Introduction
Clostridioides difficile remains a leading cause of healthcare-associated diarrhea. Reference Echenique, Penugonda, Stosor, Ison and Angarone1,Reference Mawer, Byrne and Drake2 Diagnostic algorithms have evolved with increasing adoption of multi-step testing. Multi-step testing generally couples a high-sensitivity test (e.g., nucleic acid amplification [NAAT] or glutamate dehydrogenase [GHD]) with reflex to a high-specificity test (toxin enzyme immunoassay Reference Blanco-Míguez, Beghini and Cumbo3 ) if the first test is positive. As testing algorithms influence estimation of C. difficile incidence rates, accurate accounting for test method is critical for tracking C. difficile epidemiologic trends. In our prior work, the transition from NAAT alone to NAAT followed by toxin EIA was associated with a nearly 50% reduction in reported hospital-onset C. difficile infection (HO-CDI). Reference Turner, Krishnan and Nelson4
The National Health Safety Network (NHSN) recently updated guidance on how healthcare facilities should report C. difficile test method. In previous guidance, test method was determined by the individual test method that was used most often. For example, if a facility used NAAT with reflex to EIA but > 50% of tests were negative (ending at NAAT only), NAAT would have been reported as the primary test method. In the January 2024 update, NHSN now recommends reporting the “testing method standardly performed.” Using the same example, facilities would no longer count individual test components—but would instead report “NAAT with reflex to toxin EIA.” 5 To assess the effect of the NHSN test method definition update, we compared models of HO-CDI incidence among six hospitals using the prior and current definitions.
Methods
Data collection
We previously conducted a multicenter longitudinal study of HO-CDI within the Duke Infection Control Outreach Network (DICON) from 2017 through 2022. Eight participating hospitals updated their C. difficile testing protocol during the study period. C. difficile infections were defined in accordance with the NHSN LabID definition: cases were identified according to the facility’s testing algorithm. For multi-step testing, the result of the last test performed was used: NAAT+ followed by toxin EIA+ counts as a case but NAAT+ followed by toxin EIA- does not. 5 HO-CDI cases were defined as those occurring after hospital day three. HO-CDI incidence rates were calculated quarterly, as number of HO-CDI cases per 10,000 patient-days present. Any repeat positive samples sent within 14 days of a prior positive were considered duplicates and not counted separately. Surveys were sent to participating hospitals to ascertain quarterly test method based on prior and current NHSN definitions.
Study design was reviewed by Duke University Health System’s institutional review board and determined to be exempt, with waiver of consent for data collection.
Statistical modeling
We plotted actual versus predicted HO-CDI cases based on the prior and updated NHSN definitions and using the formula provided by NHSN (Supplement). To test for statistical difference between prior and updated NHSN definitions, we constructed negative binomial models for HO-CDI using the same covariates as in the NHSN C. difficile risk adjustment calculation: inpatient community-onset C. difficile prevalence rate, medical school affiliation, number of ICU beds, facility type, facility bed size, and reporting from emergency department or 24-hour observation unit. To avoid confounding by decreasing HO-CDI incidence rates since the NHSN model was last re-calibrated, we developed locally optimized models using the same NHSN covariates but allowing each hospital to begin from its own baseline. Reference Turner, Grambow and Woods6 Two separate models were created: model 1 utilized the prior NHSN definition of test method (based on individual test method most frequently used per quarter); model 2 utilized the January 2024 NHSN update (in which test method was reported independent of how frequently individual test components were conducted). Models were constructed using glm.nb from the MASS package in R (version 4.4.1, https://cran.r.project.org). Inspection of quantile-quantile (QQ) plots suggested the negative binomial distribution to be a reasonable assumption (consistent with NHSN risk modeling). Throughout the study, a significance threshold of 0.05 was used; 95% confidence intervals were calculated by Wald method. See Supplement for additional details on modeling methods.
Results
Of the eight DICON hospitals changing test method from January 2017 to December 2022, 6 responded to our survey request. Five of the six converted from NAAT alone to NAAT with reflex to toxin EIA; one site transitioned from EIA alone to NAAT alone; all submitted a minimum of 5 years of data including at least 2 years before and after any change in C. difficile testing protocol. The final cohort included 1,814 hospital-onset C. difficile cases occurring among 4,690,841 patient-days from these six hospitals.
