Clostridium difficile infections are the most frequent hospital-acquired infection in the United States. 1 For several years, the largest hospital in our health system (hospital A) experienced higher-than-expected numbers of healthcare-facility–onset C. difficile laboratory identified events (HO-CDI LabID), per the National Healthcare Safety Network (NHSN), with standardized infection ratios (SIRs) consistently above 1.0. These infections occurred despite several interventions, including heightened environmental disinfection with sporicidal agents, an ultraviolet automatic disinfection system, antimicrobial stewardship, and hand hygiene campaigns. Until recently, our 3 hospitals used the nucleic acid amplification test (NAAT) as the sole clinical test for C. difficile infections; however, in February 2017, we instituted an enzyme immunofluorescence assay (EIA) after any NAAT-positive result. Given that the NHSN reporting definition for HO-CDI LabID events was based solely on NAAT-positive results, our SIRs remained unchanged after implementation of reflex EIA. On January 1, 2018, NHSN changed the reporting methodology for institutions performing a 2-step test (ie, NAAT with reflex EIA). Instead of all NAAT-positive tests, only NAAT-positive/EIA-positive tests were included in the new LabID definition. 2 We describe the immediate impact of these NHSN methodology changes in the number of HO-CDI LabID events observed, the number of predicted events, and the final impact on C. difficile SIRs across our health system during the first 2 quarters of 2018.
Methods
The Froedtert Health System comprises 3 hospitals in the greater Milwaukee area: hospital A is a 565-bed urban inpatient teaching affiliated facility, hospital B is a 206-bed hospital in a suburban area (which as of January 1, 2018, switched its reporting category from community nonteaching affiliated to community teaching-affiliated), and hospital C is a 70-bed facility in a rural area 33 miles from downtown Milwaukee. In hospital A, 110 of 565 beds (19.5%) are allocated to hematology-oncology patients and solid-organ transplant recipients. The definition of HO-CDI LabID events and the respective methodology changes were obtained directly from NHSN. The number of predicted HO-CDI LabID events, monthly patient days, and quarterly SIRs were downloaded from the NHSN database. No new C. difficile interventions were implemented in either hospital during 2018.
Clostridium difficile clinical tests
Unformed stool samples underwent NAAT C. difficile testing using hospital specific methodologies in accordance with the manufacturer’s product insert. Hospital A used Xpert C. difficile (Cepheid, Sunnyvale, CA); hospital B used Meridian Illumigene C.diff (Meridian Bioscience, Cincinnati, OH); and hospital C used Nanosphere Verigene C.diff (Luminex, Austin, TX). In February 2017, all stool specimens testing C. difficile positive by any of the NAAT tests were automatically reflexed for TcdA and TcdB EIA testing (C.diff Quik Chek Complete, Alere, Lowell, MA).
Statistics
Counts of events were analyzed using Poisson regression while offsetting for the natural-log of patient days. Due to the low number of events in hospitals B and C, these 2 hospitals counts were combined for all analyses. Changes in SIRs, which were calculated as observed events divided by predicted events, were analyzed using a binomial distribution. All analyses were performed with SAS version 9.4 software (SAS Institute, Cary, NC).
Results
From January 1, 2017, to June 30, 2018, our health system experienced 241 HO-CDI LabID events. Among these 241 events, 98 (40.7%) were NAAT-positive/EIA-positive and the remaining 143 (59.3%) were NAAT-positive/EIA-negative (Table). During the first 2 quarters of 2018, we identified 130 NAAT-positive events, of which 31 (23.8%) were EIA-positive and thus were reported as HO-CDI LabID events; the remaining 99 (76%) were EIA-negative. These changes resulted in a major decrease in HO-CDI LabID events in all 3 hospitals from 52.5 to 15.5 events per quarter before and after the NHSN methodology change, respectively. Both the NHSN methodology change (P=.0003) and the hospital (A vs. B+C; P=.0083) had an impact on the number of HO-CDI LabID events (interaction not detected). Hospital A went from 47.75 HO-CDI LabID events per quarter in 2017 to 14 events per quarter in the first half of 2018 (P<.001) (Figure 1). Hospitals B and C went from a combined mean of 2.37 to 1.5 HO-CDI LabID events per quarter (P=.049). The number of NAAT-positive/EIA-positive events remained stable before and after the NHSN definition change in all 3 hospitals (P>.05; Table 1 and Figure 1).
Fig. 1 The impact of the NHSN methodology changes on HO-CDI LabID events, predicted HO-CDI LabID events, and NAAT positive (+)/EIA positive (+) events. Rates were calculated based on 10,000 patient days. Note. NHSN, National Healthcare Safety Network; HO-CDI LabID, healthcare facility onset Clostridium difficile laboratory identified event as per NHSN definition; NAAT, nucleic acid amplification test; EIA, toxin enzyme immunoassay; SIR, standardized infection ratio. NAAT+/EIA+ events were determined using the monthly counts of HO-CDI LabID events and removing all the events that were NAAT+/EIA−. January 2017 shows a zero for NAAT+/EIA+ events as that month reflex EIA had not yet been instituted.
