The Drivers of Acute and Long-term Care Clostridium difficile Infection Rates: A Retrospective Multilevel Cohort Study of 251 Facilities

Abstract

Background

Drivers of differences in Clostridium difficile incidence across acute and long-term care facilities are poorly understood. We sought to obtain a comprehensive picture of C. difficile incidence and risk factors in acute and long-term care.

Methods

We conducted a case-cohort study of persons spending at least 3 days in one of 131 acute care or 120 long-term care facilities managed by the United States Veterans Health Administration between 2006 and 2012. Patient (n = 8) and facility factors (n = 5) were included in analyses. The outcome was the incidence of facility-onset laboratory-identified C. difficile infection (CDI), defined as a person with a positive C. difficile test without a positive test in the prior 8 weeks.

Results

CDI incidence in acute care was 5 times that observed in long-term care (median, 15.6 vs 3.2 per 10000 person-days). History of antibiotic use was greater in acute care compared to long-term care (median, 739 vs 513 per 1000 person-days) and explained 72% of the variation in C. difficile rates. Importation of C. difficile cases (acute care: patients with recent long-term care attributable infection; long-term care: residents with recent acute care attributable infection) was 3 times higher in long-term care as compared to acute care (median, 52.3 vs 16.2 per 10000 person-days).

Conclusions

Facility-level antibiotic use was the main factor driving differences in CDI incidence between acute and long-term care. Importation of acute care C. difficile cases was a greater concern for long-term care as compared to importation of long-term care cases for acute care.

Clostridium difficile infection, antimicrobials, inter-facility patient transfer, transmission, healthcare associated infection

Clostridium difficile infection (CDI) is transmitted by the fecal–oral route and is one of the most common and deadly infections among patients in acute care facilities [1] and residents of long-term care facilities [2]. Individual-level risk factors for the disease have been extensively studied and include age, proton pump inhibitor use, feeding tube insertion, abdominal surgical procedures, and, most importantly, exposure to antimicrobials that disrupt the gut microflora [3, 4]. Risk factors associated with the healthcare environment have been identified as well, and include room-, ward-, and facility-level infection pressure [5, 6] and antibiotic pressure [7, 8].

Rates of infection with C. difficile tend to be several times higher in acute care compared with long-term care settings (4.0–15.5 per 10000 bed-days in acute care [9] vs 1.7–6.0 per 10000 bed-days in long-term care [10]). Some argue that patient movement from outpatient settings may be partially responsible for high levels of C. difficile in acute care, via importation of infected or colonized patients [11]. Conversely, others have found that the majority of cases of community-associated disease are associated with healthcare exposures [12].

It is not known whether the same factors explain variation in C. difficile incidence in both acute and long-term care because facility-level CDI has been studied in acute and long-term care environments separately. What has been demonstrated is that among long-term care facilities, facility-level antibiotic use and importation from acute care predict C. difficile incidence rates [2]; among acute care facilities, neither facility-level antibiotic use [13] nor infection control practices [14] have been shown to predict C. difficile incidence rates. A better understanding of why this heterogeneity exists is crucial to assess whether distinct interventions may be required to reduce C. difficile rates in acute and long-term care facilities.

Therefore, we conducted a comparative analysis of the prevalence of individual- and facility-level risk factors and their associations with C. difficile risk in acute and long-term care facilities with shared populations. This study was conducted in the Veterans Health Administration (VHA), which provides healthcare services for a unique veteran population predominantly comprised of male enrollees [15]. It is the largest integrated healthcare system in the United States [16] and has a system-wide electronic medical record [17], making it an ideal ecosystem in which to study the drivers of CDI in linked acute and long-term care facilities.

METHODS

This study used a multilevel longitudinal case-cohort design: it was multilevel in that both individual- and facility-level risk factors were included in the analysis; it was longitudinal in that patients were followed before and during their stays and their exposures to intermittent risk factors (eg, antimicrobials) were assessed on a daily basis (allowing the time lag between exposure and outcome [18] to be specified); and it employed a computationally efficient case-cohort design [19] in which all cases of C. difficile and a random sample of uninfected controls were examined (as certain analysis steps were computationally demanding).

