Hospital Ward Antibiotic Prescribing and the Risks of Clostridium difficile Infection

Research Original Investigation LESS IS MORE Hospital Ward Antibiotic Prescribing and the Risks of Clostridium difficile Infection Kevin Brown, PhD; Kim Valenta, PhD; David Fisman, MD, MSc; Andrew Simor, MD; Nick Daneman, MD, MSc IMPORTANCE Only a portion of hospital-acquired Clostridium difficile infections can be traced back to source patients identified as having symptomatic disease. Antibiotic exposure is the main risk factor for C difficile infection for individual patients and is also associated with increased asymptomatic shedding. Contact with patients taking antibiotics within the same hospital ward may be a transmission risk factor for C difficile infection, but this hypothesis has never been tested. Invited Commentary page 633 CME Quiz at jamanetworkcme.com and CME Questions page 668 OBJECTIVES To obtain a complete portrait of inpatient risk that incorporates innate patient risk factors and transmission risk factors measured at the hospital ward level and to investigate ward-level rates of antibiotic use and C difficile infection risk. DESIGN, SETTING, AND PATIENTS A 46-month (June 1,, through March 31, 14) retrospective cohort study of inpatients 18 years or older in a large, acute care teaching hospital composed of 16 wards, including intensive care units and 11 non intensive care unit wards. EXPOSURES Patient-level risk factors (eg, age, comorbidities, hospitalization history, antibiotic exposure) and ward-level risk factors (eg, antibiotic therapy per patient-days, hand hygiene adherence, mean patient age) were identified from hospital databases. MAIN OUTCOMES AND MEASURES Incidence of hospital-acquired C difficile infection as identified prospectively by hospital infection prevention and control staff. RESULTS A total of 2 of 34 298 patients developed C difficile (incidence rate,.9 per patient-days; 9% CI,.26-6.73). Ward-level antibiotic exposure varied from 21.7 to 6.4 days of therapy per patient-days. Each % increase in ward-level antibiotic exposure was associated with a 2.1 per (P <.1) increase in C difficile incidence. The association between C difficile incidence and ward antibiotic exposure was the same among patients with and without recent antibiotic exposure, and C difficile risk persisted after multilevel, multivariate adjustment for differences in patient-risk factors among wards (relative risk, 1.34 per % increase in days of therapy; 9% CI, 1.16-1.7). CONCLUSIONS AND RELEVANCE Among hospital inpatients, ward-level antibiotic prescribing is associated with a statistically significant and clinically relevant increase in C difficile risk that persists after adjustment for differences in patient-level antibiotic use and other patient- and ward-level risk factors. These data strongly support the use of antibiotic stewardship as a means of preventing C difficile infection. JAMA Intern Med. 1;17(4):626-633. doi:.1/jamainternmed.14.8273 Published online February 23, 1. Author Affiliations: Division of Epidemiology, Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada (Brown, Fisman); Department of Anthropology and School of the Environment, McGill University, Montreal, Quebec, Canada (Valenta); Sunnybrook Health Sciences Centre, University of Toronto, Toronto, Ontario, Canada (Simor, Daneman). Corresponding Author: Nick Daneman, MD, MSc, Division of Infectious Diseases and Clinical Epidemiology, Sunnybrook Health Sciences Centre, University of Toronto, 7 Bayview Ave, Toronto, ON, M4N 3M, Canada (nick.daneman@sunnybrook.ca). 626 (Reprinted) jamainternalmedicine.com

Clostridium difficile Infection Original Investigation Research Antibiotic exposure represents the principal risk factor for Clostridium difficile infection, and existing research estimates that inpatients taking antibiotics are, on average, 6% more likely to acquire the infection. 1 Prolonged antibiotic exposure and exposure to larger antibiotic doses are associated with increased C difficile infection risk, 2 and some antibiotics (clindamycin, cephalosporins, and fluoroquinolones) are associated with a greater risk relative to other antibiotic classes. 3,4 Risk may increase over time with increased prescribing of certain antimicrobials. Important gaps in knowledge remain with respect to the natural history of how C difficile bacteria are transmitted among hospitalized patients. Hospital environments are persistently contaminated with C difficile spores, and surfaces in rooms of infected patients are contaminated before, during, and after treatment for C difficile infection. 6 Exposure to symptomatic patients with C difficile infection has been identified as an independent risk factor for transmission. 7 However, exposure to spores from other symptomatic patients may not explain most new cases of C difficile infection acquired in hospitals. 8 In a C difficile outbreak in a long-term care facility, almost half of the residents had asymptomatic colonization, and antibiotic exposure was the primary risk factor for asymptomatic colonization. 