Case‐Control Study of Antibiotic Use and Subsequent Clostridium difficile–Associated Diarrhea in Hospitalized Patients
Objective. To determine which antibiotics increase or decrease the risk of Clostridium difficile–associated diarrhea (CDAD).
Design. Retrospective case‐control study.
Setting. Nonprofit, integrated healthcare delivery system in Northern California.
Patients. Study participants included patients with cases of hospital‐acquired CDAD that occurred during the period from 1999 through 2005 (
) and control patients (
) matched for facility, calendar quarter during which hospitalization occurred, diagnosis related group for the index hospitalization, and length of hospital stay. All case and control patients had received antibiotics in the 60 days before the index date. For each antibiotic, the risk of CDAD was examined in relation to whether the patient received the antibiotic, after adjustment for use of other antibiotics, demographic characteristics, selected health conditions, and use of healthcare services.
Results. The following antibiotics were associated with a significantly increased risk of acquiring CDAD: imipenem‐cilastin (odds ratio [OR], 2.77), clindamycin (OR, 2.31), cefuroxime (OR, 2.16), moxifloxacin (OR, 1.88), ceftazidime (OR, 1.82), cefpodoxime (OR, 1.58), ceftizoxime (OR, 1.57), and ceftriaxone (OR, 1.49). Metronidazole and doxycycline were associated with a significantly reduced risk of CDAD (OR for metronidazole, 0.67; OR for doxycycline, 0.41). Other factors associated with an increased risk of CDAD were older age, longer hospital stays, use of proton pump inhibitors, prior gastrointestinal disease, and prior infection (not including C. difficile infection.)
Conclusions. Some antibiotics appear to increase the risk of acquiring CDAD, notably clindamycin, third‐generation cephalosporins, and carbapenems, whereas metronidazole and doxycycline appear to be protective, compared with other antibiotics.
Received May 22, 2007; accepted August 22, 2007; electronically published November 21, 2007.
Clostridium difficile is the most common cause of hospital‐acquired, antibiotic‐associated diarrhea. Carriage of the organism increases markedly with debilitation and hospitalization.1 The incidence of this disease is increasing in many areas of the world,2,3 and a new strain with increased virulence is emerging as a major cause of hospital‐associated morbidity throughout the Western hemisphere.4‐6
The most consistently reported risk factors for C. difficile–associated diarrhea (CDAD) are advanced age, underlying disease severity, frailty, duration of hospitalization, and exposure to antibiotics.7‐15 Numerous articles have addressed the question of which antibiotics are associated with increased risk of CDAD,8,12,15‐25 but concern exists about the validity of these studies. Previous studies have had small sample sizes, used inappropriate or inadequate controls, and/or have not controlled well for comorbidity and other confounders.11 In their comprehensive review of the literature, Thomas, et al.11 concluded that, among the 33 studies that met final inclusion criteria, only 2 provided valid evidence of the role of antibiotics in hospital‐acquired CDAD. McFarland et al.12 found an increased risk for CDAD after cephalosporin exposure and penicillin exposure, and Chang and Nelson26 found an increased risk associated with clindamycin exposure. More recent studies have continued to implicate clindamycin24 and cephalosporins5,27 and indicate that fluoroquinolone exposure may also be a risk factor for CDAD.5,16,19,21,28 This study aims to present a better model of the risk of acquiring CDAD that is associated with various antibiotics.
Methods
Setting
Kaiser Permanente of Northern California (KPNC) is a nonprofit, integrated healthcare delivery system that provides care to more than 3 million members. KPNC operates 16 hospitals throughout Northern California, each providing intensive care services and ranging in size from approximately 106 to 330 beds.
Case Patient Ascertainment
All specimens tested for C. difficile toxin were identified by KPNC’s microbiology laboratory. Testing was performed using a combination A and B toxin enzyme immunoassay (Premier Toxin A and B; Meridian Bioscience). The laboratory only performs this test on diarrheal stool samples. We extracted data on all tests ordered from January 1, 1998, through December 31, 2005, that yielded a positive result. To select patients with new C. difficile infection, we required that they not have had another test result positive for C. difficile within 1 previous year. We studied only those patients for whom the C. difficile test was ordered on or after the third day of hospitalization in a KPNC hospital. We assumed that the C. difficile test occurred within 2 days of the onset of symptoms and thus defined the index date as the date of the C. difficile test minus 2 days. Because data on antibiotic use were incomplete for people who had had stays in non‐KPNC hospitals or skilled nursing facilities, we excluded people who had received care in these settings within 60 days before the index date. Similar to other studies,8,18 we assumed that antibiotic use in the 60 days before the index date could be a risk factor for CDAD. Our goal was to determine which antibiotics are risk factors for CDAD, not whether antibiotic use versus nonuse was a predictor. Therefore, case patients were excluded if they had not received any antibiotics (as an inpatient or as an outpatient) in the 60 days before the index date. Finally, to be certain that we had complete ascertainment of other risk factors, we required case patients to have had continuous membership in KPNC in the year before the index date.
