Study design

This was a case-control study conducted at Henry Ford Health (HFH), a five-hospital health-system located in southeast Michigan. Cases included patients who developed CDI progression, controls were patients who did not develop CDI progression. This study was approved by the local institutional review board (IRB #15244) with a waiver of consent.

Patients were included if they were hospitalized at HFH from 1/1/2018 to 2/22/2022, were ≥65 years, had an initial microbiologically confirmed CDI episode, were classified as having non-severe CDI at symptom onset based on national guideline criteria, ^{Reference McDonald, Gerding and Johnson15} and received therapy with vancomycin (oral/rectal) or fidaxomicin for ≥72 hours. Persons ≥65 years of age were selected due to evidence suggesting a higher risk of poor outcomes (i.e., disease progression) when compared to those <65 years of age. ^{Reference Abou Chakra, Pepin, Sirard and Valiquette5} Patients were excluded if they met any criteria for severe or fulminant CDI at disease onset, ^{Reference McDonald, Gerding and Johnson15} had missing or inadequate data related to CDI diagnosis (i.e., laboratory values), had an initial CDI diagnosis from an outside hospital, or received CDI-directed treatment (i.e., oral/rectal vancomycin, fidaxomicin, metronidazole, eravacycline) within 1 week prior to CDI diagnosis. Only patients experiencing their first CDI episode were included.

Data source

Patient data were manually reviewed and extracted from the electronic health record (EHR) using Microsoft SQL Server Management Studio (Microsoft Corp., Redmond, WA, USA). Data extraction parameters were set to identify patients with a positive C. difficile test who were older than 65 years old at time of positive test during the prespecified timeframe. The identified patients were then randomized and screened for inclusion. Data collection included patient demographics (ie, immunosuppressed status), comorbid conditions, C. difficile severity, and infection characteristics (ie, community- or hospital-onset, stool counts per day), and CDI treatment selection. Patient outcomes included hospital length of stay, odds of CDI progression, and requirement of colectomy. Data were captured by a single investigator using a standardized electronic case report form.

Key study definitions

CDI progression was defined as disease progression from non-severe to severe or fulminant disease from initial documentation of symptoms within the EHR, or time zero. C. difficile severity (ie, non-severe, severe, fulminant) was classified according to criteria from national guidelines. ^{Reference McDonald, Gerding and Johnson15}

C. difficile infection characteristics (i.e., serum creatinine [SCr], white blood cell [WBC] count, number of stools per day, fulminant disease clinical indicators) were evaluated from the time of symptom onset to patient discharge or 10 days following initiation of CDI-directed antimicrobial therapy. Patients were classified as experiencing CDI progression if they met any of the following after time zero: (1) WBC >15,000 cells/mm³, SCr >1.5 mg/dL (severe CDI), or (2) clinical presentation hypotension or shock with systolic blood pressure <90 mmHg or MAP <65 mmHg, ileus on imaging, or the presence of megacolon on imaging with colon diameter at cecum >12 cm (fulminant CDI). ^{Reference McDonald, Gerding and Johnson15} The primary exposure variable of interest was time-to-CDI treatment, defined as the time from initial CDI symptom documentation onset to CDI treatment, in hours.

C. difficile onset was categorized as community-acquired CDI (CA-CDI), or within 48 hours of admission, and hospital-acquired CDI (HA-CDI), or greater than 48 hours after admission. For CA-CDI, onset was defined as the same criteria as HA-CDI, with the addition of including the emergency department triage time if symptoms existed at presentation. For HA-CDI, onset was defined as the time the first symptom occurred (≥3 unformed stools or documentation of loose stools in the medical record) prior to C. difficile testing being ordered. Patients were considered immunosuppressed if they met any of the following criteria: history of solid organ transplant, human immunodeficiency virus with a CD4 count <200 cells/mm³, or receiving prednisone >20 mg/day for >2 weeks.

Clinical cure was defined as treatment completion with symptom resolution and without disease progression while inpatient or treatment continued outpatient for patients discharged within 10 days of treatment initiation. Other secondary endpoints were hospital length of stay (LOS) and ICU LOS from symptom onset, time from treatment initiation to discharge, and requirement of colectomy. Stool counts were obtained using nursing assessment flowsheets, intake/output logs (including rectal tube and bag), and nursing shift notes.

Microbiology and C. difficile testing

C. difficile testing is performed using a multistep algorithm at HFH. The first step involves using an enzyme immunoassay for C. difficile glutamate dehydrogenase antigen and enzyme immunoassay for C. difficile toxins A and B. Glutamate dehydrogenase positive and toxin positive results were considered diagnostic of CDI. If there is discordance between the results of glutamate dehydrogenase and enzyme immunoassay, polymerase chain reaction (PCR) amplifying the gene target for toxin B is performed to confirm a positive or negative test and interpreted in clinical context. Only symptomatic patients with a positive C. difficle test as outlined above were evaluated.

Statistical analysis

Descriptive statistics (proportion [%], median [IQR]), were used to describe the patients with CDI progression and those without CDI progression. In bivariate analyses, categorical variables were compared using the Pearson χ² or Fisher’s exact test, and continuous variables were compared using the Mann-Whitney U test. To determine variables independently associated with CDI disease progression, variables associated with case/control status (P < .25) in the bivariate analyses were entered into a multivariable logistic regression model. Using a backward, stepwise approach, variables were retained in the final model if they were significant and/or their inclusion/exclusion from the model changed the measure of association for the exposure of interest, time-to-CDI-directed antimicrobial therapy, by more than 10%. ^{Reference Hosmer, Lemeshow and Sturdivant21} Given the lack of literature specifically examining time-to-CDI-directed antibiotic therapy, the variable was evaluated in three distinct ways. First, we assessed time-to-CDI-directed therapy as a continuous variable and compared the median values between progressors and non-progressors using the Mann–Whitney U test. Second, we analyzed time-to-CDI-directed antimicrobial therapy as a categorical variable (<48 hours, 48–72 hours and >72 hours) and compared this between progressors and non-progressors using a χ² test. Finally, we used classification and regression tree (CART) analyses to identify a threshold at which time-to-CDI-directed antimicrobial therapy was significantly different between those who experienced progression versus lack of progression. This breakpoint served as a marker to determine the threshold of categorizing patients into the “early” CDI treatment or “delayed” CDI treatment groups. For the multivariable regression analyses, we evaluated all three ways of examining time-to-CDI-directed therapy and presented all three in the results. Variables included in the model were restricted to an event-to-variable ratio of 10:1; model fit was assessed using the Hosmer-Lemeshow goodness of fit test. The categorical variables were evaluated for collinearity using the χ² test prior to model inclusion; variable selection was prioritized based on the magnitude of association. All statistics were performed with SPSS Statistics for Windows v.26.0 (IBM, Armonk, NY).