Artificial intelligence (AI) technology can accurately identify cases of healthcare-associated infections (HAI) even in complex clinical situations, a study has found.
The study, published in the American Journal of Infection Control, highlights the need for clear and consistent language when using AI tools for this purpose.
The findings also illustrate the potential for incorporating artificial intelligence technology into a cost-effective component of routine infection surveillance programs.
According to the latest HAI hospital prevalence survey conducted by the Centers for Disease Control and Prevention, there were 687,000 (680,000) HAI cases in U.S. acute care hospitals in 2015, and there were 72,000 HAI-related deaths among hospital patients.
Researchers say about 3% of hospital patients have at least one nosocomial infection at any given time.
Implementation of infection surveillance programs and other infection prevention protocols have reduced the incidence of hospital infections, they said, but they remain a risk, especially for critically ill hospitalized patients who have devices such as central lines, catheters or breathing tubes inserted.
Many hospitals and other healthcare facilities have nosocomial infection surveillance programs to monitor for increased risk of infection, but these programs require significant resources, training and expertise to maintain, the researchers said.
In resource-limited settings, cost-effective alternatives may help strengthen surveillance programs and better protect high-risk patients, they said.
In the new study, researchers at Saint Louis University and the University of Louisville School of Medicine evaluated the performance of two artificial intelligence tools to accurately identify HAIs.
One of the tools was built using OpenAI’s ChatGPT Plus, and the other was developed using an open source large language model called Mixtral 8x7B.
The tools were tested against two types of HAI: central line-associated bloodstream infection (CLABSI) and catheter-associated urinary tract infection (CAUTI).
CLABSI is a serious infection that occurs when germs, usually bacteria or viruses, enter the bloodstream through a catheter. CAUTI occurs when bacteria, usually bacteria, enter the urethra through the catheter and cause an infection.
The AI tool was presented with descriptions of six fictional patient scenarios of varying complexity and asked whether the descriptions represented CLABSI or CAUTI.
These descriptions include information such as the patient’s age, symptoms, date of admission, and dates when central lines or catheters were inserted and removed. AI answers are compared to expert answers to determine accuracy.
For all six cases, both AI tools accurately identified HAIs when given explicit prompts, the researchers said.
They found that missing or vague information in descriptions could prevent AI tools from producing accurate results.
For example, the description used by the AI tool did not include the date the catheter was inserted and could not give the correct response.
Abbreviations, a lack of specificity, the use of special characters and reporting dates in numerical format rather than in months all contributed to inconsistent responses, the researchers said.
“Our findings are the first to demonstrate the power of AI-assisted HAI monitoring in a healthcare setting, but they also underscore the need for this technology,” said Timothy L. Wiemken, associate professor at Saint Louis University. technology for human supervision,” the study’s lead author.
Wiemken added: “As the role of AI in medicine rapidly evolves, our proof-of-concept study confirms the need to continue developing AI tools using real patient data to support infection preventionists.”
(Except for the headline, this story has not been edited by NDTV staff and is published from a syndicated feed.)
