Antibiotics and hospital-acquired Clostridium difficile infection: update of systematic review and meta-analysis

Abstract

Objectives

To update the evidence for associations between antibiotic classes and hospital-acquired Clostridium difficile infection (HA-CDI).

Methods

Electronic databases of journal articles, scholarly theses and conference proceedings using subject headings and keywords related to CDI and antibiotic exposure were searched. Observational epidemiological studies measuring associations between antibiotic classes and HA-CDI were eligible for inclusion. Pooled ORs and 95% CIs were calculated using a random effects model. Study factors identified a priori were examined as sources of heterogeneity. The quality of the studies was assessed using the Newcastle-Ottawa Scale.

Results

Of 569 citations identified, 13 case–control and 1 cohort study (15 938 patients) were included. The strongest associations were found for third-generation cephalosporins (OR = 3.20, 95% CI = 1.80–5.71; n = 6 studies; I² = 79.2%), clindamycin (2.86, 2.04–4.02; n = 6; I² = 28.5%), second-generation cephalosporins (2.23, 1.47–3.37; n = 6; I² = 48.4%), fourth-generation cephalosporins (2.14, 1.30–3.52; n = 2; I² = 0.0%), carbapenems (1.84, 1.26–2.68; n = 6; I² = 0.0%), trimethoprim/sulphonamides (1.78, 1.04–3.05; n = 5; I² = 70%), fluoroquinolones (1.66, 1.17–2.35; n = 10; I² = 64%) and penicillin combinations (1.45, 1.05–2.02; n = 6; I² = 54%). The study population and the timing of measurement of antibiotic exposure were the most common sources of heterogeneity. Study quality scored high for seven studies, moderate for six studies and low for one study.

Conclusions

The risk of HA-CDI remains greatest for cephalosporins and clindamycin, and their importance as inciting agents should not be minimized. The importance of fluoroquinolones should not be overemphasized, particularly if fluoroquinolone-resistant epidemic strains of C. difficile are absent.

diarrhoea, antimicrobials, risk factors, healthcare facilities

Introduction

Clostridium difficile is the leading cause of healthcare facility (HCF)-associated diarrhoea, estimated to cost the USA more than $3 billion per year.¹ Susceptibility to infection with C. difficile is induced by exposure to factors that disrupt gut microflora, most usually antibiotics that are commonly used in HCFs. Our earlier systematic review of the published literature up to 2001 found that clindamycin and third-generation cephalosporins were most strongly associated with HCF-associated C. difficile infection (HA-CDI).²

The rates of CDI in industrialized countries have risen with the emergence of the NAP1/RT027 strain in 2002, responsible for outbreaks of severe disease in North America and Europe.^3,4 Although rates of community-associated CDI (CA-CDI) are also increasing worldwide, 70%–80% of cases are associated with exposure to an HCF.⁵ The aim of this study was to evaluate the associations between antibiotic classes and the risk of HA-CDI over the period January 2002 to December 2012.

Methods

A systematic review was conducted in March 2013. The PRISMA statement was used to guide the methodology and reporting of the study.⁶

Search strategy and selection criteria

Searches of the primary literature were conducted using Medline (PubMed, OvidSP) and Embase (OvidSP). Dissertations and theses were searched using WorldCat (www.worldcat.org), ProQuest Dissertations and Theses (PQDT; www.proquest.com), Electronic Theses Online Service (EThOS), Networked Digital Library of Theses and Dissertations (NDLTD; www.ndltd.org) and the National Library of Australia (Trove; www.trove.nla.gov.au). The Conference Proceedings Citation Index was searched using the Web of Science (incorporated 2003). The reference lists of the articles were examined.

All the searches were limited to studies published from 1 January 2002 to 31 December 2012 that reported on human participants. Primary literature searches were conducted using MeSH and Emtree terms and keywords in all fields: Clostridium difficile, diarrhea, diarrhoea, colitis, antibacterial agents, anti-infective agents, antibiotic, antimicrobial, case–control studies, cohort studies, retrospective studies and prospective studies (see the full search strategy in the Supplementary data at JAC Online). Searches of theses and dissertations used the terms Clostridium difficile, epidemiology and public health. Conference abstracts were searched using the terms Clostridium difficile, antibiotic, antimicrobial and risk factors.

