In the early months of 1975, as a fourth-year surgical resident, I had the unique experience of having a rotation at the Hospital Albert Schweitzer in Deschapelles, nestled in the remote Artibonite Valley of Haiti. The goals of my surgical program were to expose us to a remote culture in great need of surgical care and to allow us to use our surgical knowledge, judgment and skills to hone independent thinking even though we were supported by a more senior surgeon. It became obvious after only a few days in the country that we would also be exposed to diseases, injuries and surgical conundrums that would challenge our neophyte surgical heuristic and previous experiences that defined our traditional surgical decision making. I quickly experienced an example of this concept!
Early in my Haitian experience, I was called to evaluate a relatively healthy young man with a classic acute abdomen: diffuse tenderness with rebound, absent bowel sounds and abdominal wall spasm. What could be clearer? To my amazement, during my exploration via a midline incision, I found diffuse small white excrescences throughout the peritoneal cavity without any evidence of visceral perforation, classic inflammation or obstruction. Appropriate cultures identified acid-fast bacilli. Yes, this was “classic” tuberculous peritonitis!
Having never seen a case, the heuristic that I had come to believe in regarding the “acute abdomen” had led me astray. Rather than a midline incision, this patient would have been most appropriately treated with anti-tuberculous therapy. It took me two additional cases before I stopped making this error and appreciated that tuberculous peritonitis should be high on my differential.
This brings me to the meaning of this editorial’s title. Since virtually every current healthcare publication including General Surgery News (B. Ramshaw, “AI Can Never Do It Alone,” September 2023, page 1) highlights the potential role and benefits of artificial intelligence, I wonder whether AI, had it been available in 1975, would have helped direct me to a more appropriate diagnosis for my Haitian patients who presented with a surgical acute abdomen. Perhaps the AI platform would not have suffered from my culture-bound heuristic. Maybe the AI algorithm would have filtered in the important informational concepts of geography, culture, local infections and other associations more relevant to the environment in which I found myself. On the other hand, would I have believed the AI-generated diagnosis? Would the conclusions generated by AI have been convincing enough to assuage my reaching for a scalpel?
As AI concepts are introduced to enhance diagnostic acumen and surgical decision making, will surgeons have the ability and fortitude to follow AI dictates when these computer-generated recommendations are in stark contrast to his or her surgical heuristic, diagnostic skills and judgment? If in the future, treatment is based on nonhuman recommendations, what liability issues will the surgeon face should potential morbidity and mortality result from incorrect AI renderings? In contrast, what will be the litigious consequences should the surgeon not follow AI diagnostic and treatment recommendations and a poor outcome ensues?
These are important collateral issues to consider. If I had the benefit of an AI-generated diagnosis for my young Haitian patient, would I have neglected to follow all of my teachings regarding the acute abdomen and treated him with anti-tuberculous therapy? Would my prior experience and heuristic have been too compelling? I am not sure. What would you have done?
Dr. Greene is the senior medical advisor for General Surgery News.