Acute and Long-term Care Cases and Controls

We built 2 datasets structured in person-day format with 1 record for each acute and long-term care person-day between January 2006 and December 2012. Within these datasets, we identified cases with CDI, defined as a person-day on which an individual had a positive C. difficile toxin test result. The C. difficile toxin test results were ascertained from all VHA microbiology records including acute care and long-term care. Person-days with a recent positive C. difficile test in the prior 56 days were then excluded (exclusion 1) as these would represent recurrent C. difficile cases [20] whose risk factors are likely distinct from new onset cases. We then identified controls as person-days without a positive C. difficile toxin test. Due to the computational burden of working with a large number of controls (>20 million), a 1% simple random sample of controls was then selected, and the remaining 99% of controls were excluded (exclusion 2). This design, known as a case-cohort design, enables the calculation of accurate parameter estimates from drastically smaller datasets [19]. Next, for both cases and controls, we excluded all those having <3 person-days within an acute or long-term care facility in the prior 56 days (exclusion 3). As such, all cases met the definition of facility-associated infection [20]. Finally, based on this dataset, we excluded acute and long-term care facilities with either <10 cases of C. difficile or <10000 weighted person-days, where weights of 1 were applied to case-days and 100 were applied to control-days (exclusion 4) to account for the case-cohort sampling. See the Supplementary Appendix for more information on the application of the exclusion criteria.

Individual-Level Risk Factors

We assessed 8 individual-level risk factors. Demographic variables included patient sex and age. Charlson comorbidity count was based on International Classification of Diseases, Ninth Revision codes documented during inpatient and outpatient stays that occurred in the 2-year period prior to a given person-day [21]. VHA acute care stay, long-term care stay, antibiotic exposure, antibiotic risk index, and proton pump inhibitor (PPI) exposure were assessed in the prior 56 days. All pharmaceutical exposure variables were assessed from pharmacy databases that covered both inpatient and outpatient care. For the antibiotic risk index, persons were classified according to the highest-risk antibiotic received in the prior 56 days. The antibiotic risk strata consisted of (1) low (tetracyclines); (2) medium (penicillins, macrolides, or sulfonamides); (3) high (cephalosporins or fluoroquinolones); and (4) very high (lincosamides) [22]. Including persons with no antibiotic receipt in the prior 56 days as a separate group, the antibiotic risk index had a total of 5 strata.

Facility-Level Risk Factors

We assessed 5 facility-level risk factors, including mean antibiotic use (persons with an antibiotic exposure in the prior 56 days, per 1000 person-days), mean PPI use (persons with a PPI exposure in the prior 56 days, per 1000 person-days), mean age, and mean facility census. We defined importation in acute care as the number of persons with a positive C. difficile test in the prior 56 days attributable to long-term care, divided by the number of patients in the acute care facility. Similarly, in long-term care, importation was defined as the number of persons with a positive C. difficile test in the prior 56 days attributable to acute care divided by the number of residents in the long-term care facility. Attribution of C. difficile cases was based on enrollee location 3 days prior to the positive test, irrespective of enrollee location at the time of the test. As such, in contrast to our case definition, our importation measures included both facility and outpatient onset cases. Our measures of importation included only those C. difficile cases that moved between acute and long-term care facilities; our measure did not include other potentially important routes of importation such as between different acute care facilities, between different long-term care facilities, and from outpatient sources. For each of these 5 variables, the facility-level means were calculated for each facility [19].

Statistical Analysis

Our analyses were divided into individual-level, facility-level, and multilevel components. Individual-level regression analyses used patient-day-level Poisson models with the outcome corresponding to whether an at-risk individual had a positive test on a given day. To measure confidence intervals (CIs) that accounted for clustering within facilities, we used generalized estimating equations (GEEs) with clusters corresponding to the facility and using the independence covariance structure. Sampling weights were also applied. For the acute and long-term care cohorts separately, 8 patient-level bivariate models were estimated, one for each of the covariates.

We described variation in facility characteristics with the median, 10th percentile (p10), and 90th percentile (p90). The median difference in facility characteristics for acute vs long-term care facilities was measured, and 95% CIs were calculated by using the appropriate percentiles from 10000 bootstrap replicates [23].

We created Poisson regression models to estimate facility-level C. difficile rates, with the outcome corresponding to the number of CDI case-days and the offset corresponding to the logarithm of the weighted number of person-days [19]. Robust sandwich estimators were used to calculate 95% CIs while accounting for clustering of outcomes within facilities. For the acute and long-term care cohorts separately, and combined, 5 bivariate models were estimated, 1 for each of the facility-level covariates, and then 1 multivariate model that included all 5 facility-level covariates. The deviance explained (D²) was measured as the null model deviance minus the deviance of the model in question, divided by the null model deviance.

