Community-associated Clostridium difficile Infections, Monroe County, New York, USA

Judicious use of antimicrobial drugs will reduce infections.

emergent disease affecting patients recently exposed to antimicrobial drugs. Community strains are similar to those found in health care settings. C lostridium diffi cile is an anaerobic, spore-forming, gram-positive bacillus that produces 2 major toxins (TcdA and TcdB). Illness caused by toxigenic C. diffi cile varies from mild diarrhea to fulminant disease and death. Infection occurs commonly in the health care setting because of concomitant exposure to the organism and antimicrobial drugs in patients with severe illnesses and concurrent conditions. Over the past several years, the incidence (1), severity and mortality rate (especially in elderly persons) (2), and treatment failure rate of C. diffi cile infection (CDI) (3) have increased. In addition, CDI has been more commonly observed in healthy persons often without known CDI risk factors (4).
The changing pattern of disease is in part being caused by the emergence of a new epidemic hypervirulent C. diffi cile strain identifi ed as North American pulsed-fi eld 1 (NAP1) by pulsed-fi eld gel electrophoresis (PFGE), BI by restriction endonuclease analysis, and 027 by PCR ribotyping (5). NAP1 strains often demonstrate resistance to quinolones, and increased use of these drugs may provide a positive selection pressure for NAP1 relative to other strains (6). The incidence, risk factors, and mode of transmission of CDI in hospital-associated (HA) disease are well described. However, few studies have examined the role of the hypervirulent NAP1 strain, antimicrobial drugs, proton pump inhibitors (PPI), and foodborne transmission on the emergence of CDI (7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19).
To defi ne the magnitude of CDI across the continuum of care (hospital, long-term care, and the community) and assess the relative incidence and possible risk factors for community-associated disease, a 6-month surveillance program for laboratory-diagnosed CDI cases was initiated in Monroe County, New York, USA, in 2008. This program was undertaken in 2 sentinel laboratories in preparation for population-based surveillance of CDI in several US states through the Emerging Infections Program of the Centers for Disease Control and Prevention (CDC). A secondary goal of this study was to compare C. diffi cile recovery rates between refrigerated fecal swab and frozen fecal specimens.

Setting
During March 1-August 31, 2008, surveillance offi cers reviewed all medical records associated with C. diffi cile toxin-positive fecal samples from 2 of 3 hospital laboratories in Monroe County, New York. These laboratories service long-term care facilities, doctors' offi ces, and the inpatient population. Electronic inpatient and outpatient medical records for all patients with fecal samples positive for C. diffi cile toxin by enzyme immunoassay were reviewed.

