Clinical and microbiological characteristics of community-onset Clostridium difficile infection in The Netherlands
Corresponding author and reprint requests: M. P. Bauer, Department of Infectious Diseases, Leiden University Medical Centre, Albinusdreef 2, PO Box 9600, 2300 RC Leiden, The Netherlands
To elucidate the prevalence, characteristics and risk factors of community-onset Clostridium difficile infection (CO-CDI), an uncontrolled prospective study was performed. For 3 months in 2007–2008, three laboratories in The Netherlands tested all unformed stool samples submitted by general practitioners (GPs) for C. difficile by enzyme immunoassay for toxins A and B, irrespective of whether GPs specifically requested this. Patients with positive results were asked to complete a questionnaire. Positive stool samples were cultured for C. difficile, and isolates were characterized. In all, 2443 stool samples from 2423 patients were tested, and 37 patients (1.5%) with positive toxin test results were identified. Mixed infections were not found. Age varied from 1 to 92 years, and 18% were under the age of 20 years. Diarrhoea was typically frequent and watery, sometimes with admixture of blood or fever. Eight of 28 patients (29%) suffered recurrences. Among 31 patients with toxin-positive stool samples for whom information was available, 20 (65%) had not been admitted to a healthcare institution in the year before, 13 (42%) had not used antibiotics during the 6 months before, and eight (26%) had neither risk factor. A separate analysis for patients whose samples were both toxin-positive and culture-positive produced similar results. Cultured C. difficile isolates belonged to 13 different PCR ribotypes, and 24% of the isolates were non-typeable (rare or new) PCR ribotypes. In conclusion, CO-CDI can affect all age groups, and many patients do not have known risk factors. Several PCR ribotypes not encountered in hospital-associated outbreaks were found, suggesting the absence of a direct link between outbreaks and community-onset cases.
Previously identified risk factors for Clostridium difficile infection (CDI) include admission to hospital or nursing home, old age, chronic comorbidity, longer hospital stay, antibiotic usage and prior chemotherapy , use of gastric acid suppressants, and nasogastric tubes. Since early 2003, both the incidence and the severity of CDI appear to have increased. This has been ascribed at least partly to the emergence of the new strain, PCR ribotype 027. In 2007, PCR ribotype 027 was found in stool samples of 25% of patients with nosocomial CDI in The Netherlands .
The incidence of CDI occurring outside healthcare facilities, usually termed community-onset CDI (CO-CDI), may be rising as well [3–10]. Some of the reported cases of CO-CDI may be truly community-acquired, but many cases may actually be linked to healthcare institutions. CO-CDI has never been investigated with detailed characterization of C. difficile isolates. In the present study, we aimed to investigate both the clinical characteristics and the source of CO-CDI in three areas in The Netherlands where outbreaks of nosocomial CDI due to PCR ribotype 027 had recently occurred. We screened all stool samples submitted by general practitioners (GPs) for C. difficile, characterized the cultured C. difficile isolates, and obtained patient information using a standardized questionnaire.
Materials and Methods
Three medical microbiological laboratories in The Netherlands participated, namely SALTRO Artsenlaboratorium in Utrecht (providing services for 900 GPs), Public Health Laboratory Kennemerland in Haarlem (400 GPs), and the Laboratory for Medical Microbiology and Infectious Diseases in Zwolle (195 GPs). All unformed stool samples submitted by GPs during a period of 3 months were tested for C. difficile, using a commercially available rapid enzyme immunoassay (EIA) for C. difficile toxins A and B (ICTAB; Meridian Bioscience, Cincinnati, OH, USA). Samples were assayed irrespective of the diagnostic tests requested by the GP. If a stool sample gave positive results and the corresponding patient had no earlier positive stool sample, this patient was included in our study.
If a stool sample gave positive results, the sample was cultured at the regional laboratory and isolates were sent to the reference laboratory at the Leiden University Medical Centre. All isolates were genetically identified as C. difficile by an in-house PCR for the presence of the gluD gene specific for C. difficile . C. difficile isolates were characterized by PCR ribotyping . The presence of tcdA, tcdB and binary toxin genes was investigated as described previously . Antimicrobial susceptibility was determined by Etest for erythromycin, clindamycin, moxifloxacin and ciprofloxacin, using the breakpoints recently described .