Finally, we built acute and long-term care multilevel models that were analogous to the individual-level Poisson GEE models, but included both individual- and facility-level covariates. In these multilevel models, the inclusion of facility-level covariates meant that 2 persons with the same individual-level risk factors residing in different facilities likely had distinct risk estimates [24]. Each multilevel model included 6 of the 8 individual-level covariates (sex, age, history of acute or long-term care stay, comorbidity count, antibiotic use, and PPI use) and all of the 5 facility-level covariates. Two individual-level variables were excluded: (1) history of stay in the same facility type (as all acute care patients had a history of acute care stay, and all long-term care residents had a history of long-term care stay); and (2) antibiotic risk index (due to its redundancy with the antibiotic use covariate).

Ethics Statement

Study approval was obtained from the Research Ethics Board of the University of Utah and the Veterans Affairs Salt Lake City Health Care System. The Board waived the need for consent because there was no contact with patients, and anonymity was assured.

RESULTS

Cohort

Of 22.5 million person-days of follow-up in acute care, 193500 met the inclusion criteria (Figure 1). This corresponded to 156169 patients (mean, 1.2 days per patient) in 131 acute care facilities. Similarly, in long-term care, of 21.5 million person-days of follow-up, 199200 patient-days met our inclusion criteria. This corresponded to 67612 residents (mean of 2.9 days per resident) in 120 long-term care facilities. We observed 35754 case-days of C. difficile, of which 28615 (80%) were diagnosed during acute care stays and 7139 (20%) were diagnosed during long-term care stays.

Figure 1.

Flowchart of patients and residents included and excluded, Veterans Health Administration facilities, 2006–2012. acase- and control-days were not identified at these stages; ^bthe 10,000 person-day facility criterion was based on the weighted sample size. Abbreviation: d, day.

Individual-Level Analyses

In unadjusted individual-level models, most variables, including male sex, comorbidity count, and pharmaceutical risk factors had a similar direction of association with C. difficile infection in acute and long-term care, but the magnitude tended to be larger in long-term care (Table 1). A recent history of acute care exposure was a common (64.5% of cases and 27.3% of controls) and strong risk factor among long-term care patients (incidence rate ratio [IRR], 4.84 [95% CI, 4.34–5.41]) whereas a history of long-term care exposure was a relatively rare (7.4% of cases and 4.0% of controls) and moderate risk factor among acute care patients (IRR, 1.89 [95% CI, 1.75–2.05]).

Table 1.

Individual Demographic and Clinical Characteristics and Clostridium difficile Infection Risk

Characteristic	Acute Care			Long-term Care
	Cases	Controls	IRR (95% CI)	Cases	Controls	IRR (95% CI)
Sex
Female	899 (3.1)	5680 (3.4)	Reference	168 (2.4)	6110 (3.2)	Reference
Male	27716 (96.9)	159205 (96.6)	1.10 (1.02–1.19)	6971 (97.6)	185951 (96.8)	1.36 (1.14–1.63)
Age
<60	6305 (22.0)	45728 (27.7)	Reference	1067 (14.9)	31943 (16.6)	Reference
60–69	8756 (30.6)	53421 (32.4)	1.19 (1.13–1.25)	1956 (27.4)	49325 (25.7)	1.19 (1.09–1.30)
70–79	6680 (23.3)	34668 (21.0)	1.40 (1.32–1.48)	1668 (23.4)	42108 (21.9)	1.19 (1.06–1.33)
≥80	6874 (24.0)	31068 (18.8)	1.60 (1.49–1.73)	2448 (34.3)	68684 (35.8)	1.07 (.95–1.20)
Comorbidity count
0	9599 (33.5)	78096 (47.4)	Reference	1583 (22.2)	84296 (43.9)	Reference
1–2	11024 (38.5)	53334 (32.3)	1.68 (1.62–1.75)	3078 (43.1)	66632 (34.7)	2.46 (2.21–2.73)
≥3	7992 (27.9)	33455 (20.3)	1.94 (1.85–2.04)	2478 (34.7)	41133 (21.4)	3.21 (2.89–3.56)
Long-term care stay in the prior 56 d
No	26507 (92.6)	158253 (96.0)	Reference
Yes	2108 (7.4)	6632 (4.0)	1.89 (1.75–2.05)
Acute care stay in the prior 56 d
No				2536 (35.5)	139724 (72.7)	Reference
Yes				4603 (64.5)	52337 (27.3)	4.84 (4.34–5.41)
Proton pump inhibitors in the prior 56 d
No	6990 (24.4)	55319 (33.6)	Reference	2151 (30.1)	88678 (46.2)	Reference
Yes	21625 (75.6)	109566 (66.4)	1.56 (1.48–1.65)	4988 (69.9)	103383 (53.8)	1.99 (1.84–2.15)
Antibiotics in the prior 56 d
No	2660 (9.3)	42904 (26.0)	Reference	836 (11.7)	97247 (50.6)	Reference
Yes	25955 (90.7)	121981 (74.0)	3.43 (3.25–3.62)	6303 (88.3)	94814 (49.4)	7.73 (6.88–8.68)
Low riska	57 (0.2)	1189 (0.7)	0.77 (.58–1.04)	24 (0.3)	2377 (1.2)	1.17 (.79–1.75)
Medium riskb	3918 (13.7)	24719 (15.0)	2.55 (2.40–2.72)	882 (12.4)	28044 (14.6)	3.66 (3.27–4.09)
High riskc	19544 (68.3)	86930 (52.7)	3.62 (3.43–3.83)	4560 (63.9)	58596 (30.5)	9.05 (7.97–10.3)
Very high riskd	2436 (8.5)	9143 (5.5)	4.29 (3.91–4.70)	837 (11.7)	5797 (3.0)	16.8 (14.5–19.4)

