Incidence and Outcomes Associated With Clostridium difficile Infections: A Systematic Review and Meta-analysis

Key Points español 中文 (chinese)

Question  What is the incidence of hospital-onset Clostridium difficile infection (CDI) and its associated length of stay?

Findings  This systematic review and meta-analysis of 13 studies using patient-days as the denominator found that the incidence of hospital-onset CDI was 8.3 cases per 10 000 patient-days. Among propensity score–matched studies of the length of stay, the mean difference in length of stay between patients with and those without CDI varied from 3.0 to 21.6 days.

Meaning  Pooled estimates from currently available literature suggest that CDI is associated with a large burden on the US health care system.

Importance  An understanding of the incidence and outcomes of Clostridium difficile infection (CDI) in the United States can inform investments in prevention and treatment interventions.

Objective  To quantify the incidence of CDI and its associated hospital length of stay (LOS) in the United States using a systematic literature review and meta-analysis.

Data Sources  MEDLINE via Ovid, Cochrane Library Databases via Wiley, Cumulative Index of Nursing and Allied Health Complete via EBSCO Information Services, Scopus, and Web of Science were searched for studies published in the United States between 2000 and 2019 that evaluated CDI and its associated LOS.

Study Selection  Incidence data were collected only from multicenter studies that had at least 5 sites. The LOS studies were included only if they assessed postinfection LOS or used methods accounting for time to infection using a multistate model or compared propensity score–matched patients with CDI with control patients without CDI. Long-term-care facility studies were excluded. Of the 119 full-text articles, 86 studies (72.3%) met the selection criteria.

Data Extraction and Synthesis  Two independent reviewers performed the data abstraction and quality assessment. Incidence data were pooled only when the denominators used the same units (eg, patient-days). These data were pooled by summing the number of hospital-onset CDI incident cases and the denominators across studies. Random-effects models were used to obtain pooled mean differences. Heterogeneity was assessed using the I² value. Data analysis was performed in February 2019.

Main Outcomes and Measures  Incidence of CDI and CDI-associated hospital LOS in the United States.

Results  When the 13 studies that evaluated incidence data in patient-days due to hospital-onset CDI were pooled, the CDI incidence rate was 8.3 cases per 10 000 patient-days. Among propensity score–matched studies (16 of 20 studies), the CDI-associated mean difference in LOS (in days) between patients with and without CDI varied from 3.0 days (95% CI, 1.44-4.63 days) to 21.6 days (95% CI, 19.29-23.90 days).

Conclusions and Relevance  Pooled estimates from currently available literature suggest that CDI is associated with a large burden on the health care system. However, these estimates should be interpreted with caution because higher-quality studies should be completed to guide future evaluations of CDI prevention and treatment interventions.

Clostridium difficile (also known as Clostridioides difficile) is the most common pathogen causing health care–associated infections in the United States, accounting for 15% of all such infections.¹ A Centers for Disease Control and Prevention report on antibiotic resistance threats categorized C difficile as an urgent threat.² Antibiotic treatment for C difficile infection (CDI) is often followed by recurrent infection, leading to nontraditional treatments, such as fecal transplant and oral administration of nontoxigenic C difficile spores.^3,4

Information about the burden of CDI in the United States could inform investments in prevention and treatment interventions. This information should include the incidence of CDI, how this incidence has changed over time, and poor outcomes associated with CDI. Although prior studies have shown that CDI is associated with poor outcomes, such as recurrence, long hospital length of stay (LOS), mortality, and high treatment costs, these results vary by study location and patient population.^2,5 In addition, many current estimates of the poor outcomes and costs associated with CDI do not take into account the underlying severity of illness among patients who develop CDI and may overestimate the true attributable outcomes.⁶

To address gaps in our understanding of the current burden associated with CDI in the United States, we conducted a systematic literature review of studies conducted in the United States and published after 2000 that evaluated the incidence of CDI and associated LOS. The goals were to describe the recent incidence of CDI and to evaluate LOS attributable to CDI.

This systematic review and meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA)⁷ and Meta-analysis of Observational Studies in Epidemiology (MOOSE)⁸ reporting guidelines. An experienced health sciences librarian (A.B.) conducted systematic searches in MEDLINE via Ovid, Cochrane Library Databases via Wiley, Cumulative Index of Nursing and Allied Health Complete via EBSCO Information Services, Scopus, and Web of Science to identify articles published from the inception of the database to February 2019. Citations published before 2000 were excluded. A combination of keywords and subject headings were used for “Clostridium difficile,” “length of stay,” and “incidence.” The full search strategies can be found in eAppendix 1 in the Supplement.

