Yup, I'm wrong in the specifics here. I think I'm conflating the stats for negative appendectomy and the willingness to go to surgery without positive imaging i

Yup, I'm wrong in the specifics here. I think I'm conflating the stats for negative appendectomy and the willingness to go to surgery without positive imaging in that setting with the ~30% of cholecystectomy cases that don't improve with surgery. (Of course, the risk/benefits of missing an infected appendix vs taking out a healthy one are very different to gallbladders with stones.) That said, I do think the underlying point I'm trying to make is still accurate. Most of the audience here are software engineers. Unlike debugging software, where we can often eliminate possibilities with 100% accuracy through objective tests and come to a conclusion with complete confidence, or at least run experiments with no consequences, humans are messy, poorly observable, and heterogeneous. A relevant history helps speed things up when trying to fix software, but it can be vital when fixing humans. Given the rate of asymptomatic gallstones visible to imaging in the population, if these lentils are causing acute liver injury, I think it's reasonable to expect that some people have had unnecessary surgery as a result. If you knew your patient had eaten these lentils, would you still send to surgery with positive imaging, or wait a few days and see if things improved?