Inclusion and exclusion criteria were developed using the Patients, Interventions, Comparisons and Outcomes approach.⁷ Studies were included in the review if they met the following criteria: observational studies conducted among hospital inpatients, measurement of antibiotic exposure, inclusion of a comparison group and outcome of HA-CDI. We excluded studies of cases only, studies assessing risk factors for severe disease, relapsing CDI or asymptomatic colonization and studies addressing CA-CDI or that did not properly exclude cases of community-acquired infection. Studies that did not examine specific antibiotics or antibiotic classes were also excluded. The review was limited to English-language studies, but non-English-language articles were included in searches and their abstracts were assessed for eligibility. Solely paediatric study populations were also excluded.

Screening and data abstraction

The abstracts were screened to eliminate irrelevant reports. Full-text articles of the remaining studies were assessed for eligibility. The authors of one study for which insufficient detail was presented were contacted; they were unable to provide the information and the study was excluded.⁸ Data were extracted by a single reviewer (C. S.). Study characteristics (citation, location, study period, study design, study population, case definition, antibiotic exposure, matching and adjustment method) were recorded. Data on the number of subjects (cases, non-cases, exposed or unexposed), unadjusted and adjusted effect sizes and 95% CIs were entered into an Excel spreadsheet. Antimicrobial agents were classified according to their therapeutic class: penicillins, cephalosporins, tetracyclines, trimethoprim/sulphonamides, macrolides, quinolones, aminoglycosides, lincosamides (clindamycin) and carbapenems. In studies that reported more than one exposure within each class, effect estimates were combined by taking the weighted average of the log ORs, with inverse variance weights.

Quality assessment

The quality of included studies was assessed based on the Newcastle-Ottawa Scale⁹ (see Table S1, available as Supplementary data at JAC Online), modified to represent issues specific to CDI.¹⁰ Study power was calculated from unadjusted effect sizes.

Statistical analysis

Analyses were undertaken using STATA v12.1 using the ‘metan’, ‘metabias’ and ‘metareg’ packages.¹¹

A pooled random effects analysis using the DerSimonian-Laird approach¹² was used to assess the risk of HA-CDI associated with antibiotic classes relative to no exposure. The most fully adjusted OR was used. A secondary analysis considered the risk associated with penicillin and cephalosporin subclasses. Estimates were weighted by the inverse of the variance. Heterogeneity of pooled effects was assessed using the I² statistic;¹³ a P value <0.1 was defined as the presence of heterogeneity.⁷

Study factors identified a priori were explored as sources of heterogeneity using meta-regression consisting of: location (North America, Europe, Australia), study period (up to and including 2002, post-2002), whether it was conducted during a CDI outbreak, the time period of measurement of antibiotic exposure (during admission only, prior to admission and during admission), study quality (low, medium or high), number of key confounders considered (≤3, >3), the study sample (antibiotic exposed versus all inpatients) and the diagnostic test used [enzyme immunoassay (EIA) or other]. Sensitivity analyses were undertaken to explore causes of heterogeneity identified from meta-regression (P < 0.10).

Publication bias was assessed by generating funnel plots and tested used Egger's asymmetry test.

Results

A total of 569 non-duplicate records were screened for eligibility; 14 met the inclusion criteria, covering 15 938 patients (Figure 1).

Figure 1.

PRISMA flow diagram.

Description of included studies

There were 13 case–control studies and one cohort study (Table 1). Ten studies were conducted in North America, three in Europe and one in Australia, spanning the period 1996–2009; eight studies were conducted after 2002 and three of these were conducted during CDI outbreaks.^4,14,15 The studies varied in size from 15 to 1142 cases of CDI. Most studies used hospital inpatients as their study population; four studies used an antibiotic exposed subpopulation.^14,16–18 Although studies specifically based on paediatric populations were excluded, one study¹⁹ included patients aged >2 years.

Table 1.

