Risk Factors for Acquisition and Loss of Clostridium difficile Colonization in Hospitalized Patients

Abstract

Asymptomatic colonization may contribute to Clostridium difficile transmission. Few data identify which patients are at risk for colonization. We performed a prospective cohort study of C. difficile colonization and risk factors for C. difficile acquisition and loss in hospitalized patients. Patients admitted to medical or surgical wards at a tertiary care hospital were enrolled; interviews and chart review were performed to determine patient demographics, C. difficile infection (CDI) history, medications, and health care exposures. Stool samples/rectal swabs were collected at enrollment and discharge; stool samples from clinical laboratory tests were also included. Samples were cultured for C. difficile, and the isolates were tested for toxins A and B and ribotyped. Chi-square tests and univariate logistic regression were used for the analyses. Two hundred thirty-five patients were enrolled. Of the patients, 21% were colonized with C. difficile (toxigenic and nontoxigenic) at admission and 24% at discharge. Ribotype 027 accounted for 6% of the strains at admission and 12% at discharge. Of the patients colonized at admission, 78% were also colonized at discharge. Cephalosporin use was associated with C. difficile acquisition (47% of patients who acquired C. difficile versus 25% of patients who did not; P = 0.03). β-lactam–β-lactamase inhibitor combinations were associated with a loss of C. difficile colonization (36% of patients who lost C. difficile colonization versus 8% of patients colonized at both admission and discharge; P = 0.04), as was metronidazole (27% versus 3%; P = 0.03). Antibiotic use affects the epidemiology of asymptomatic C. difficile colonization, including acquisition and loss, and it requires additional study.

INTRODUCTION

Clostridium difficile infection (CDI) is the most common infectious cause of hospital-associated diarrhea, and while most CDI infections are mild, severe CDI can lead to outcomes, such as toxic megacolon, colectomy, and death (1–4). Current estimates of the excess health care costs associated with CDI in the United States are $4.8 billion per year (5). The emergence of the epidemic NAP1/BI/027 strain of C. difficile (6–8) and increased infection rates (9, 10) have generated a renewed interest in this pathogen.

Despite this interest in CDI, optimal methods for preventing CDI remain poorly understood. Traditionally, symptomatic CDI patients have been considered the primary reservoir for C. difficile transmission because they shed more C. difficile in their stool than asymptomatically colonized patients (11–13). However, recently published studies indicate only one-third or fewer of new CDI cases in the hospital setting can be attributed to transmission from another CDI case (14–17). It is possible that asymptomatic carriers are an important source of C. difficile transmission (18, 19). Curry et al. (20) found that the percentage of CDI cases related to symptomatic CDI patients and that related to asymptomatic C. difficile carriers was almost identical (30% versus 29%, respectively), even though most patients were not screened for asymptomatic carriage (20, 21). Asymptomatic colonization is more common than symptomatic CDI (22), and these patients may represent a previously underappreciated reservoir for transmission.

Given the potential role for asymptomatic carriers in C. difficile transmission and CDI, it is important to understand the epidemiology of asymptomatic colonization. Several previous studies have examined risk factors for C. difficile colonization at hospital admission (18, 23–27), but risk factors specific for C. difficile acquisition during hospitalization, in the absence of CDI, have been less commonly studied (18, 27, 28). To our knowledge, no studies have reported risk factors for the loss of C. difficile colonization while hospitalized. We performed a prospective cohort study of patients admitted to medical and surgical wards to determine risk the factors for acquisition of C. difficile colonization and loss of C. difficile colonization during hospitalization.

MATERIALS AND METHODS

This study was performed at Barnes-Jewish Hospital (BJH), a 1,250-bed tertiary care hospital in St. Louis, MO, from July 2010 through July 2012. Adult patients admitted to general medical or surgical wards with an anticipated length of stay of >48 h were eligible for enrollment. Patients were excluded from the study if they were pregnant, had diarrhea on admission, or were deemed to be a poor study candidate (e.g., were unable to roll over for a rectal swab or provide informed consent). The Washington University Human Research Protection Office approved this study.

