Background: Clostridioides difficile infection (CDI) is a common, often nosocomial infection associated with substantial morbidity and mortality. Antibiotics are the most important modifiable risk factor, but empiric antibiotics remain appropriate for many patients with severe acute illness. Which antibiotics minimize the risk of CDI remains an important unanswered question. Because protective equipment and isolation were nearly universal for COVID-19, patients admitted for COVID-19 would be an ideal cohort in which to study the risk of CDI with specific antibiotics.
Methods: Using data from HCA Healthcare, a geographically diverse health system accounting for approximately 5% of all US healthcare encounters, we assembled a retrospective cohort of adults aged ≥18 years admitted for COVID-19 between March 2020 and February 2021, excluding hospital stays of less than 24 hours and admissions including CDI treatment in the first 24 hours of hospitalization.We theorized that patients would be more likely to be prescribed antibiotics if they were older, less clinically stable, or required more respiratory support. For each antibiotic studied, we created a multilevel logistic regression model predicting the probability of receipt, based on predictor variables from the first 24 hours of hospitalization: age, white blood cell (WBC) count, quick Sepsis-related Organ Failure Assessment (qSOFA), WHO clinical progression scale, and flow rate of supplemental oxygen. Each model was clustered by hospital.We then used coarsened exact matching to control for the probability of antibiotic receipt and estimated the causal effect of one daily dose equivalent of each antibiotic on risk of CDI (based on ICD-10 code) both during the index hospitalization (using a generalized linear model with a log-link function) and following discharge (using a Cox regression of time-to-event data among patients discharged alive).
Results: Of 93,715 index hospitalizations, 274 (0.29%) included a diagnosis of CDI over a total of 877,726 patient-days (incidence rate 3.1/10,000 patient-days). Of 78,636 patients discharged alive, 131 (0.17%) developed CDI between discharge and censoring over a median of 89 days of observation (IQR: 37-198 days). The most prescribed antibiotics were azithromycin (50.0%) and ceftriaxone (47.7%). The incidence rate ratios for one daily dose-equivalent of each antibiotic on rates of CDI in the index hospitalization appear in Table 1. Hazard ratios for one daily dose-equivalent of each antibiotic on CDI after discharge appear in Table 2.
Conclusions: Each dose of an antibiotic covering anaerobic organisms is associated with increased risk of inpatient and post-discharge CDI; cephalosporins are associated with lesser risk. When empiric antibiotics are necessary and risk of anaerobic infection is low, a cephalosporin may be preferable. The duration of antibiotics should be minimized, as every dose is associated with increased risk.