Table 1 Effect of the NHSN Methodology Changes on HO-CDI LabID Events and Standardized Index Ratios
Note. NHSN, National Healthcare Safety Network; HO-CDI LabID, healthcare-facility–onset Clostridium difficile laboratory identified event per the NHSN definition; NAAT, nucleic acid amplification test; EIA, toxin enzyme immunoassay; SIR, standardized infection ratio.
a Hospital A: 565-bed teaching affiliated hospital with 110 beds assigned to hematology-oncology and solid organ transplant patients.
b Hospital B: 206-bed community hospital with a new teaching affiliation denomination as of Q1 2018.
c Hospital C is a 70-bed rural hospital nonteaching affiliated. During 2017, 57 of 194 HO-CDI LabID events were located in the hematology oncology units. All hospitals are located in the Greater Milwaukee area.
d Represents 2 months of EIA testing.
e These events were not included in the HO-CDI LabID events for that quarter due to the change in NHSN methodology.
Because the adjustment for calculating the predicted HO-CDI LabID events changed from a NAAT-based test to an EIA-based test in January 1, 2018, the predicted number of events decreased in hospital A from 37.15 predicted events per quarter to 23.93 predicted events in the first 2 quarters of 2018 (P=.0057) (Figure 1). Hospitals B and C combined did not experience an impact in their predicted number of infections, even though hospital B went from community-nonteaching affiliated to community-teaching affiliated (6.9 vs 6.1 per quarter; P>.05). Regarding SIRs, hospital A experienced a drastic drop in SIRs after the methodology change in definition from 1.29 to 0.59 (P=.03); similarly, hospital B’s SIR dropped from 0.69 to 0.22, and hospital C’s SIR decreased from 0.64 to 0.37, before and after the methodology change, respectively (P>.5, both hospitals).
Discussion
The change in HO-CDI LabID event methodology, despite adjustment of the predicted number of events, had a major impact in the number of HO-CDI LabID events and SIRs in one of our hospitals; however, NAAT-positive/EIA-positive events remained unchanged. This change occurred in the absence of any new prevention interventions late 2017 and in 2018. This study demonstrates how changes in the interpretation of the clinical tests can play a major role in the reporting of hospital-acquired infections despite the lack of fluctuations in the underlying number of NAAT-positive/EIA-positive cases. Several years ago, more than half of the hospitals in the United States transitioned to NAAT-based testing for C. difficile with a concomitant increase in the number of CDI infections.Reference Lessa, Winston and McDonald 3 Not surprisingly, we found that a reflex-EIA decreased the number of reported HO-CDIs by more than half. However, are we underreporting clinically significant CDI cases by omitting EIA-negative results? The literature is not clear: a couple of studies showed that EIA results do not correlate with disease severity.Reference Humphries, Uslan and Rubin 4 , Reference Guerrero, Chou, Jury, Nerandzic, Cadnum and Donskey 5 In contrast, a large prospective observational cohort study showed that virtually all C. difficile-associated complications and deaths occurred in patients with positive toxin immunoassay.Reference Polage, Gyorke and Kennedy 6 A recent C. difficile guideline recommended the use of a multistep testing algorithm, such as NAAT followed by EIA.Reference McDonald, Gerding and Johnson 7 However, this guidance clearly recommends testing stewardship by foregoing tests in patients who had received laxatives. Even though we did not present test stewardship data, contrary to hospital A, hospitals B and C were successful in the past couple of years at decreasing inappropriate C. difficile testing. These testing stewardship initiatives included testing early on admission and only when clinically indicated, not testing with laxative use, and rejecting formed stools. Also, hospital A has a large number of beds assigned to immunocompromised patients (19.5%), which are populations at higher risk for toxigenic C. difficile colonization.Reference Cannon, Musuuza and Barker 8 , Reference Jain, Croswell and Urday-Cornejo 9
In summary, a change in HO-CDI LabID-event methodology had an immediate and drastic impact in the SIR at a large academic hospital with large number of immunocompromised patients. Future studies should evaluate the impact of these methodological changes elsewhere and using longer observations times.
Acknowledgments
Financial support
No financial support was provided relevant to this article.
Conflicts of interest
L.S.M.P. has received grant funding from Cepheid. N.L. and B.B. have both been consultants for Luminex, have served as speakers for Biofire, and have received grant funding from Cepheid, Alere/TechLab, and GenePOC. All other authors report no conflicts of interest relevant to this article.