Selection of Control Patients
For each patient, we selected up to 8 matched control patients. The pool of potential control patients was all persons hospitalized in a KPNC hospital who had no test result positive for C. difficile during the study period. Control patients were matched to case patients with respect to being hospitalized at the same hospital during the same calendar year and quarter and receiving the same diagnosis related group code. Length of stay for the control patient had to be at least as long as the interval between the case patient's admission date and the date the C. difficile test was ordered for the case patient, and the day the test was ordered (minus 2 days) became the control patient’s index date. (Case patients with index dates that occurred more than 14 days into their hospitalization were matched to control patients with stays at least 14 days long.) Control patients could not have had any hospitalizations at non‐KPNC hospitals or skilled nursing facilities in the 60 days before the index date; they also had to have used antibiotics in the 60 days before the index date and had to have been continuous members of KPNC in the year before the index date.
Measures of Antibiotic Use
For each study subject, we identified which antibiotics were ordered or prescribed for the patient (as either an inpatient or an outpatient) in the 60 days before their index date, using KPNC’s pharmacy system database. Antibiotics that were not used by at least 2% of either the case or control patients were grouped together as “other antibiotics.”
Measures of Other Potential Confounders
For all study participants, we identified the number of days spent in the hospital in the 60 days before their index date, as well as whether they spent a day or more in an intensive care unit (ICU) during that time and whether they used proton pump inhibitors (PPIs) (as inpatients or outpatients). We used the Johns Hopkins ACG Case‐Mix (version 7.1) Major Expanded Diagnosis Clusters to identify patients who received clinical diagnoses in the year before their index dates (excluding diagnoses received during the index hospitalizations) that fell into clusters of conditions thought to be risk factors (or proxies for risk factors) for CDAD. As a more general measure of health status and contact with the health system, we estimated overall healthcare costs for each patient in the year before their index date. Healthcare costs are the costs (to the health plan) of all medical services (inpatient, outpatient, and pharmacy) provided to the patient and covered by the health plan, excluding only the costs of the index hospitalization and antibiotic use. Costs were estimated in a manner similar to that described in other studies conducted at KPNC.29,30
Data Analysis
We performed a sequence of analyses, incrementally adjusting for more covariates to see how the inclusion of these additional variables affected the results. Our initial analysis was a conditional logistic regression (one for each antibiotic) that adjusted only for the matching criteria (matched OR 1). We then ran a second analysis that was similar to the first analysis except that we adjusted for the patient’s use or nonuse of other antibiotics in the 60 days before the index date (matched OR 2).
In our final (and primary) analysis, we included measures of demographics, health status, and health services use, as well as antibiotic exposures (ie, age; sex; hospital‐days in the 60 days before the index date; stay in intensive care unit; use of proton pump inhibitors in 60 days before index date; log‐transformed medical costs in the year before the index date; and the presence or absences of diagnoses relating to cardiovascular disease, gastrointestinal disease, general surgery, infections, malignancies, and renal disease in the year before the index date). As in the earlier models, the dependent variable was a dichotomous variable that indicated if the person had CDAD; the key independent variables were dichotomized variables (use vs nonuse) corresponding to each antibiotic the patient used in the 60 days before the index date.
