New research suggests that an Artificial Intelligence (AI)-guided infection prevention bundle may provide valuable support in both controlling Clostridioides difficile infection (CDI) incidence and advancing antimicrobial stewardship within hospital environments. This recent study, conducted at Michigan Medicine, linked to the University of Michigan, set its focus on comparing patient outcomes before and after the introduction of an institution-centric previously validated AI model explicitly designed to predict CDI risk.
Findings, published on our affiliated portal, HCPLive.com, reveal a somewhat complicated picture. Despite the AI bundle making no significant contribution to decreasing CDI incidence, it demonstrated a distinct association with drastic reductions in CDI-related antimicrobial use. Artificial Intelligence models, developed as a means of predicting CDI risk in patients currently in hospital care, with their design aimed at identifying patients most likely to benefit from targeted infection prevention resources. Still, up until now, their relationship with patient results in a clinical context has remained unclear.
The bacterium causing diarrhea and colitis, Clostridioides difficile, is a significant health concern in the U.S., with almost half a million infections reported annually. Tackling this prominent problem necessitates targeted prevention initiatives, given the constraint on most hospitals’ resources. The investigation of this AI-guided infection prevention bundle integrated into Michigan Medicine’s clinical workflows focused on comparing the adult inpatient hospitalizations pre- and post-AI implementation.
The study periods include a pre-AI period from September 2021 to August 2022 and a post-AI period from January 2023 to December 2023, with the four-month period of AI implementation excluded from the analysis. The mainly female and White patient groups, in both the pre-and post-AI periods, served to highlight significant differences in factors such as age, COVID-19 related hospitalizations and recent history of hospitalization. The CDI incidence rate was the primary outcome of the study with secondary outcomes encompassing antimicrobial use and qualitative assessments of bundle use.
Notably, investigators identified significant reductions in the use of certain antibiotics, more prominent in high-risk hospitalizations flagged by the AI tool. Furthermore, the fieldwork also revealed varied experiences with AI-guided workflows amongst healthcare staff, indicating a scope for future research in the domain. Future investigations should aim at devising novel strategies for continuous education and stakeholder engagement, exploring scalability across diverse healthcare settings, identifying optimal alert thresholds balancing sensitivity and alert burden, and assessing effect on clinical outcomes via randomized clinical trials.