Video-Based Assessment and Prediction of Surgical Outcomes in Sleeve Gastrectomy

Surgical complications impact quality of life and escalate healthcare costs. Intraoperative performance, accounting for 40% to 60% of adverse events, is often judged through visual cues. The accuracy of this in predicting postoperative outcomes remains underexplored. We sought to assess surgeon accuracy in predicting outcomes from robotic sleeve gastrectomy (RSG) videos.

Fellowship-trained surgeons reviewed RSG clips collected from a high-volume tertiary bariatric center. Twenty-five cases were chosen: 13 with complications and 12 without. Surgeons predicted which cases would result in complications and assigned Global Evaluative Assessment of Robotic Skills scores (GS) for technical proficiency. The primary outcome was the accuracy of predicting postoperative outcomes, including bleeding, leak, stenosis, or no complications. Secondary outcomes involved correlations between GS, comment themes, predicted outcomes, and actual outcomes.

Ninety-one clips were reviewed by 19 surgeons. Accuracy for predicting specific complications was 28.6%, whereas "complicated" vs "uncomplicated" accuracy was 47.3%. Mean GS showed no significant differences between procedures with or without complications (19.97 ± 3.98 vs 21.27 ± 3.47). Videos predicted as uncomplicated had a higher GS (22.15 ± 2.97 vs 19.93 ± 3.89, p < 0.05). Comment analysis showed correlations between predicted outcomes and themes (eg bleeding) as well as staple line or suturing issues and stenosis.

Experienced bariatric surgeons exhibited low accuracy in predicting complications from edited RSG videos. Visual cues failed to reliably predict outcomes, highlighting limitations of using video-based methodology for risk assessment. As this could impact surgical training and medico-legal proceedings, further research is needed to improve predictive accuracy and address cognitive biases in the surgical environment.