Publications were included if they evaluated the incidence of CDI or LOS associated with CDI. Studies were excluded if they did not contain original data, did not have a control group, were published outside the United States, were published in a language other than English, or were published before 2000. The year 2000 was chosen as the beginning of this systematic literature review because that was when the epidemic BI/NAP1/027 strain of C difficile emerged, after which CDI increased in prevalence and became less responsive to treatment.⁴ We excluded studies if they assessed only a specific subset of patients, unless that population could be categorized as 1 of the following subsets: immunocompromised patients, patients in the intensive care unit, patients with cancer, patients with end-stage renal disease, patients undergoing hemodialysis, surgical patients, solid-organ transplant recipients, patients with high-risk gastrointestinal conditions, or peripartum women. We excluded studies with a study period of less than 1 year. We also excluded studies of long-term care facilities. Incidence data were collected only from multicenter studies that had at least 5 sites, because single-site or small studies may be biased by outbreaks or other local conditions. We included incidence studies with denominators of patient-days or person-years, known timing of the CDI such as after surgery or after admission (ie, hospital onset [HO]), or exclusion of patients with a history of CDI.

Studies were included in the LOS analysis only if they provided data on postinfection LOS, if they used methods accounting for time to infection using a multistate model, or if propensity score–matched patients with CDI were compared with uninfected controls.^5,9 Studies were excluded if they did not have an uninfected control group or a denominator that included patients without CDI.

Titles and abstracts of all articles were screened to assess inclusion criteria. Two of 9 independent reviewers (M.L.S., M.A.W., M.F.K., H.-Y.C., M.L.C., L.A.H., D.J.D., A.R.M., and E.N.P.) abstracted data for each article. Reviewers resolved disagreements by consensus.

The reviewers abstracted data on study design, study population, setting and years, inclusion and exclusion criteria, number of patients included, description of control group, definition of CDI, outcomes (eg, incidence and LOS), and an assessment of the potential risk of bias. Risk of bias was assessed using the Downs and Black scale.¹⁰ Reviewers followed all questions from this scale as written except for question 27 (a single item on the Power subscale, which was scored 0-5), which was changed to a yes or no. Two of us (A.R.M. and M.L.S.) performed component quality analysis independently, reviewed all inconsistent assessments, and resolved disagreements by consensus.¹¹

Data analysis was performed in February 2019. Excel spreadsheet software version 2007 (Microsoft Corp) and RevMan statistical software version 5.3 (Cochrane Community) were used for statistical analysis. Incidence data were pooled only when the denominators used the same units (eg, patient-days). These data were pooled by summing the number of HO-CDI incident cases and the denominators across studies. Pooled incidence was reported as the number of incident cases per the given denominator (eg, 10 000 patient-days).¹² No P values were calculated.

Of the 34 775 articles identified (Figure), 119 were full-text articles, and 86 (72.3%) of those articles met the selection criteria and were included in the systematic literature review.^13-93 Among these, 66 articles evaluated incidence,^13-78 and 20 articles evaluated LOS.^{16,54,66,79-95} One-fifth of the studies that assessed LOS (4 studies)^84,87,91,94 scored 18 or more points of the 28 points possible on the Downs and Black scale¹⁰ and, thus, were considered to be of higher quality.

Sixty-six studies^13-78 measured CDI incidence. Thirteen of those 66 studies^13-25 used patient-days as the denominator (Table 1). Among these studies, the CDI incidence varied from 2.8 CDI cases per 10 000 patient-days²² to 15.8 CDI cases per 10 000 patient-days.²⁰ Three studies^13,17,23 were conducted by the Centers for Disease Control and Prevention. Three studies^17,18,21 were done in New York State. One study²⁴ from Southern California found that the incidence of community-onset, health care facility (HCF)–associated CDI (11.1 cases per 10 000 patient-days) was almost 2-fold higher than that for HO, HCF-associated CDI (6.8 cases per 10 000 patient-days). The pooled incidence of HO-CDI among the 13 studies^13-25 (Table 1) that used patient-days as the denominator was 8.3 CDI cases per 10 000 patient-days. Four studies^13,15,18,21 included more than 100 facilities.

