A progressive study published in the American Journal of Infection Control (AJIC) unveils that artificial intelligence (AI) technologies possess significant potential in the accurate detection of healthcare-associated infections (HAI), even in multifaceted clinical scenarios. This illuminating research underscores the pressing need for unambiguous and steady language when employing AI resources in this regard. It opens up a vista of possibilities for incorporating AI technology as a budget-friendly component in systematic infection surveillance programs.
Based on the latest HAI Hospital Prevalence Survey carried out by the Centers for Disease Control and Prevention, approximately 687,000 HAIs were reported in acute care hospitals across the U.S. in 2015. Further alarming is the fact that 72,000 HAI-linked fatalities occurred among hospital patients the same year. About 3% of all hospital patients are threatened with at least one HAI at any particular time.
While infection surveillance programs and other infection-prevention protocols have curbed the occurrence of HAIs, they still pose significant risks, especially for critically ill hospitalized patients with inserted medical devices like central lines, catheters, or breathing tubes. HAI surveillance programs implemented by many hospitals and healthcare facilities require extensive resources, continuous training, and specialized expertise. In settings where resources are constrained, a budget-friendly alternative could bolster surveillance programs, ensuring improved protection for high-risk patients.
The recently published study, undertaken by researchers at Saint Louis University and the University of Louisville School of Medicine, investigates the proficiency of two AI-powered tools in the accurate identification of HAIs. The tested tools, built using OpenAI’s ChatGPT Plus and the open-source language model named Mixtral 8x7B, were evaluated on two types of HAIs: central line-associated bloodstream infection (CLABSI) and catheter-associated urinary tract infection (CAUTI).
Six fictional patient scenarios featuring varying complexity levels were presented to the AI tools to ascertain whether the descriptions indicated a CLABSI or a CAUTI. Ambiguities or missing information in the descriptions might hinder the AI tools from delivering precise results. The pressing need for human supervision of these technologies is thus underscored.
This groundbreaking research paves the way for continued development of AI tools leveraging real-world patient data to support infection preventionists. The study suggests that AI-powered tools may offer a cost-effective route to enhance our surveillance programs, bolstering infection preventionists in their daily responsibilities.