The results for any given antibiotic were adjusted for the use of all other antibiotics used by the patient. The following covariables were included in the final model: sex, age, number of hospital‐days in the 60 days before the index date, whether the patient spent time in the ICU in the 60 days before the index date, PPI use versus nonuse in the 60 days before the index date, receipt of a code for any of the 6 Major Expanded Diagnosis Cluster groupings in the year before the index date, and the patient’s medical costs in the year before the index date. Age and number of hospital‐days were treated as continuous variables. Medical costs were log‐transformed to reduce skewness. In multivariate analysis, the odds ratio (OR) for each antibiotic estimated the relative risk of CDAD when that antibiotic was used, compared with when that antibiotic was not used. We tested using number of days of use as the measure of antibiotic use rather than the dichotomous measure, and results were similar. However, the overall model fit (by the log‐likelihood χ2 test) was better with the dichotomous measures. To investigate possible changes over time in the relationship between CDAD and antibiotics, we performed a post hoc analysis in which we included time and drug‐interaction effects, time was a dichotomous variable that indicated whether hospitalization occurred in the years 1999‐2002 or 2003‐2005. We also ran 2 separate models, 1 for each period. All analyses were performed using SAS software (SAS Institute). All differences discussed in the text are significant at P of .05 or less.
Results
During the period from 1999 through 2005, nearly 1.1 million patients were admitted to a KPNC hospital. A total of 2,355 patients met this study’s inclusion criteria for CDAD case patients, of which 2,134 had used antibiotics in the prior 60 days. We found at least 1 matched control for 1,142 (54%) of these patients. A total of 412 case patients were matched to 1 control, 220 to 2 controls, 150 to 3 controls, 104 to 4 controls, and 256 to 5 to 8 controls. Case patients were more likely to be male (51% vs 47%) (Table 1) and older (69 vs 68 years of age). In the 60 days before their index dates, case patients had spent more days in the hospital (mean, 13; SD, 12; vs mean, 9; SD, 7), were more likely to have been admitted to the ICU (37% vs 31%), were more likely to have used PPIs (40% vs 28%), and used more types of antibiotics (mean ± SD, 
vs 
different antibiotics) than did controls. In the 60 days before the index date, case patients were more likely to have received diagnoses relating to cardiovascular disease, gastrointestinal disease, surgery, infections, and renal disease. Case patients also had significantly higher medical costs than controls in the year before their index dates.
Table 2 presents data relating to case and control patients' use of antibiotics in the 60 days before their index dates, as well as the results of the series of analyses estimating the relationship between the receipt of antibiotics and CDAD. The results of the 3 sets of analyses are fairly consistent with one another—both with respect to the relative effect size of the different antibiotics and with respect to the absolute magnitude of the ORs. We infer from this that the matching criteria eliminated much of the potential confounding, and not much residual confounding occurred that was related to concurrent antibiotic use or to the demographic and health‐related covariables. We found that the use of imipenem‐cilastin, clindamycin, cefuroxime, moxifloxacin, ceftazidime, ceftizoxime, and/or ceftriaxone was a predictor of increased odds of CDAD. The use of metronidazole and/or doxycycline was a predictor of decreased odds of CDAD.
Other risk factors that were predictors of CDAD in the multivariate model were age (OR, 1.01 [95% confidence interval {CI}, 1.01‐1.02]), time spent in the hospital during the 60 days before the index date (OR, 1.04; [95% CI, 1.03‐1.06]), having received a diagnosis of gastrointestinal disease (OR, 1.21 [95% CI, 1.01‐1.45]) or infection (OR, 1.31 [95% CI, 1.06‐1.62]) in the year before the index date, and the use of PPIs in the 60 days before the index date (OR, 1.23 [95% CI, 1.03‐1.48]).
In the post hoc analysis, we found that the inclusion of time and drug‐interaction terms did not significantly improve the overall fit of the model (
). However, the association with CDAD differed for a few antibiotics between the earlier period (1999‐2002) and the later period (2003‐2005). Most notably, the association of meropenem with CDAD was higher in the later period, compared with the earlier period (
). Both imipenem‐cilastin and meropenem were positively associated with CDAD when we restricted the analysis to patients with index dates in the later period (OR for imipenem‐cilastin, 5.1 [95% CI, 1.4‐18]; OR for meropenem, 2.6 [95% CI, 1.1‐6.7]). Levofloxacin, on the other hand, was significantly associated with CDAD in the earlier period (OR, 1.5 [95% CI, 1.0‐2.2]), but not in the later period (OR, 0.7 [95% CI, 0.3‐1.4]).
Discussion
This large case‐control study found that use of certain antibiotics (imipenem‐cilastin, clindamycin, cefuroxime, moxifloxacin, ceftazidime, ceftizoxime, and/or ceftriaxone) is associated with increased risk of CDAD. The results are also consistent with prior research findings that hospital‐acquired CDAD is related to age, time spent in the hospital (before the index date), use of PPIs, and gastrointestinal disease.
