Introduction
Among healthcare-associated infections, Clostridioides difficile infections (CDI) are a leading cause of morbidity and mortality. Reference Guh, Mu and Winston1 While prior studies have linked risk factors such as older age, use of antibiotics and proton-pump inhibitors, and recent hospitalization with CDI risk, less is known about the association between race and ethnicity and CDI risk or outcomes. Reference Loo, Bourgault and Poirier2 Prior studies suggest that CDI may differentially affect individuals based on race and ethnicity. Reference Yang, Rider, Baehr, Ducoffe and Hu3,Reference Lee, Zhou, Ortiz-Gratacos, Al Isso, Tan and Abdul-Mutakabbir4 In this study, we examined the association between healthcare outcomes and race and ethnicity in CDI patients to characterize these disparities.
Methods
A retrospective, descriptive study was conducted on C. difficile tests performed at a major academic hospital from July 20, 2021 to December 31, 2023. Testing was performed using a two-step reverse algorithm: C. difficile polymerase chain reaction (PCR, Biofire FilmArray Gastrointestinal Panel or Diasorin Molecular Simplexa C. difficile Direct rtPCR assay) was performed on all specimens and if positive a toxin enzyme immunoassay (EIA, Techlab Quik Chek Complete EIA) test was performed. All positive C. difficile PCR tests were initially included. Patients with a recurrent test (within 8 weeks of another positive test), patients under 18 years old, patients with unknown race or ethnicity and patients with results performed after discharge were subsequently excluded (see supplementary Figure 1).
Data was managed and statistical analysis were performed in SAS version 9.4 software (SAS Institute, Cary, NC). Patients who self-reported non-Hispanic and White were classified as non-racially and ethnically minoritized (REM) and everyone else was classified as REM. This method was based on a similar data analysis from Lee et al. Reference Lee, Zhou, Ortiz-Gratacos, Al Isso, Tan and Abdul-Mutakabbir4
The tests were separated into concordant (PCR+/EIA+) and discordant (PCR+/EIA−) positive results and analyses were performed separately. For comparisons between REM in these groups, χ2 tests were used, unless numbers were low enough to require Mantel Hansel. Shapiro-Wilk was used for comparison of medians.
To evaluate testing rates, 9,682 C. difficile tests collected between January 1, 2022 and December 31, 2023 were included. The shortened time frame accounted for an early institutional recommendation encouraging duplicate ordering. Patients with more than one test in one day or whose tests were collected in an outpatient setting were excluded from the testing rate results. Since the denominator for the rate was patient-days (as determined by National Healthcare Safety Network methodology), only patients with a patient day at the time of testing were included.
The study was approved by the Institutional Review Board of the University of Michigan. Primary clinical outcomes include median length of stay (LOS), need for ICU admission, all-cause mortality during admission, C. difficile complications, and need for surgical intervention.
Results
Between July 2021 to December 2023, a total of 16,620 C. difficile tests (10,259 hospitalized patients) were identified, 2,366 of which were C. difficile PCR positive (PCR+), of which 1,765 PCR + tests remain after exclusions. Subsequent EIA testing of the positive PCR specimens yielded 457 (25.9%) that were concordant positive (PCR+/EIA+), while 1,308 (74.1%) were discordant positive (PCR+/EIA−).
Testing and infection rates were examined in a subset of 9,682 C. difficile tests from January 2022 to December 2023. Testing rates were 16.8 per 1,000 patient days in non-REM versus 14.3 tests per 1,000 patient days in REM. Furthermore, tests had 3.7% positivity in non-REM and 3.8% positivity in REM (see Table 2).
Overall, most tests were obtained from non-REM, which aligns with hospital baseline demographics (see Tables 1 and 2).
Demographics
Table 1. Long description
The table presents demographic data for a population sample of 85,111 individuals. It includes categories such as race, age, sex, and Medicaid enrollment. The table has 15 rows and 2 columns. The first column lists demographic categories: Non-Hispanic White, Hispanic, Black/African American, Asian, Pacific Islander/Native Hawaiian, American Indian/Alaska Native, Middle Eastern/North African, More than one race, Other, Unknown, Age, median (year), Sex, male, and Medicaid. The second column provides the corresponding numbers and percentages for each category. For example, Non-Hispanic White individuals make up 76.5 percent of the sample, while Hispanic individuals account for 3.7 percent. The median age is 58 years, 42.2 percent of the sample is male, and 2.2 percent are enrolled in Medicaid. The table highlights the racial and ethnic composition, median age, gender distribution, and Medicaid enrollment within the population sample.
Outcome and demographic differences between concordant and discordant positive Clostridioides difficile (C. diff) results
Table 2. Long description
The table presents a comparison of outcomes and demographic differences between concordant positive and discordant positive Clostridioides difficile (C. diff) results. It includes data on demographics, length of stay, clinical outcomes, and other relevant factors. The table has 45 rows and 18 columns, with headers such as Total, Hispanic, Black/African American, Age, Sex, Medicaid, Hospital-onset CDI, Length of stay, LOS, ICU, Other clinical outcomes, CDI-associated surgical intervention, Specialist consultation, Comorbid conditions, and CDI risk factors. Notable trends include differences in testing rates and positivity between non-REM and REM groups, with testing rates of 16.8 per 1,000 patient days in non-REM versus 14.3 tests per 1,000 patient days in REM, and positivity rates of 3.7% in non-REM and 3.8% in REM.
