Clostridiodes difficile infection (CDI) is the most common healthcare-associated infection in the United States and is considered an “urgent” threat by the Centers of Disease Control and Prevention. Reference Theriot and Young9–Reference Allegretti, Marcus and Storm11 CDI in particular impacts females, the elderly, and the immunocompromised at higher rates. Reference Lessa, Mu and Bamberg12 There are nearly half a million cases of CDI per year costing an estimated $5 billion. Reference Ma, Brensinger, Wu and Lewis10,Reference Guh, Mu and Winston13 CDI management may require treatment across multiple healthcare settings, impacting those in the community setting, hospitalized patients, or older individuals in long-term care facilities because these patients cycle in and out of healthcare institutions with recurrent infections. Reference Guh, Mu and Winston13
Approximately 25% of patients with primary CDI have a recurrent infection after standard-of-care treatment, and those with first recurrence have a high risk of future recurrences. Recurrences are common because current therapies are effective at relieving symptoms by killing the toxin-producing bacteria or binding the toxin, but none repair gut microbiome dysfunction. In addition, C. difficile-targeted antibiotics do not have any effect on dormant C. difficile spores. Reference Cornely, Crook and Esposito14,Reference Gerding, Kelly and Rahav15 Once a patient has a recurrence, signaling a disrupted microbiome, their risk of having subsequent recurrences rises substantially and further exacerbates the impact of CDI. Reference Theriot and Young9,Reference Lessa, Mu and Bamberg12
We conducted a systematic review of that evidence to gain a better understanding of the implications of CDI on healthcare costs and resource use. We also evaluated the quantity and quality of studies that have examined the economics and healthcare use for patients with CDI and rCDI. Here, we summarize our findings, describe gaps in the literature detailing the differential impact of CDI and rCDI on patients and the US healthcare system, and identify gaps in the evidence for further investigation.
Methods
We searched the abstracting services of PubMed, Embase, and the Cochrane Collaboration to identify research that evaluated the resource use and costs to care for patients with CDI and rCDI. Specific inclusion and exclusion criteria are listed below, but the focus of this review was on those articles reporting utilization of healthcare services and/or economic end points. Search terms and filters used for Embase and PubMed literature searches are summarized in Table 1. In addition to conducting searches of electronic abstracting databases, the references of relevant primary studies, guideline documents, published meta-analyses, and authoritative clinical reviews were examined to identify other possible articles. The Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines were followed in this analysis. Reference Page, McKenzie and Bossuyt16
Literature Search Terms Utilized
Studies were eligible for inclusion if they met the following criteria: evaluated either primary or rCDI; included end points of healthcare use (eg, hospitalizations, clinic visits, etc.) or economic outcomes (eg, cost of care) for services provided in the United States; published after 2010 due to rapidly changing epidemiology of the disease, new therapies, and evolving treatment strategies; and published in English. Cost analyses from other countries were excluded because resource unit costs vary widely between countries. Economic models (eg, cost-effectiveness analyses) were beyond the scope of this review because these models are typically constructed using other published data and the focus of this review was on the primary literature.
After completing the searches from abstracting services, the titles and abstracts of identified studies were examined for relevance. Full manuscripts of potentially relevant articles were retrieved to verify eligibility and undergo data extraction. Two reviewers (D.C.M. and E.P.A.) independently conducted the literature searches and independently reviewed article titles and abstracts for inclusion in the analysis. Differences between reviewers regarding including or excluding a study were resolved by evaluation and discussion of the article by the 2 reviewers.
Each study was evaluated with respect to study design, incorporation of a comparison and/or control group, data source, end points, and sample size. For study design, we evaluated whether a comparison group was used and the use of propensity score matching (PSM). PSM is a common technique to analyze observational research because it incorporates many variables when matching cases and controls. Generally, studies using PSM have higher internal validity because the matching approach is robust and overcomes small cell sizes that plague traditional matching.