Characteristics of 14 case–control and 1 cohort study assessing HA-CDI

Study citation	Data source	Study period	Study population	Age, years cases/ controls, mean (SD)	Male, %	Case definition	Control definition	Matching	Adjustment method	Antibiotic exposure	Definition of HCF-acquisition	N cases/ non-cases
Asha 200619	Leeds Teaching Hospitals, UK	Jun 2001– Apr 2002	inpatients >2 years	N/A	N/A	Diarrhoea and positive test for C. difficile toxin (cell culture cytotoxicity)	1. Diarrhoea and negative assay for C. difficile toxin	age, gender, date, ward location	conditional logistic regression	source not stated; 28 days prior to index date	≥3 days after admission	503/118a
							2. Inpatients without diarrhoea (antibiotics previous 28 days)		1: LOS, feeding tubes
									2: LOS, antacids, feeding tubes, other antibiotics
Baxter 200816	Kaiser Permanente of Northern California, USA	Jan 1998– Dec 2005	inpatients; antibiotics previous 60 days	69/68	48	Diarrhoea and positive test for C. difficile toxin; no history of CDI for 1 y [toxin A/B EIA (Meridian Premier)]	inpatients with no positive test result for C. difficile	hospital, date, DRG, LOS	conditional logistic regression.	pharmacy database; 60 days prior to index date	≥3 days after admission	1142/3351
	16 hospitals, 106–330 beds								Age, sex, days in hospital, other antibiotics, proton pump inhibitors, medical costs, comorbidity
Hensgens 201222	9 hospitals, Netherlands	Mar 2006 –May 2009	inpatients	62 (21)/59 (21)	52	Diarrhoea and positive test for C. difficile toxin (various EIA)	1. Diarrhoea and negative test for C. difficile toxin	hospital, ward, date	conditional logistic regression.	electronic medical information system; 3 months prior to index date	≥2 days after admission or physicians request	337/337a
							2. inpatients without diarrhoea		Age, sex, comorbidity, other antibiotics
Kallen 200914	Community hospital, USA	Jun 2005– May 2007	inpatients; antibiotics previous 3 months	77 (31–97)/ 68 (2–95)b	44	Diarrhoea and positive test for C. difficile toxin [toxin A/B EIA (Meridian Premier)]	patients hospitalized for at least 48 h; no CDI up to 30 days post discharge	none	logistic regression.	pharmacy records; 3 months prior to index date	≥48 h after admission	100/100
									Age, comorbidity, LOS, PPI use, NG tube use, tube feeding, other antibiotics
Loo 20054	12 hospitals, Quebec, Canada	Jan–Jun 2004	inpatients	75 (66–72)c	51	Diarrhoea and positive test for C. difficile toxin and/or a clinical diagnosis (various EIA and culture cytotoxicity)	inpatients admitted and discharged during the same period; no known history of CDI	age, Charlson Index, date, ward, length of time at risk	conditional logistic regression.	source not reported; 6 weeks prior to diagnosis	≥72 h after admission or within 1 month of previous admission	237/237
	256–705 beds								age, sex, other antibiotics, no. of days at risk of CDI, comorbidity, chemotherapy, PPIs, histamine H2-blockers, enteral feeding
McCusker 200324	Veterans Affairs Maryland Healthcare System, USA	Jan–Jun 2001	inpatients	72 (66–79) / 67 (56–79)c	98	Diarrhoea and positive test for C. difficile toxin [toxin A EIA (Wampole))	patients admitted for at least 48 h during the same 6 month period as the cases; no history of CDI and no receipt of oral metronidazole	unit of admission, length of time at risk	conditional logistic regression.	electronic medical records; 6 weeks prior to index date	≥72 h after admission	30/60
	778 beds over four sites								other antibiotics, days at risk
Minson 200720	615 bed non-academic urban medical facility, Nashville, USA	Oct 2005– Mar 2006	inpatients > 18 years	69/63	41	Diarrhoea and a positive test for C. difficile toxin [toxin A/B EIA (Meridian Premier)]	inpatients with no history of CDI (or metronidazole/ vancomycin use during admission)	age, gender, length of time at risk, LOS in ICU	none	source not reported; previous 8 weeks	≥72 h after admission	27/54
Modena 200517	Temple University Hospital, Philadelphia, USA	Aug 2003– Jun 2004	inpatients; ≥5 days antibiotics	57 (18)/60 (17)	47	Diarrhoea and positive test for C. difficile toxin (no history of CDI) [toxin A/B EIA (Wampole)]	inpatients with other infections	none	logistic regression.	medical and pharmacy records during index admission	antibiotics >5 days	50/200
									ICU, LOS
Muto 200525	600 bed tertiary care teaching hospital, Pittsburg, USA	Jan 2000 – Apr 2001	inpatients	64 (17–95)/ 59 (16–93)b	52	signs and symptoms of CDI and positive test for C. difficile toxin (cell culture cytotoxicity)	inpatients with no positive test for C. difficile	date of admission, service, LOS	conditional logistic regression.	electronic medical records; 28 days prior to diagnosis	≥72 h after admission	203/203
									age, diabetes mellitus, transplantation, H2 blockers, PPIs
Polgreen 200715	Small, rural hospital, Iowa, USA	Jan 2004– Apr 2004	inpatients	83b	39	Diarrhoea and positive test for C. difficile toxin [toxin A EIA (Becton Dickinson)]	inpatients admitted during the study period	age, sex	conditional logistic regression.	medical records; 6 weeks prior to admission	≥48 h after admission	15/45
									LOS
Sundram 200921	520 bed district general hospital, Surrey, England	Mar 2006– Mar 2007	inpatients (excluding paediatric)	81(9)/ 80 (9)	49	Diarrhoea and positive test for C. difficile toxin [toxin A/B EIA (Meridian Premier)]	inpatients without diarrhoea and no history of CDI	age, sex, ward, ASA score, LOS	conditional logistic regression.	medical records; 6 weeks prior to onset	>48 h after admission	97/97
									Enteral feeding, any antibiotic exposure
Thomas 200323	560 bed tertiary teaching hospital, Perth, Australia	Jan 1996 – Oct 1998	inpatients	72 (15–97)b	46	Diarrhoea and positive test for C. difficile toxin (cell culture cytotoxicity or culture)	inpatients without diarrhoea and no history of CDI	age, gender, date	conditional logistic regression	medical records; index admission	>48 h after admission (or <48 h if previous admission within 7 days)	149/310
									GI procedures, comorbidity, LOS, other antibiotics
Vesta 200526	225 bed teaching hospital, 400 bed community hospital, 400 bed Veterans Affairs hospital, Oklahoma, USA	Feb 2000– Dec 2002	inpatients	56 (18–93)d	57	Diarrhoea and positive test for C. difficile toxin (no details given)	inpatients with negative C. difficile toxin test	hospital, sex, age, LOS	none	medical and pharmacy records; 42 days prior to diagnosis	≥72 h after admission	72/72
Stevens 201018 e	Tertiary care academic hospital, Rochester, NY, USA	Jan 2005– Dec 2005	inpatients ≥18 years; ≥2 days of antibiotics	59 (18–102)b	49	Positive test for C. difficile toxin; up to 60 days post discharge [toxin A/B EIA (Meridian Premier))	N/A	N/A	marginal Cox proportional hazards	electronic pharmacy records; index admission	>2 days after admission and no CDI in 60 days prior to admission	240/7552
									age, comorbidity, GI procedures, units, LOS, antacids, laxatives, enemas

