GI Endoscopy Training: Validation of a Computer-Based Endoscopy Simulator

Abstract

Background: Conventional endoscopy training allows trainee access to real patients under supervision. This method may decrease diagnostic accuracy and increase patient discomfort, procedure-related morbidity and mortality. Computer-based endoscopy simulation permits more standardized acquisition and evaluation of skills. Aim: To establish construct validity of a computer-based endoscopy simulator by determining whether it differentiates novices from experienced endoscopists. Methods: An expert group comprised 4 experienced gastroenterologists, each having performed over 1000 colonoscopies. A novice group comprised a cohort of 30 first-year Canadian GI trainees (< 10 colonoscopies each) attending a 2-day endoscopy course hosted by McMaster University in July 2006. All participants received identical pretest instruction, and were asked to undertake the same colonoscopy simulation on the GI Mentor II VR simulator (Simbionix, USA, Cleveland, OH), with a redundant sigmoid and transverse colon but no significant pathology other than melanosis coli. Endoscopic skills were assessed by time to cecum, total procedure time, mucosa visualization (%), efficiency of examination, time with a clear view (%), number of times caused excessive local pressure, time patient was in pain (%), total time with loop formation, number of times a 3-D map was used, and number of times with lost view. Differences in mean scores among the 2 groups were assessed by independent sample t-tests. P-values were two-tailed, with statistical significance evaluated at 0.05 level. Results: Experts were more efficient than novices (91.3% vs 62.2%, p < 0.05). There was a significantly shorter procedure time (451.0s vs 886.4s, p < 0.05) and time to cecum (172.3s vs 531.6s, p < 0.05), larger proportion of time spent with a clear view of the lumen (90.5% vs 84.1%, p < 0.05), and less time with loop formation (1.8s vs 12.6s, p = 0.01) or excessive local pressure (0 vs 3.3, p < 0.05) in the hands of the experts. Percentage of mucosa visualized (85.3% vs 82.8%) and pain (0.25% vs 1.1%) were not statistically different. 3D map was used more often by novices than experts (17.2s vs 0 s, p < 0.05). Experts showed greater consistency in their performance than the novices, as demonstrated by a smaller SDs across all measures. Conclusion: The computer based endoscopy simulator can distinguish between expert and novice endoscopists, and thus has construct validity. Simulation may have a useful role in endoscopy training curricula as an objective assessment of skill acquisition. Future studies should validate the simulator as a predictor of performance in real procedures.