Patients were interviewed at study enrollment to determine demographics, comorbidities, recent health care exposures, medications received in the previous 90 days, and history of CDI. Patient interviews were supplemented with a chart review. During the patient hospitalizations, study personnel followed the patients to determine white blood cell (WBC) counts and maximum temperature within 48 h of admission stool collection, medication exposures, and C. difficile testing and/or CDI diagnosis. Study personnel contacted participants via phone 60 days after discharge to determine if the patient had been diagnosed with CDI.

Specimen collection and microbiological analysis.

Each patient's first stool specimen after enrollment was collected for analysis. Rectal swabs were obtained if a patient was unable to provide a stool specimen within 48 h of admission. Diarrheal stool specimens sent to the BJH laboratory for C. difficile testing from study subjects were also collected. A final stool sample or rectal swab was collected as close to discharge as possible.

The microbiological examination methods used have been published elsewhere (23, 29, 30). Briefly, specimens were stored at −30°C until cultured. Stool/swab specimens were heat shocked at 80°C for 10 min, followed by inoculation into cycloserine-cefoxitin mannitol broth with taurocholate lysozyme cysteine (CCMB-TAL broth; Anaerobe Systems, Morgan Hill, CA). The broths were incubated anaerobically at 35°C for up to 7 days. When growth was detected, 1 ml of broth was centrifuged, and the pellet was plated to prereduced tryptic soy agar (TSA) II with 5% sheep blood (blood agar plate [BAP]) (BBL BD and Co., Sparks, MD).

The culture methods used have been shown to detect as little as 10 CFU of C. difficile per gram of stool (29). C. difficile identification was based on colony morphology, Gram stain, and biochemical testing. C. difficile isolates were tested for toxin A and B production and glutamate dehydrogenase (C. diff Quik Chek; Techlab, Blacksburg, VA) after incubation in brain heart infusion broth. PCR ribotyping was performed on all C. difficile isolates (30), and the ribotyping banding patterns were analyzed using the DiversiLab Bacterial BarCodes software. A similarity of ≥95% was required for isolates to be considered identical. All unique strains were compared with the Cardiff-European Center for Disease Control (ECDC) collection of C. difficile strains; a Washington University (WU) strain number was assigned to unique strains that did not match any in the Cardiff-ECDC collection. All C. difficile isolates were also tested by PCR for the presence of tcdA and tcdB and binary toxin genes by PCR, as previously described (23, 31).

Statistical analyses.

Only study subjects with at least two stool or rectal swab specimens collected were included for analysis. Medications were classified as those received in the 90 days before study enrollment or after enrollment but before the first C. difficile-positive specimen collection or discharge specimen collection date. There were no differences noted in patient characteristics, medication exposures, or colonization acquisition/loss between toxigenic and nontoxigenic C. difficile; therefore, analyses ignored the toxin status of the C. difficile isolates. Chi-square analyses (Pearson chi-square and Fisher's exact test, as appropriate) and univariate logistic regression were used to determine univariate risk factors for acquisition or loss of C. difficile colonization while hospitalized. Continuous data were assigned to clinically relevant categories. All statistical tests were two-tailed, and a P value of ≤0.05 was considered significant. Analyses were performed with SPSS for Windows, version 21 (Armonk, NY).

RESULTS

Enrollment and demographics.

Two hundred forty-six patients enrolled in the study had ≥2 specimens collected. Nine patients were excluded from analyses because they did not have a discharge stool specimen available, and two were excluded because they were subsequently determined to have had CDI on admission; thus, 235 patients were included in the analyses. A total of 559 specimens were collected from these 235 patients. The majority of the patients were admitted to the medicine ward (92% [n = 215]). Most patients (68% [n = 160]) were admitted from home, and 29% (n = 68) were admitted from an outside hospital. Females comprised 53.2% (n = 125) of the patients, and 74% (n = 173) were white.

C. difficile colonization at admission.

Forty-nine (21%) patients were colonized with C. difficile on admission, 37 (76%) of whom were colonized with a toxigenic isolate (Table 1). Thirty-eight patients (78%) were colonized at both their admission and discharge stool specimens. There were 22 unique strains isolated, the most common of which was ribotype 014/020 (29% of all admission strains). Ribotype 027 was the fifth most common strain (6% of all admission strains). One strain (WU #67) was isolated on an admission specimen but not on a discharge specimen.

TABLE 1.