Figure 1A plots hospital-onset C. difficile incidence over time adjusting for test method as defined in prior NHSN guidance (model 1). None of the facilities met prior criteria for reporting test type as NAAT with reflex to EIA for any quarter. Figure 1B plots hospital-onset C. difficile incidence rates adjusting for test method as defined in the January 2024 NHSN update (model 2). With the updated definition, 5 of the 6 facilities now met criteria for a change to NAAT with reflex to EIA (indicated in blue).
Hospital Onset CDI Incidence Rate by Testing Strategy. Each panel represents one hospital. Each point plots the actual HO-CDI incidence per quarter. Lines represent predicted HO-CDI incidence rates, where Column A uses the prior NHSN definition (based on the most commonly performed test) and Column B uses the 2024 NHSN definition (based on the facility’s test strategy). Both points and lines are color coded to represent the NHSN-defined test method used by each hospital at each given point in time.
Table 1 compares select model parameters. Model 2, based on the NHSN’s updated test method definition, outperformed model 1 by both Akaike information criterion (AIC, an estimator of prediction error) and likelihood ratio test (P < .001). Model 2 showed closer concordance between predicted versus reported HO-CDI incidence (Figure S1), was more likely to capture when sites actually transitioned to two-step testing (Figure 1A/B), and more likely to detect the expected effect of changing to a more specific test. Test type did not meaningfully contribute to Model 1 – in part due to limited contrast among recorded test types. In model 2, NAAT with reflex to EIA was associated with roughly half (incidence rate ratio: 0.52) the reported HO-CDI incidence relative to NAAT testing alone.
Comparison of hospital-onset C. difficile models using prior and January 2024 National Health Safety Network test method definitions
1 Akaike information criterion—a model selection parameter that balances goodness of fit against a penalty for over-fitting. Lower values indicate a more favorable model fit.
The standard NHSN prediction model tended to predict more HO-CDI cases than were observed in our region, consistent with trends toward declining rates since 2017. The prior NHSN test definition consistently predicted more HO-CDI cases than either the updated definition or the actual reported number of HO-CDI cases (Figure S1). Reference Turner, Grambow and Woods6
Discussion
Accurately capturing a facility’s testing algorithm is critical for tracking C. difficile trends and properly adjusting for interfacility comparisons. The NHSN recently updated recommendations for reporting C. difficile test method, emphasizing standard test strategy over prior definitions based on the individual test component conducted most often. Using longitudinal data from a cohort of six hospitals that recently changed their testing protocol, we compared models of HO-CDI incidence using the prior and updated NHSN definitions. Within this cohort, the updated NHSN definition was less likely to over-estimate HO-CDI rates and resulted in superior model performance. Whereas prior definitions meant that facilities rarely met criteria for reporting test method as NAAT with reflex to EIA, the updated definition correctly captured the adoption of a multi-step testing algorithm by five sites. Not only did model performance improve with the new definition, but test type was significantly associated with HO-CDI incidence.
Our study carries several limitations. It is relatively small – including just 6 hospitals –limiting generalizability. C. difficile testing practices were also fairly uniform: five of the six transitioned from NAAT to NAAT with reflex to toxin EIA, limiting our ability to assess other testing strategies. Additionally, the study was conducted within a broader quasi-experimental study aimed at reducing hospital-onset C. difficile rates – thus there is risk of potential confounding from other interventions. Reference Turner, Krishnan and Nelson7
In summary, our results validate the NHSN’s updated recommendations related to C. difficile test method, suggesting the updated approach is likely to improve risk adjustment and modeling of HO-CDI incidence. Given the significant anticipated improvement in model performance, re-baselining following the adoption of the new definition may help to assure facility-wide HO-CDI incidence rates are properly adjusted for current laboratory testing practices.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/ice.2025.10222.
Data availability statement
All authors had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. The content of this manuscript is original, and has not been previously published.
Financial support
Research reported in this publication was supported by the Duke Clinical and Translational Science Institute (CTSI). The content is solely the responsibility of the authors and does not necessarily reflect the official view of Duke University or the Duke CTSI. Although not directly related to this grant, Nicholas A. Turner was also supported in part by an Antibacterial Resistance Leadership Group Fellowship [National Institute of Allergy and Infectious Diseases UM1Al104681].
Competing interests
NAT reports grant funding from CDC, NIH; research contract funding from PDI, Purio, and Basilea; consulting for Techspert. DJA reports grant funding from CDC, AHRQ and royalties from UpToDate; DJA is an owner of Infection Control Education for Major Sports, LLC. All other authors report no potential conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.