N/A, not applicable; LOS, length of stay; DRG, diagnostic-related group; PPI, proton pump inhibitor; NG, nasogastric; ICU, intensive care unit; ASA, American Society of Anesthesiologists; GI, gastrointestinal.

^aThe numbers of non-cases relate to control group 2.

^bMedian (range).

^cMedian (IQR).

^dMean (range).

^eCohort study.

All studies included symptomatic cases with a positive C. difficile assay. The most common diagnostic test was toxin A/B EIA in six studies.^{14,16–18,20,21} Definitions of HCF acquisition were >48 h (n = 6)^{14,15,18,21–23} and >72 h after admission (n = 7).^{4,16,19,20,24–26} Two studies incorporated previous contact with an HCF in their definition.^4,23

Most studies (12/14) used asymptomatic controls. Three studies measured exposure during admission only,^17,18,23 while the remainder measured antibiotics received prior to and during the admission (28 days to 3 months). Only two studies evaluated all nine main antibiotic classes of interest.^16,25 Ten studies reported various associations for subclasses of penicillins (n = 4),^14,19,21,24 cephalosporins (n = 1)²² or both (n = 5).^{4,16,20,23,25}

Study quality was scored high for seven studies^{14,16,18,22–25} (see Table S2 available as Supplementary data at JAC Online). Studies with low to moderate scores experienced limitations across all three domains. All studies took into account at least some important confounders. Six studies addressed four or more confounders but not for every association examined.^{4,14,16,22,23,25} The most common confounders assessed were age, sex, length of stay and exposure to other antibiotics. Only five studies considered comorbidities.^{4,16,18,22,23} Studies commonly failed to report details of a priori sample size calculations, and six studies^{15,17,19–21,26} either had insufficient power to detect any differences or did not report sufficient information to allow calculation.

