Joshua Lang, Cardiologist and Instructor in Medicine at Brigham and Women’s Hospital, shared a post on LinkedIn:
“The worst medical mistake I’ve ever made began at 2 in the morning on the overnight oncology service. I was a senior resident. Up on the 7th floor in a windowless room, sitting alone, clicking through lab results, and probably eating stale cookies. My pager went off that there was a new patient in the emergency room.
Daniel, a 45 year old man with metastatic pancreatic cancer had fluid building up in his lungs and was coming in short of breath. (Details changed for privacy.) I looked into his chart and there was cancer everywhere, abdomen, lungs, liver. He had been dealing with this for years and had been on a couple lines of chemotherapy. Right before I went down to meet him I saw that his code status was full, meaning he had opted for chest compressions and intubation in the event that his heart and lungs gave out. Had anyone talked to him about his prognosis?
He was on a stretcher accompanied by his elderly parents, who held his hand. We went through all of his symptoms, medications, and checklists to get him safely into the hospital. As part of procedure, we went over his code status. I felt that it was a righteous thing to do to review his prognosis and discuss code status.
I asked him, ‘Has anyone discussed prognosis with you?’ No, not really, he said. (There was this sentiment among the residency and at large in medicine at the time that many oncologists only presented the rosiest possibilities and ignored the hard facts and that patients ended up in hospice too late. See all the writing by Atul Gawande on mortality.)
I told him that with metastatic pancreatic cancer that we would hope for the best but that in all likelihood his prognosis was less than 6 months. What were his values? Would aggressive CPR and intubation make sense in light of all that? The family cried and I did my best to be supportive. All there in the emergency room at 3am.
At some point he stopped me with his parents and they said, ‘We are just so surprised, this is so different than what we have heard so far.’
Alarm bells went off in my head. I asked them for a moment to step outside, went back into his chart, and searched the oncology notes. My heart sank, and I felt sick. He had metastatic pancreatic NEUROENDOCRINE tumor. The prognosis can be years on treatment. I WAS DEAD WRONG. I went back in, apologized, said I spoke too soon, and that the team would come in first thing. They were horrified but calm. I emailed everyone on his care team and my bosses, and cried.
My mistake was one of hubris and lack of attention to detail. That’s why it was so bad. It wasn’t just a mistake… it was over confidence. AND it’s exactly the kind of mistake that AI can make with its probabilistic models. It’s exactly the mistake Sam Altman made when he demo’d a cancer patient navigating care with our early LLMs. Pancreatic neuroendocrine tumors are very rare compared to adenocarinoma. I wondered today, would ChatGPT running GPT5 make the same error?”
More posts featuring Pancreatic Cancer.