Clinical and epidemiological information
The laboratories collected demographic data from all patients whose stool samples were submitted by GPs. If a stool sample gave positive results, the GP who had submitted the sample was asked to give the patient an envelope containing information on the background and aim of our study, together with a request to complete a web-based or printed questionnaire.
The patients were asked for information concerning their symptoms, treatment and possible risk factors. The question about stool consistency on the day of maximal illness was illustrated by drawings from the Bristol Stool Scale . We chose to enquire about antibiotic use during the 6 months prior to diarrhoea instead of 3 months, because we wanted to rule out damage to the colonization barrier persisting for longer than 3 months after the use of antibiotics. It is unclear how long this damage may persist but, in animal models, persisting susceptibility to C. difficile colitis 74 days after one dose of clindamycin has been described . If the patient could not or did not respond, we asked the GPs for the most essential patient information.
In June 2008, we asked the GPs of all included patients for information on persistent diarrhoea attributed to CDI, recurrences and deaths. The diagnosis of a recurrence was left to the judgement of the GPs.
Data were analysed with SPSS 14.0 for Windows (SPSS Inc., Chicago, IL, USA). Descriptive statistics and Maentel–Haenszel-adjusted ORs were used to examine possible correlations. Non-normally distributed continuous variables were compared with the Mann–Whitney U-test.
In total, 2443 stool samples from 2423 patients were submitted by GPs. Thirty-seven (1.5%) patients with positive samples were identified. Of all 419 samples from patients aged 65 years or older, 19 (4.5%) were toxin-positive. The laboratories in Utrecht and Zwolle registered whether GPs explicitly requested testing for C. difficile. This was the case in 12 of 32 positive stool samples.
Twenty-one patients completed a questionnaire. Information on ten of the remaining 16 patients was obtained from their GPs. We obtained follow-up information on 25 patients from their GPs in June, 2008.
Co-infection and characterization of the isolates
Co-infection of CDI with other enteropathogens was not found. Thirteen different PCR ribotypes were found, and seven strains could not be characterized by PCR ribotyping (Table 1). No C. difficile could be cultured from stool samples of five patients (14%), and the stools of three patients (8%) were not cultured, because of logistical errors. As these eight patients may have had false-positive stool toxin test results, we performed the analysis for all patients and for patients who had a positive culture. In spite of the fact that outbreaks due to the strain PCR ribotype 027 had occurred in all regions, this PCR ribotype was not found in the community.
Table 1. PCR ribotypes found at various regional laboratories, followed by number (%) of isolates that contained binary toxin genes and number (%) that were resistant to various antibiotics
|002||1||0||0||1 (100)||1 (100)||0|
|014||3||0||0||3 (100)||1 (33)||1 (33)|
|023||1||1 (100)||0||1 (100)||0||0|
|025||1||0||0||1 (100)||1 (100)||0|
|044||1||0||0||1 (100)||1 (100)||0|
|078||4||3a (75)||0||4 (100)||2 (50)||1 (25)|
|081||1||0||1 (100)||1 (100)||0||0|
|110||1||0||0||1 (100)||1 (100)||0|
|117||2||0||1 (50)||2 (100)||0||0|
|Unknown ribotype||7||0bc||0||6 (100)b||1 (17)c||0|
|No Clostridium difficile was cultured||5|| |
|No culture was performed||3|
Different PCR ribotypes were not clearly linked to an age group or region, except for PCR ribotype 078, which was found four times in the region of Zwolle but not in either of the other regions. The numbers of patients in each specific PCR ribotype group were very small, limiting the possibility of finding associations with clinical characteristics.
Five of seven unknown PCR ribotypes belonged to patients who had not been admitted to a hospital or nursing home and who were not employed in healthcare. A sixth patient had been admitted both to a hospital and to a nursing home, and for the seventh patient this information was not available.
The isolates were tested for antimicrobial susceptibility and production of binary toxin (Table 1). Genes for binary toxin production were found in four isolates, which belonged to PCR ribotype 023 or 078, both of which have been associated with binary toxin production.