Results
Figure 1 displays the PRISMA flow diagram describing the article selection process. Initially >10,000 articles were identified, but after applying restrictions on date and language, 3,866 remained. Once title and abstracts were screened, 253 full reports were obtained and evaluated for inclusion, leading to a total of 28 published studies for review. Reference Nelson, Scott and Boules1–Reference Dubberke, Schaefer, Reske, Zilberberg, Hollenbeak and Olsen8,Reference Duhalde, Lurienne, Wingen-Heimann, Guillou, Buffet and Bandinelli17–Reference Kulaylat, Rocourt and Podany36 Across these studies, various data sources were utilized to examine the impact of CDI and/or rCDI (Table 2). The most frequent source of data evaluated was Medicare (n = 6), Reference Nelson, Scott and Boules1,Reference Zilberberg, Shorr, Jesdale, Tjia and Lapane6,Reference Shorr, Zilberberg, Wang, Baser and Yu23–Reference Drozd, Inocencio and Braithwaite25,Reference Stewart and Hollenbeak35 followed by a variety of commercial medical and pharmacy databases (eg, Truven Health MarketScan, HealthCore Integrated Research, PharMetrics Plus, Kaiser Permanente Northern California, Cerner). Reference Feuerstadt, Stong, Dahdal, Sacks, Lang and Nelson2–Reference Kuntz, Baker and Kipnis4,Reference Duhalde, Lurienne, Wingen-Heimann, Guillou, Buffet and Bandinelli17,Reference Mollard, Lurienne, Heimann and Bandinelli20,Reference Palli, Quimbo and Strauss28–Reference Quimbo, Singer, Strauss and Thomas30 Two publications used the Premier hospital database, Reference Gallagher, Reilly, Navalkele, Downham, Haynes and Trivedi26,Reference Magee, Strauss, Thomas, Brown, Baumer and Broderick27 which includes hospital services but does not capture healthcare services received beyond the hospital.
Flow diagram for systematic review processes.
Databases Used by the Identified Studies
In total, 20 studies evaluated primary CDI or did not specify, whereas 8 studies Reference Nelson, Scott and Boules1–Reference Dubberke, Schaefer, Reske, Zilberberg, Hollenbeak and Olsen8 evaluated both primary CDI and rCDI. The study objectives and outcomes of interest varied across the articles (Table 3), but common study end points included total costs (inpatient and outpatient), hospital costs or charges, healthcare resources, hospital length of stay (LOS), patient mortality, and hospital readmission rates. We noted significant heterogeneity in the methods applied to evaluate the impact of CDI.
Use of Comparator Group, Study End Points, and Study Results
Note. CDI, Clostridioides difficile infection; GI, gastrointestinal; ID, infectious disease; LOS, length of stay; NA, not applicable; SD, standard deviation.
Quality of CDI studies
We observed significant variability across the studies. One issue included the parameters for defining an episode of CDI. In general, studies used the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis codes. However, a few studies defined CDI based on exposure to the oral antibiotics vancomycin and fidaxomicin, which are primarily used to treat CDI. Reference Nelson, Scott and Boules1,Reference Gallagher, Reilly, Navalkele, Downham, Haynes and Trivedi26,Reference Magee, Strauss, Thomas, Brown, Baumer and Broderick27,Reference Pakyz, Carroll, Harpe, Oinonen and Polk34 Another issue was the variability in the duration of time used to define another episode of rCDI, although most studies of rCDI used a relatively narrow window of time after the initial episode to define recurrence. Moreover, 2 of 8 studies defined recurrence as within 42 days of the primary CDI episode, 4 studies defined recurrence as within 56 to 60 days of the primary CDI episode, and 2 studies defined recurrence as within 84 to 90 days of the primary episode. Reference Nelson, Scott and Boules1–Reference Dubberke, Schaefer, Reske, Zilberberg, Hollenbeak and Olsen8 Despite the various definitions, most studies evaluating recurrences did so using a relatively narrow window of time after the initial episode to account for the inherent rapidity of recurrences. Another difference across the studies was the time horizon during which resources were captured. Studies using the Agency for Healthcare Research and Quality’s data sets (n = 4) were typically cross-sectional at a point in time because tracking patients across multiple admissions was not possible. In contrast, studies using payer data, such as Medicare (n = 6) and commercial insurance claims (n=10), were able to follow patients over time. When analyzing the strength of the evidence in the identified studies, 16 (57.1%) of the 28 studies used a comparator group. Of these 16 studies, 9 (56.2%) used PSM Reference Zhang, Prabhu and Marcella3,Reference Zilberberg, Shorr, Jesdale, Tjia and Lapane6,Reference Mehrotra, Jang, Gidengil and Sandora22,Reference Shorr, Zilberberg, Wang, Baser and Yu23,Reference Drozd, Inocencio and Braithwaite25,Reference Sammons, Localio, Xiao, Coffin and Zaoutis31,Reference Tabak, Zilberberg, Johannes, Sun and McDonald32,Reference Stewart and Hollenbeak35,Reference Kulaylat, Rocourt and Podany36 to select controls.
