Patient: M.T., 44-year-old warehouse supervisor 
Chief concern (Day 1): “Bad low-back pain after lifting a box.” 
Home AI tool: Patient uses the health system’s portal symptom-checker chatbot. After entering “low back pain,” 
“both legs tingling,” “can’t feel when wiping,” and “peeing less than usual,” the bot outputs “likely muscular 
strain home care; clinic visit in 3–5 days.” No red-flag alert is displayed. 
Course: 
• Day 2–3: Pain worsens; bilateral sciatica, saddle numbness, and urinary retention (“I haven’t peed since 
last night unless I strain”). The bot again recommends home care. 
• Day 4 (ED): Patient febrile, hypotensive, confused. Bladder scan >1200 mL; purulent urine after 
catheterization. MRI lumbar: massive central L4–L5 disc herniation compressing cauda equina. 
• Labs: Leukocytosis, lactate 5.2 mmol/L, creatinine bump. 
• Outcome: Despite urgent decompression and broad-spectrum antibiotics/ICU care, the patient develops 
uroseptic shock with multi-organ failure and dies on Day 6. 
Studies show online/AI symptom-checkers vary widely in diagnostic/triage accuracy and may miss 
emergencies; WHO and AHRQ urge caution and human oversight for clinical AI