C. difficile strains isolated on admission and discharge (n = 33 strains)

Strain	No. (%) isolated on:		Strain positive for:
	Admission	Discharge	tcdA and/or tcdB?	Binary toxin?
014/020	14 (29)	10 (18)	Yes	No
106/174	4 (8)	5 (9)	Yes	No
010	3 (6)	2 (4)	No	No
015/046	3 (6)	5 (9)	Yes	Yes
027	3 (6)	7 (12)	Yes	Yes
001	2 (4)	0	Yes	No
077	2 (4)	1 (2)	Yes	No
053	0	3 (5)	Yes	No
017	0	2 (4)	Yes	No
WU #20	2 (4)	1 (2)	No	No
WU #38	1 (2)	2 (4)	No	No
WU #42	2 (4)	1 (2)	Yes	Yes
WU #45	1 (2)	2 (4)	No	No
WU #46	0	2 (4)	No	No
WU #48	2 (4)	1 (2)	No	No
Other strains	10 (20)	13 (23)	11/18 (61%)	0/18 (0%)

C. difficile colonization at discharge.

Fifty-seven (24%) patients were colonized with C. difficile at discharge, 42 (74%) of whom were colonized with a toxigenic isolate (Table 1). There were 27 unique ribotypes isolated, the most common of which was ribotype 014/020 (18% of all discharge isolates). Ribotype 027 was the second most commonly identified ribotype (12% of discharge strains). Of the ribotypes identified in an admission specimen, 74% (20 of 27) were also present in a discharge specimen. The ribotypes identified only in discharge specimens were 053, 017, WU #46, 05/053, WU #49, WU #58, and WU #69.

Risk factors for C. difficile acquisition.

One hundred eighty-six patients were not colonized with C. difficile at admission. Nineteen of these patients (10%) acquired C. difficile while hospitalized, 13 (68%) of whom acquired toxigenic strains. The risk factors for C. difficile acquisition are given in Table 2. Patients newly colonized at discharge were more likely than patients not colonized at discharge to have received a cephalosporin during admission (47% versus 25%, respectively; P = 0.03). For several variables, there was a trend toward significance. Of the patients colonized at discharge, 16% were admitted for >14 days, compared to 5% of patients not colonized at discharge (P = 0.09). Seventy-nine percent of patients colonized at discharge had one or more admissions to acute care hospitals in the previous 90 days compared with 56% of patients not colonized at discharge (P = 0.08). More patients colonized at discharge had >14 days since last inpatient discharge than patients not colonized at discharge (32% versus 20%, respectively; P = 0.08).

TABLE 2.

Risk factors for C. difficile acquisition while hospitalized (n = 186 patients not colonized on admission)