Case Defi nitions
Cases were subcategorized according to published surveillance defi nition guidelines (20). The date of CDI onset was defi ned as the date of the positive fecal test result, not the date of diarrhea onset. The date of diarrhea onset was not readily available in the electronic medical records. An incident case was defi ned as disease in a patient with a C. diffi cile toxin-positive fecal sample and no positive assay result in the preceding 8 weeks. Cases were considered recurrent if there was a positive assay result within 2-8 weeks of the most recent toxin-positive specimen. Positive assay results obtained within 2 weeks of the prior positive assay result were considered duplicates and excluded. Possible case-patients were also excluded if they had no documented diarrhea, an initial fecal sample representing recurrence of an episode before the start date of surveillance, were <12 months of age at the time of testing, or were not residents of Monroe County.
Cases were classifi ed into 3 categories. The fi rst category was health care facility onset (HCFO) cases. These cases were in CDI case-patients who had C. diffi cile toxin-positive fecal specimens obtained >48 hours of hospital admission or during residence in a long-term care facility.
The second category was community-onset health care-associated (CO-HCA) cases. These cases were in CDI case-patients who had C. diffi cile toxin-positive fecal specimens obtained <48 hours of hospital admission or as an outpatient and who had documented exposure to health care in the previous 12 weeks. Health care exposure was defi ned as a >24-hour stay in a hospital or 48 hours in an emergency department, residence in a long-term care facility, receipt of chronic hemodialysis, or regular hospital visits for intravenous infusion. No differentiation was made between case-patients with disease onset within 4 weeks of discharge and those with disease onset 4-12 weeks after discharge from the health care facility (i.e., indeterminate case) (20). Cases in the HCFO and CO-HCA categories were referred to as health care-associated (HA) CDI.
The third category was potential community-associated (potential CA) cases. These cases were in CDI case-patients who had C. diffi cile toxin-positive fecal specimens obtained within 48 hours of hospital admission or in an outpatient setting and who had no documented health care exposure in the 12 weeks before the positive test result. Potential CA case-patients were interviewed to confi rm the lack of health care exposures, and to assess contact with health care personnel, medication use, travel, food consumption, and exposure to animals in the previous 12 weeks. Potential CA case-patients who were interviewed and reported exposure to health care in the previous 12 weeks during the interview were reclassifi ed as having CO-HCA; those who denied such exposures were confi rmed as having CA-CDI. Patients who died, could not be reached, did not speak English, refused consent, or whose doctor refused consent were classifi ed as having probable CA-CDI.
Information was obtained on demographics; previous positive assay results; dates of admission to acute-care facilities, emergency departments, and long-term care facilities; inpatient mortality rates; and CDI complications, such as toxic megacolon, renal failure, or intensive care unit admission. The study was approved by the institutional review boards at CDC, the University of Rochester Medical Center, Rochester General Hospital, and the New York State Department of Health.

Laboratory Methods
Fecal samples (unformed) were tested at clinical laboratories for C. diffi cile toxin by using the Premier Toxins A & B enzyme immunoassay (Meridian Bioscience, Inc., Cincinnati, OH, USA). Aliquots of fecal specimens were stored at −20°C until cases were reviewed. All available fecal samples of patients classifi ed as potential CA and a random sample of the CO-HCA and HCFO (1 of each classifi cation each week) were submitted to the New York State Public Health Laboratory (Wadsworth Center Laboratory, Albany, NY, USA) for culture. For 31 randomly chosen stool specimens, an additional specimen was obtained by using a culture swab (Copan 159C or similar). To avoid germination of C. diffi cile, no enrichment or anaerobic transport media were used. Fecal swab specimens were stored at 4°C to compare the recovery rate from a swab stored at 4°C with that for a frozen fecal aliquot. Both specimen types were stored for several weeks before shipment and were shipped on dry ice.

Culture and Molecular Characterization Methods
Fecal specimens were placed on cycloserine-cefoxitin fructose agar plates and incubated at 35°C under anaerobic conditions for 96 hours. Culture-negative fecal samples and swab specimens were treated by using alcohol shock (21) and recultured on cycloserine-cefoxitin fructose agar containing 0.1% sodium taurocholate (22). Plates were examined daily for characteristic colonies. Isolates were shipped to CDC for molecular characterization, which included PCR for binary toxin and major tcdC gene deletions, toxinotyping (22,23), and SmaI PFGE. PFGE banding patterns were analyzed by using BioNumerics version 5.10 (Applied Maths, Austin, TX, USA) and compared with the CDC C. diffi cile database (24). NAP types were assigned to patterns with ≈80% similarity to established NAP clusters. MICs for clindamycin, metronidazole, moxifl oxacin, levofl oxacin, and vancomycin were determined by using the agar dilution method, and results were interpreted by using the Clinical and Laboratory Standard Institute M11-A7 breakpoint criteria (25). For surveillance purposes only, levofl oxacin MICs were interpreted by using criteria for moxifl oxacin, and vancomycin MICs were interpreted by using criteria for Staphylococcus aureus.

