Contrasting methods of measurement of antibiotic exposure in clinical research: a real-world application predicting hospital-associated Clostridioides difficile infection

The goal of this article is to summarize common methods of antibiotic measurement used in clinical research and demonstrate analytic methods for selection of exposure variables. Variable selection was demonstrated using three methods for modeling exposure, using data from a case-control study on Clostridioides difficile infection in hospitalized patients: 1) factor analysis of mixed data, 2) multiple logistic regression models, and 3) Least Absolute Shrinkage and Selection Operator (LASSO) regression. The factor analysis identified 9 variables contributing the most variation in the dataset: any antibiotic treatment; number of classes; number of treatments; dose; and classes monobactam, {beta}-lactam {beta}-lactamase inhibitors, rifamycin, carbapenem, and cephalosporin. The regression models resulting in the best model fit used predictors any antibiotic exposure and proportion of hospitalization on antibiotics. The LASSO model selected 22 variables for inclusion in the predictive model, exposure variables including: any antibiotic treatment; classes {beta}-lactam {beta}-lactamase inhibitors, carbapenem, cephalosporin, fluoroquinolone, monobactam, rifamycin, sulfonamides, and miscellaneous; and proportion of hospitalization on antibiotics. Investigators studying antibiotic exposure should consider multiple aspects of treatment informed by their research question and the theory on how antibiotics may impact the distribution of the outcome in their target population.