Community-associated or -acquired CDI (CA-CDI) was first reported in 1984 by Stergachis et al. (10) who found that the attack rate of CA antimicrobial drug–associated colitis requiring hospitalization was 1.4/100,000 population. In a review by Riley et al. (11) of 580 C. difficile toxin–positive stool samples submitted from patients with diarrhea and a clear history of recent (<4 weeks) antimicrobial drug exposure, 10.7% were from patients who did not have a recent history of hospitalization. Additional reports indicated incidences from 7.7–12.2 cases of CA-CDI per 100,000 persons in US communities (7,12) to 25 primary episodes per 100,000 persons per year in Sweden (8). In December 2005, the US Centers for Disease Control and Prevention (CDC) reported a series of severe CDI cases in populations previously considered at low risk, including generally healthy persons living in the Philadelphia, Pennsylvania, USA, area (13).

In January 2006, CDC was notified by a regional Veterans Affairs (VA) hospital in North Carolina that surveillance conducted from October 2004 through December 2005 indicated that 35% of patients with CDI experienced onset of disease in the community. Fewer than 10% of these patients had stayed overnight in a healthcare facility, and only half had received antimicrobial drugs in the previous 2 months (14). In response to apparent emerging CDI in the community, CDC was invited by the North Carolina Department of Health and Human Services to investigate these findings and to recommend control measures.

The estimated incidence of CA-CDI in the VA catchments and county adult population was 21 and 46 per 100,000 persons per year, respectively (Table 1). Among men in the VA and in Durham County, those 45–64 years of age had a higher incidence than those 18–44 and >64 years of age combined (p<0.05). Overall, incidence was higher among county female case-patients than male case-patients (62 vs. 28/100,000 persons per year, respectively; p = 0.0005).

The incidence rate we found was highest for the 45–64 years of age category (28.5/100,000) for the VA catchments. In Durham County, men 45–64 years of age had a higher rate than other those in age categories combined (p<0.05). This finding is notable because CDI is usually associated with older patients (9,19). Women had a higher rate than males in the community (62 vs. 28 cases/100,000 persons/year, respectively; p = 0.0005), as was also recently noted in a report from Connecticut (19). Although some data suggest an increased risk for multiple CDI recurrences in healthy women (24) and recent reports have noted severe CDI in pregnant women (25), female gender has not been previously a well-documented risk factor for CDI.

Previous reports have shown that an age >65 years is a risk factor for hospital-onset CDI (2,26,27). In comparison, the median age of the case-patients in this community appeared to be younger, 61 years in Durham County and 63 years in VA. In addition, the proportion and severity of fever, leukocytosis, and renal insufficiency for our case-patients were lower (26,28,29). No case-patients were admitted to the intensive care unit for CA-CDI, and none underwent colectomy. One case-patient in each population died within 10 days of diagnosis, and the death was attributable to CDI. Nonetheless, 15% had disease severe enough to require a visit to the emergency department, and another 59% required hospital admission for CDI management.

Antimicrobial drug exposure has long been known as a risk factor for healthcare-associated CDI. However, among CA cases in our study, 49% were not exposed to antimicrobial drugs. This percentage was slightly higher than the recent findings from Philadelphia (13) and Connecticut (19), where 24% and 36%, respectively, were not exposed to antimicrobial drugs. In contrast, Dial et al. found the absence of antimicrobial drug exposure >90 days to range from 60% to 70% (30). However, their analysis was limited to a clinical research database in which some hospitalization and antimicrobial drug exposures may not have been included. Two prospective studies have recently been conducted in the community. Bauer et al. (31) found that 42% of CA-CDI case-patients had not been exposed to antimicrobial drugs during the prior 6 months, and Wilcox et al. (20) found that 84% case-patients had not received antimicrobial drugs during the month before C. difficile detection. One hypothesis to explain the absence of antimicrobial drug exposure is that there are unmeasured factors affecting the epidemiology of CDI. For example, remote antimicrobial drug exposure, or exposure to other medications with antimicrobial activity, may be increasing the risk of disease; alternatively, increased awareness of CDI may be leading to increased testing and documentation of C. difficile in patients not previously tested. Another possibility is that strains with new virulence properties (e.g., binary toxin) that enable disease in the absence of prior antimicrobial drug use have emerged. Despite antimicrobial drug exposure being absent in many patients, we found that this exposure remained the most important modifiable risk factor for CA-CDI.

Additional risk factors included markers of chronic disease such as outpatient visits in the VA population and GERD and cardiac failure in the county population. In the VA population, frequent outpatient visits could reflect transmission in ambulatory care settings and could be a marker of a more severe underlying disease. Unlike other recent studies (9), we did not find proton pump inhibitors or H2 blockers were a major risk factor for CA-CDI. However, the finding of GERD as a risk factor suggests the possibility that undocumented use of over-the-counter proton pump inhibitors could have increased risk in these patients. Alternatively, there may be factors in the pathogenesis of GERD that increases the risk for CDI.

Our study has several limitations. First, it is likely that there was incomplete case ascertainment, especially in those who underwent testing by outside laboratories so that, as high as these population incidence estimates are, they are likely underestimates of the true incidence. However the degree of underestimate is less likely in the VA system as there is financial incentive for patients to undergo testing within the system. The community physician survey conducted in Durham County indicated that there were 2 commercial laboratories other than Durham hospital laboratories used for testing. Although we were unable to determine the number of C. difficile tests performed at 1 laboratory, only 14 case-patients were identified from the other. It is also possible that patients received empiric therapy for CDI without a test being performed. However, the same survey of Durham County physicians indicated this was not a common practice. Some potential case-patients were categorized as unknowns when little or no medical records were available. We did not collect data on laboratory testing performed for any other enteric pathogens besides C. difficile nor did we perform cultures for C. difficile, and therefore no isolates were available.

Instead, case confirmation was limited to toxin immunoassay testing using the C. DIFFICILE TOX A/B II TECHLAB test. In an independent review (32), the sensitivity of this test was 83.3% and the specificity was 98.7%. To address the concern of inadequate sensitivity in the toxin immunoassay and to avoid any misclassification bias in our case–control studies, we excluded controls who had diarrhea. Despite the high specificity of this test, there are valid concerns that if a low-prevalence population, such as relatively asymptomatic persons without prior antimicrobial drug exposures, is tested, the likelihood of a false-positive result may be unacceptably high. Although this is an insurmountable obstacle to a retrospective analysis of current clinical testing practice for CDI, the fact that all study laboratories had rejection criteria to prevent testing formed stool, near uniform medical record documentation of diarrhea (i.e., patients with documented absence of diarrhea were excluded), and a median duration of diarrhea symptoms of 1 week suggests a reasonable pretest likelihood of CDI among these patients.

Another limitation was that few interviews were performed with case-patients. However, the adequacy of records indicating exposure to inpatient healthcare or antimicrobial drugs was verified among 53 case-patients who were interviewed by telephone. Only 1 case-patient was reclassified on the basis of an undocumented healthcare exposure, which was discovered during the interview process. Five of the 11 case-patients for whom antimicrobial drug exposure was not identified in their available medical records reported antimicrobial drug use. However, 3 of these were VA case-patients for whom medical records did not document such use, which suggests that some patients may have been mistaken about their antimicrobial drug exposure.

Another limitation is that our use of outpatient controls for the VA case–control study may have resulted in bias toward the null with regard to outpatient healthcare-related risk factors. Although we attempted to contact >400 candidate controls from the voter registration list for the Durham County case–control study, we encountered difficulty in reaching persons by phone and eliciting their participation. This resulted in only 48 controls being available and limited the power of this analysis.