Variablea	No. (%) of patients:		P value
	Colonized with C. difficile on discharge stool (n = 19)	Not colonized with C. difficile on discharge stool (n = 167)
Surgical ward (vs medical)	3 (16)	11 (7)	0.16
Age ≥65 yr	9 (47)	53 (32)	0.17
Nonwhite race	4 (21)	48 (29)	0.60
Female	12 (63)	85 (51)	0.31
History of CDI	3 (16)	10 (6)	0.13
Health care worker	1 (5)	17 (10)	0.70
Lives with health care worker	3 (16)	17 (10)	0.44
Spends ≥2 h/wk visiting health care facility	1 (5)	15 (9)	1.00
Length of hospital stay (days)
1–2	2 (11)	29 (17)	Reference
3–7	10 (53)	97 (58)	0.62
8–14	4 (21)	33 (20)	0.53
>14	3 (16)	8 (5)	0.09
Admitted from:
Home	12 (63)	115 (69)	Reference
Health care facility	7 (37)	52 (31)	0.61
Reason for admission
Infection	5 (26)	44 (26)	Reference
Exacerbation of chronic condition	8 (42)	59 (35)	0.77
New medical/surgical problem	6 (32)	64 (38)	0.76
Comorbidities
Diabetes	7 (37)	58 (35)	0.86
Congestive heart failure	3 (16)	32 (19)	1.00
Liver disease	1 (5)	25 (15)	0.48
Chronic renal insufficiency	2 (11)	21 (13)	1.00
Chronic lung disease	7 (37)	34 (20)	0.10
HIV	0 (0)	5 (3)	1.00
Solid organ transplant	0 (0)	11 (7)	0.61
Stem cell transplant	1 (5)	0 (0)	0.10
Solid malignancy	1 (5)	24 (14)	0.48
Hematologic malignancy	1 (5)	3 (2)	0.35
Other immunocompromised	3 (16)	21 (13)	0.72
Inflammatory bowel disease	1 (5)	5 (3)	0.48
Fever at admission	1 (5)	18 (11)	0.70
Usual no. of bowel movements per day
<1 or ostomy bag	4 (21)	41 (25)	Reference
1–2	12 (63)	111 (67)	0.87
≥3	3 (16)	15 (9)	0.38
Procedure(s) in previous 90 days
Upper endoscopy	0 (0)	7 (4)	1.00
Lower endoscopy	0 (0)	5 (3)	1.00
Surgery	4 (21)	15 (9)	0.11
Any HCF exposure	19 (100)	145 (87)	0.14
No. of admissions to acute care hospitals in previous 90 days
0	4 (21)	74 (44)	Reference
≥1	15 (79)	93 (56)	0.08
No. of days since last inpatient discharge
No hospitalizations	4 (21)	74 (44)	Reference
0–14	9 (47)	59 (35)	0.10
>14	6 (32)	34 (20)	0.08
No. of outpatient clinic visits in previous 90 days
0	5 (26)	37 (22)	Reference
1–2	5 (26)	56 (34)	0.53
3–5	6 (32)	48 (29)	0.90
>5	3 (16)	26 (16)	0.84
Other health care exposures in previous 90 days
Inpatient PT/OT rehab	1 (5)	6 (4)	0.55
Reside in nursing home/LTCF	0 (0)	4 (3)	1.00
Reside in subacute care facility	0 (0)	1 (1)	1.00
Outpatient hemodialysis	1 (5)	5 (3)	0.50
Antimicrobials
Aminoglycosides
90 days before admit/enroll	0 (0)	5 (3)	1.00
After enroll, before index date	1 (5)	1 (1)	0.19
β-lactam–β-lactamase inhibitor combination
90 days before admit/enroll	3 (16)	19 (11)	0.48
After enroll, before index date	1 (5)	20 (12)	0.70
Carbapenem
90 days before admit/enroll	1 (5)	8 (5)	1.00
After enroll, before index date	1 (5)	7 (4)	0.59
Cephalosporin
90 days before admit/enroll	3 (16)	28 (17)	1.00
After enroll, before index date	9 (47)	41 (25)	0.03
Clindamycin
90 days before admit/enroll	0 (0)	4 (2)	1.00
After enroll, before index date	1 (5)	8 (5)	1.00
Daptomycin
90 days before admit/enroll	0 (0)	1 (1)	1.00
After enroll, before index date	1 (5)	5 (3)	0.48
Doxycycline
90 days before admit/enroll	0 (0)	8 (5)	1.00
After enroll, before index date	0 (0)	7 (4)	1.00
Linezolid
90 days before admit/enroll	0 (0)	2 (1)	1.00
After enroll, before index date	1 (5)	1 (1)	0.19
Macrolide
90 days before admit/enroll	1 (5)	15 (9)	1.00
After enroll, before index date	4 (21)	17 (10)	0.24
Metronidazole
90 days before admit/enroll	2 (11)	11 (7)	0.63
After enroll, before index date	1 (5)	16 (10)	1.00
Penicillins
90 days before admit/enroll	0 (0)	2 (1)	1.00
After enroll, before index date	0 (0)	2 (1)	1.00
Fluoroquinolone
90 days before admit/enroll	4 (21)	32 (19)	0.77
After enroll, before index date	2 (11)	20 (12)	1.00
TMP-SMX
90 days before admit/enroll	1 (5)	11 (7)	1.00
After enroll, before index date	0 (0)	9 (5)	0.60
Vancomycin i.v.
90 days before admit/enroll	5 (26)	38 (23)	0.78
After enroll, before index date	6 (32)	40 (24)	0.57
Other medications
Proton pump inhibitors
90 days before admit/enroll	8 (42)	58 (35)	0.52
After enroll, before index date	10 (53)	72 (43)	0.43
H2 blockers
90 days before admit/enroll	3 (16)	12 (7)	0.19
After enroll, before index date	2 (11)	15 (9)	0.69
Narcotic or antimotility
90 days before admit/enroll	10 (53)	86 (52)	0.93
After enroll, before index date	14 (74)	95 (57)	0.22
Laxative
90 days before admit/enroll	6 (32)	43 (26)	0.59
After enroll, before index date	7 (37)	75 (45)	0.50
Chemotherapy
90 days before admit/enroll	0 (0)	8 (5)	1.00
After enroll, before index date	0	0

