Challenges in exposure characterization for healthcare-associated infections: applications in the clinical epidemiology of Clostridioides difficile infection

Jessica Lynn Webster

doi:10.17918/00010808

Healthcare-associated infections (HAI) place a substantial burden on hospitalized patients and healthcare systems due to the range of complications associated with infection. Prevention of HAI include efforts to reduce a patient's likelihood of infection (environmental transmission) as well as progressing to symptomatic disease once infected (patient susceptibility). There is a need for an improved understanding of both patient-specific and environmental risk factors for HAI, to inform prevention measures and improve patient outcomes. In this dissertation, we identified three challenges associated with characterizing exposure to HAI, focusing on healthcare-associated Clostridioides difficile infection: 1) in measuring and operationalizing antibiotic exposure and identifying relevant aspects of exposure that may predict disease, 2) in evaluating antibiotic use as a time-varying exposure within a causal model for C. difficile infection, and 3) in teasing apart the impact of patient-specific and environmental risk factors for C. difficile infection. We use data from two hospital-based case-control studies to demonstrate epidemiologic methods that can be used to answer these questions, or similar questions in the context of other HAI. In Aim 1 we use factor analysis of mixed data (FAMD), logistic regression, and Least Absolute Shrinkage and Selection Operator (LASSO) regression to explore the distribution of antibiotic exposure variables within our dataset and identify the characteristics most predictive of C. difficile infection. In Aim 2 we use inverse probability weighting (IPW) and marginal structural models (MSM) to evaluate the causal effect of time-varying antibiotic exposure on C. difficile infection accounting for exposure and covariate history. And in Aim 3 we used hierarchical and cross-classified multilevel models to quantify the impact of patient-level and environmental risk factors for C. difficile infection accounting for patient movements throughout the hospital. When evaluating characteristics of antibiotic exposure, we found that there are benefits to exploring different aspects of exposure, such as medication or class of antibiotic, number of unique courses, or the amount of time on antibiotics. However, we found that a more simplistic measure representing any antibiotic exposure was among the most relevant characteristics for describing the variation in our dataset as well as for predicting C. difficile infection. When considering the causal effect of time-varying antibiotic exposure and C. difficile infection, we found that by using IPW and MSM, we were able to account for history of exposure and covariate values and produce more accurate, unbiased estimates of the causal effect. And finally, we found using multilevel models that while antibiotic use is the most important risk factor in patients that developed healthcare-associated C. difficile infection, environmental risk factors are additionally important and should be considered in research involving hospitalized patients and HAI. This research explored challenges in evaluating complex exposures in clinical research, focusing on applications in healthcare-associated C. difficile infection. These results will move the field forward by providing methodological examples for researchers conducting research in similar environments, as well as provide evidence for infection prevention activities such as antibiotic stewardship and environmental cleaning.

Challenges in exposure characterization for healthcare-associated infections: applications in the clinical epidemiology of Clostridioides difficile infection

Files and links (1)

Abstract

Metrics

Details

Challenges in exposure characterization for healthcare-associated infections: applications in the clinical epidemiology of Clostridioides difficile infection

Files and links (1)

Abstract

Metrics

Details

Drexel University Social media