Purpose Physicians-in-training often perform bedside thoracenteses in academic medical centers, and complications are more common among less experienced clinicians. Simulation-based mastery learning (SBML) is one potential solution to this problem. This study evaluated the effects of a randomized trial of thoracentesis SBML on patient complications: iatrogenic pneumothorax (IP), hemothorax, and reexpansion pulmonary edema (REPE). Method The authors randomized internal medicine residents to undergo thoracentesis SBML at a tertiary care academic center from December 2012 to May 2016. They subsequently compared thoracentesis complications from procedures performed by SBML-trained residents, traditionally trained residents (no simulation training), and those referred to pulmonary medicine or interventional radiology (IR). Results During the study period, 917 thoracenteses were performed on 709 patients. IP occurred in 60 (6.5%) procedures, of which 7 (11.6%) were clinically meaningful. SBML-trained residents performed procedures with a trend toward lower combined clinically meaningful complications (IP, hemothorax, REPE) compared with traditionally trained residents (7.9% vs. 0%; P =.06). SBML-trained residents caused fewer clinically meaningful IPs compared with traditionally trained residents, pulmonary, and IR referrals (P =.02). Hemothorax occurred after 4 (0.4%) thoracenteses, and SBML-trained residents had a trend toward lower hemothorax (0) compared with other groups (P =.07). REPE occurred after 3 (0.3%) procedures, with no differences between groups. SBML-trained residents performed procedures with lower combined clinically meaningful complications compared with other groups (P =.008). Conclusions Residents randomized to an SBML intervention performed thoracenteses with low rates of clinically meaningful complications. Rigorous education represents a successful quality improvement strategy.