Development of performance metrics for automated evaluation of surgeons' performance under simulation-based training

General surgery residency is continuously changing to accommodate eighty-hour work weeks, increasing patient expectations, and a malpractice environment. These factors contribute to decreased autonomy during the chief residency year, which creates a wider distribution of skills among graduating residents. While hospital administrators are concerned with quality assurance, patient safety, and cost reduction, patients continue to increase their expectations and awareness of surgeons' skills. To address the concerns of both employees and patients, it is important to ensure that residents are adequately prepared to perform surgery. The University of Virginia Health System does not use a standardized method to evaluate residents' readiness to perform all necessary surgical skills before graduating. Our research into performance metrics contributes to this end goal: the automation of the resident evaluation process. We have hypothesized a series of key performance metrics that are transferrable from the testing laboratory to the operating room. This research included basic task element analysis based on twenty-nine videos of novice residents and surgical experts performing vessel ligation on a bench-top simulator. Additional data was collected from ten first-year residents performing vessel ligations for the first time while wearing inertial sensors and eye-tracking goggles. The residents' data was examined from physical and cognitive performance perspectives, with metrics including time, hand movements, and visual gaze behaviors. The cross-examination of novice and expert data is essential for the furtherance of metric development that will ultimately lead to automated evaluation.