Pooled effects

Figure 2 presents the pooled effects for exposure to the nine main antibiotic classes from case–control studies using non-diarrhoeal controls. Overall exposure to antibiotics was associated with a 60% (95% CI = 1.31–1.87) increased risk of CDI, but there was a substantial variation between antibiotic classes. The strongest associations were seen for clindamycin (OR = 2.86, 95% CI = 2.04–4.02), cephalosporins (OR = 1.97, 95% CI = 1.21–3.23), carbapenems (1.84, 95% CI = 1.26–2.68), quinolones (fluoroquinolones, OR = 1.66, 95% CI = 1.17–2.35) and trimethoprim/sulphonamides (OR = 1.78, 95% CI = 1.04–3.05). There was no association with aminoglycosides, tetracyclines or macrolides. There was evidence of heterogeneity in the pooled ORs for all classes except clindamycin (I² = 28.5%, P = 0.22), carbapenems (I² = 0.0%, P = 0.45) and tetracyclines (I² = 48.5%, P = 0.14).

Figure 2.

Antibiotic classes and the risk of hospital-acquired CDI.

Stratification by antibiotic subclass

There was a 50% (95% CI = 1.05–2.24) increased risk of CDI associated with penicillin combination antibiotics but not for other penicillin subgroups (Figure 3). Second-, third- and fourth-, but not first-generation cephalosporins were also associated with two to three times the risk of CDI (Figure 4); however, heterogeneity persisted, particularly for third-generation cephalosporins (I² = 79.2%, P < 0.001).

Figure 3.

Penicillins and the risk of hospital-acquired CDI.

Figure 4.

Cephalosporins and the risk of hospital-acquired CDI.

Meta-regression and sensitivity analyses

The study population was the most common source of heterogeneity (Table 2), seen for trimethoprim/sulphonamides, third-generation cephalosporins and macrolides. Excluding studies based on antibiotic-exposed inpatients reduced the heterogeneity, except for macrolides, for which further analysis indicated that studies conducted after 2002 or those using antibiotic-exposed participants had ∼30% smaller effect sizes than other studies.

Table 2.

Sensitivity analysis to explore heterogeneity

Study subgroup	Exposure	No. of studies	No. of patients	OR (95% CI)	Heterogeneity (I2), % (P value)
Study population: all inpatients	Third-generation cephalosporins	5	1757	4.04 (2.81–5.81)	0.0% (0.72)
	Trimethoprim/sulphonamides	4	1283	2.22 (1.53–3.23)	0.0% (0.59)
	Macrolides	6	2378	1.64 (0.91–2.93)	77% (0.001)
Antibiotic exposure assessment: prior to hospitalization	Fluoroquinolones	9	6335	1.39 (1.09–1.76)	29% (0.19)
	Aminoglycosides	5	5791	0.89 (0.70–1.14)	0.0% (0.72)
HA-CDI definition >72 h	Penicillin combinations	4	5463	1.20 (0.99–1.47)	0.0% (0.74)
HA-CDI definition >48 h	Penicillin combinations	2	659	3.20 (1.75–5.91)	0.0% (0.35)
Toxin EIA test	First-generation cephalosporins	3	4911	1.04 (0.88–1.24)	0.0% (0.90)
Studies conducted post-2002	Second-generation cephalosporins	3	892	4.08 (2.27–7.34)	0.0% (0.68)
Studies conducted up to 2002	Second-generation cephalosporins	3	5358	1.72 (1.27–2.31)	7.8% (0.34)
Post-2002 and all inpatients	Macrolides	3	892	1.00 (0.56–1.79)	0.0% (0.56)

The timing of antibiotic exposure measurement was a significant heterogeneity source for fluoroquinolones and aminoglycosides. The risk associated with fluoroquinolones fell from 66% to 39% after excluding one study that measured antibiotic exposure during hospital admission only.²³

The strongest predictor of heterogeneity for penicillin combinations was the definition of hospital acquisition: there was no significant association when analyses were restricted to 4/6 studies that used a definition of >72 h compared with studies using a definition of >48 h (Table 2).

The choice of diagnostic test was weakly associated with heterogeneity for first-generation cephalosporins, where studies that used toxin EIA tests had smaller effect sizes (OR = 1.04, 95% CI = 0.88–1.24) than studies using other methods (2.08, 1.37–3.17, n = 2 studies). For second-generation cephalosporins, there were higher associations for studies conducted after 2002 (4.08, 2.27–7.34).

