This conference uses three short (non-infectious) FUO-style cases to explore fever patterns that initially appeared infectious but ultimately required a broader diagnostic frame. The cases include aseptic meningitis after dental work and medication exposures, prolonged fever in a patient with controlled HIV and prosthetic valves ultimately raising concern for drug fever from eplerenone, and fever/SIRS after ECMO decannulation. Together, the cases emphasize the importance of medication timelines, stopping rules for antimicrobials, and recognizing inflammatory or iatrogenic syndromes that can mimic infection.
The teaching portion reviews drug-induced aseptic meningitis, drug fever, and ECMO decannulation fever, with attention to how clinicians decide when negative infectious testing should meaningfully change management. The final section applies the same cases to AI tools, comparing how large language models approached the differential diagnosis, what they prioritized correctly, and where they risked overcalling infection or missing key temporal clues. The session highlights both the promise and limitations of AI as a diagnostic reasoning aid: useful for broad differentials and pattern recognition, but still dependent on careful clinical framing, medication reconciliation, and clinician judgment.