9 Although asymptomatically colonized individuals contribute less to environmental contamination at an individual level, asymptomatic carriers outnumber symptomatic patients by a ratio of 3:1 and as such could represent an important source of C difficile infection transmission. In the absence of reliable measures of patient colonization and environmental contamination, transmission risks can potentially be estimated as a function of aggregated measures of patient risk factors for colonization, such as mean ward- or hospital-level antibiotic prescribing. 11 Multilevel models can be used to tease apart the effect of individual-level risk factors that affect patient susceptibility (direct effects) from group-level effects that affect transmission risks that are independent of individual-level effects (indirect effects). 12 We sought to establish the effect of ward antibiotic-prescribing rate on ward C difficile infection incidence and whether the effects observed extended beyond the direct antibiotic effects on patients infection risk. Methods Ethics Statement Study approval was obtained from the Research Ethics Board of Sunnybrook Health Sciences Centre. The board waived the need for patient consent because there was no contact with patients and patient anonymity was assured. Study Design and Participants A retrospective cohort study design was used to assess the association of individual- and ward-level risk factors with the incidence of C difficile infection among patients admitted to Sunnybrook Hospital, a large, acute care teaching hospital located in Toronto, Ontario, Canada. The source cohort consisted of all patients older than 18 years without a previous C difficile infection diagnosis who were hospitalized in an acute care ward at Sunnybrook Hospital from June 1,, through March 31, 14. We excluded patients in the hospital s psychiatry, obstetrics, neonatal, and long-term care wards given a low expected event rate of C difficile infection. Case Definition Patients infected with C difficile were identified by the Infection Prevention and Control Department via active surveillance during the study period. A C difficile infection case was defined as any hospitalized patient with laboratory confirmation of a positive toxin assay result together with diarrhea or visualization of pseudomembranes on sigmoidoscopy, colonoscopy, or histopathologic analysis. 13,14 For the purposes of case identification, diarrhea was defined as 3 or more loose or watery bowel movements in a 24-hour period, which was unusual or different for the patient, and with no other recognized cause. When a patient developed a C difficile infection, the remaining hospitalized days were excluded from the atrisk patient-days. Toxin assays at the hospital have been performed by polymerase chain reaction (BD GeneOhm Cdiff; Becton, Dickinson and Company) since September 9, which includes the entire study period. For C difficile infection case admissions, event time was the number of days from hospital admission to symptom onset or positive toxin assay result for rare cases (<1%) in which symptom onset was missing. For noncase admissions, censoring time was the number of days from hospital entry until discharge, study termination, or death. In addition to excluding hospitalized days after C difficile infection, the first 2 days of each hospital admission were also excluded because patients are not at risk of nosocomial infection at the beginning of a hospital stay. Antimicrobial Exposure Assessment Patient antibiotic exposures were drawn from pharmacy dispensing records. We examined records for receipt of any antibiotic in the prior days but excluded exposure to metronidazole, oral vancomycin hydrochloride, or fidaxomicin because these may be treatments for C difficile infection. 1 Antibiotic receipt was classified according to the Anatomical Therapeutic Chemical (ATC) Classification System, 17th edition. 16 As per previous work, 4 we classified individual patients according to whether they had received a high-risk antibiotic (defined as receipt of cephalosporins or carbapenems, fluoroquinolones, or clindamycin and other lincosamides; ATC codes: J1D, J1M, and J1FF), had received a mediumrisk antibiotic but not a high-risk antibiotic (defined as penicillins, sulfonamides and trimethoprim, macrolides and streptogramins, or aminoglycosides; ATC codes: J1C, J1E, J1FA, J1FG, and J1G), or had received no antibiotics or a low-risk antibiotic only (defined as receipt of tetracyclines; ATC code: J1A). Patient Risk Factors Patient age, sex, admission unit (classified as medical, surgical, or oncologic), and number of previous admissions were retrieved from hospital administrative records. Any patient re- jamainternalmedicine.com (Reprinted) JAMA Internal Medicine April 1 Volume 17, Number 4 627

