Beyond the Bots: Human Factors Assessments and Interventions in Robotic-Assisted Surgeries Using Mixed Methods Approaches

An operating room (OR) is a complex, adaptive, and safety-critical system where various elements and agents continuously interact. The introduction of technologies, such as surgical robots, into ORs modifies these interactions and presents challenges for surgical teams, particularly surgeons. While robotic-assisted surgeries (RAS) improve precision and patient outcomes, many errors stem from deficiencies in non-technical aspects of performance (e.g., communication failures and poor decision-making) rather than technical skills alone. Human factors considerations focus on aligning technologies with human capabilities and limitations to enhance safety, efficiency, and usability. However, assessing these factors is challenging due to the subtle, often intangible nature of human behaviors and decision-making processes. Traditional subjective methods, such as observations and self-reports, offer valuable insights but are prone to bias. Objective methods, including sensor-based monitoring and data-driven models, offer reliable alternatives but require technical expertise, can be computationally intensive, and often rely on subjective ground truth, limiting their objectivity. This dissertation adopts a mixed-methods approach, combining qualitative insights from surgeons with objective, data-driven methodologies to explore human factors in RAS. First, a sensor-based framework was developed and tested in live RAS to assess inefficiencies in OR layout, workflow, and team dynamics, demonstrating that certain human factors can be objectively evaluated. Next, electroencephalogram-based neural insights determined intraoperative variations in surgeons’ cognitive workload during robotic teleoperation. These measures highlighted the need for targeted interventions to improve the non-technical aspects of surgeons’ performance. To address this, a dyadic video-reflection coaching framework was developed and effectively enhanced surgeons’ non-technical skills (NTS). This coaching solution also revealed rich, qualitative insights into the challenges surgeons face in RAS. While effective, dyadic NTS coaching was resource intensive. To scale this solution, machine learning and large language models were leveraged to predict surgeons’ behaviors and provide automated, coaching-style feedback for NTS training. Through these integrated approaches, this dissertation offers practical insights for human factors considerations in robotic integration, with implications for surgical education, surgical robot design, and the broader organizational and sociotechnical contexts in which RAS occur. This work offers new directions for improving surgeons’ intraoperative performance, team dynamics, and patient safety in RAS.