Clinical characteristics and follow-up information
Clinical patient characteristics are listed in Table 2. Median ages of the patients with positive and with negative toxin stool tests were significantly different (54 years; (range, 1–92 years) and 37 years (range, 0–97 years), respectively; p <0.001). Symptoms were serious, with watery consistency, high stool frequency and often fever (24%) and admixture of blood (33%), and patients were usually treated (86%). The recurrence rate was high, with eight (29%) of patients suffering recurrences and one patient (4%) still suffering from diarrhoea on follow-up. Of those eight patients who suffered recurrence, six suffered one recurrence, one suffered two recurrences, and one suffered four recurrences. Four of 32 patients had died. Three deaths were deemed by the GPs not to be attributable to CDI, and for one death this information was not available.
Table 2. Clinical characteristics of Clostridium difficile infection (CDI) (sums of percentages may amount to more than 100, because of rounding)
|Characteristic (continuous variables)|| |
| Age (years), median (range)||54 (1–92)||69 (1–92)|
| Interval between start of diarrhoea and stool test (days)a||10 (5–65)||12 (7–65)|
|Characteristic (categorical variables), proportion (%)|| |
| Age category (years)|
| 0–4||2/37 (5)||1/29 (3)|
| 5–9||1/37 (3)||1/29 (3)|
| 10–14||2/37 (5)||2/29 (7)|
| 15–19||2/37 (5)||2/29 (7)|
| 20–39||4/37 (11)||2/29 (7)|
| 40–59||8/37 (22)||4/29 (14)|
| 60–79||8/37 (22)||8/29 (28)|
| ≥80||10/37 (27)||9/29 (31)|
| Female sex||20/37 (54)||16/29 (55)|
| Stool consistency on the day of maximal illness|
| Formed||1/21 (5)||1/16 (6)|
| Mushy||1/21 (5)||1/16 (6)|
| Watery||19/21 (90)||14/16 (88)|
| Stool frequency on the day of maximal illness (times per day)|
| 1–3||2/21 (10)||1/16 (6)|
| 4–6||7/21 (33)||5/16 (31)|
| 7–10||5/21 (24)||4/16 (25)|
| >10||7/21 (33)||6/16 (38)|
| Admixture of blood with stools on any day||7/21 (33)||4/16 (25)|
| Abdominal pain on any day||14/21 (67)||9/16 (56)|
| Temperature over 38°C on any day||5/21 (24)||3/16 (19)|
| Metronidazole||16/21 (76)||13/16 (81)|
| Metronidazole, followed by vancomycin||2/21 (10)||2/16 (13)|
| No treatment||3/21 (14)||1/16 (6)|
| Course of diarrhoea|
| Recovery from diarrhoea without antibiotics||9/28 (32)||7/22 (32)|
| Recovery from diarrhoea after one treatment||10/28 (36)||8/22 (36)|
| Recovery from treatment after ≥1 recurrences||8/28 (29)||7/22 (32)|
| Persistent diarrhoea||1/28 (4)||0/22 (0)|
| Died||4/32 (13)||4/25 (16)|
| Death partially attributable to CDI||0/31 (0)||0/24 (0)|
The risk factors that we investigated are listed in Table 3. Only 35% of patients had been admitted to healthcare facilities, and only 58% had used antibiotics during the 6 months before diarrhoea developed. This percentage was similar among those who had been admitted (55%) and those who had not (60%). The antibiotics mentioned most often were amoxycillin–clavulanic acid (nine patients) and amoxycillin (four other patients).
