR. Risk-adjusted RMSTs were 6.02[5.77,6.26] versus 5.57[5.33,5.81] years after CABG+MVr and CABG+MVR, respectively(difference 5.4[1.2,9.4] months, p=0.01). Among 1,118 surgeons, 223 PA(performed 2,191 surgeries
89.5% MVr rate) and 235 PR(performed 1,930 surgeries
23.0% MVr rate). Risk-adjusted RMSTs were 5.76[5.36,6.15] versus 5.77[5.40,6.14] years among beneficiaries undergoing surgery by PA surgeons and PR surgeons, respectively(difference 0.1[-6.6,6.6] months, p=0.964). CONCLUSIONS: In Medicare beneficiaries with iMR undergoing CABG+MVS, CABG+MVr was associated with improved survival, even after risk-adjustment for measured confounders. This may be due to unmeasured confounding variables affecting the decision to perform MVr or MVR, such as valvular pathology and/or severity of regurgitation. After endeavoring to account for unmeasured confounders using surgeon preference as an instrumental variable, surgeons who preferred CABG+MVr or CABG+MVR achieved similar long-term survival for their patients.