The definitions of C difficile used to identify cases varied. Three studies^17,18,21 used clinical findings and results of laboratory tests for C difficile, 3 studies^13,14,23 used the Centers for Disease Control and Prevention surveillance definition to identify C difficile, 2 studies^20,22 applied infection preventionist evaluations for C difficile surveillance, and 2 studies^24,25 used only results of laboratory tests for C difficile. The remaining studies used a variety of ways to identify CDI, including International Classification of Diseases, Ninth Revision (ICD-9) codes or other billing codes,^15,16,19 laboratory test results,^15,16,20,23 clinical findings,^15,23 and initial doses of C difficile antibiotic therapy.¹⁹ When we examined incidence by time period, we found that the early studies from 2000 to 2008 had a range from 2.8 to 12.2 CDI cases per 10 000 patient-days, studies from 2008 to 2009 had a range from 6.3 to 9.6 CDI cases per 10 000 patient-days, and the later studies after 2010 reported a range from 6.8 to 15.8 CDI cases per 10 000 patient-days (Table 1).

Fourteen studies^26-39 included both inpatients and outpatients (Table 2), reflected in a denominator of person-years in 8 studies.^{27-30,32,34,36,39} Seven of those 14 studies^{27-30,32,34,39} used only ICD-9 codes to define CDI. In a study³⁶ of adult and adolescent patients with HIV/AIDS that included more than 100 hospitals, during 10 years of study, the peak incidence of CDI was 9.59 cases per 1000 person-years among patients with clinical AIDS. A study²⁸ of the Armed Forces Health Surveillance Center in Maryland over the course of 12 years found the incidence of community-associated CDI to be 5.5 cases per 100 000 person-years. In a study²⁹ evaluating the annual incidence of CDI and multiply recurrent CDI per 1000 person-years, the incidences increased by 42.7% and 188.8%, respectively, during a decade (2001-2012) in the United States. In another study³⁰ with 12 years of data from 5 administrative databases, elderly people (ie, aged >65 years) had a CDI rate of 677 cases per 100 000 person-years. In contrast, a managed-care organization in Colorado found that the CDI incidence in 2007 was 14.9 CDI cases per 10 000 patient-years.³² These studies were too diverse to pool together into 1 estimate.

Three studies^40-42 included only inpatients (Table 2). Two of these studies^41,42 assessed the Agency for Healthcare Research and Quality (AHRQ) National Inpatient Sample (NIS). One evaluated infant patients from the AHRQ NIS cohort,⁴¹ and the other study evaluated adult patients from the AHRQ NIS cohort.⁴² Both studies documented substantial increases in CDI incidence between 2000 and 2005, from 2.8 to 5.1 cases per 10 000 hospitalizations, and from 5.5 to 11.2 cases per 10 000 hospitalizations, respectively.^41,42 The third study,⁴⁰ which was from the US National Hospital Discharge Survey between 2001 and 2010, found that the incidence of CDI in the pediatric population was 1.2 CDI discharges per 1000 total discharges.

Twenty-six studies^43-68 documented HO-CDIs, which we assumed were incident cases (Table 3). Of these studies, the AHRQ NIS was the main data set, represented by 10 included studies.^{43,45,47,50,51,56,58,59,61,68} These studies assessed diverse patient populations with different comorbidities, including peripartum women⁶⁸ and patients with inflammatory bowel disease,⁴³ lymphoma,⁴⁵ leukemia,⁵⁸ subarachnoid hemorrhage treated with microsurgical or endovascular aneurysm repair,⁴⁷ chronic liver disease,⁵⁰ hematopoietic stem cell transplant,⁵¹ megacolon,⁵⁶ or heart failure.⁵⁹ Thus, the results of these studies were also too diverse to pool together. One study⁶⁸ found that the CDI incidence among peripartum women increased from 0.36 cases per 10 000 in 1998 to 0.70 cases per 10 000 in 2006. The US National Hospital Discharge Survey database was represented in 6 included studies.^{49,52,53,55,64,65} These studies also assessed diverse patient populations, including children⁵² and adults with different comorbidities, such as cancer^49,52 and inflammatory bowel disease.⁶⁵ In 1 of these studies,⁶⁵ the overall incidence of HO-CDI was 369.8 cases per 10 000 hospitalizations for inflammatory bowel disease. In that same study,⁶⁵ the HO-CDI incidence was 445.6 cases per 10 000 hospitalizations for ulcerative colitis and 220.3 cases per 10 000 hospitalizations for Crohn disease.

