A nurse at St. Rose Dominican Hospital refused to follow an AI sepsis alert that recommended IV fluids for an elderly patient with kidney problems, potentially preventing life-threatening complications.
At St. Rose Dominican Hospital in Henderson, Nevada, nurse Adam Hart encountered an AI-generated sepsis alert for an elderly female patient with dangerously low blood pressure. The hospital's AI system flagged the patient for sepsis protocol, prompting the charge nurse to order immediate IV fluids. However, Hart observed that the patient had a dialysis catheter indicating compromised kidneys, and warned that routine IV fluids could overwhelm her system and cause fluid to accumulate in her lungs. When Hart refused to follow the AI-prompted protocol, a physician intervened and ordered dopamine instead of fluids to raise blood pressure without adding volume. The incident highlights broader concerns about AI implementation in healthcare, where hospitals have widely adopted algorithmic models including Epic's sepsis-prediction algorithm, which later evaluations found substantially less accurate than marketed. Nurses across the country have expressed wariness about unvalidated AI tools, with some staging demonstrations in California and New York City. The report describes multiple examples of AI monitoring systems generating frequent false alarms and alerts that are difficult for nurses to interpret, leading to workflow disruptions and mistrust among healthcare workers.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that fail to perform reliably or effectively under varying conditions, exposing them to errors and failures that can have significant consequences, especially in critical applications or areas that require moral reasoning.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed