Bariatric Surgeons Are Unable to Predict Complications Based on Surgical Video

Ballo MD, Profant C, Tindal EW, et al. Video-Based Assessment and Prediction of Surgical Outcomes in Sleeve Gastrectomy. J Am Coll Surg. 2025; in press. 

Fellowship-trained bariatric surgeons were recruited to review videos of robotic sleeve gastrectomy at a tertiary bariatric center to predict the procedures that would result in complications. 25 cases were used; 13 had a complication and 12 did not. Complications included five with bleeding, three with leaks, and five with stenosis.

The surgeons’ accuracy for predicting the specific complication was 28.6%, and when this was converted to predict complication versus no complication, this accuracy was 47.3%, showing their prediction was about the same as a coin flip. However, for stenosis specifically, the prediction rate was higher than the combined prediction accuracy of 28.6% at 50%. While the predication rate is higher for stenosis, 50% is still not a reliable prediction rate and highlights our lack of understanding of the intra-operative factors that will lead to complication.

Reviewing surgeons predicted more complications for operating surgeons that had lower rated technical proficiency scores (GEARS scores), which indicates an expected higher likelihood of complication in surgeons with lower robotic technical proficiency that was not present in this set of 25 videos.

This study also poses an important question over whether an artificial intelligence model would be able to identify factors that are imperceivable to the human eye or of underestimated importance that may be more impactful in predicting the likelihood of postoperative complications.