Research Original Investigation Clostridium difficile Infection ceiving insulin or an antidiabetic medication (ATC code: A) at any point during any hospitalization was considered diabetic. We also examined the use of antacids (ATC code: A2), chemotherapeutic agents (ATC code: L1), and feeding tubes (gastric, nasogastric, or jejunostomy). To account for the time delay between transient pharmaceutical exposures and C difficile infection risk, we measured receipt in any of the previous days rather than receipt on a given day. Hospital Ward Risk Factors Using hospital bed assignment information, we identified the ward occupants for each inpatient day; when a patient was located in multiple wards on a given day, we considered that patient to be an occupant of the ward on which he or she was located at noon. We calculated ward-level risk factors that represented mean characteristics of the ward patient population during the 46-month study period. The following wardlevel measures were retrieved from the hospital information system: age (mean age), antibiotic use in days of therapy (DOTs) per patient-days, antacid use (DOTs per patientdays), chemotherapeutic agent use (DOTs per patientdays), and feeding tube use (tube in situ per patientdays). Within each ward, observer nurses measured hand hygiene adherence at specific hand hygiene moments (before entering patient room, after leaving patient room, before aseptic procedure, and after body fluid contact) on a quarterly basis through the study period, as per provincial guidelines. 17 Adherence was pooled across periods and hand hygiene moments and was reported as a percentage of total hand hygiene opportunities. Statistical Analysis Patient Risk Factors To estimate the effect of individual risk factors on C difficile infection risk, we developed a Poisson regression model that aimed to predict the time elapsed from hospital admission to the occurrence of a first C difficile infection. Our data were structured in counting process format with one record for each patient-day. The crude incidence rate ratio was assessed in a Poisson regression for each of the 12 individual-level risk factors. Hospital Ward Risk Factors Using the ward-level risk factors above in addition to wardlevel C difficile infection incidence, we developed bivariate inverse-variance weighted linear mixed-effects regression models to estimate the effect of each ward-level factor on C difficile infection incidence, which were fitted using the Hartung-Knapp-Sidik-Jonkman method. 18 We also considered the best-fitting, 2-covariate, ward-level model by comparing model Akaike Information Criterion for all 1 two-covariate models. As a sensitivity analysis, we examined the association between ward-level antibiotic use and C difficile infection incidence among the 11 non intensive care unit (ICU) wards separately, excluding the ICU wards. To clearly distinguish patient-level and ward-level antibiotic effects, we measured the association between wardlevel antibiotic prescribing and C difficile risk separately in patients with and without direct recent antibiotic exposure. We tested whether there was a difference in association of wardlevel antibiotic use and C difficile risk between the 2 groups using the Δ method. Multilevel Model To assess the independent effect of individual exposures and aggregate ward-level antibiotic exposure, we developed a multilevel Poisson regression model with random intercepts corresponding to wards. The multilevel model included 8 individual-level risk factors: time since admission (modeled as a spline with a knot at days for first admission and readmission separately), patient age (per -year increase), sex, diabetes mellitus, and individual exposure to antibiotics, gastric acid inhibitors, chemotherapeutic agents, and presence of a feeding tube. The number of adjustment factors was restricted to ensure at least events per covariate, 19 and the selection of covariates was based on established associations with C difficile infection risk. 2, Analyses were conducted using R statistical software, version 3..2 (R Foundation for Statistical Computing); the glm, rma, and glmer functions were used for the unadjusted, bivariate mixed-effects and the multivariate mixed-effects statistical models, respectively. Results Inpatient Cohort We identified 34 298 patients who had an acute care hospital stay that exceeded 2 days at Sunnybrook Hospital from June 1,, through March 31, 14. These patients spent 428 88 patient-days in the 16 study wards during the 46-month study period. The median age of the cohort was 68.4 years (interquartile range, 4.3-81. years), whereas 9718 (28.3%) of the 34 298 patients had additional admissions. Patients received antibiotics in 21 239 (4.