PCR, polymerase chain reaction; EIA, enzyme immunoassay; REM, racially and ethnically minoritized; CDI, Clostridioides difficile infection; LOS, length of stay; ICU, intensive care unit; IV, intravenous.
Median age was significantly younger in REM versus non-REM, and REM patients had higher proportion of select comorbidities (see Table 2). REM patients with CDI were more likely to be insured through Medicaid compared to non-REM, although this was only significant in patients with concordant positive tests. No significant difference in sex or hospital-onset of CDI were observed between the groups. See Table 2 for additional details.
Median length of stay (LOS) was slightly but significantly shorter for REM versus non-REM in both concordant positive tests and discordant positive tests There was no difference in prolonged length of stay (>10 days) between REM vs non-REM. There was no significant difference in all-cause ICU admission; however, median ICU LOS was significant longer in REM versus non-REM in both concordant and discordant positive tests. Other outcomes, including all-cause mortality, incidence of ileus or megacolon, incidence of CDI-associated surgical intervention, CDI recurrence, choice of CDI treatment, need for infectious disease or gastroenterology consultation, were not significantly different.
In adjusted analyses clustered by patient MRN, REM status was not associated with ICU stay in either the PCR+/EIA + group (adjusted RR 0.95, 95% CI 0.69–1.32; P = .77) or the PCR+/EIA− group (adjusted RR 0.97, 95% CI 0.75–1.25; P = .82). Similarly, REM status was not associated with LOS>10 days in the PCR+/EIA + group (adjusted OR 1.11, 95% CI 0.63–1.96; P = .72) or the PCR+/EIA− group (adjusted OR 0.86, 95% CI 0.63–1.17; P = .33).
Discussion
Overall, we found few differences in CDI outcomes between REM and non-REM in the setting of high C. difficile testing rates. While other studies have demonstrated REM patients were more likely to have a prolonged hospitalization or require an ICU stay, Reference Yang, Rider, Baehr, Ducoffe and Hu3 our findings revealed a shorter median LOS in REM compared with non-REM and no significant difference in all-cause ICU admission. When examining outcomes, such as choice of CDI treatment or need for subspecialty consultation or surgical intervention, there were no differences in clinical care. Additionally, our C. difficile testing rates were higher than those previously reported at other hospitals, Reference Krouss, Israilov and Alaiev5 and on par with hospital systems who reported concerns with significant over-testing of their complex patient populations. Reference Kamboj, Brite and Aslam6 A plausible explanation for our lack of outcome disparities may be related to high baseline testing rates, as infections may be diagnosed and managed with less discretion in all individuals. Efforts to improve antibiotic stewardship for discordantly positive tests will need to be partnered with review of demographic data to ensure disparities are not being unmasked.
Our findings demonstrated that REM patients were significantly younger, more likely to have Medicaid, and were more likely to have certain comorbidities. Older age has previously been demonstrated to be a risk factor for CDI, Reference Loo, Bourgault and Poirier2 but contrary to this, we found that REM patients had a younger age of diagnosis. Consistent with prior studies, REM patients were more likely to have Medicaid and were more likely to have diabetes with complications and CKD. Prior studies have shown that CKD may partially explain the relationship between race and ethnicity and CDI. Reference Lee, Zhou, Ortiz-Gratacos, Al Isso, Tan and Abdul-Mutakabbir4
One limitation was that our study evaluated a population from one Midwestern academic hospital, which is notably located in a suburban setting compared to major urban centers in other studies. Reference Krouss, Israilov and Alaiev5,Reference Kamboj, Brite and Aslam6 It is difficult to assess the socioeconomic status of REM patients admitted to our center, but the proportion of patients with Medicaid insurance is notably different from other studies. Reference Lee, Zhou, Ortiz-Gratacos, Al Isso, Tan and Abdul-Mutakabbir4 This suggests that stratifying by race and ethnicity may not be the optimal way to investigate disparities at our institution and makes a stronger argument for use of alternatives like Social Vulnerability Index for future studies. Additionally, direct chart review was not performed. Future studies would benefit from evaluating multicenter data from multiple geographic areas and corroboration with direct chart review.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/ice.2026.10492.
Acknowledgments
None.
Financial support
None.
Competing interests
Daniel Li, MD has no financial disclosures or conflicts of interest; Katie Moore, MPH has no financial disclosures or conflicts of interest; Noreen Mollon, MS has no financial disclosures or conflicts of interest; Amanda Valyko, MPH has no financial disclosures or conflicts of interest; Laraine Washer, MD has no financial disclosures or conflicts of interest; Anastasia Wasylyshyn, MD has no financial disclosures or conflicts of interest.
Open access