Healthcare burden of CDI and rCDI
Although some studies specifically assessed the initial CDI episode or rCDI episodes, the majority of studies compared any history of CDI (current or past) to non-CDI controls. Reference Duhalde, Lurienne, Wingen-Heimann, Guillou, Buffet and Bandinelli17,Reference Mehrotra, Jang, Gidengil and Sandora22–Reference Drozd, Inocencio and Braithwaite25,Reference Magee, Strauss, Thomas, Brown, Baumer and Broderick27,Reference Campbell, Dean, Nathanson, Haidar, Strauss and Thomas29–Reference Tabak, Zilberberg, Johannes, Sun and McDonald32,Reference Pakyz, Carroll, Harpe, Oinonen and Polk34–Reference Kulaylat, Rocourt and Podany36 Other comparisons evaluated a primary CDI episode versus rCDI Reference Rodrigues, Barber and Ananthakrishnan5–Reference Shah, Aitken and Barragan7 and 1 study examined both non-CDI and rCDI compared to any CDI or a single episode. Reference Zhang, Prabhu and Marcella3 Few studies went beyond classifying patients based on the number of recurrences.
A study by Feurerstadt et al Reference Feuerstadt, Stong, Dahdal, Sacks, Lang and Nelson2 examined multiple rCDI episodes using administrative claims from ∼100 commercial insurance programs across the United States among 46,000 patients aged 18–64 years. Patients were classified as having only an initial CDI or 1, 2, or ≥3 recurrent episodes. Patients with recurrences had substantially higher total all-cause direct costs ranging from $132,000–$207,000, compared with $71,980 for those individuals without rCDI. Over a 12-month period, the percentage of persons who required hospitalization was 29% among those who did not have a CDI recurrence, whereas this percentage ranged from 53% to 69% among those with 1 to ≥3 CDI recurrences.
Nelson et al Reference Nelson, Scott and Boules1 also examined the associated healthcare costs among 269,000 individuals aged ≥65 years enrolled in Medicare fee for service with an index CDI episode. In this study, 35% of patients had ≥1 episodes of rCDI over 12 months of follow-up. During this time, the all-cause direct costs (Medicare and direct payments) were $76,000 for those who did not have a subsequent recurrence compared to a range of $96,000–$99,000 for those who had ≥1 recurrences. However, unlike Feuerstadt et al, subsequent episodes were not substantially more costly than the initial episode, likely due to the effect of diagnosis-related group (DRG)–based payment rules used by Medicare. These rules are designed to reduce hospital readmissions by capping reimbursements in instances of excessive healthcare resource utilization. Thus, combined with lower reimbursement rates in general despite potential readmissions, cost associated with CDI under the Medicare program may be different than that for commercial plans.