Ten studies^69-78 evaluated surgical patients (Table 3), and, thus, we assumed that the CDI cases were incident cases. Five studies^73,75-78 used data from AHRQ NIS. These AHRQ NIS studies analyzed a variety of surgical procedures, including spine surgery⁷⁶; hip,⁷³ knee,⁷⁷ or lower-extremity⁷⁸ arthroplasty; and elective colon resections.⁷⁵ One of them had CDI occurring in 1.4% of patients, for a rate of 144.99 cases of C difficile colitis per 10 000 elective colon resections, and the incidence increased from 1.31% in 2004 to 1.67% in 2006.⁷⁵

Twenty studies^{16,54,66,79-94} (Table 4) evaluated CDI-associated LOS. Sixteen studies^{54,66,79-89,92,94,95} used propensity score matching to evaluate LOS associated with CDI, 2 studies^16,93 used postinfection LOS, 1 study⁹⁰ matched on LOS from admission until either positive C difficile test results or discharge, and 1 study⁹¹ accounted for time to infection using a multistate model. Also, one of the propensity score matched–studies applied multistate modeling to account for timing of infection.⁸⁸ Pediatric patients were included in 3 of these studies.^66,86,87

Among the 13 propensity score–matched studies of adults, the CDI-associated mean difference in LOS (in days) between patients with CDI and patients who did not have CDI varied greatly from 3.0 days (95% CI, 1.44-4.63 days)⁷⁹ to 10.3 days.⁵⁴ Among the 3 pediatric propensity score–matched studies,^66,86,87 the highest CDI-associated mean difference in LOS (in days) was 21.6 days (95% CI, 19.29-23.90 days).⁶⁶

Among the studies that used multistate models to account for timing of infection, a study⁹¹ performed in the Veterans Affairs health care system found that the magnitude of its estimated impact was smaller when methods were used to account for the time-varying nature of infection. That study estimated a CDI-attributable LOS of only 2.27 days (95% CI, 2.14-2.40 days).⁹¹ The other study⁸⁸ that performed propensity score matching and used a multistate model converged on similar excess LOS estimates of 3.1 days (95% CI, 1.7-4.4 days) and 3.3 days (95% CI, 2.6-4.0 days), respectively.

Four studies^84,87,91,94 that evaluated LOS earned 18 or more points on the Downs and Black scale.¹⁰ One study⁹¹ also used multistate modeling. Another was also performed in the Veterans Affairs health care system^84,91 and found a mean difference between patients with and without CDI of 7.5 days.⁸⁴ One study⁸⁷ of pediatric patients found that those with CDI had a longer LOS (adjusted odds ratio, 4.34; 95% CI, 3.97-4.83). Another study⁹⁴ of adult patients in Pennsylvania hospitals showed an attributable hospital LOS difference of 2.4 days (95% CI, 0.7-4.4 days; P < .01) between patients with and without CDI.

National epidemiological investigations have demonstrated recent marked increases in CDI in the United States.³⁴ Thus, a national public health response to this increase requires current estimates of the CDI incidence.^96-98 Our systematic review of the literature found that the CDI incidence varied by study and that the investigators used different denominators when they calculated the incidence for specific study populations. In our meta-analysis of studies that used patient-days as the denominator, we estimated the incidence of CDI in the United States to be 8.3 CDI cases per 10 000 patient-days.

Variation in CDI incidence may be due, in part, to advances in diagnostic technology and variations in diagnostic practices.^99-101 Nucleic acid amplification tests are more sensitive than traditional C difficile stool tests (eg, toxin enzyme immunoassay). Nucleic acid amplification tests have been used more frequently in clinical practice since 2009, when the first commercial polymerase chain reaction was approved by the US Food and Drug Administration.¹⁰² The topic of CDI testing methods and risk adjustment is complex.^103,104 Concerns have been expressed about the adequacy of risk adjustment to account for different CDI testing methods (toxin enzyme immunoassay alone, polymerase chain reaction alone, toxin enzyme immunoassay plus glutamate dehydrogenase followed by polymerase chain reaction for discrepancies, polymerase chain reaction followed by toxin enzyme immunoassay, and other diagnostic options) across HCFs. The choice of testing methods substantially affects the performance of these testing algorithms.^99-101

In addition, the CDI incidence found by these studies likely varied because of the different database structures adopted by the various hospitals.^13-78 Some analyses were based on health care systems databases, but most used large infection control surveillance, state, or national discharge databases.^13-25 Beginning in January 2013, the Centers for Medicare & Medicaid Services began requiring public reporting of CDI rates via the National Healthcare Safety Network for those hospitals participating in the Inpatient Prospective Payment System.⁶⁴ Specifically, 1 study²⁹ demonstrated an increase in the annual incidence of CDI and multiply recurrent CDI per 1000 person-years by 42.7% and 188.8%, respectively, between 2001 and 2012. Another CDI surveillance study³³ in 7 US states reported an increase not only in community-associated CDI incidence rates but also an increase in health care–associated CDI incidence rates. Furthermore, CDI can complicate comorbid conditions and result in the need for additional hospital resources.³⁴ Included studies detected an increase in the CDI incidence in patients with inflammatory bowel disease,⁴³ patients with cancer,⁵² those undergoing surgery,^75,76 and even infants.⁴¹ The results of our systematic review of literature and meta-analysis emphasize the need to perform C difficile surveillance and direct resources to the prevention of CDI in order to reduce the incidence across the United States.

