Predicting C. Difficile Infection (CDI)

Written By Phil Jacokes

Overview

Reducing the incidence and transmission of healthcare-associated infections is a clinical quality improvement initiative within Michigan Medicine. The focus of this project is on assessing whether prevention strategies guided by patient-level risk of C. Difficile Infection (CDI) and prevent and identify patients early, and whether early identification can lead to better clinical outcomes.

Principal Investigator(s)

Jenna Wiens

Krishna Rao

Digital Health Innovation Support

Digital Health Innovation provided EHR data to develop the model. This model runs on a Digital Innovation supported pipeline, Near-Real Time RDW. Retrospective and prospective validation of the model was supported using this pipeline. With newly developed IT infrastructures, the model generates a CDI risk score on a daily basis at MM. The generated score helps clinicians identify which patients should receive interventions, and what interventions should be applied. The Model Deployment team also engaged with the Health Information Technology and Services (HITS) to integrate survey questions in MiChart during the order entry process for a C. difficile order. This integration enabled the researchers to assess reasoning behind provider suspicion of CDI and to use this qualitative data to improve the predictive algorithm.

Partnerships

College of Engineering 

Clinical Intelligence Committee (CIC)

Health Information Technology and Services (HITS)

Office of Clinical Informatics (OCI)

Learning Health Sciences (LHS)

Publications

Oh et al 2018: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6421072/

Li et al 2019: https://pubmed.ncbi.nlm.nih.gov/31139672/

Otles et al 2021: https://static1.squarespace.com/static/59d5ac1780bd5ef9c396eda6/t/60fb3ba110343004004f24ba/1627077538209/Performance_Gap___Prospective_Validation.pdf

Phil Jacokes
Previous

Projecting IV Fluid Usage
Next

Racial Bias in Pulse Oximetry Measurement