%) of 46 661 admissions and had feeding tubes in 476 (.2%) of 46 661 admissions. Patients Developing C difficile Infection We identified 2 patients developing a new-onset C difficile infection during the 46-month study period (incidence rate,.9 per patient-days; 9% CI,.26-6.73). Cases were distributed across 3 types of admitting services, with 111 among patients admitted via surgery services, 1 admitted via medicine services, and 34 admitted via oncology services. Individual Patient Characteristics and the Risk of Infection The incidence rates for patients with and without individuallevel risk factors are given in Table 1. The individual-level risk factors associated with C difficile infection were age, readmission, direct exposure to antibiotics, and use of a feeding tube. Each -year increase in age was associated with a 1.7-fold increase in C difficile infection risk (9% CI, 1.-1.17). Having a previous admission was associated with a 1.42-fold increase in risk (9% CI, 1.-1.82). 628 JAMA Internal Medicine April 1 Volume 17, Number 4 (Reprinted) jamainternalmedicine.com

Clostridium difficile Infection Original Investigation Research Table 1. Individual-Level Risk Factors and Clostridium difficile Infection Incidence Risk Factor Age, y No. of Cases No. of Patient-days Incidence per Patient-days (9% CI) < 3 66 3.3 (3.8-7.3) -9 24 321 4.3 (2.9-6.) 6-69 43 8 493.3 (4.-7.2) 7-79 63 93 691 6.7 (.3-8.6) Relative Risk (9% CI) 1.7 (1.-1.17) a 8 9 132 3 6.8 (.-8.3) Sex Male 14 23 784.9 (.-7.) 1. (.78-1.27) Female 11 192 84 6. (.-7.2) 1 [Reference] Hospitalization First hospitalization 286 846.2 (4.-6.1) 1 [Reference] Additional 141 742 7.4 (6.1-9.) 1.42 (1.-1.82) admission(s) Admission service Medicine 1 188 869.8 (4.8-7.) 1 [Reference] Oncology 34 48 93 6.9 (.-9.7) 1.19 (.81-1.88) Surgery 111 19 766.8 (4.8-7.) 1. (.77-1.34) Diabetes mellitus No 176 297 978.9 (.1-6.8) 1 [Reference] Yes 79 13 6 6. (4.9-7.) 1.2 (.79-1.34) Any antibiotic No 63 191 64 3.3 (2.6-4.2) 1 [Reference] Yes 192 236 984 8.1 (7.-9.3) 2.46 (1.8-3.28) Antibiotic risk index None or low 6 196 73 3.3 (2.6-4.2) 1 [Reference] Medium 23 47 663 4.8 (3.2-7.3) 1.46 (.9-2.34) High 167 184 82 9. (7.8-.) 2.73 (2.-3.63) Antacid exposure in previous d No 6 11 764.6 (4.4-7.2) 1 [Reference] Yes 19 312 824 6.1 (.3-7.) 1.8 (.82-1.43) Chemotherapeutic agent exposure in previous d No 214 37 439.7 (.-6.) 1 [Reference] Yes 41 3 149 7.7 (.7-.) 1.3 (.97-1.89) Feeding tube in situ in previous d No 176 33 733.3 (4.6-6.2) 1 [Reference] Yes 79 97 8 8.1 (6.-.1) 1.2 (1.16-1.98) a Per -year increase in age. Ward Characteristics and C difficile Infection Incidence The 16 study wards included 2 level II ICUs (patients requiring detailed observation) and 3 level III ICUs (patients requiring advanced respiratory support), 2 cardiology wards, 4 internal medicine wards, 4 surgery wards, and 1 oncology ward. Ward-level characteristics are given in Table 2. The rate of antibiotic use in wards varied from 21.7 DOTs per patient-days in ward 6 to 6.4 DOTs per patient-days in ICU 3 (median, 3.6 DOTs per patient-days; interquartile range, 26.6-36.9 DOTs). Mean antibiotic use in the ICUs was 47.2 DOTs per patient-days compared with 3.9 DOTs per patient-days in non-icu wards (P <.1). At the ward level, antibiotic use was the strongest predictor of C difficile infection incidence (Figure 1). Each % increase in ward-level antibiotic use was associated with an increased incidence of C difficile infection of 2.1 per patient-days (slope = 2.1, P <.1, R 2 =.). The largest negative outlier in the association was ICU, which was the hospital burn ICU. Rate of ward-level feeding tube exposure was marginally associated with C difficile infection (slope =.9, P =.,R 2 =.11). Other ward-level factors, including hand hygiene adherence, mean inpatient age, and rates of antacid use and chemotherapeutic agent use, were not significantly associated with C difficile infection incidence. The addition of any of the other 4 ward-level factors to the model with ward-level antibiotic use did not alter the association between wardlevel antibiotic use and C difficile infection incidence (data not shown). When we examined the association between jamainternalmedicine.com (Reprinted) JAMA Internal Medicine April 1 Volume 17, Number 4 629

Research Original Investigation Clostridium difficile Infection Table 2. Variation in Ward-Level Characteristics and Clostridium difficile Infection Incidence Intensive Care Unit Non Intensive Care Unit Ward Variable 1 2 3 4 a 6 7 8 9 11 12 13 14 1 16 Patient-days, in 12.4 7.6.8 13.1 11.2 18.4 36.9 26.8 39.8 33.9.7 38.6 39.6 39.1 36.3 33.6 thousands Hand hygiene, % 87.2 89. 84.6 8.4 92.9 84.8 8. 88.1 88.4 8.6 86.9 9.4 86.3 87. 86.2 87.3 Mean age, y 6.4 68.4 63.8 69..3 7.9 71.4 76.7 76.7 74.6 76.1 64. 67.9 8.2 66. 64. Medication receipt, DOTs per patient-days Antibiotics 34.3.7 6.4 47.4 44.6 21.7 27.9 33.8 26.6 3.6 33. 39.6 26.3 29.2 3.1 36.9 Antacids.9 3.8 94.4 93.7 76.1 71.3 8.9 9.6. 66.8 6.1 62.8 71.2 8.6 6.1 49.6 Chemotherapeutic 13.6 8. 6.7 2.6 3..8 1.9 1.9 1. 1.4 1.9 9.8.4 14. 12.3 14. agents Feeding tube, tube-days per patient-days 29.6 23. 8.2 2. 38. 12. 4. 1.2 13.8 4.8 4.6 4.2. 1.4 3.3 6. C difficile infections No. 8 1 22 14 3 9 11 1 18 23 27 12 19 27 22 Incidence per patient-days 6.4 19.7.6.7 2.7 4.9 3..6 4. 