^aHCF, health care facility; PT, physical therapy; OT, occupational therapy; LTCF, long-term care facility; TMP-SMX, trimethoprim-sulfamethoxazole.

Risk factors for loss of C. difficile colonization.

Among the 49 patients colonized with C. difficile on admission, 11 (22%) were not colonized at discharge and thus lost colonization during the admission. The risk factors for loss of C. difficile colonization are given in Table 3. Patients who lost C. difficile colonization were more likely to report spending ≥2 h per week visiting a health care facility (27% versus 3%; P = 0.03). Patients who lost colonization were more likely to have received a β-lactam–β-lactamase inhibitor combination (36% versus 8%; P = 0.04) or metronidazole (27% versus 3%; P = 0.03) than patients who did not lose colonization. Congestive heart failure was more common among patients who lost colonization, but the difference was of borderline significance (36% versus 11%; P = 0.06).

TABLE 3.

Risk factors for loss of C. difficile colonization (n = 49 patients colonized on admission)

Variablea	No. (%) of patients:		P value
	Not colonized with C. difficile on discharge stool (n = 11)	Colonized with C. difficile on discharge stool (n = 38)
Surgical ward (vs medical)	3 (27)	3 (8)	0.12
Age ≥65 yr	2 (18)	7 (18)	1.00
Nonwhite race	1 (9)	9 (24)	0.42
Female	8 (73)	20 (53)	0.31
History of CDI	1 (9)	4 (11)	1.00
Health care worker	2 (18)	3 (8)	0.31
Lives with health care worker	1 (9)	3 (8)	1.00
Spends ≥2 h/wk visiting health care facility	3 (27)	1 (3)	0.03
Length of hospital stay (days)
1–2	2 (18)	5 (13)	Reference
3–7	5 (46)	20 (53)	0.63
8–14	1 (9)	10 (26)	0.30
>14	3 (27)	3 (8)	0.43
Admitted from:
Home	6 (55)	27 (71)	Reference
Health care facility	5 (46)	11 (29)	0.47
Reason for admission
Infection	3 (27)	12 (32)	Reference
Exacerbation of chronic condition	4 (36)	13 (34)	0.81
New medical/surgical problem	4 (36)	13 (34)	0.81
Comorbidities
Diabetes	3 (27)	14 (37)	0.73
Congestive heart failure	4 (36)	4 (11)	0.06
Liver disease	1 (9)	11 (29)	0.25
Chronic renal insufficiency	3 (27)	5 (14)	0.36
Chronic lung disease	3 (27)	10 (26)	1.00
HIV	0	0
Solid organ transplant	2 (18)	2 (5)	0.21
Stem cell transplant	0	0
Solid malignancy	0 (0)	5 (13)	0.57
Hematologic malignancy	1 (9)	1 (3)	0.40
Other immunocompromised	2 (18)	4 (11)	0.61
Inflammatory bowel disease	0 (0)	2 (5)	1.00
Fever at admission	1 (9)	4 (11)	1.00
Usual no. of bowel movements per day
<1 or ostomy bag	4 (36)	6 (16)	Reference
1–2	7 (64)	24 (63)	0.29
≥3	0 (0)	8 (21)	Undef
Procedure(s) in previous 90 days
Upper endoscopy	0 (0)	3 (8)	1.00
Lower endoscopy	0 (0)	3 (8)	1.00
Surgery	2 (18)	6 (16)	1.00
Any HCF exposure	11 (100)	36 (95)	1.00
No. of admissions to acute care hospitals in previous 90 days
0	2 (18)	10 (26)	Reference
≥1	9 (82)	28 (74)	0.71
No. of days since last inpatient discharge
No hospitalizations	2 (18)	10 (26)	Reference
0–14	6 (55)	13 (34)	0.36
>14	3 (27)	15 (40)	1.00
No. of outpatient clinic visits in previous 90 days
0	1 (9)	5 (13)	Ref
1–2	5 (46)	13 (34)	0.59
3–5	2 (18)	10 (26)	1.00
