Considered one of the most urgent microbial threats by the Centers for Disease Control and Prevention (CDC), estimates of the excess costs of C. difficile infection (CDI) to the healthcare system range from $897 million to over $4 billion. Our long-term goal is to develop tools to identify patients at risk for CDI that could reduce its incidence, decrease transmission, improve patient outcomes, and reduce healthcare expenditures. We have developed and validated an algorithm using the electronic health record (EHR) to identify patients at high risk for CDI several days in advance of their diagnosis. However, there is a gap in knowledge as to whether real- world data-driven risk models can improve outcomes by guiding interventions in a clinical setting. To fill this gap in knowledge and improve CDI prevention efforts in hospitals, we propose the following specific aims: 1) to prospectively deploy an institution-specific daily risk prediction model for CDI and assess how elevated risk relates to colonization with C. difficile; 2a) to conduct a quality improvement study assessing a hospital-wide intervention bundle that incorporates patient risk for CDI; and 2b) to identify heterogeneous intervention effects across different subgroups (e.g., colonized versus not colonized; specific ribotypes) and secondary outcomes (e.g., reduced severity/complications). We will apply our model to daily extracts of EHR data, collect discarded rectal swabs and stool after standard clinical testing is completed to determine colonization status / ribotypes, and assess our model with respect to colonization status, potentially incorporating it to further improve the model. Using rates of hospital-acquired CDI, we will also assess the impact of a hospital-wide, risk-based prevention bundle rolled out for each ward in stepped-wedge, cluster- randomized fashion. The bundle will include both infection prevention and antimicrobial stewardship components. This project?s successful completion would provide a model for improving the prevention of CDI and other healthcare associated infections in hospitals and health centers.
Among HAIs, Clostridioides difficile infection (CDI) is responsible for approximately 500,000 infections per year, of which 66% are healthcare-associated, and undiagnosed patients who are colonized with C. difficile are increasingly recognized as a potential source of infection in the hospital setting. The overall objective of this proposal is to prospectively deploy our validated machine learning risk model for hospital-acquired CDI to gain an increased understanding of 1) how the risk of CDI relates to colonization status and 2) its utility in guiding management and improving outcomes. This can lead to a paradigm shift in how CDI is prevented in hospitals and serve as a model to combat other infections.