6.8 4.8 7. 3. 4.9 7.4 6.6 Abbreviation: DOTs, days of therapy. a Burn intensive care unit. Figure 1. Association of Ward-Level Exposures With Ward Clostridium difficile Infection (CDI) Incidence A B 2 2 CDI Incidence per Patient-days 1 Burn intensive care unit Other intensive care units Non-intensive care wards CDI Incidence per Patient-days 1 2 3 3 4 4 6 Ward Antibiotic Use, DOTs per Patient-days 84 86 88 9 92 94 Ward Hand Hygiene Adherence, % C 2 D 2 CDI Incidence per Patient-days 1 CDI Incidence per Patient-days 1 4 6 7 8 9 Ward Antacid Use, DOTs per Patient-days 4 6 8 Ward Feeding Tube Use, Tubes In Situ per Patient-days A, Antibiotic use; B, hand hygiene; C, antacid use; and D, feeding tube use. Each symbol represents a hospital ward. The size of the symbols is proportional to the amount of follow-up time on each ward. DOTs indicates days of therapy. 63 JAMA Internal Medicine April 1 Volume 17, Number 4 (Reprinted) jamainternalmedicine.com

Clostridium difficile Infection Original Investigation Research ward-level antibiotic use and C difficile infection incidence among the 11 non-icu wards, the association remained statistically significant (slope = 4.1, P =.3, R 2 =.41). We separately measured the association between C difficile infection incidence and ward antibiotic exposure rate among patients recently exposed and those not recently exposed to antibiotics (Figure 2). Each % increase in wardlevel antibiotic use was associated with a 1.8 per increase (slope = 1.8, P =., R 2 =.) in the incidence of C difficile infection among patients without direct recent exposure and a 1.6 per increase (slope = 1.6, P =., R 2 =.14) among patients with direct recent exposure to antibiotics. The effect of ward-level antibiotic exposure on C difficile infection incidence did not differ significantly between patients directly using or not directly using antibiotics (P =.16). Ward-Level Antibiotic Use and C difficile Infection Incidence: Multilevel Model After adjustment for patient characteristics, the ward-level antibiotic exposure remained associated with C difficile infection risk (Table 3). Each % increase in ward antibiotic exposure rate was associated with a 1.34-fold increase in C difficile infection risk (9% CI, 1.16-1.7). Discussion In this 46-month cohort study of C difficile infection risk, we found that ward-level antibiotic exposure is the main risk factor for infection. The effect of antibiotic prescribing reaches beyond individual-level antibiotic use, such that all patients, irrespective of whether they receive antibiotics directly, are at higher risk of C difficile infection in high antibioticprescribing wards. Ward-level C difficile infection risk was not confounded by other ward-level aggregate patient characteristics, including antacid use, chemotherapy, feeding tube presence, age, or crowding, or by individual-level patient comorbidities or antibiotic exposures. This is the first study, to our knowledge, to consider wardlevel antibiotic exposure as a risk factor for C difficile infection. In a previous multilevel study considering individual- and hospital-level risk factors, Pakyz et al 11 found that hospital-level antibiotic exposure rates were not a significant predictor of hospital-level C difficile infection incidence. This finding suggests that hospital-level antimicrobial use may not differ meaningfully across centers or that factors that were not considered, such as infection control practices or C difficile diagnostic testing rate, 21 may have confounded an underlying association. We hypothesize that the marked effects of ward-level antibiotic exposure rate are likely explained by an increase in the number of patients colonized with, and shedding, C difficile in wards with high rates of antibiotic use. This high prevalence of antibiotic use would increase environmental contamination and the incidence of C difficile infection. This mechanism is supported by research indicating that antibiotic exposure is the principal risk factor for C difficile colonization and that approximately half of C difficile strains among C difficile infection cases in hospitals cannot be genetically linked Figure 2. Ward Clostridium difficile Infection (CDI) Incidence and Antibiotic Use Across Hospital Wards and Among Patients With and Without Direct Antibiotic Exposure CDI Incidence per Patient-days 2 1 6 Patients with direct antibiotic exposure Patients without direct antibiotic exposure 9 14 13 7 14 8 13 1 6 7 9 11 11 1 1 16 8 1 16 2 3 3 4 4 6 Ward Antibiotic Use, DOTs per Patient-days Each pair of numbered symbols represents the incidence of C difficile infection among the subset of patients who received antibiotics (diamonds) and those who did not (circles) within a given ward. For correspondence of ward identifiers, see Table 2. DOTs indicates days of therapy. Table 3. Patient- and Ward-Level Risk Factors for Clostridium difficile Infection From a Multilevel Model Risk Factor Relative Risk (9% CI) Patient-level risk factors Age (per -y increment) 1.12 (1.3-1.21) Male sex.98 (.76-1.2) Diabetes mellitus 1. (.76-1.31) Admission unit Oncology.97 (.64-1.46) Surgery 1. (.7-1.32) Medication history in previous d Antibiotics 2.2 (1.