>5	3 (27)	10 (26)	0.75
Other health care exposures in previous 90 days
Inpatient PT/OT rehab	0 (0)	3 (8)	1.00
Reside in nursing home/LTCF	0	0
Reside in subacute care facility	0	0
Outpatient hemodialysis	1 (9)	2 (6)	0.56
Antimicrobials
Aminoglycosides
90 days before admit/enroll	0 (0)	1 (3)	1.00
After enroll, before index date	0	0
β-lactam–β-lactamase inhibitor combinations
90 days before admit/enroll	2 (18)	10 (26)	0.71
After enroll, before index date	4 (36)	3 (8)	0.04
Carbapenem
90 days before admit/enroll	2 (18)	4 (11)	0.61
After enroll, before index date	0	0
Cephalosporin
90 days before admit/enroll	3 (27)	9 (24)	1.00
After enroll, before index date	3 (27)	5 (13)	0.36
Clindamycin
90 days before admit/enroll	0 (0)	2 (5)	1.00
After enroll, before index date	0	0
Daptomycin
90 days before admit/enroll	1 (9)	1 (3)	0.40
After enroll, before index date	0	0
Doxycycline
90 days before admit/enroll	0 (0)	1 (3)	1.00
After enroll, before index date	0	0
Linezolid
90 days before admit/enroll	0 (0)	1 (3)	1.00
After enroll, before index date	0 (0)	1 (3)	1.00
Macrolide
90 days before admit/enroll	1 (9)	8 (21)	0.66
After enroll, before index date	1 (9)	3 (8)	1.00
Metronidazole
90 days before admit/enroll	2 (18)	4 (11)	0.61
After enroll, before index date	3 (27)	1 (3)	0.03
Penicillins
90 days before admit/enroll	0 (0)	1 (3)	1.00
After enroll, before index date	0	0
Fluoroquinolone
90 days before admit/enroll	2 (18)	7 (18)	1.00
After enroll, before index date	0 (0)	5 (13)	0.57
TMP-SMX
90 days before admit/enroll	0 (0)	8 (21)	0.17
After enroll, before index date	0 (0)	3 (8)	1.00
Vancomycin i.v.
90 days before admit/enroll	3 (27)	11 (29)	1.00
After enroll, before index date	4 (36)	7 (18)	0.24
Other medications
Proton pump inhibitors
90 days before admit/enroll	4 (36)	14 (37)	1.00
After enroll, before index date	3 (27)	12 (32)	1.00
H2 blockers
90 days before admit/enroll	3 (27)	5 (13)	0.36
After enroll, before index date	2 (18)	3 (8)	0.31
Narcotic or antimotility
90 days before admit/enroll	5 (46)	23 (61)	0.49
After enroll, before index date	4 (36)	13 (34)	1.00
Laxative
90 days before admit/enroll	2 (18)	10 (26)	0.71
After enroll, before index date	3 (27)	7 (18)	0.67
Chemotherapy
90 days before admit/enroll	1 (9)	0 (0)	0.22
After enroll, before index date	0	0

^aHCF, health care facility; PT, physical therapy; OT, occupational therapy; LTCF, long-term care facility; TMP-SMX, trimethoprim-sulfamethoxazole.

CDI and C. difficile colonization.

Three patients had a self-reported history of CDI within the 180 days before enrollment; only one of these cases was verified with laboratory results. Eleven additional patients reported a remote history of CDI. Twenty-four patients had a diarrheal stool specimen collected and tested for CDI by the BJH laboratory during admission; 22 (9%) patient specimens were negative, and 2 (1%) patients were diagnosed with CDI (17 and 43 days after last negative stool). Neither of these patients was colonized with C. difficile at admission. No patients developed CDI in the 60 days after hospital discharge.