Table 3. Proportions (%) of patients with risk factors for Clostridium difficile infection
|Hospital admission in the year prior to diarrhoea||9/31 (29)||8/25 (32)|
|Hospital admission in the year prior to diarrhoea and/or admission to a nursing home in the year prior to diarrhoea||6/31 (19)||6/25 (24)|
|No admission to healthcare institutions in the year prior to diarrhoea||20/31 (65)||15/25 (60)|
|Employment in healthcare||1/31 (3)||0/25 (0)|
|No admission to healthcare institutions in the year prior to diarrhoea or employment in healthcare||19/31 (61)||15/25 (60)|
|Family members employed in healthcare||4/21 (19)||3/16 (19)|
|Hospital admission of family members in the year prior to diarrhoea||2/21 (10)||2/16 (13)|
|Visit to a nursing home in the year prior to diarrhoea||4/21 (19)||3/16 (19)|
|No link to healthcare institutions (as assessed by the above variables)||9/21 (43)||6/16 (38)|
|Antibiotics during the 6 months prior to diarrhoea||18/31 (58)||16/25 (64)|
|Antibiotics during the 6 months prior to diarrhoea of those not admitted to healthcare institutions in the year prior to diarrhoea||12/20 (60)||11/15 (73)|
|No admission to healthcare institutions in the year prior to diarrhoea or employment in healthcare or antibiotics during the 6 months prior to diarrhoea||7/31 (23)||4/25 (16)|
|Family members who experienced diarrhoea during the month prior to diarrhoea||5/19 (26)||4/14 (29)|
|Use of medication||21/31 (68)||15/25 (60)|
|Use of medication compatible with relevant comorbidity||18/31 (58)||14/25 (56)|
|Use of corticosteroids||0/31 (0)||0/25 (0)|
|Use of antiperistaltic agents||1/31 (3)||0/25 (0)|
|Use of gastric acid suppressants||8/31 (26)||7/25 (28)|
|Monitoring by a medical specialist (including nursing home physician)||13/21 (62)||12/16 (75)|
|Pet ownership||6/21 (29)||4/16 (25)|
|Professional contact with farm animals||0/31 (0)||0/25 (0)|
Most patients had comorbidity, as judged by the fact that 21 of 31 (68%) used medication and 13 of 21 (62%) reported being monitored by a medical specialist. When two patients who used only a selective serotonin re-uptake inhibitor for a mood disorder and one patient who used only acetaminophen for pain because of osteoporosis were excluded, the percentage of patients who used medication fell to 58. Use of medication was associated with older age groups, but not restricted to these groups (lowest age quartile, none of six patients; both middle quartiles, two of nine patients; highest quartile, seven of seven patients). Gastric acid suppressants were used by 26% of patients. No patient was found to have a profession involving contact with farm animals.
In this study of clinical and microbiological characteristics of CO-CDI, the prevalence of CDI among patients with community-onset diarrhoea for which microbiological diagnostics were requested amounted to 1.5%. In most cases, there was no specific request to test for C. difficile, which would have caused six of ten cases to be missed. The clinical picture of the disease was severe, with a high recurrence rate. We found no link to healthcare facilities in the majority of cases. Moreover, of the patients who were not admitted to healthcare institutions, 40% had not used antibiotics during the 6 months prior to the development of diarrhoea. Furthermore, 42% of all patients did not use medication compatible with relevant comorbidity, and 18% were under 20 years of age. Finally, most of the PCR ribotypes found were not associated with outbreaks in healthcare institutions. In particular, PCR ribotype 027 was not found, in spite of the fact that, in all of these areas, outbreaks with this PCR ribotype had recently occurred.
Methodological issues might have affected the results of this study. First, our study population was based on stool samples that were submitted by GPs, which may have led to referral bias. In The Netherlands, GPs are encouraged by their guidelines to culture stools when there is serious illness . Therefore, it is possible that CO-CDI can run a much milder self-limiting course, in which no diagnostic tests are performed. Second, we screened for cases with an EIA for toxins A and B, and test characteristics will have influenced the population identified. EIAs may be relatively insensitive in comparison with stool culture and cytotoxicity assays . We used an immunochromatography assay (ICTAB; Meridian) that has been shown to have a sensitivity of 91%, a specificity of 97%, a positive predictive value of 70% and a negative predictive value of 99% in comparison with the cytotoxicity assay used as the reference standard . However, the characteristics of this assay were determined in a population of hospitalized patients, and it is unknown whether these characteristics may be extrapolated to a community setting. The design of the study was not optimized for a high recovery rate of C. difficile cultures, as each centre was allowed to apply its own culture protocol. This may have resulted in the 14% toxin-positive and culture-negative stool samples. Alternatively, EIAs of these samples may have been falsely positive. Therefore, we analysed results from toxin-positive and culture-positive samples in a separate analysis. Third, bias may have been introduced by the manner in which clinical data were gathered. Part of the information came from questionnaires, which were completed by 21 of 37 patients. Possibly, the severity of diarrhoea or comorbidity of patients who completed the questionnaire differed from those who did not.