In the previously published literature, clindamycin and cephalosporins appear to be the antibiotics most frequently implicated in relation to CDAD,11 with fluoroquinolones emerging in more recent studies.5,16,19,21,28 These studies, however, have numerous limitations, chief among which is their small sample size (only 4 of the 39 studies we identified had more than 100 case patients, and no analysis addressing research questions comparable to ours had more than 300 case patients). Most studies analyzed antibiotics by class rather than individually, although some evidence indicates that not all antibiotics within a class have similar effects with respect to the acquisition of CDAD.21 Our sample of 1,142 case patients made it by far the largest study of its kind, to our knowledge, and allowed us to study individual antibiotics and control for the individual’s use of other antibiotics during the same period.
Another criticism of previous studies is inadequate control for confounders.11 Many risk factors have been posited for CDAD other than antibiotic use.10,12 In numerous studies, the relationship between general underlying health conditions (or disease severity) and acquisition of CDAD has also been shown.9,12,28,31,32 Given gaps in our understanding of the etiology of C. difficile infection and the development of CDAD, it is difficult to know whether these various factors are important in and of themselves or whether they are proxies for other factors and/or for each other. Nevertheless, we believe that we have controlled for confounders better than prior studies, through both the case‐control matching process and regression.
Consistent with a number of previous studies, we found the use of clindamycin and certain cephalosporins (both second and third generation) to be significant risk factors for CDAD.11,12,20,28,33 We also found the use of imipenem‐cilastin, a broad‐spectrum β‐lactam, to be a significant predictor of CDAD, although meropenem, another carbapenem similar to imipenem‐cilastin, was not found to be significant during all the years of the study included in the main analysis. However, in the post hoc analysis, we found that the relationship between the use of carbapenems (both imipenem‐cilastin and meropenem) and CDAD was significant in the more recent period and appeared to be growing. The use of moxifloxacin was also an important predictor of CDAD, although the use of 2 other fluoroquinolones (ciprofloxacin and levofloxacin) was not. In the post hoc analysis, the use of levofloxacin was positively associated with CDAD in the early period but not in the later period. These periods correspond to a 2002 KPNC formulary decision to replace levofloxacin with moxifloxacin, with the result that moxifloxacin use was increasing at the same time that levofloxacin use was decreasing.
We found some evidence that the association of particular antibiotics with CDAD may change over time, possibly as a result of changes in the overall patterns of use for those antibiotics in the hospital and community and changes in the proliferation of antibiotic‐ resistant clones.34 Previous studies suggest that the relationship between antibiotic use and the acquisition of CDAD may be linked to the susceptibility profile of the organism, in addition to antibiotic effects on the resident flora.35 Altered bowel flora, generally caused by the use of antibiotics, allows the establishment of C. difficile colonization and the escalation of toxin production to produce symptoms. The resistance of C. difficile to an antibiotic may allow it to multiply in vivo, competing more effectively against normal flora. So we would expect antibiotics with broad‐spectrum activity that is not effective against C. difficile to be more likely to result in disease. A newly introduced antibiotic might not have an association with CDAD if it is initially active against C. difficile, but eventually resistance develops (or a resistant clone emerges), and then association with CDAD would occur. If use of a particular antibiotic declines significantly, more susceptible clones may reemerge. We found evidence to support this supposition.
Some antibiotics, including ampicillin, trimethoprim‐sulfamethoxazole, and azithromycin, did not elevate the risk of CDAD. The appearance of a protective effect for doxycycline and metronidazole may be the result of intrinsic antimicrobial activity against the causative agent.
The patients included in this study met specific inclusion criteria (such as having to have at least 1 matched control) and are not necessarily representative of the general population at risk. Although we think it reasonable to assume that the relationships between various antibiotics and the acquisition of CDAD would be similar in patients who did not meet our inclusion criteria, this study does not address that question.
The number of antibiotic combinations used was too great to assess possible interaction effects among different antibiotics. We also did not look in detail at the potential relationship between the quantity of antibiotics received and CDAD. When we substituted our estimate of days of use for the dichotomous variables used in our primary model, we found no substantive difference in the results. Nevertheless, duration of use or quantity used might have some independent effect on the risk of acquiring CDAD.
