Clostridium difficile (also known as Clostridioides difficile) infection (CDI) is one of the most common health care–associated (HCA) infections and is a significant cause of morbidity and mortality, especially among older adult hospitalized patients. Although the incidence of HCA CDI and its attributable hospital length of stay (LOS) have appeared to be increasing nationally in the United States, they are highly dependent on the frequency of epidemic or outbreak strains, which influence both the rate and severity of CDI and changes over time. In addition, variable definitions of CDI influenced by a changing diagnostic paradigm can lead to both overdiagnosis and underdiagnosis. Other factors that contribute to the uncertainty are the inclusion of patients younger than 2 years (because they have high rates of asymptomatic colonization), the inclusion of patients only with specific underlying illnesses (eg, cancer or stem cell transplant) or at specific hospital locations (eg, intensive care unit residence), failure to exclude recurrent and multiply recurrent CDI cases from primary CDI totals, and a multiplicity of denominators used to calculate incidence in person-years.
Nonetheless, despite the daunting task, Marra et al1 provide a meta-analysis of the published incidence of HCA CDI and increased LOS attributed to CDI. The authors wisely included studies of HCA incidence data from only 13 multicenter studies that included at least 5 sites and used a standardized 10 000 patient-days as the denominator. The 13 studies were published between 2004 and 2014 and include incidence data from 1987 through 2012, encompassing the increase in incidence of the NAP1/BI/027 epidemic strain of C difficile in the United States.2 The increased incidence of NAP1/BI/027 was first reported in 2005 in a study3 documenting isolates found from 2001 to 2003 at 8 widely dispersed health care facilities in 6 states, which suggested that this strain was already entrenched in multiple US geographic sites by the early 2000s. Rates of CDI increased progressively during the early 2000s and were likely attributed, at least in part, to the presence of the epidemic NAP1/BI/027 strain. Results of the meta-analysis by Marra et al1 likely reflect this contribution of a uniquely epidemic strain to the overall HCA CDI incidence, which is high by current standards at 8.3 cases per 10 000 patient-days, with a wide reported range of 2.8 cases per 10 000 patient-days to 15.8 cases per 10 000 patient-days, the latter in a cancer center. Not only was the NAP1/BI/027 strain epidemic in distribution, it was associated with extraordinarily severe disease, including the need for colectomy and increased mortality that can lead to increased LOS for patients with CDI.4
Marra et al1 also report the increase in LOS associated with CDI in 16 studies that used propensity matching to adjust for a variety of confounders. These studies were published between 2008 and 2018 and report data from 1997 to 2016, well into the period when CDI laboratory testing switched from predominantly enzyme immunoassay (EIA) detection of toxins A and B in stool to nucleic acid amplification (NAAT) testing for the presence of a toxin gene–containing C difficile organism in stool. Diagnostic testing using EIA for toxins A and B began in the late 1990s and was the primary test modality until NAAT testing became available in 2009.5 The importance of this change is that, compared with EIA, NAAT increased the rate of positive testing from 43% to 67% and correspondingly increased the CDI surveillance rates.6 The use of NAAT as any part of CDI testing increased in the Centers for Disease Control and Prevention’s Emerging Infection Program laboratories in 10 states from 11% in 2010 to 79% in 2016 (written communication, Alice Y. Guh, MD, Centers for Disease Control and Prevention, January 18, 2018). The use of NAAT testing alone for CDI diagnosis has been questioned in several studies,7 suggesting that it is overdiagnosing CDI by detecting both colonized patients and those with CDI when compared with the outcomes of patients whose diagnosis is made by toxin EIA. This disparity in diagnosis was recognized in the report by Lessa et al,2 who reported that NAAT was used in 52% of Emerging Infection Program sites in 2011 and performed a sensitivity analysis to determine the effect of NAAT testing percentage on CDI rates. Marra et al1 also confirm that later studies published after 2010 had a higher rate range than earlier published studies.
Nucleic acid amplification testing for CDI can clearly increase the reported incidence, but how might it affect LOS? If, as suspected, more patients with colonization or mild disease are detected by NAAT than by toxin testing, these patients may have a lower attributable LOS because they either may have less-severe CDI or no CDI at all and are simply colonized. Marra et al1 report widely differing excess LOS due to CDI, ranging from 3.0 to 10.3 days in adults to 21.6 days in a pediatric study. One of their cited studies by Pak et al8 in a single center reported data for LOS based on positive toxin EIA test (used for 3 years) and positive NAAT test (used for the next 4 years): the LOS associated with CDI diagnosed by EIA toxin was 10.1 days, and the LOS associated with CDI diagnosed by NAAT was 6.6 days. These LOS differences, although not statistically significant, suggest a trend toward shorter LOS when NAAT testing is performed, consistent with less-severe or overdiagnosed CDI.
What can the reader conclude from this plethora of CDI rate and LOS studies? Marra et al1 are to be complimented for assembling this multiplicity of data into a meta-analysis; however, these reported HCA incident rates from largely 2000 to 2012 are somewhat old news, reflecting the apex of the NAP1/BI/027 CDI epidemic in the United States, as moderated by use of the less-sensitive EIA toxin testing for diagnosis in the early reported years, and possibly reflecting NAAT testing in the publications from 2010 and later. Beginning around 2010, CDI rates reflect increasing use of the much more sensitive NAAT tests, resulting in reports of progressively increasing CDI rates. Currently, there is increasing test utilization of algorithms that include the use of EIA toxin tests rather than NAAT alone.5,9 At the same time, the incidence of the NAP1/BI/027 strain has been decreasing in the United States, according to Emerging Infection Program site data from the Centers for Disease Control and Prevention,2,9 which shows this decrease from an incidence of NAP1/BI/027 of 30.7% in a report of data from 2011 to an incidence of 9.8% for 2014 to 2015.
Taken together, decreasing rates of NAP1/BI/027 strains and the current swing back to increased use of toxin testing should translate into lower CDI rates. Indeed, data from the Centers for Disease Control and Prevention Antibiotic Resistance and Patient Safety Portal10 confirm a 29% decrease in the CDI standardized infection ratio in US hospitals from 2015 to 2018. This is good news, but stay tuned. Neither the next diagnostic test for CDI nor the next epidemic C difficile strain is predictable in terms of when it may occur or its effect on CDI rates and LOS.
Published: January 8, 2020. doi:10.1001/jamanetworkopen.2019.18599
Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 Gerding DN. JAMA Network Open.
Corresponding Author: Dale N. Gerding, MD, Research and Development, Edward Hines Jr VA Hospital, 5000 S 5th Ave, Bldg 1, Room B347, Hines, IL 60141 (email@example.com).
Conflict of Interest Disclosures: Dr Gerding reported receiving personal fees from Merck, Rebiotix/Ferring, Actelion, DaVolterra, Summit, Pfizer, MGB Biopharma, Sanofi Pasteur, and Medpace and reported having a patent for a treatment for Clostridium difficile infection issued.
Gerding DN. Incidence and Outcome of Clostridium difficile Infection—Beware of Strain Type and Diagnostic Tests. JAMA Netw Open. 2020;3(1):e1918599. doi:10.1001/jamanetworkopen.2019.18599
Customize your JAMA Network experience by selecting one or more topics from the list below.