Simulation-based training intervention using artificial intelligence to improve clinical bronchoscopy performance: a pre-postintervention study.

Abstract

Does a simulation-based training intervention with an artificial intelligence (AI) navigation system improve their clinical bronchoscopy performance? And can the AIs outcome measures be used to evaluate clinical performance? Pre-postintervention study. Odense University Hospital of Southern Denmark, pulmonary endoscopy suite. Nine bronchoscopists (4 experienced, >500 bronchoscopies and 5 intermediates, 10-500 bronchoscopies). Diagnostic completeness (DC), structured progress (SP), procedure time (PT) and procedure efficiency (DC/PT). The primary outcome measures showed no statistically significant difference between the pre- and postintervention bronchoscopies DC: 53% versus 59%, p=0.16, SP: 29% versus 32%, p=0.35 and PT: 219 s versus 181 s, p=0.22. The experienced outperformed the intermediates regarding DC: 73% versus 43%, p<0.001, SP: 47% versus 13%, p<0.001 and procedure efficiency: 533 s/full inspection versus 274 s/full inspection, p<0.001 but not on PT: 189 s versus 208 s, p=0.53). DC, SP and PT showed no statistically significant difference after a simulation-based training intervention. DC, SP and procedure efficiency differentiated between experienced and intermediate bronchoscopists and can be used to evaluate clinical bronchoscopy performance.