The emergence of a new, more virulent, strain of C. difficile (BI/NAP1) increases the importance of identifying risk factors for CDAD. Because this was a retrospective study and strain typing is not routinely ordered by clinicians, we do not know which of our cases were caused by the new epidemic strain. However, the epidemic strain has been identified at KPNC in patients enrolled in a separate clinical study that was conducted from August 2005 through June 2006. During this period, 36% of C. difficile isolates recovered from stool samples collected as part of that trial were of the epidemic strain (Deborah Molrine, Massachusetts Biologic Laboratories; personal communication). The association of the epidemic strain with the use of particular antibiotics in our patient population is an area to be addressed.
Acknowledgments
Financial support. This study was supported by the Permanente Medical Group.
Potential conflicts of interest. All authors report no conflicts of interest relevant to this article.
References
- 1. Yassin SF, Young‐Fadok TM, Zein NN, Pardi DS. Clostridium difficile‐associated diarrhea and colitis. Mayo Clin Proc 2001; 76:725‐730.
- 2. Archibald LK, Banerjee SN, Jarvis WR. Secular trends in hospital‐acquired Clostridium difficile disease in the United States, 1987‐2001. J Infect Dis 2004; 189:1585‐1589.
- 3. Thomas C, Stevenson M, Williamson DJ, Riley TV. Clostridium difficile–associated diarrhea: epidemiological data from Western Australia associated with a modified antibiotic policy. Clin Infect Dis 2002; 35:1457‐1462.
- 4. McDonald LC, Killgore GE, Thompson A, et al. An epidemic, toxin gene‐variant strain of Clostridium difficile. N Engl J Med 2005; 353:2433‐2441.
- 5. Kazakova SV, Ware K, Baughman B, et al. A hospital outbreak of diarrhea due to an emerging epidemic strain of Clostridium difficile. Arch Intern Med 2006; 166:2518‐2524.
- 6. Hubert B, Loo VG, Bourgault AM, et al. A portrait of the geographic dissemination of the Clostridium difficile North American pulsed‐field type 1 strain and the epidemiology of C. difficile–associated disease in Quebec. Clin Infect Dis 2007; 44:238‐244.
- 7. Gerding DN, Johnson S, Peterson LR, Mulligan ME, Silva J Jr. Clostridium difficile–associated diarrhea and colitis. Infect Control Hosp Epidemiol 1995; 16:459‐477.
- 8. Bignardi GE. Risk factors for Clostridium difficile infection. J Hosp Infect 1998; 40:1‐15.
- 9. Kyne L, Sougioultzis S, McFarland LV, Kelly CP. Underlying disease severity as a major risk factor for nosocomial Clostridium difficile diarrhea. Infect Control Hosp Epidemiol 2002; 23:653‐659.
- 10. McFarland LV. Epidemiology, risk factors and treatments for antibiotic‐associated diarrhea. Dig Dis 1998; 16:292‐307.
- 11. Thomas C, Stevenson M, Riley TV. Antibiotics and hospital‐acquired Clostridium difficile–associated diarrhoea: a systematic review. J Antimicrob Chemother 2003; 51:1339‐1350.
- 12. McFarland LV, Surawicz CM, Stamm WE. Risk factors for Clostridium difficile carriage and C. difficile–associated diarrhea in a cohort of hospitalized patients. J Infect Dis 1990; 162:678‐684.
- 13. Dial S, Delaney JA, Barkun AN, Suissa S. Use of gastric acid‐suppressive agents and the risk of community‐acquired Clostridium difficile–associated disease. JAMA 2005; 294:2989‐2995.
- 14. Bartlett JG. Narrative review: the new epidemic of Clostridium difficile–associated enteric disease. Ann Intern Med 2006; 145:758‐764.
- 15. Climo MW, Israel DS, Wong ES, Williams D, Coudron P, Markowitz SM. Hospital‐wide restriction of clindamycin: effect on the incidence of Clostridium difficile–associated diarrhea and cost. Ann Intern Med 1998; 128:989‐995.
- 16. Pepin J, Saheb N, Coulombe MA, et al. Emergence of fluoroquinolones as the predominant risk factor for Clostridium difficile–associated diarrhea: a cohort study during an epidemic in Quebec. Clin Infect Dis 2005; 41:1254‐1260.
- 17. Modena S, Bearelly D, Swartz K, Friedenberg FK. Clostridium difficile among hospitalized patients receiving antibiotics: a case‐control study. Infect Control Hosp Epidemiol 2005; 26:685‐690.