-2.72) Antacids.84 (.62-1.13) Chemotherapeutic agents 1.48 (1.4-2.11) Feeding tube in situ in previous d 1.14 (.82-1.8) Ward-level risk factors Antibiotic exposure rate (per % increase) 1.34 (1.16-1.7) to previously identified symptomatic patients. 8 The hospital burn center was the only outlier, with lower-than-expected C difficile infection incidence given its high ward-level antibiotic use. The burn center is unique in that it had a low nursepatient staffing ratio, single-bed rooms, and a younger patient population, which is consistent with findings from a prior study. 22 Our multilevel statistical model revealed that younger age and patient pharmaceutical exposures did not completely account for the lower-than-expected incidence in the burn ICU, suggesting that other patient or ward characteristics may have been be at play. The independent association of ward antibiotic exposure with C difficile infection risk most likely reflects the nonindependence of communicable disease cases. Communicable diseases differ from other classes of disease because a case is also a risk factor. 23 In the context of C difficile infection, this state- 12 12 4 4 2 2 3 3 jamainternalmedicine.com (Reprinted) JAMA Internal Medicine April 1 Volume 17, Number 4 631

Research Original Investigation Clostridium difficile Infection ment means that an increase in disease-related force of infection could occur via antibiotic exposure in individuals who never themselves become symptomatic cases. 24 Such indirect effects are well recognized with communicable disease control interventions, and indeed this effect may be conceptualized as an inverse of herd immunity seen with vaccines. 2 Analogously, beneficial herd effects would logically be seen in wards with reduced antibiotic prescribing, as was observed in our study. A previous meta-analysis 26 of antimicrobial stewardship interventions lends credibility to this explanation because these interventions have produced substantial reductions in C difficile infection incidence with only small reductions in antibiotic prescribing. As such, the principal clinical implication of this study is that aggregate ward-level antibiotic use should be subject to surveillance by infection control and stewardship personnel. Hospital antimicrobial stewardship programs consistently achieve substantial reductions (22%-36%) in overall antibiotic use, 27 and such interventions reduce C difficile infection incidence by %. 26 Because almost all antibiotics are associated with increased C difficile infection risk, 4 antimicrobial stewardship initiatives aiming to reduce infection incidence should aim to reduce overall antimicrobial exposure in addition to reducing use of specific high-risk agents. Furthermore, our results suggest that hand hygiene with soap and water should be considered before and after caring for patients using antibiotics, especially in ICU wards with high levels of antibiotic use. Like any observational study, ours was subject to a number of limitations, including confounding by unmeasured patient characteristics, such as comorbidities, and outcome ascertainment bias related to potential systematic differences in physicians vigilance for detecting milder cases of infection. This was a single-hospital study, and the overall number of wards at our study hospital was small (n = 16). Furthermore, our study was subject to limitations because of incomplete follow-up information on patients after hospital discharge. We considered patients who were discharged as censored, but patient censoring may not have been independent of the study outcome. 28 Conclusions Our 46-month study of inpatient C difficile infection risk across 16 wards of a large tertiary care hospital found a strong association between ward antibiotic prescribing and C difficile infection incidence that affected patients with and without recent antibiotic exposure. Future studies of C difficile infection etiology should seek to quantify patient, ward, and airborne contamination with C difficile spores to more clearly describe the mechanisms that link ward-level antimicrobial use and infection incidence. These findings strongly support the further funding and development of hospital antibiotic stewardship programs. ARTICLE INFORMATION Accepted for Publication: November 8, 14. Published Online: February 23, 1. doi:.1/jamainternmed.14.8273. Author Contributions: Drs Brown and Daneman had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Brown, Valenta, Fisman, Daneman. Acquisition, analysis, or interpretation of the data: Brown, Simor, Daneman. Drafting of the manuscript: Brown. Critical revision of the manuscript for important intellectual content: All authors. Statistical analysis: Brown. Obtained funding: Brown, Fisman. Administrative, technical, or material support: Brown, Valenta, Simor, Daneman. Study supervision: Fisman, Daneman. Conflict of Interest Disclosures: Dr Simor reported receiving honoraria for speaking on behalf of Optimer Pharmaceuticals Inc Canada (now Cubist Pharmaceuticals). No other disclosures were reported. Funding/Support: This study was supported by a doctoral research award from the Canadian Institutes for Health Research (Dr Brown). Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication. Additional Contributions: Marion Elligsen, BScPhm, developed the infection prevention and control database and aided in antimicrobial coding. Dariusz Pajak, BASc, CPHI(C), provided important information on hospital ward configuration. REFERENCES 1. Owens RC Jr, Donskey CJ, Gaynes RP, Loo VG, Muto CA. Antimicrobial-associated risk factors for Clostridium difficile infection. Clin Infect Dis. 8; 46(suppl 1):S19-S31. 2. Stevens V, Dumyati G, Fine LS, Fisher SG, van Wijngaarden E. Cumulative antibiotic exposures over time and the risk of Clostridium difficile infection.clin Infect Dis. 11;3(1):42-48. 3. Slimings C, Riley TV. Antibiotics and hospital-acquired Clostridium difficile infection: update of systematic review and meta-analysis. J Antimicrob Chemother. 14;69(4):881-891. 4. Brown KA, Khanafer N, Daneman N, Fisman DN. Meta-analysis of antibiotics and the risk of community-associated Clostridium difficile infection. Antimicrob Agents Chemother. 13;7 ():2326-2332.. Baxter R, Ray GT, Fireman BH. Case-control study of antibiotic use and subsequent Clostridium difficile associated diarrhea in hospitalized patients. Infect Control Hosp Epidemiol. 8;29 (1):44-. 6. Sethi AK, Al-Nassir WN, Nerandzic MM, Bobulsky GS, Donskey CJ. Persistence of skin contamination and environmental shedding of Clostridium difficile during and after treatment of C. difficile infection. InfectControlHospEpidemiol. ;31(1):21-27. 7. Dubberke ER, Reske KA, Olsen MA, et al. Evaluation of Clostridium difficile associated disease pressure as a risk factor for C difficile associated disease. Arch Intern Med.7; 167():92-97. 8. Eyre DW, Cule ML, Wilson DJ, et al. Diverse sources ofc. difficile infection identified on whole-genome sequencing. N Engl J Med. 13;369 (13):119-1. 9. Riggs MM, Sethi AK, Zabarsky TF, Eckstein EC, Jump RLP, Donskey CJ. Asymptomatic carriers are a potential source for transmission of epidemic and nonepidemic Clostridium difficile strains among long-term care facility residents. Clin Infect Dis. 7;4(8):992-998.. Guerrero DM, Becker JC, Eckstein EC, et al. Asymptomatic carriage of toxigenic Clostridium difficile by hospitalized patients. JHospInfect.13; 8(2):1-18. 11. Pakyz AL, Jawahar R, Wang Q, Harpe SE. Medication risk factors associated with healthcare-associated Clostridium difficile infection: a multilevel model case-control study among 64 US academic medical centres. J Antimicrob Chemother. 14;69(4):1127-1131. 12. Diez Roux AV, Aiello AE. Multilevel analysis of infectious diseases. J Infect Dis. ;191(suppl 1): S2-S33. 13. Provincial Infectious Diseases Advisory Committee. Best Practices Document for the Management ofclostridium difficile in All Health Care Settings Protecting Patients and Staff. Toronto, ON: Ministry of Health and Long-Term Care; 7. 14. McDonald LC, Coignard B, Dubberke E, Song X, Horan T, Kutty PK; Ad Hoc Clostridium difficile 632 JAMA Internal Medicine April 1 Volume 17, Number 4 (Reprinted) jamainternalmedicine.com

Clostridium difficile Infection Original Investigation Research Surveillance Working Group. Recommendations for surveillance of Clostridium difficile associated disease. Infect Control Hosp Epidemiol. 7;28(2): 14-14. 1. Zar FA, Bakkanagari SR, Moorthi KMLST, Davis MB. A comparison of vancomycin and metronidazole for the treatment of Clostridium difficile associated diarrhea, stratified by disease severity. Clin Infect Dis.7;4(3):32-37. 16. WHO Collaborating Centre for Drug Statistics Methodology, Norwegian Institute of Public Health. Guidelines for ATC classification and DDD assignment 14. Oslo, Norway: WHO Collaborating Centre for Drug Statistics Methodology, Norwegian Institute of Public Health; 13. http://www.whocc.no /atc_ddd_publications/guidelines/. Accessed June, 14. 17. Sax H, Allegranzi B, Chraïti M-N, Boyce J, Larson E, Pittet D. The World Health Organization hand hygiene observation method. Am J Infect Control. 9;37():827-834. 18. IntHout J, Ioannidis JP, Borm GF. The Hartung-Knapp-Sidik-Jonkman method for random effects meta-analysis is straightforward and considerably outperforms the standard DerSimonian-Laird method. BMC Med Res Methodol. 14;14(1):2. 19. Peduzzi P, Concato J, Feinstein AR, Holford TR. Importance of events per independent variable in proportional hazards regression analysis, II: accuracy and precision of regression estimates. J Clin Epidemiol. 199;48(12):3-1.. Bignardi GE. Risk factors for Clostridium difficile infection.jhospinfect. 1998;4(1):1-1. 