Statistical Analysis
All statistical analyses were performed with SAS version 9.1 (SAS Institute, Cary, NC, USA) assuming a 2-tailed α of 0.05. Univariate analysis was conducted to summarize the demographic and clinical characteristics of case-patients. Bivariate analyses were used to compare these characteristics across preliminary and fi nal classifi cations by using Wilcoxon rank sum, χ 2 , and Fisher exact tests as appropriate.

Surveillance
During the study, 366 incident CDI cases were identifi ed after excluding 558 C. diffi cile toxin-positive stool assays for patients who did not meet eligibility criteria. The distribution of cases is shown in Figure 1. Of these cases, 196 (54%) were categorized as HCFO and 170 (46%) as CO cases. Eighty-three cases (22% of all cases) were potentially community-associated, 58 (72%) of patients with these cases were interviewed, and 16 (20%) cases were reclassifi ed as CO-HCA. Overall, 67 cases (18% of all cases) were classifi ed as CA; 42 of these cases were confi rmed by interview and 25 were considered probable CA. Review of available electronic inpatient and outpatient records for the probable case-patients showed no previous exposure to health care. Therefore, we believe that most cases were truly CA. The probable and defi nite CA casepatients had similar ages, race distributions, and outcomes.

Clinical Characteristics
Clinical characteristics of CDI cases by epidemiologic classifi cation are shown in Table 1. Compared with HCFO and CO-HCA case-patients, CA case-patients (confi rmed and probable) were younger (median age 53 vs. 78 and 69 years, respectively; p<0.001). Illness among CA casepatients was milder; only 13 (19%) of patients required hospitalization compared with 39 (38%) CO-HCA casepatients (p = 0.02) ( Table 2). Duration of hospitalization was 7.0 days vs. 3.5 days (p = 0.06) for CO-HCA and CA case-patients, respectively. None of the hospitalized CA case-patients died or had any complications. Laboratory confi rmation of recurrence was documented in 22% of the HCFO and CO-HCA case-patients and 12% of the CA case-patients (Table 2).

Interview Results
Forty-two case-patients were confi rmed as CA-CDI case-patients by interview. Median (SD) age of CA-CDI case-patients was 53 (22.6) years. CA casepatients appeared healthy, and only 30% of case-patients had an underlying illness that required regular physician visits. Severity of reported symptoms varied; the median number of bowel movements per day was 10 (range 0-50 movements), and 1 patient had syncope before diarrhea onset. Other than diarrhea, 86% of patients reported abdominal pain, 48% nausea, 40% fever and 17% bloody feces. Average time from diarrhea onset to diagnosis was 14 days (range 0-92 days). Clinical recurrence requiring a repeat course of treatment was reported by 21%; some recurrences were treated without laboratory confi rmation.
Medication and health care exposures of CA casepatients in the 12 weeks before CDI are shown in Table 3. Thirty-two (76%) of the interviewed case-patients reported using antimicrobial drugs, and 20% took >1 type of drug. The most commonly taken drugs were penicillins, followed by clindamycin and cepholosporins (Table 3). Indications for antimicrobial drug use included upper respiratory or ear infection (26%), bronchitis/pneumonia (17%), and dental abscess treatment or dental prophylaxis (17%). Twenty-six percent of CA case-patients reported using PPIs, and only 2% reported using H 2 blockers.
Of the 42 interviewed CA case-patients, 35 (83%) reported receiving care at an outpatient offi ce in the 12 weeks before the C. diffi cile toxin-positive test result; 13 (31%) received care only at a dental offi ce. Nine (21%) case-patients reported visiting health care facilities without receiving care in the 12 weeks before diagnosis. Five (12%) case-patients reported no exposure to any outpatient or inpatient health care facility. Eight (19%) case-patients did not report exposure to a health care facility or to antimicrobial drugs.

