Study design and practice environment

This was a retrospective, single community health system, pragmatic, propensity score pair matched cohort study in hospitalized patients ≥ 18 years old treated for a first occurrence of CDI between June 18, 2018 and December 21, 2021. The study site, Norton Healthcare, is located in Louisville, Kentucky, U.S.A. and has four adult hospitals totaling >1500 inpatient beds. The study received exempt status from the University of Louisville Institutional Review Board.

During the study period, there were four infectious diseases pharmacists and one nurse practitioner who dedicated time to a clinical transition of care program called “C. diffiCARE”. Patients were candidates for tapered vancomycin if they were high risk for recurrence defined as ≥ 65 years, immunocompromised, or had severe/fulminant infection. Patients or providers could decline the recommendation to utilize a taper. The C. diffiCARE team assisted in obtaining discharge vancomycin prescriptions and provided outpatient follow-up. Any patient with CDI could utilize the C. diffiCARE program for medication assistance and follow-up, regardless of the vancomycin regimen utilized.

The standard vancomycin cohort included patients prescribed enteral vancomycin regimens for 10–14 days, and the intervention arm included patients prescribed a standardized 12-week taper of enteral vancomycin (Figure 1). Patients could have received partial treatment for CDI with non-vancomycin therapy (e.g., fidaxomicin) prior to switching to enteral vancomycin. Patients were only included once in the study period. CDI diagnosis was defined as a positive C. difficile polymerase chain reaction test (Cepheid Xpert® C. difficile/Epi Assay) with or without a positive toxin enzyme immunoassay (Techlab Quik Chek®) and prescribed enteral vancomycin. Exclusion criteria were previous CDI within the past year, death or transition to hospice during hospitalization, previous colectomy, colectomy during or within 30 days of completing enteral vancomycin, diagnosed prior to admission, and clinical failure defined as a change in CDI therapy prior to the planned stop date of the enteral vancomycin course.

Figure 1.

Tapered Vancomycin Regimen.

All eligible patients who received tapered enteral vancomycin were included in the experimental arm. Standard of care controls were identified using exact matching. Controls were identified based on a random number generator and selected if they had an exact propensity match in the experimental arm based on age (< 65 or ≥ 65 years) and receipt of any non-CDI antibiotics during hospitalization or the 6-month post discharge follow up period; patients were not required to receive the same risk category of antibiotics in order to be matched. A matched patient could only be included once. Matched patients were screened for eligibility criteria. If an exclusion criteria was found, another round of matching occurred. 1:1 matching was used based on the anticipated sample size and rigorous criteria of exact matching.

Patient demographics extracted from the electronic medical record included age, sex, race, C. difficile associated risk of death score (CARDS) ^{Reference Kassam, Cribb Fabersunne and Smith22} , dates of hospital admission and discharge, first positive CDI test, start and stop dates of enteral vancomycin including discharge prescriptions, and dates of recurrence and death within 6 months of hospital discharge when relevant. Additional factors collected associated with high risk of recurrence were immunocompromised status and severity of illness per IDSA guidelines. ^{Reference Johnson, Lavergne and Skinner3} Immunocompromised status was defined as receiving chemotherapy, hematopoietic stem cell/solid organ transplant, HIV with CD4 < 200 cells/mm³, B-cell/T-cell depleting therapies, interleukin inhibitor therapy, absolute neutrophil count < 500 cells/mm³ or high-dose corticosteroids (≥20 mg/day of prednisolone equivalents for ≥ 2 weeks). Potential cofounders were collected during hospitalization and within 6 months of discharge including receipt of bezlotoxumab, FMT, rifaximin, metronidazole, fidaxomicin, cholestyramine or colestipol, PPIs, and non-CDI antibiotics. Non-CDI antibiotics were categorized as high, medium or low risk for CDI. ^{Reference Brown, Khanafer, Daneman and Fisman23–Reference Webb, Subramanian and Lopansri26}

Outcomes

The primary outcome was CDI recurrence within 6 months post hospital discharge. Recurrences were defined as meeting both of the following criteria: 1. new diarrhea or positive C. difficile microbiologic test; 2. prescription for fidaxomicin, enteral vancomycin every 6 hours or 4 times daily or metronidazole monotherapy. Secondary outcomes were time to CDI recurrence, all-cause mortality within 6 months of discharge, and time to mortality.

Sample size and analysis

Prior to the study, it was estimated that 100 patients received tapered enteral vancomycin. With alpha equal to 0.05 and assuming a 25% CDI recurrence rate in the standard vancomycin arm and a 15% reduction in CDI recurrences in the taper arm, it was estimated there would be 76% power to detect a difference in the primary outcome. The observed rate and number of days to the composite outcomes were summarized between cohorts. If a patient experienced recurrence before mortality, days to recurrence was reported. Multivariate analysis included an inverse probability weight Cox regression for time to recurrence or death stratified by matched pairs. Additionally, a sensitivity analysis was performed on a composite outcome of CDI recurrence or mortality to represent potential cases of recurrence among individuals who died post-hospitalization.

Inverse propensity score weights (IPSWs) were assigned to each observation to further balance covariates between study arms after subject matching. To produce IPSWs, an initial propensity score was calculated from a logistic regression model to estimate the probability of receiving a vancomycin taper given a set of covariates. Covariates in the logistic regression model included sex, race, immunocompromised status, admission to the intensive care unit, acute renal failure, liver disease, cardiopulmonary disease, malignancy, inflammatory bowel disease, diabetes mellitus, and the following confounding treatments during the follow-up period: bezlotoxumab, fidaxomicin, FMT, rifaximin, metronidazole, cholestyramine or colestipol, non-CDI antibiotics, and PPIs. Individual IPSWs were calculated as the inverse probability of being assigned to the vancomycin taper. For subjects who were in the taper arm, IPSWs were calculated as 1/propensity score. For subjects in the control, IPSWs were calculated as 1/(1-propensity score). ^{Reference Chesnaye, Stel and Tripepi27}

Comparisons of baseline characteristics included Mann–Whitney U and χ² tests as appropriate. Balance between study arms was assessed with standardized mean differences (SMDs) for baseline patient information. A cutoff of 0.2 was used to identify characteristics that remained imbalanced after matching. ^{Reference Rubin28} Multivariate analysis included conditional logistic regression models to account for subject matching and IPSWs, and reported as conditional odds ratios (cORs) with 95% confidence interval (CIs). P-values of less than 0.05 were statistically significant. All statistical analyses were performed using R version 4.3.1. ²⁹