In spite of this possible bias, we feel that the strength of the study is the detail of the information that we did obtain. Most previous studies lack this detail, and no other study has investigated the follow-up of patients with CO-CDI. Moreover, we characterized C. difficile isolates by ribotyping, which serves as an extra tool with which to investigate epidemiological associations.
Most studies on CO-CDI lack a clear definition of what is to be considered community-acquired. Often, CDI is designated as community-acquired when stool samples were collected in the community without knowledge of the patient’s prior healthcare contacts. The European Centre for Disease Prevention and Control and the CDC have arbitrarily divided CO-CDI (and nosocomial CDI during the first 48 h of the admission) into community-onset healthcare facility-associated (CO-CDI occurring within 4 weeks after discharge from a healthcare facility) and community-acquired (occurring after 12 weeks after discharge) [18,19], leaving an intermediate period. Using these definitions, Kutty et al.  found many CO-CDI cases to be community-onset healthcare facility-associated, suggesting that they were not actually acquired in the community, but in healthcare facilities. Only 17% of CO-CDI cases in a Dutch hospital-based surveillance study were community-acquired when the definitions of the European Centre for Disease Prevention and Control were applied . The detail of the clinical information that we obtained allows for a clear distinction between CDI that is truly community-acquired and CDI that may have been acquired in healthcare facilities.
Furthermore, studies investigating CO-CDI seldom use molecular characterization of C. difficile isolates as an additional epidemiological tool. A Canadian study  characterized 17% of C. difficile strains from community sources as PCR ribotype 027, but no clinical data were available to verify that the patients had not been recently admitted to healthcare institutions.
A recent surveillance study by the CDC  found results very similar to ours. However, in this investigation, unlike ours, patients were not systematically surveyed and PCR ribotyping was not performed.
Finally, a recent case–control study in the UK  investigated the prevalence and clinical characteristics of patients with cytotoxin-positive stools submitted by GPs. The proportion of positive samples (2.1%) was consistent with that in our study. The proportions of patients who used antibiotics in the previous 4 weeks and who were hospitalized in the last 6 months were 52% and 45%, respectively. Unfortunately, no information was provided on comorbidity, animal contacts, follow-up and clinical characteristics of the diarrhoeal illness other than stool frequency. Also, the authors mentioned the frequent occurrence of PCR ribotype 001, but did not provide information on other PCR ribotypes found and whether these are associated with hospital outbreaks.
The incidence rate of CO-CDI cannot be estimated from our data, as it is unclear how many cases have been missed because patients did not visit their GPs or GPs did not perform diagnostic tests. Therefore, it is difficult to compare our findings with the results of surveillance studies of gastroenteritis in Dutch general practices [23,24], which did not test for C. difficile.
Interestingly, our data suggest that CO-CDI does not directly result from the effects of healthcare-associated outbreaks. We did not find support for the hypothesis that an animal reservoir plays a major role in CO-CDI. However, in the region of Zwolle, PCR ribotype 078 was the most frequently encountered strain. This strain has frequently been found in recent surveillance studies of nosocomial CDI in The Netherlands. It has also been found in farm animals and meat products, and transmission from animals to humans seems possible. It was isolated from stools of diarrhoeal piglets in The Netherlands . The city of Zwolle is situated in a rural part of The Netherlands, and one could speculate that a link between humans and animal cases exists in this area.
In conclusion, the prevalence of C. difficile in stools of patients with community-onset diarrhoea in The Netherlands for which diagnostics are requested by their GPs is 1.5%. All age groups can be affected, and many patients have not been admitted to healthcare institutions or used antibiotics. Many PCR ribotypes of C. difficile that are not encountered in hospital-associated outbreaks are found. Physicians, including GPs, should be aware of the possibility of CDI outside of the known risk factors.
We would like to thank S. Peeters, D. Mink and J. van Es for toxin testing and cultures of stools and processing of demographic data, and I. Sanders and C. Harmanus for characterization and antibiotic susceptibility testing of isolates.
Genzyme Corporation provided an unrestricted grant for this study. All authors declare that they do not have any commercial or other association that might pose a conflict of interest regarding this manuscript.