Isolates
Of 145 fecal specimens cultured, 127 (87%) grew C. diffi cile. The recovery rates for HA and CA samples were 93% and 85%, respectively. The recovery of C. diffi cile was the same (80%) from fecal swab specimens and aliquots (100% agreement). One hundred nineteen C. diffi cile isolates underwent molecular and antimicrobial drug susceptibility testing. Six toxinotypes were identifi ed; ≈50% of the isolates in each of the 3 epidemiologic classifi cations were wild-type toxinotype 0 (Table 4). Of the other variants, toxinotype III was the most commonly identifi ed in all 3 epidemiologic classifi cations. Toxinotype V, a strain reported in food animals (7,26) was identifi ed in 2.5%-5% of the samples and varied depending on the epidemiologic classifi cation.
The PFGE types of 68% of the isolates were previously named NAP types; the remaining types were classifi ed into 20 different and unnamed PFGE types ( Table 5). The primary PFGE strain type among the 3 epidemiologic classes was NAP1, characteristically toxinotype III and carrying binary toxin and an 18-bp deletion in the tcdC gene. Three isolates identifi ed as NAP1 by PFGE differed from the epidemic strain by having toxinotype IX/XIII and lacking the 18-bp deletion. Two of the strains from the CA isolates had characteristics that were typical of NAP1 but  were <80% related to NAP1 by PFGE; these strains were classifi ed as NAP1-related. NAP8, a strain associated with toxinotype V, was seen exclusively in CA isolates and at low numbers. NAP7 and NAP8 toxinotype V strains have been isolated from food animals in the United States (7,26). A comparison of antimicrobial drug susceptibilities of the epidemic strain NAP1/toxinotype III and other strains is shown in Table 6. The NAP1 epidemic strain is more resistant to quinolones and has a slightly higher MIC 50 to metronidazole ( Figure 2).