This systematic literature review has some limitations. First, the results of systematic literature reviews and meta-analyses are only as valid as the results of the studies evaluated. Most studies included in this systematic literature review were of moderate-to-low quality and may have overestimated the outcomes. We need more high-quality studies so that we can accurately determine postinfection LOS, because LOS before the infection should not be attributed to C difficile.⁵ Second, we included studies that used ICD-9 codes to define CDI. The ICD-9 codes are used for billing purposes and are not ideal for surveillance. However, a prior meta-analysis¹⁰⁵ found that the ICD-9 code for C difficile had good sensitivity, specificity, positive predictive value, and negative predictive value compared with clinical definitions. Third, we only included studies conducted in the United States and published in English, which limits the external validity of this research. We used these inclusion criteria because our goal was to evaluate the burden of CDI in the United States. Future systematic literature reviews should be performed to evaluate this burden in other countries. Fourth, we found heterogeneity in all LOS-stratified analyses (eAppendix 2 and eTable in the Supplement). We found that the higher-quality studies that used advanced statistical methods to attempt to account for time-dependent bias found lower CDI-attributable LOS compared with other studies that did not use advanced methods. In addition, our incidence estimates were derived from multicenter studies only. Incidence rates in small studies may be variable and subject to bias; thus, this criterion was established a priori to determine representative incidence rates. From incident cases of CDI (36 studies), we were unable to exclude recurrent and multiply recurrent CDI cases if the study did not exclude those cases. For this meta-analysis, we decided to calculate the incidence rate with studies with a similar denominator (patient-days), with a result of 8.3 CDI cases per 10 000 patient-days.

Pooled estimates from the currently available literature suggest that C difficile is associated with a large burden on the US health care system. However, these estimates should be used with caution, and higher-quality studies should be completed to guide future evaluations of C difficile prevention and treatment interventions.

Accepted for Publication: October 26, 2019.

Published: January 8, 2020. doi:10.1001/jamanetworkopen.2019.17597

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 Marra AR et al. JAMA Network Open.

Corresponding Author: Marin L. Schweizer, PhD, Carver College of Medicine, Department of Internal Medicine, University of Iowa, 200 Hawkins Dr, Iowa City, IA 52242-1071 (marin-schweizer@uiowa.edu).

Author Contributions: Drs Marra and Schweizer had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Marra, Perencevich, Nelson, Samore, Diekema, Schweizer.

Acquisition, analysis, or interpretation of data: Marra, Perencevich, Samore, Khader, Chiang, Chorazy, Herwaldt, Kuxhausen, Blevins, Ward, McDanel, Nair, Balkenende, Schweizer.

Drafting of the manuscript: Marra, Chiang, Ward.

Critical revision of the manuscript for important intellectual content: Marra, Perencevich, Nelson, Samore, Khader, Chorazy, Herwaldt, Diekema, Kuxhausen, Blevins, McDanel, Nair, Balkenende, Schweizer.

Statistical analysis: Marra, Kuxhausen, Nair, Schweizer.

Obtained funding: Perencevich, Nelson, Samore, Schweizer.

Administrative, technical, or material support: Samore, Chiang, Chorazy, Diekema, Kuxhausen, Blevins, Ward, Balkenende, Schweizer.

Supervision: Herwaldt, Diekema, Schweizer.

Conflict of Interest Disclosures: Dr Samore reported receiving an Epicenter grant from the Centers for Disease Control and Prevention (CDC) and grants from the Department of Veterans Affairs (VA), Agency for Healthcare Research and Quality, National Institutes of Health, Western Institute for Biomedical Research, and Pfizer outside the submitted work. Ms Ward and Dr Nair reported receiving Epicenter grants from the CDC during the conduct of the study. No other disclosures were reported.

Funding/Support: This work was funded by the CDC’s Safe Healthcare, Epidemiology, and Prevention Research Development Program under contract 200-2011-42039 (principal investigator: Dr Samore). This work was also supported in part by Center of Innovation funding grant CIN 13-412 (principal investigator: Dr Perencevich) from the VA Health Services Research and Development Service. Dr Nelson was supported by VA Health Services Research and Development Career Development Award 11-210. Dr Schweizer was supported by VA Health Services Research and Development Career Development Award 11-215.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the VA or the US government. Dr Perencevich, a JAMA Network Open associate editor, was not involved in the editorial review of or the decision to publish this article.