This retrospective validation study compared conventional hard boundary objective performance indicators (OPIs) with probabilistic expected OPIs that model uncertainty in surgical step timing. Using 1,016 da Vinci Xi robotic cases across seven procedures, including 238 lobectomies, the authors analyzed 5,408 single instance annotated steps and 27,040 paired OPIs. Examples included tool path length, endowrist angular path length, camera clutch rate, energy use rate, and hand clutch rate during specific surgical steps.
Expected OPIs were less often biased than hard boundary OPIs (49.37 percent vs 56.63 percent) and showed lower variance in 40.56 percent of comparisons, vs 8.52 percent favoring hard boundary OPIs. The authors conclude that probabilistic OPIs are more robust to annotation variability in surgical data analysis.