Data are presented as No. (%) unless otherwise indicated. The IRR in this case-cohort design is calculated as Cases1 / (Cases1 + Controls1 × 100) ÷ Cases2 / (Cases2 + Controls2 × 100) where group 1 is the comparator group and group 2 is the referent group and 100 is 1 ÷ the probability of selection.

Abbreviations: CI, confidence interval; IRR, incidence rate ratio.

^aTetracyclines.

^bPenicillins, macrolides, sulfonamides.

^cCephalosporins, carbapenems, fluoroquinolones.

^dLincosamides (eg, clindamycin).

Facility-Level Analyses

Clostridium difficile incidence and patient characteristics were highly variable (Table 2). CDI incidence in acute care (median, 15.6 per 10000 person-days; p10 = 7.7, p90 = 23.7) was >5 times that observed in long-term care (median, 3.2 per 10000 person-days; p10 = 1.3, p90 = 9.4). History of antibiotic exposure was universally high among acute care facilities but was lower and more variable among long-term care facilities. Among acute care facilities, importation of C. difficile cases from long-term care facilities was generally low whereas among long-term care facilities, importation of C. difficile cases from acute care facilities was >3-fold higher, and varied substantially more than in acute care facilities.

Table 2.

Characteristics of Acute and Long-term Care Facilities in the Veterans Health Administration

Characteristic	Acute Care Facilities (n = 131), Median (p10, p90)	Long-term Care Facilities (n = 120), Median (p10, p90)	Difference (AC-LTC), Median (95% CI)
Clostridium difficile incidence, per 10000 patient-days	15.6 (7.7, 23.7)	3.2 (1.3, 9.4)	12.4 (11.2, 14.1)
Importationa, per 10000 patient-days	16.2 (1.5, 43.0)	52.3 (15.4, 153.2)	–36.1 (–48.0, –22.9)
Antibiotic use in prior 56 d, per 1000 patient-days	739 (692, 787)	513 (359, 681)	225 (202, 253)
Proton pump inhibitor use in prior 56 days, per 1000 patient-days	651 (541, 755)	562 (419, 654)	89 (65115)
Average patient or resident age	66.6 (64.0, 69.8)	72.3 (68.2, 75.3)	–5.8 (–6.7, –5.1)
Average patient or resident census	44.5 (11.6, 92.2)	56.5 (25.2, 113.4)	–12.0 (–24.7, 2.3)

Abbreviations: AC, acute care; CI, confidence interval; LTC, long-term care; p10, 10th percentile; p90, 90th percentile.

^aIn acute care, defined as the proportion of persons with a recent C. difficile test attributable to long-term care. In long-term care, it is defined as the proportion of persons with a recent test attributable to acute care. Attribution of C. difficile cases is based on patient location 3 days prior to their positive test.