The impact of healthcare resource utilization of CDI and rCDI
A common theme that continued to be observed was the higher healthcare resource usage by patients with recurrent infection than primary CDI. Kuntz et al Reference Kuntz, Baker and Kipnis4 assessed the utilization of healthcare services using data from the Kaiser Foundation Health Plan. They used an “optimal matching algorithm” to identify a comparator group for CDI cases and found that 50% of patients with rCDI had at least 1 overnight hospitalization during the 12-month follow-up period compared to 38% among matched nonrecurrent CDI comparators. Reference Kuntz, Baker and Kipnis4 Patients with rCDI consistently had higher healthcare resource utilization compared to patients with primary CDI as well as patients without CDI, including a higher risk for emergency department visits, longer hospital stays, increased use of intensive care unit services, and higher 1-year mortality than patients with primary CDI.
Rodrigues et al Reference Rodrigues, Barber and Ananthakrishnan5 conducted an observational study of patients with rCDI and reported that there were 15.4 outpatient office visits per patient during the year following the rCDI episode. Additionally, they noted that 94% of patients required another hospitalization and that 84% had at least 1 CDI-related hospitalization after the recurrent episode. Among hospitalized patients, 40% were discharged from the hospital to short-term rehabilitation facilities and 13% of patients died. Reference Rodrigues, Barber and Ananthakrishnan5
One common outcome evaluated by 15 of the included studies was hospital LOS, which was significantly higher among patients with CDI compared to those without CDI. For example, in a recent analysis utilizing the Truven Marketscan database, Mollard et al Reference Mollard, Lurienne, Heimann and Bandinelli20 reported that individuals who were admitted with CDI as a primary diagnosis had an average LOS of 5.9 days; for those with CDI as a secondary diagnosis, the presence of CDI as a complication increased the overall LOS by 4.4 days compared to those without CDI. In another large study, Magee et al Reference Magee, Strauss, Thomas, Brown, Baumer and Broderick27 assessed the burden of CDI in hospitalized patients using the Premier hospital database and PSM. These researchers reported that the adjusted hospital LOS was 13.2 days for patients with CDI compared to 8.5 days for patients without CDI (P < .01), an additional 5.7 days were attributable to CDI. This increase in LOS was also observed in studies that focused on specific populations. For example, Duhalde et al Reference Duhalde, Lurienne, Wingen-Heimann, Guillou, Buffet and Bandinelli17 evaluated individuals with cancer and CDI and found that the average LOS was 23.1 additional days compared to cancer patients without CDI.
Three studies specifically assessed the impact of rCDI on LOS compared to a primary CDI episode. In a single-center study, Shah et al Reference Shah, Aitken and Barragan7 reported that recurrent CDI was associated with a median hospital LOS that was more than twice the duration associated with primary CDI (24 days vs 11 days, respectively). In another study evaluating nursing-home residents, an additional 20.3 hospital days were attributable to rCDI compared to primary CDI alone. Reference Mehrotra, Jang, Gidengil and Sandora22 Zhang et al Reference Zhang, Prabhu and Marcella3 utilized PSM within a large national healthcare claims database to match patients with rCDI to comparators without a recurrence and found that rCDI contributed an additional 2 hospital days over a 6-month period. Based on Zhang et al and other articles, the LOS attributable to primary CDI appears to be at least 7 days (up to 23 days for some studies), and for patients with rCDI, the cumulative LOS is typically >9 days. Reference Zhang, Prabhu and Marcella3,Reference Rodrigues, Barber and Ananthakrishnan5,Reference Duhalde, Lurienne, Wingen-Heimann, Guillou, Buffet and Bandinelli17,Reference Magee, Strauss, Thomas, Brown, Baumer and Broderick27
The cost impact of CDI and rCDI
The total healthcare cost for patients who experience CDI is significant, and for those with rCDI, costs are higher compared to patients with a primary CDI episode. Zhang et al Reference Zhang, Prabhu and Marcella3 reported that patients with CDI had $24,205 (95% CI, $23,436–$25,013) higher costs compared to patients without CDI over a 6-month follow-up period. Reference Zhang, Prabhu and Marcella3 In addition, these researchers used PSM a second time to compare patients with primary CDI only to patients with one rCDI episode. Recurrent CDI contributed an additional $10,580 (95% CI, $8,849–$12,446) compared to those with primary CDI only. Yu et al Reference Yu, Baser and Wang24 studied nursing home residents and reported that 2-month follow-up costs were an additional $14,977 for patients with CDI compared to patients without CDI. Medicare paid the majority ($13,277) of this increased cost.