Discussion
CA-CDI was fi rst described in the 1980s in patients receiving outpatient antimicrobial drug treatments (27)(28)(29). In 2005, the emergence of the NAP1 epidemic strain was associated with an increase in the incidence of HA-CDI and an increase in reports of CDI in low-risk populations, such as persons living in the community, children, and peripartum women (4). Our laboratory surveillance showed that 18% of the CDI cases were CA, a fi nding that is similar to other surveillance studies reporting percentages of 20%-30% (9,(30)(31)(32). CA-CDI case-patients were younger and healthier than those with HA exposure (median age 53 vs. 72 years). Although 20% of CA case-patients had illness severe enough to require hospitalization, no CDI-related complications or deaths were reported.
In a population-based surveillance study in Durham North Carolina, USA, in 2005 (9), 59% of the CA-CDI case-patients required hospitalization, and 15% reported an emergency department visit. Similar to our fi ndings, none of those case-patients required admission to intensive care units or surgical interventions, such as colectomy. Surveillance fi ndings for CA-CDI in Connecticut, USA, in 2006 (13) showed that 111 (46%) of 241 CA-CDI case-patients required hospitalization, 29 (12%) required admission to intensive care units, 5 (2%) had toxic megacolon or colectomy, and 5 (2%) died of CDI complications. However, surveillance in Connecticut was conducted on the basis of preferential reporting by physicians and infection prevention specialists, which may have resulted in identifi cation of the most severe disease and hospitalized case-patients. In addition, interviews were not performed to confi rm lack of health care exposure.
We observed that 76% of CA-CDI case-patients were exposed to antimicrobial drugs in the 12 weeks before diagnosis. This percentage is higher than previously reported estimates of 40%-61% (9,10,13,14,31,33,34) and may refl ect more complete information obtained during detailed case interviews. For example, several patients received antimicrobial drugs from their dentist, and such information is likely unavailable in outpatient medical records. These drugs were prescribed for common outpatient indications, and several patients received clindamycin for dental prophylaxis or infection. The role of PPI in CA-CDI remains controversial. Some studies have reported  increased risk for disease associated with their use (12,35). Twenty-six percent of CA-CDI case-patients interviewed reported PPI use. However, it is diffi cult to assess if this is a major risk without a control group comparison. Molecular testing showed a similar distribution of strains between HA and CA cases, and the percentage of cases with the NAP1 epidemic strain ranged from 46% in HA cases to 32% in CA-CDI cases. The percentage of CA cases with the NAP1 strain was similar to that in other reports (18%-37%) (8,(36)(37)(38). Similar strain distribution in health care facilities and the community suggests that in contrast to the emergence of CA-methicillin-resistant S. aureus strains, there was no preferential transmission of particular strains within or outside the health care setting. Health care facilities might act as a reservoir for CA disease or that the community might act as a reservoir for HA-associated disease.
Our study examined potential exposure routes for C. diffi cile acquisition in the community. There are 4 postulated sources for exposure to C. diffi cile spores (39): consumption of contaminated food and water, animal-to-person contact, person-to-person contact, and environment-to-person contact. Foodborne acquisition has been hypothesized as a source of CA infections on the basis of recovery of C. diffi cile spores from food products and similarities between strains recovered from animals and those known to cause disease in humans (15)(16)(17)(18)(19). However, there is currently insuffi cient evidence to support foodborne acquisition as a common source of CA-CDI (7).
We assessed food and animal exposure during interviews with CA-CDI case-patients and did not fi nd any specifi c association. However, we were unable to compare our observations with food and animal exposure patterns among persons in the community without CDI. Other possible sources of exposures include environments contaminated by C. diffi cile spores, such as hospitals and long-term care facilities; 21% of CA-CDI case-patients reported visiting or accompanying a family member to a health care facility in the 12 weeks before diagnosis. Contact with an ill or C. diffi cile-colonized family member or a household member who worked in a health care setting (i.e., someone who might have carried C. diffi cile spores on their hands or clothes) is another possible exposure. Two case-patients reported that a family member had diarrhea or was given a diagnosis of CDI, and several had a household member who worked in a health care setting. We also observed an excellent C. diffi cile recovery rate from refrigerated stool swabs, indicating that this method could be used in epidemiologic studies in which storage and processing of C. diffi cile specimens are required.
Our fi ndings need to be interpreted in light of several limitations. We were unable to calculate the incidence of CA-CDI because surveillance did not include all laboratories servicing the Monroe County population. This study is descriptive, and the lack of a control group prevents us from estimating the risk for various exposures in development of CDI. At the time of this study, diagnosis of C. diffi cile relied on testing with the toxins A and B enzyme  immunoassay, which has a sensitivity of 60%-90% and specifi city of 90%-95%. In low-prevalence populations, such as outpatients, the positive predictive value is low and the likelihood of false-positive results is higher, which might have biased some results by including patients who did not have CDI (40). However, this bias was minimized by laboratory refusal of formed (i.e., nondiarrheal) fecal specimens and exclusion of cases without diarrheal symptoms. We did not review medical records from physician and dental offi ces. Therefore, patient-reported antimicrobial drug and PPI use was not confi rmed. We attempted to interview all persons with potential CA-CDI but were unable to do so in 29% of the cases. These cases were defi ned as probable CA and included in our clinical summary. The small number of CA-CDI cases and isolates also limited our capacity to assess difference between NAP1 and other strains in severity and outcome of CDI.
In conclusion, CA-CDI represented 18% of CDI cases in Monroe County. CA-CDI case-patients were younger and healthier than HA-CDI case-patients. Use of antimicrobial drugs in outpatient settings remains a serious exposure, and even limited exposure to the health care environment or to persons in contact with health care facilities might play a crucial role in acquisition of CDI in the community. Prevention of CA-CDI will require further studies to understand risk factors leading to CDI in patients not exposed to antimicrobial drugs and the role of various potential exposures to C. diffi cile, such as food, animals, and household environment. Our results suggest that educating outpatient clinicians, including dentists, about the risk for community-associated CDI following use of oral antimicrobial drugs and that promoting judicious use of these drugs are potentially important interventions for the prevention of CDI in the outpatient setting.