The bivariate association between facility-level antibiotic use and C. difficile incidence (Figure 2, upper panel) was strong in long-term care facilities (D² = 61%; IRR, 1.67 per increase of 100 [95% CI, 1.54–1.80]), but this association was not significant in acute care (D² = 0%; IRR, 0.92 per increase of 100 [95% CI, .73–1.16]). In models that included both facility types, antibiotic use was the strongest predictor (D² = 72%; IRR, 1.82 per increase of 100 [95% CI, 1.72–1.92]). Importation (Figure 2, lower panel) was a stronger predictor of C. difficile incidence in long-term care (D² = 57%; IRR, 1.78 per increase of 100 [95% CI, 1.65–1.92]) compared with acute care facilities (D² = 10%; IRR, 1.97 per increase of 100 [95% CI, 1.41–2.74]). Our multivariate model of facility CDI incidence revealed that, taken together, the 5 covariates we examined explained 74% of variation in CDI incidence in long-term care, but only 13% in acute care facilities.

Figure 2.

The association between acute care facility (black) and long-term care facility (gray) antibiotic use (upper panel) and importation (lower panel) and Clostridium difficile infection incidence in Veterans Health Administration facilities, 2006–2012. Note that importation is, in acute care, defined as the proportion of persons with a recent positive C. difficile test attributable to long-term care. In long-term care, it is defined as the proportion of persons with a recent test attributable to acute care. Attribution is based on patient location 3 days prior to their positive test.

Multilevel Analyses

Our multilevel analyses included both individual- and facility-level risk factors (Table 3). In long-term care, individual-level antibiotic use was strongly associated with risk (IRR, 4.40 [95% CI, 3.80–5.10]); however, facility-level importation of acute care cases (IRR, 1.35 per increase of 100 [95% CI, 1.18–1.53]) and facility-level antibiotic use (IRR, 1.20 per increase of 100 [95% CI, 1.02–1.41]) also contributed to CDI risk. In the acute care model, individual-level antibiotic use (IRR, 3.16 [95% CI, 2.99–3.34]) and facility-level importation of long-term care cases were associated with risk (IRR, 1.88 per increase of 100 [95% CI, 1.19–2.96]) whereas facility-level antibiotic use was not (IRR, 0.88 per increase of 100 [95% CI, .70–1.12]).

Table 3.

Multilevel Model Including Individual- and Facility-Level Predictors of Clostridium difficile Infection Incidence in Veterans Health Administration Acute and Long-term Care Facilities

Level	Acute Care, IRR (95% CI)	Long-term Care, IRR (95% CI)
Individual-level
Male sex	0.96 (.88–1.04)	1.33 (1.00–1.77)
Age
<60	Reference	Reference
60–69	1.13 (1.07–1.18)	1.19 (1.03–1.37)
70–79	1.32 (1.26–1.39)	1.31 (1.12–1.53)
≥80	1.56 (1.47–1.66)	1.47 (1.28–1.69)
Comorbidities
0	Reference	Reference
1–2	1.53 (1.47–1.60)	1.07 (.95–1.21)
≥3	1.64 (1.56–1.73)	1.23 (1.09–1.38)
Long-term care stay in the prior 56 d	1.28 (1.18–1.38)
Acute care stay in the prior 56 d		1.49 (1.29–1.72)
Pharmaceutical exposures in prior 56 d
Antibiotic	3.16 (2.99–3.34)	4.40 (3.80–5.10)
PPI	1.35 (1.30–1.40)	1.31 (1.21–1.41)
Facility level
Importation, per 10000 patient-days, per increase of 100	1.88 (1.19–2.96)	1.35 (1.18–1.53)
Antibiotic use, per 1000 patient-days, per increase of 100	0.88 (.70–1.12)	1.20 (1.02–1.41)
PPI use, per 1000 patient-days, per increase of 100	0.97 (.87–1.07)	0.92 (.82–1.02)
Average patient or resident age, per increase of 10	0.96 (.59–1.55)	0.61 (.38–.97)
Average patient or resident census per increase of 10	1.14 (.92–1.43)	0.80 (.62–1.04)

Abbreviations: CI, confidence interval; IRR, incidence rate ratio; PPI, proton pump inhibitor.

DISCUSSION

In this study of acute and long-term care CDI, which included 251 facilities from across the United States, we found that (1) interfacility variation in antibiotic use is considerably larger in long-term care compared with acute care; (2) facility-level antibiotic use was the main facility-level factor driving differences in CDI incidence between acute and long-term care; and (3) importation of acute care C. difficile cases is a more significant concern for long-term care facilities than importation of long-term care cases is for acute care facilities.

