Protocol for a retrospective, comparative effectiveness study of the association between transesophageal echocardiography (TEE) monitoring used in coronary artery bypass graft (CABG) surgery and clinical outcomes

AbstractBackgroundCoronary artery bypass graft (CABG) surgery is the most widely performed adult cardiac surgery in the US. Transesophageal echocardiography (TEE) is an ultrasound-based cardiac imaging modality used in CABG surgery for hemodynamic monitoring and management of complications related to cardiopulmonary bypass. However, there are no comparative effectiveness studies (randomized or non-randomized) that have investigated the relationship between TEE monitoring and clinical outcomes among patients undergoing CABG surgery. Because of this lack of evidence, recommendations for TEE in CABG surgery remain indeterminate (Class II). We aim to compare the clinical outcomes of patients undergoing CABG surgery with vs without TEE monitoring. This protocol will detail how we plan to investigate the hypothesis that TEE monitoring in CABG surgery will be associated with improved clinical outcomes.Methods and AnalysisThis investigation will be an observational retrospective, comparative effectiveness, cohort study using Centers for Medicare and Medicaid Services (CMS) claims data from January 1, 2013 to October 15, 2015. The aim is to determine if TEE monitoring during CABG surgery is associated with improved 30-day survival, lower incidence of stroke, shorter length of hospitalization, and incidence of esophageal perforation. To alleviate the potential bias from unmeasured confounding, we propose leveraging hospitals’ (or surgeons’) preference for TEE in CABG surgery as an instrumental variable (IV). We will combine this IV technique with statistical-matching-based methods by pairing hospitals (or surgeons) with similar observed confounding variables but considerably different preference for TEE monitoring in CABG surgery. Our research design is meant to emulate a cluster-randomized encouragement experiment. The following a priori protocol will detail how we plan to execute this analysis.