- 18. Palmore TN, Sohn S, Malak SE, Eagan J, Sepkowitz KA. Risk factors for acquisition of Clostridium difficile–associated diarrhea among outpatients at a cancer hospital. Infect Control Hosp Epidemiol 2005; 26:680‐684.
- 19. McCusker ME, Harris AD, Perencevich E, Roghmann MC. Fluoroquinolone use and Clostridium difficile–associated diarrhea. Emerg Infect Dis 2003; 9:730‐733.
- 20. Anand A, Bashey B, Mir T, Glatt AE. Epidemiology, clinical manifestations, and outcome of Clostridium difficile–associated diarrhea. Am J Gastroenterol 1994; 89:519‐523.
- 21. Gaynes R, Rimland D, Killum E, et al. Outbreak of Clostridium difficile infection in a long‐term care facility: association with gatifloxacin use. Clin Infect Dis 2004; 38:640‐645.
- 22. Zimmerman RK. Risk factors for Clostridium difficile cytotoxin–positive diarrhea after control for horizontal transmission. Infect Control Hosp Epidemiol 1991; 12:96‐100.
- 23. Dhalla IA, Mamdani MM, Simor AE, Kopp A, Rochon PA, Juurlink DN. Are broad‐spectrum fluoroquinolones more likely to cause Clostridium difficile–associated disease? Antimicrob Agents Chemother 2006; 50:3216‐3219.
- 24. Raveh D, Rabinowitz B, Breuer GS, Rudensky B, Yinnon AM. Risk factors for Clostridium difficile toxin–positive nosocomial diarrhoea. Int J Antimicrob Agents 2006; 28:231‐237.
- 25. Wistrom J, Norrby SR, Myhre EB, et al. Frequency of antibiotic‐associated diarrhoea in 2462 antibiotic‐treated hospitalized patients: a prospective study. J Antimicrob Chemother 2001; 47:43‐50.
- 26. Chang VT, Nelson K. The role of physical proximity in nosocomial diarrhea. Clin Infect Dis 2000; 31:717‐722.
- 27. Starr JM, Martin H, McCoubrey J, Gibson G, Poxton IR. Risk factors for Clostridium difficile colonisation and toxin production. Age Ageing 2003; 32:657‐660.
- 28. Muto CA, Pokrywka M, Shutt K, et al. A large outbreak of Clostridium difficile–associated disease with an unexpected proportion of deaths and colectomies at a teaching hospital following increased fluoroquinolone use. Infect Control Hosp Epidemiol 2005; 26:273‐280.
- 29. Ray GT, Levine P, Croen LA, Bokhari FA, Hu TW, Habel LA. Attention‐deficit/hyperactivity disorder in children: excess costs before and after initial diagnosis and treatment cost differences by ethnicity. Arch Pediatr Adolesc Med 2006; 160:1063‐1069.
- 30. Ray GT, Mertens JR, Weisner C. The excess medical cost and health problems of family members of persons diagnosed with alcohol or drug problems. Med Care 2007; 45:116‐122.
- 31. Vesta KS, Wells PG, Gentry CA, Stipek WJ. Specific risk factors for Clostridium difficile–associated diarrhea: A prospective, multicenter, case control evaluation. Am J Infect Control 2005; 33:469‐472.
- 32. Changela U, Cannon JP, Aneziokoro C, Shah PS, Thottapurathu L, Lentino J. Risk factors and mortality associated with Clostridium difficile–associated diarrhoea at a VA hospital. Int J Antimicrob Agents 2004; 24:562‐566.
- 33. Loo VG, Poirier L, Miller MA, et al. A predominantly clonal multi‐institutional outbreak of Clostridium difficile–associated diarrhea with high morbidity and mortality. N Engl J Med 2005; 353:2442‐2449.
- 34. Gerding DN. Clindamycin, cephalosporins, fluoroquinolones, and Clostridium difficile–associated diarrhea: this is an antimicrobial resistance problem. Clin Infect Dis 2004; 38:646‐648.
- 35. Johnson S, Samore MH, Farrow KA, et al. Epidemics of diarrhea caused by a clindamycin‐resistant strain of Clostridium difficile in four hospitals. N Engl J Med 1999; 341:1645‐1651.
-
Presented in part: 2006 Interscience Conference on Antimicrobial Agents and Chemotherapy; San Francisco, California.