Further evidence concerning the burden of rCDI is provided by Shah et al, Reference Shah, Aitken and Barragan7 who reported that rCDI led to an additional $24,445 in healthcare costs over a 3-month period compared to CDI patients who did not experience a recurrence. Reference Shah, Aitken and Barragan7 A separate single-center study by Rodrigues et al Reference Rodrigues, Barber and Ananthakrishnan5 estimated that the mean total rCDI-associated healthcare costs per patient was $34,104 over a 12-month period, with hospitalizations accounting for a majority of costs. Reference Rodrigues, Barber and Ananthakrishnan5 Furthermore, the number of recurrences is associated with increasing costs. For example, Feuerstadt et al Reference Guh, Mu and Winston13 reported that each additional recurrence substantially increased healthcare costs by as much as $27,159 to $59,973. Reference Guh, Mu and Winston13 Notably, these studies did not report CDI-related costs nor control for potential confounding factors. In studies with 12-month follow-up, it is difficult to confidently attribute the increase in healthcare costs solely to rCDI without controlling for underlying comorbidities. As described above, Nelson et al Reference Nelson, Scott and Boules1 assessed the number of episodes of CDI using Medicare data and found that the mean total all-cause direct costs were $76,024 with no CDI recurrence, $99,348 with 1 recurrent episode, $96,148 with 2 recurrent episodes, and $96,517 with 3 or more CDI recurrent episodes over a 12-month period. Reference Nelson, Scott and Boules1 Also, subgroups with other comorbidities may be particularly affected by CDI. For example, Duhalde et al Reference Duhalde, Lurienne, Wingen-Heimann, Guillou, Buffet and Bandinelli17 reported that among cancer patients with CDI, the additional cost of CDI was $60,159 compared to those without CDI. Reference Duhalde, Lurienne, Wingen-Heimann, Guillou, Buffet and Bandinelli17
Discussion
This systematic review found strong evidence supporting the negative impact of primary and recurrent CDI on healthcare costs and resource use, especially hospital services. Patients who develop CDI have longer lengths of stay, consume more resources, and have higher total costs compared to matched controls. Reference Zhang, Prabhu and Marcella3,Reference Rodrigues, Barber and Ananthakrishnan5,Reference Shah, Aitken and Barragan7 In addition, CDI recurrence further increases all-cause healthcare costs, by estimates of $10,000 to $60,000, compared to a primary CDI episode, and rCDI can potentially be even more costly among patients with significant comorbidities, who are the most vulnerable population for rCDI. Reference Zhang, Prabhu and Marcella3,Reference Duhalde, Lurienne, Wingen-Heimann, Guillou, Buffet and Bandinelli17
Despite the clear impact on healthcare resource utilization, there seemed to be little consensus in how cost data should be recorded and reported among studies. A number of reports used PSM to identify relevant control groups, addressing some of the potential bias observed in real-world analyses. Reference Ma, Brensinger, Wu and Lewis10 However, 2 studies attempted to assess the impact of the number of recurrent episodes on healthcare resources and costs. The presence of a control group in many studies would facilitate comparisons between patients with and without rCDI, and any studies without this comparison should be interpreted with caution. Furthermore, observational studies that do not attempt to control for likely confounders among groups should be framed such that there are many likely factors contributing to additional healthcare utilization beyond CDI. Some of these risk factors are confounders of patients who are already more likely to regularly utilize healthcare across a variety of settings, even further enhancing the risk of rCDI. Reference Enoch, Murray-Thomas and Adomakoh37 Those risk factors include a LOS for the initial hospital admission, a previous admission through the emergency department, as well as comorbidities of hypertension, moderate or severe liver disease, renal disease, dementia, cancer, and use of lipid-lowering therapy. Reference Enoch, Murray-Thomas and Adomakoh37 The strongest predictor of a recurrent CDI episode is a prior history of CDI, signaling that patients with recurrence are uniquely susceptible and should be of particular focus for careful treatment and follow up. Reference Rodrigues, Barber and Ananthakrishnan5,Reference Zilberberg, Shorr, Jesdale, Tjia and Lapane6
