Lygia Stewart, Lawrence W. Way, UCSF and SF VA Medical Center, San Francisco, CA, UCSF, San Francisco, CA

Purpose: Laparoscopic bile duct injuries (BDI) continue to be a problem. We previously utilized human error analysis to elucidate mechanisms for prevention. In this study we used principles of cognitive psychology to examine impediments to injury identification (ID) during the index operation. Specifically how confirmation bias impedes injury ID. Confirmation bias is selective thinking where findings that confirm pre-set beliefs are observed and contradictory findings ignored. Methods: 280 BDI following laparoscopic cholecystectomy were studied. Stewart-Way injury distribution was Class I, 7%; Class II, 21%; Class III, 61%; and Class IV, 10%. Records were examined for: 1) intra-operative cholangiography: routine (IOC), or of an abnormal duct (DIC), 2) perceived anatomic anomalies (additional ducts, ductal anomalies, vascular anomalies, other), 3) additional input (conversion to open procedure, specimen exam). Cases identified at the index operation or post-op were compared. Results: 25% of injuries were recognized at the index operation. In 38% the OP report noted nothing unusual, including 9 cases where the CHD was clipped. Routine IOC did not improve injury ID (22% with IOC vs 19% without, P=0.318), but with suspected injury, DIC improved injury ID (94% with DIC vs 21% without, P<0.0001). Identification of vascular anomalies correlated with decreased injury ID (13% with vs 46% without, P<0.0001). Noting an additional duct improved injury ID (76% with a duct vs 5% without, P<0.0001), but in 24% of these cases the duct was classified as an accessory duct (reflecting the power of bias). No injuries were identified in 23 cases where ductal anomalies, but not an additional duct, were noted. Additional input most improved injury ID when an injury was suspected (ID 44% in cases opened due to poor visibility vs 89% specimen exam for suspected BDI, P=0.032). More abnormalities were reported in cases with bile duct injury ID (2.5 cases ID vs 0.8 cases not ID, P<0.0001). Number of abnormalities noted vs ID was; 0, 0%; 1, 16%; 2, 33%; 3, 74%; and 4, 82%. Thus, it took 3 abnormalities to derail confirmation bias. Observations in the context of a suspected injury improved ID, 88% suspected vs 18% unsuspected, P<0.0001. Conclusions: This study elucidates how confirmation bias impairs the operative recognition of laparoscopic bile duct injuries. Awareness of these principles should improve injury identification at the index operation, and possibly the prevention of laparoscopic bile duct injuries.