Discussion

The implementation of a multi-step testing algorithm for CDI had a considerable impact on HO-CDI rates and patient management in our healthcare system. We observed a significant reduction in HO-CDI cases per 10,000 patient days from 5.92 to 2.36 (P < 0.001). This finding is consistent with previous studies that have shown that multi-step algorithms can lower CDI rates by improving diagnostic specificity and reducing the likelihood of detecting asymptomatic colonization rather than true infection. ^{Reference Goodenough, Sefton and Overton10–Reference Turner, Krishnan and Nelson12} The reduction in HO-CDI rates can be attributed to the increased specificity of the multi-step algorithm, which incorporates both PCR and EIA toxin A/B. This approach mitigates the risk of overdiagnosis associated with standalone molecular tests that may detect colonization. The NHSN definition of HO-CDI, which relies on the final test result, underscores the importance of using a diagnostic strategy that balances sensitivity and specificity. ³ In the post-intervention period, 54% of PCR-positive patients that previously would have met the NHSN definition of HO-CDI were negative for EIA toxin, thus reducing the number of reportable cases and potential financial penalties levied against our healthcare system. Our findings support current guidelines recommending multi-step algorithms to improve CDI diagnosis and reporting accuracy. ^{Reference McDonald, Gerding and Johnson9}

Despite the significant reduction in reportable HO-CDI cases, our study found no significant differences in clinical outcomes between the pre- and post-intervention periods (Table 2). Key metrics such as hospital LOS, ICU admissions, maximum WBC count, abdominal pain, bowel movement frequency, repeat testing within 90 days, and subsequent treatment for CDI in the next 90 days remained comparable. There was a significantly lower all-cause 30-day readmission rate in the post-intervention group, though readmission related to CDI were similar. This suggests that the multi-step algorithm did not adversely affect patient care quality or outcomes, aligning with previous research indicating that more specific diagnostic criteria do not compromise clinical management. ^{Reference Turner, Krishnan and Nelson12,Reference Hogan, Hitchcock and Frost13} Interestingly, we found that clinicians continued to treat PCR-positive patients at the same rate between the pre- and post-intervention time periods (87% vs 83%, P = 0.205), which may be an important factor why clinical outcomes remained similar. This may have been due to a combination of unfamiliarity with a new test and the clinician fear of withholding an antibiotic, as well as the multi-step algorithm recommending treatment of EIA toxin-negative patients where strong clinical suspicion for CDI infection persisted. However, significantly less EIA toxin-negative patients in the post-intervention arm were treated compared to EIA toxin-positive patients (73.5% vs 100%, P < 0.001). Thus, though the treatment rate of PCR-positive patients stayed the same between the pre- and post-intervention time periods, utilizing the multi-step algorithm to identify toxin-negative patients in the post-intervention period helped providers identify which patients they could safely defer treatment.

There was an increase in ID and GI consultations in the post-intervention period, suggesting that the multi-step algorithm prompted more thorough clinical evaluations to assist with interpretation of indeterminate results. In the post-intervention period, both ID and GI consultants were more likely to document positive PCR results as indicative of colonization rather than infection (26 of the 27 cases documented as colonization had a negative (EIA) toxin). Despite consultants documenting more PCR-positive cases as colonization in the post-intervention period, the proportion of those cases considered colonized that received (CDI) treatment remained the same between the 2 groups (41.7% vs 40.7%, P = 1)

Our analysis of patients with positive PCR results in the post-intervention period revealed significant differences in management based on EIA toxin results (Table 3). Patients with positive EIA toxin tests were more likely to receive CDI treatment, reflecting the higher specificity and clinical application of toxin detection. ID consultants were more likely to label toxin-negative patients as colonized compared to toxin-positive patients (P < 0.001) and did so at a higher rate than GI consultants (P = 0.025). This stratification underscores the utility of the multi-step algorithm in refining CDI diagnosis and guiding appropriate treatment decisions. The subgroup analysis also revealed that repeat testing within 90 days was performed similarly regardless of initial toxin status, with some toxin-negative patients converting to toxin-positive on subsequent testing, and vice versa. This finding highlights the potential for intermittent toxin shedding and the need for ongoing clinical vigilance in managing patients with suspected CDI. Our study also found that the treatment of toxin-negative patients, compared to toxin-negative patients that did not receive CDI treatment, did not reduce the need for subsequent treatment in the next 90 days (8.3% vs 11.5%, P = 0.696), nor reduce 30-day readmission rates (18.1% vs 7.7%, P = 0.341). This suggests that withholding antibiotics in toxin-negative patients deemed colonized may not adversely affect outcomes. However, there may still be some PCR-positive/toxin-negative patients that may benefit from treatment. ^{Reference Prosty, Hanula, Katergi, Longtin, McDonald and Lee14} As such, it is always important to use any diagnostic strategy as just one part of a comprehensive approach to clinical decision-making.

Our study had some important limitations. The retrospective design may have introduced biases related to data collection and interpretation. The study was conducted over a relatively short period and involved a specific patient population from two academic hospitals in one geographic region. A larger sample size and longer study duration could provide more comprehensive insights into the long-term effects of the multi-step algorithm, as clinicians may be slow to adopt withholding treatment of select PCR-positive patients. The decision to treat patients was left to the discretion of the treating medical providers, which could introduce variability in treatment approaches. This variability might affect the consistency of clinical outcomes and the interpretation of the results. The study also highlighted the potential for intermittent toxin shedding, as some toxin-negative patients converted to toxin-positive on subsequent testing, and vice-versa. This complexity in toxin detection underscores the need for ongoing clinical vigilance but also introduces a layer of complexity in interpreting the results. Our study also did not assess for any mortality differences. Finally, our findings may not be relatable to centers that have implemented a multi-step algorithm utilizing different tests or order of tests, particularly those that do not include NAAT testing as the initial step. Addressing these limitations in future studies could enhance the understanding of the multi-step testing algorithm’s impact on CDI diagnosis and management.

The shift to a multi-step testing algorithm for CDI in our healthcare system led to a significant reduction in reportable HO-CDI cases without compromising clinical outcomes. Our findings reinforce the value of multi-step testing algorithms in managing CDI and highlight the importance of ongoing education and collaboration among healthcare providers to optimize patient care.