DISCUSSION

While many studies have been published on risk factors for CDI in hospitalized patients, data on C. difficile colonization, including risk factors for acquisition and loss, are much more limited. We previously examined C. difficile colonization among patients admitted to the medical and surgical wards at BJH and found that 21% of patients were colonized at admission (15% with toxigenic C. difficile and 6% with nontoxigenic strains) (23). The current analysis included data from an additional 9 months of admission to the medical and surgical wards, and more than one-third of the patients in the current analysis were not included in the previous cohort, but the prevalence of C. difficile colonization at admission was consistent, at 21%. The admission prevalence was consistent despite that patients in the current cohort, but not the previous one, were required to have been admitted for ≥48 h. This is higher than previously reported estimates of <1% to 14% (11, 18, 24–26, 28, 32–36). This may be due to the high frequency of recent health care exposures in our population; it is also possible this is due to the highly sensitive culture methodology used in this study.

We found similar levels of C. difficile strain variability among colonized patients at admission (22 strains among 49 patients; ratio, 2.2) and discharge (27 strains among 57 patients; ratio, 2.1). However, the percentage of ribotype 027 strains doubled, from 6% of strains isolated at admission to 12% of strains isolated at discharge. While the actual numbers involved in this increase are small (3 patients at admission versus 7 patients at discharge), the results are nonetheless intriguing and are consistent with those of Loo et al. (28) and Didelot et al. (14). Loo et al. (28) found that 13.3% of patients colonized with C. difficile on admission were colonized with the NAP1/027 strain, but among patients who acquired C. difficile colonization while hospitalized, 36.1% were colonized with the NAP1 strain (28). In addition, the NAP1 strain was responsible for 62.7% of CDI cases that occurred in the hospitalized cohort (28). Using multilocus sequence typing, Didelot et al. (14) found that 63% of ST1 isolates (the NAP1/BI/027 strain) had a common ancestor within the previous 6 months, compared to only 19% of all other C. difficile sequence types. The reasons for this disproportionate number of acquisitions of the NAP1/BI/027 strain are unclear. It may be attributed to the purported enhanced sporulation capacity of this ribotype, selective pressures from the overuse of fluoroquinolones in hospitals, a possible yet-to-be defined mechanism for more efficient transmission, or it may simply cause more CDI and/or more severe CDI (53, 54). Some data indicate the NAP1/BI/027 strain is associated with higher severity of illness in CDI patients, particularly during outbreaks (4, 6, 37–40).

Data on the risk factors for C. difficile acquisition and loss during hospitalization are limited, although recent antibiotic use has been identified as a risk factor for acquisition (18). We identified cephalosporin use while hospitalized to be a risk factor for C. difficile acquisition (P = 0.03). This relationship is not surprising, since cephalosporin use is a well-known risk factor for CDI (41); we have noted this previously at BJH (42). In addition, several variables with biological plausibility measuring overall health care exposure (length of stay >14 days, ≥1 inpatient admission in the previous 90 days) were of borderline significance. Several previous studies have noted the association between recent hospitalization and increased C. difficile colonization at admission (18, 24, 26, 27). Health care-related exposures may be a surrogate for severity of illness, and these patients may be at greater risk for C. difficile colonization due to underlying comorbidities or antimicrobial exposures. However, we found no difference in the median number of antimicrobial types received between patients who acquired C. difficile and those who did not (data not shown). It is possible that some patients with recent health care exposures may have had preexisting colonization present at admission but below detectable limits; exposure to antimicrobials, such as cephalosporins, would result in an environment favorable to C. difficile growth.

To our knowledge, our study is the first to report factors associated with a loss of C. difficile colonization while hospitalized. We found that receipt of β-lactam–β-lactamase inhibitor combinations while hospitalized was more common among patients who lost colonization than those who did not lose colonization (P = 0.04). This finding, along with the identification of cephalosporin use as a risk factor for C. difficile acquisition, is notable. Previous studies have suggested a relationship between β-lactam–β-lactamase inhibitor combinations, cephalosporins, and CDI. Wilcox et al. (34) reported that a formulary switch from third-generation cephalosporins to piperacillin-tazobactam resulted in a decrease in CDI rates; Wilcox et al. (43) and Alston and Ahern (44) reported that a shortage of piperacillin-tazobactam and subsequent increase in cephalosporin prescribing led to a corresponding increase in CDI rates. Our results are complementary to those of Settle et al. (45), who found a significant increase in both colonization and CDI with the use of cefotaxime compared to those with piperacillin-tazobactam. Receipt of metronidazole while hospitalized was also associated with loss of colonization (P = 0.03). In a previous analysis of risk factors for CDI at BJH, receipt of metronidazole was protective against CDI (42). Cleary et al. (46) suggested that prophylactic metronidazole use might decrease CDI in high-risk patients, and Rodriguez et al. (47) found fewer cases of CDI when metronidazole was used preoperatively.

