Reviewed by Barry Rittmann, MD, MPH, Virginia Commonwealth University
This past month, The American Journal of Infection Control published a pair of articles that assessed 2 different modalities to assist in surveillance of different metrics in infection prevention. The first evaluated artificial intelligence in identifying central line-associated bloodstream infections and catheter-associated urinary tract infections. 2 separate models were created, using ChatGPT-4 and Mixtral 8x7b-based models. They evaluated 6 NHSN training scenarios of varying complexity. In both models, the A.I. tools were able to accurately report both CLABSI and CAUTI scenarios regardless of complexity. This is a promising development in the use of A.I. technology regarding HAI surveillance and could likely be used in other scenarios as well, including surgical site infections and ventilator associated events.
In the second article, a Swedish medical center compared hand hygiene monitoring of an electronic monitoring system to direct observation. Using a 4-step protocol created for the study, they were able to evaluate hand hygiene events in all of the WHO “My five moments of hand hygiene.” In this study, they found that the hand hygiene monitoring system had high sensitivity (90.2%) and positive predictive value (95.7%), with an accuracy of 87.1%. However, specificity and negative predictive value, compared to direct observation were low, at 50.0% and 29.0%. This study illustrated several of the strengths and weaknesses of electronic monitoring systems, and uniquely was able to assess all of the “My five moments of hand hygiene”.
References:
Wiemken TL, Carrico RM. Assisting the infection preventionist: Use of artificial intelligence for health care-associated infection surveillance. Am J Infect Control. 2024 Feb 29:S0196-6553(24)00077-4. doi: 10.1016/j.ajic.2024.02.007. Epub ahead of print. PMID: 38483430.
Granqvist K, Ahlstrom L, Karlsson J, Lytsy B, Erichsen A. Hand hygiene in a clinical setting: Evaluation of an electronic monitoring system in relation to direct observations. Am J Infect Control. 2024 Jan 23:S0196-6553(24)00052-X. doi: 10.1016/j.ajic.2024.01.013. Epub ahead of print. PMID: 38272313.