21. Brown KA, Fisman DN, Daneman N. Hospital Clostridium difficile infection testing rates: is don t ask, don t tell at play? Infect Control Hosp Epidemiol. 14;3(7):911-912. 22. Crabtree SJ, Robertson JL, Chung KK, et al. Clostridium difficile infections in patients with severe burns.burns. 11;37(1):42-48. 23. Giesecke J. Modern Infectious Disease Epidemiology. 2nd ed. New York, NY: Oxford University Press; 2. 24. Vynnycky E. An Introduction to Infectious Disease Modelling. New York, NY: Oxford University Press;. 2. Halloran ME, Longini IM Jr, Struchiner CJ. Design and interpretation of vaccine field studies. Epidemiol Rev. 1999;21(1):73-88. 26. Feazel LM, Malhotra A, Perencevich EN, Kaboli P, Diekema DJ, Schweizer ML. Effect of antibiotic stewardship programmes on Clostridium difficile incidence: a systematic review and meta-analysis. J Antimicrob Chemother. 14;69(7):1748-174. 27. Dellit TH, Owens RC, McGowan JE Jr, et al; Infectious Diseases Society of America; Society for Healthcare Epidemiology of America. Infectious Diseases Society of America and the Society for Healthcare Epidemiology of America guidelines for developing an institutional program to enhance antimicrobial stewardship. Clin Infect Dis.7;44 (2):19-177. 28. Schumacher M, Allignol A, Beyersmann J, Binder N, Wolkewitz M. Hospital-acquired infections: appropriate statistical treatment is urgently needed!int J Epidemiol. 13;42():2-8. Invited Commentary Uncovering the Role of Antibiotics in the Transmission of Multidrug-Resistant Organisms L. Clifford McDonald, MD Conventional wisdom has suggested 2 distinct categories of epidemiologic risk factors in the development of Clostridium difficile infection (CDI): factors that increase the risk of transmission of C difficile and factors that disrupt the patient s lower intestinal microbiota, a major host defense against infection. This host defense func- Related article page 626 tion may be best understood in terms of the expression of these microorganisms collective and representative genome, known as the microbiome. Although antibiotics appear to be the major disruptive force of the microbiome in hospitalized patients, evidence indicates that other medications, such as proton pump inhibitors and antidepressants, and chronic conditions, such as obesity, 1 may also be associated with microbiome disruption and/or CDI. In addition to increasing the risk of infection, the microbiome disruption from antibiotics may also increase C difficile transmission via increased likelihood of asymptomatic colonization and, once colonized, increasing clonal expansion and domination of the microbiota by C difficile. Meanwhile, there is increasing evidence pointing to the importance of asymptomatic carriers in the transmission of C difficile in hospitals. However, few studies have examined the epidemiology of antibiotics affecting transmission of C difficile among patients, something Brown et al 2 have addressed in this issue of JAMA Internal Medicine. This study examined an individual acute care hospital cohort for 4 years, capturing individual-level risk factors, such as age, sex, previous admission, and inpatient medication exposures, including but not limited to antibiotic exposures. In addition, mean characteristics of the ward or unit population during the 46-month study period were recorded, including mean age, feeding tube use, and antibiotic, chemotherapeutic, and antacid medication use in days of therapy per patient-days. Other ward- and unit-level risk factors included patient density and hand hygiene adherence. Multivariable models and, most important, a multilevel model were constructed in which patient and ward factors were examined together in regard to their increasing risk of CDI. The major finding was that each % increase in overall ward or unit antibiotic exposure was independently associated with a 34% increase in CDI. Other previously described patient risk factors associated with individual CDI risk in the multilevel model included age and antibiotic, chemotherapy, and feeding tube exposures in the preceding 7 days. The main finding of this study reveals how antibiotics, by affecting the microbiomes of a subset of patients across a population (patients in wards or units of a hospital), puts the entire population, including those who do not receive antibiotics, at increased risk via increased transmission. The converse is also true; if unnecessary antibiotic use is decreased through improved stewardship, it will lead to a proportionate decrease in CDI. This same indirect effect of disrupting the microbiome of neighboring patients, rendering them more at risk for asymptomatic colonization and, once colonized, at increased risk for transmission, may be an important role for antibiotics in the epidemiology of a number of other multidrug-resistant organisms, including carbapenem-resistant Enterobacteriaceae and vancomycin-resistant enterococci. jamainternalmedicine.com (Reprinted) JAMA Internal Medicine April 1 Volume 17, Number 4 633