While these results suggest antimicrobial prophylaxis might prevent C. difficile colonization and/or CDI, caution is warranted, and we are not advocating this as an approach to CDI prevention. Despite the positive impact of β-lactam–β-lactamase inhibitor combinations on C. difficile acquisition observed in this study, these antimicrobials may still have a deleterious effect on the intestinal microbiomes of patients, as this has been demonstrated in both the human gut and mouse models (48, 49). This effect may persist after the completion of antimicrobial therapy and may increase future risk of CDI. Prophylactic treatment of C. difficile asymptomatic carriers with oral vancomycin has been shown to result in higher rates of C. difficile over time (50). Another concern with the use of prophylactic antibiotics is the emergence of resistance; Baines et al. (51) reported reduced susceptibility to metronidazole in isolates of C. difficile ribotype 001. Prophylactic use of antimicrobials might also increase selective pressure for other multidrug-resistant organisms, and more data are needed on the relationship between β-lactam–β-lactamase inhibitors, metronidazole, C. difficile colonization, and CDI. However, antimicrobial prescribing practices are potentially modifiable risk factors; antimicrobial stewardship programs that promote the use of β-lactam–β-lactamase inhibitor combinations or metronidazole over cephalosporins, when appropriate, may help reduce C. difficile acquisition and CDI.

The clinical and epidemiological importance of C. difficile colonization remains unclear. A growing body of evidence suggests that transmission from symptomatic patients cannot account for most CDI cases (12, 14–16), and asymptomatic carriers may be involved in transmission (18–20). We found that the majority of C. difficile patients colonized at admission were also colonized at discharge (78%); these results are similar to those of McFarland et al. (11), who found that 82% of patients colonized or with CDI were still positive on discharge. Although only two patients developed CDI, both cases occurred after a new C. difficile acquisition. This is consistent with past research that new C. difficile acquisitions are more strongly associated with the development of CDI than preexisting colonization (52). Kyne et al. (34) found that patients colonized with C. difficile were more likely than patients with CDI to mount an effective immune response; the immune response of these patients may protect against CDI.

There are several limitations to this study. While the overall sample size was reasonable, the subanalyses, particularly risk factors for loss of colonization, were limited by small sample sizes for these subanalyses. The borderline risk factors for C. difficile acquisition (prolonged hospitalization and recent hospitalization) may have reached statistical significance if the sample size were larger. Because the loss of C. difficile colonization appears to be a relatively uncommon event, larger studies are needed to confirm these findings or identify other risk factors for loss of colonization. Next, the high prevalence of C. difficile colonization at admission identified here might not be generalizable to other health care facilities or patient populations. Finally, we did not perform environmental sampling of the patient hospital rooms as part of this study, and thus we are unable to include this component in our assessment of C. difficile acquisition risk.

The results of our studies indicate the burden of C. difficile colonization at admission is greater than previously estimated, but the epidemiology of these patients and its clinical importance are complex. Asymptomatically colonized patients may be an important source of C. difficile transmission, as are CDI cases, but modifiable risk factors, such as antimicrobial use, for C. difficile acquisition and loss do exist. Our results have significant implications for CDI prevention. Antimicrobial stewardship programs that shift use away from cephalosporins and toward β-lactam–β-lactamase inhibitors may result in decreased C. difficile colonization and CDI rates. Our study provides evidence that the mechanism behind this relationship may be 2-fold: decreasing cephalosporin use leads to decreased acquisition and CDI, and increasing β-lactam–β-lactamase inhibitor combination use leads to decreased colonization and CDI. These results require confirmation in future studies. Effective CDI prevention is multifactorial and multidisciplinary, and future efforts combining the expertise of physicians, microbiologists, pharmacists, and infection control practitioners can reduce the burden of disease due to CDI.