Diarrhoea controls

Three studies included symptomatic patients who tested negative for C. difficile as controls. A calculation of crude effect sizes found smaller associations for the toxin-negative than the non-diarrhoeal control group in one study.²² Asha et al.¹⁹ reported a 4-fold increased risk associated with ‘broad-spectrum’ cephalosporins (OR = 3.8, 95% CI = 3.0–4.6). Vesta et al.²⁶ provided descriptive information only and, due to the matched design, precluded the calculation of effect estimates.

Cohort studies

One cohort study, consisting of ∼8000 adult inpatients exposed to antibiotics and followed up until 60 days post-discharge, was included.¹⁸ Independent associations for fluoroquinolones [hazard ratio (HR) = 4.05, 95% CI = 2.75–5.97], third-/fourth-generation cephalosporins (HR = 3.12, 95% CI = 1.85–5.25), β-lactamase inhibitor combination penicillins (HR = 2.25, 95% CI = 1.46–3.48) and trimethoprim/sulphonamides (HR = 2.03, 95% CI = 1.19–3.47) were reported. Weak associations were observed for clindamycin (HR = 1.92, 95% CI = 0.84–4.40) and macrolides (HR = 1.56, 95% CI = 0.75–3.25), and no association was apparent for aminoglycosides (HR = 0.88, 95% CI = 0.26–2.95).

Non-English-language studies

Five out of 27 non-English language studies were eligible for inclusion at abstract screening (Table S3 available as Supplementary data at JAC Online).^27–31 Of these, full texts of three articles were available, and two were eligible for inclusion following a full text review. The findings were generally consistent with those of studies published in English, with the strongest associations reported for clindamycin³¹ and third-generation cephalosporins.²⁸

Publication bias

For studies reporting results for the main antibiotic classes, the funnel plot (Figure S1 available as Supplementary data at JAC Online) showed little evidence of publication bias. The result of Egger's test of small-study effects was not significant (P = 0.57).

Discussion

A systematic review and meta-analysis was undertaken to summarize the evidence for associations between antibiotic classes and the risk of HA-CDI. The findings indicate that third-generation cephalosporins remain the strongest antibiotic risk factor. The modest association seen for fluoroquinolone antibiotics is not surprising as they are more specifically related to C. difficile infections with the NAP1/RT027 fluoroquinolone-resistant epidemic strain.³²

Since our earlier review, the quality of studies has improved, reducing the threat of bias. However, four studies restricted their study populations to inpatients with recent antibiotic exposure, which are not representative of all HA-CDI cases,^14,16–18 and one study used toxin-negative controls, which do not represent the source population of cases.²⁶ Bias relating to diagnostic suspicion remains problematic if those exposed to antibiotics are more likely to be tested for C. difficile; guidelines that recommend testing all hospitalized patients who develop diarrhoea should be followed.³³ There was a variation in the definition of HA-CDI, with only a small number of studies including recent contact with HCFs,^4,23 according to guidelines,³⁴ and 27 studies were excluded from the review due to insufficient information regarding the acquisition of CDI.

There is a lack of consensus regarding the appropriate time window to measure antibiotic exposure giving rise to heterogeneity. Recent studies suggest the greatest risk is in the first 30 days but that it remains increased for up to 90 days.^22,35 Furthermore, the number of antibiotics, dosage and duration of antibiotic exposure have all been previously identified as risk factors for CDI.³⁶ Residual confounding remains a potential issue as only a few studies took into account a sufficient number of confounding factors.

This study provides the most up-to-date and systematic synthesis of the literature in relation to the risk of HA-CDI associated with antibiotics. However, the results should be interpreted with caution as the number of studies contributing to any one analysis ranged from three to 10. Our study is limited by the availability of a single reviewer to select, extract and analyse the data. Continued improvements in the conduct of observational studies of CDI would be gained by following recommendations from the STROBE initiative.³⁷ The adoption of recently developed guidelines for the diagnosis and surveillance of CDI in future studies will also improve the quality of CDI epidemiological research.

Funding

The study was investigator-driven and was carried out as part of our routine work.

Transparency declarations

None to declare.

Author contributions

C. S. designed and planned the study, undertook the literature searches, data collection, data analysis, data interpretation and prepared the manuscript. T. V. R. contributed to the design of the study, interpreted the data and contributed to the writing of the manuscript.