We found that there was a high degree of variation in antibiotic use across long-term care facilities, with anywhere from 30% to 80% of residents having a history of antibiotic use in the prior 2 months. On the other hand, acute care facility antibiotic use history ranged between 70% and 80%. High variation in antibiotic use is thought to be driven more by clinician factors, such as tendencies for inappropriate treatment of asymptomatic bacteriuria [25] and tendencies to prescribe antibiotic courses for longer duration [26], than by patient factors [27]. Many antimicrobial stewardship interventions can bring about reductions in antibiotic use and improved clinical outcomes [28, 29], including the use of social normative feedback to prescribers [30] and interventions to reduce unnecessary treatment of asymptomatic bacteriuria [31]. A combination of such scalable strategies could be applied across an entire healthcare system to achieve large reductions of C. difficile cases.

We found that facility-level antibiotic use was a driver of differences in C. difficile incidence among long-term care facilities, but not among acute care facilities. In our study, analogous datasets were used to measure antibiotic use and CDI incidence in both acute and long-term care, adding to the robustness of these findings. These somewhat contradictory findings are not novel: previous studies have found that interfacility differences in CDI incidence can be predicted by facility-level antibiotic use in long-term care [2], but that this does not hold true across acute care facilities [13]. An interfacility association between antibiotic use and CDI incidence among acute care facilities may be masked due to a combination of factors, including confounding by unmeasured patient or facility factors, insufficient interfacility variation in antibiotic use, or a ceiling effect on the association.

This study also found that importation from acute into long-term care was >3 times more common than the converse. Previous work has shown that local, or endogenous, C. difficile pressure impacts downstream facility CDI rates [6] and that exogenous, imported infection pressure from acute care is also an important driver of long-term care rates of disease [2]. Recent work has shown that healthcare facilities that are the recipients of large numbers of interfacility transfers tend to have substantially higher rates of CDI [32]. It appears that long-term care facilities are impacted by their acute care partners, but the converse is somewhat less apparent for acute care facilities, whose inpatient populations come from a variety of sources other than long-term care facilities, including other acute care facilities and outpatient settings.

Our study has a number of limitations. First, our study considered laboratory-identified infections. As such, cases we identified were not necessarily symptomatic. However, previous work at the VHA has shown that >90% of laboratory-identified cases are clinically confirmed [33]. Second, this study only captured C. difficile tests and medical exposures that occurred within the VHA system, and as such may have substantially underestimated these variables among VHA enrollees who also have access to alternate private or public health insurance [34]. We guarded against misclassification of our outcome by limiting our analyses to facility-onset, facility-associated disease, as VHA inpatients or long-term care residents are less likely to receive non-VHA care. Nevertheless, underestimation of our exposure variables of interest, including antibiotic use and importation, likely led to an attenuation of the effect estimates we observed, due to nondifferential misclassification. Three, our study did not account for variability in diagnostic methods used across facilities. This may have led to interfacility variation in case severity as facilities introduced highly sensitive polymerase chain reaction–based tests which may identify less severe cases [35, 36]. Four, our study did not fully capture factors impacting C. difficile incidence, such as variation in invasive procedures (feeding tube use or gastrointestinal surgery) and infection control practices. Five, our measures of importation included only those C. difficile cases that moved between VHA acute and long-term care facilities; our measure did not include other potentially important routes of importation such as between different acute care facilities, between different long-term care facilities, between non-VHA and VHA facilities, and from outpatient sources.

This is the first comparative analysis of drivers of CDI incidence across acute and long-term care and provides evidence of the importance of bolstering antimicrobial stewardship programs to reduce antibiotic use. We have shown the potential for interfacility transmission of CDI, and that long-term care facilities disproportionately shoulder the burden of dealing with C. difficile cases from acute care, rather than vice versa. This suggests that it may be wise for healthcare systems to coordinate infection control practices, so that appropriate infection prevention and control strategies are put into place for incoming patients and long-term care residents with histories of CDI.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Financial support. This study was supported by the Centers for Disease Control and Prevention (intra-agency agreement 11FED1106563) and the Veterans Health Administration (Center of Innovation 14–267). K. A. B. received support from the Veterans Health Administration Advanced Fellowship in Informatics.

Disclaimer. The funders had no role in the design and conduct of the study, or in the preparation or approval of the manuscript.

Potential conflicts of interest. K. A. B. has received support from AstraZeneca. All other authors report no potential conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.