Several studies used treatment with antibiotics to identify patients with CDI, but exposure to these agents provides little insight into the patient’s CDI episode. Although vancomycin and fidaxomicin are recommended for primary and recurrent infections, bezlotuxumab is currently the only approved product specifically for preventing future recurrences. Reference Alonso and Mahoney38 In patients with a history of recurrence, bezlotuxumab achieved modest improvement in recurrence rates compared to placebo (32% vs 49%, respectively); the mechanism of action is thought to be mediated through binding of Clostridioides difficile toxin B. Reference Gerding, Kelly and Rahav15
This systematic literature review had several strengths and limitations. Broad search terms and literature searching strategies were incorporated to identify all previous analyses of healthcare resources and costs used in the treatment of CDI. Also, although there have been well regarded healthcare cost analyses conducted in other countries (eg, Europe and Asia), they were excluded here because resource unit costs vary widely between countries. Additionally, the window of time used to identify and define a rCDI episode differed across studies, although most of the data were generally consistent with the definitions put forth by the Infectious Diseases Society of America and Society for Healthcare Epidemiology of America guidelines, which define rCDI as a confirmed episode within 8 weeks of the primary infection, and some claims-based analyses expanded the time window to 12 weeks or more. Reference Page, McKenzie and Bossuyt16,Reference Mollard, Lurienne, Heimann and Bandinelli20,Reference Shrestha, Bime and Taleban21,Reference McDonald, Gerding and Johnson39 Despite this limitation, this systematic review had key strengths of identifying real-world evidence publications with the rigor of comparison groups and methodologies including PSM to demonstrate the economic implications of primary and rCDI to a range of health systems.
In conclusion, CDI is a costly disease, with hospitalizations being the main cost driver. Recurrent CDI is associated with incremental cost increases across all facets of medical care, and more studies are needed to fully understand the burden of rCDI.
Acknowledgments
Barbara McGovern and Brooke Hasson provided excellent feedback in editing the manuscript.
Financial support
This work was supported by Seres Therapeutics.
Conflict of interest
Drs. Pham, Gratie, and Amin are consultants to Seres Therapeutics. In addition, Dr. Amin has served as primary or coinvestigator of clinical trials sponsored by, NeuroRx Pharma, Pulmotect, Blade Therapeutics, Novartis, Takeda, Humanigen, Eli Lilly, PTC Therapeutics, Octapharma, Fulcrum Therapeutics, and Alexion; and as a speaker and/or consultant for BMS, Pfizer, BI, Portola, Sunovion, Mylan, Salix, Alexion, AstraZeneca, Novartis, Nabriva, Paratek, Bayer, Tetraphase, Achaogen, La Jolla, Ferring, Seres, Millennium, PeraHealth, Eli Lilly, Spero, HeartRite, AseptiScope, and Sprightly, not having to do with this study and manuscript. Drs. Malone and Armstrong are consultants to Seres Therapeutics through Strategic Therapeutics, LLC. Seres Therapeutics has an investigational product being studied in the treatment of Clostridioides difficile infections; however, no study assessed in this analysis used a Seres Therapeutics product.
Open access
