This work was presented at the 52nd ICAAC meeting, San Francisco, 9–12 September 2012.
A rapid and accurate diagnosis of Clostridium difficile infection (CDI) is essential for patient management and implementation of infection control measures. During a prospective time-series study, we compared the impact of three different diagnostic strategies on patient care. Each strategy was tested during a 3-month period: P1 (diagnosis based on the stool cytotoxicity assay and the toxigenic culture), P2 (diagnosis based on PCR) and P3 (two-step algorithm based on glutamate dehydrogenase detection followed by nucleic acid amplification test). The following criteria were used to assess the quality of patient management: (i) time for result reporting, (ii) frequency of repeat testing within 7 days, (iii) time elapsed between stool collection and beginning of treatment for patients with CDI, and (iv) frequency of empirical treatment for patients without CDI. Of 1122 stool samples (P1 n = 359, P2 n = 374, P3 n = 389), 36 (10.0%), 47 (12.3%) and 48 (12.3%) were positive for C. difficile during P1, P2 and P3, respectively. The time for reporting of a positive or a negative result was significantly shorter and the frequency of redundant stool samples within 7 days was lower during P2 and P3 than during P1. Patients with CDI were specifically treated with vancomycin or metronidazole earlier during P2 and P3 than patients from P1 (0.5 ± 0.5 days and 1.0 ± 1.8 days vs. 2.0 ± 1.7 days). The empirical therapy among patients without CDI decreased from 13.6% during P1 to 6.4% during P2 and 5.6% during P3. A rapid CDI diagnosis impacts positively on patient care.
The epidemiology of Clostridium difficile infection (CDI) has dramatically changed over the last decade. Since 2003, many outbreaks of severe CDI with high mortality rates have been reported worldwide [1, 2]. This situation coincided with the emergence and the rapid dissemination of a previously extremely rare genotype named BI/NAP1/027 [3, 4]. Currently, C. difficile is recognized as the most frequent aetiological agent of healthcare-associated diarrhoea in hospitalized adult patients.
In this context, a rapid diagnosis of CDI is a key step in the successful management of the disease. It enables the physician to initiate treatment of the patient expeditiously and to implement control measures rapidly to avoid cross-contamination. Moreover, an accurate diagnosis is also essential to obtain reliable data for surveillance, to assess the efficacy of intervention measures to reduce CDI, and to enable comparison between institutions as part of performance management. Currently there is no single ‘gold standard’ for the diagnosis of CDI. The two reference methods often used for studies are the stool cytotoxicity assay, which detects the presence of ‘free’ C. difficile toxins in stool samples (primary toxin B but also toxin A), and toxigenic culture, which detects C. difficile isolates in stool that have the potential to produce toxins. The crucial question about the clinical significance of the presence of a toxigenic C. difficile strain in the stool without any free detectable toxin is still a matter of debate . Both reference methods are long, time-consuming and require 24–72 h to produce results. The recent changes in the epidemiology of CDI have stimulated companies to find more rapid and sensitive methods for diagnosis. Since 2009, nucleic acid amplification tests based on the detection of either tcdA or tcdB genes (which encode for toxins A and B, respectively) have become commercially available. In addition, strategies based on a two-step algorithm using the detection of glutamate dehydrogenase (GDH) as a screening test followed by a confirmatory test for the GDH-positive stools were proposed by the recent American and European guidelines [6, 7]. However, the impact of these rapid methods on patient management has not been evaluated.
The aim of the study was to assess the changes in patient management after implementing a rapid diagnosis of CDI.
Materials and Methods
This study was performed at Saint-Antoine hospital, which is a public university-affiliated 750-bed hospital located in eastern Paris. It provides emergency and acute-care services. Around 30 425 patients are admitted each year, corresponding to 221 711 days of hospitalization (data from 2011).
Clostridium difficile testing is performed on physician's request and systematically for all diarrhoeic stool samples prescribed after day 3 of hospitalization (corresponding to a healthcare facility-onset, healthcare-associated diarrhoea). In our hospital, repeated stool culture within a period of 7 days following the initial C. difficile test is systematically discarded but is recorded in the microbiological database as an indicator of quality. Patients with CDI are placed in a single room and contact precautions are implemented according to a written local procedure. Isolation measures are audited by the infection control nurses during ward rounds.
A prospective study comparing three consecutive periods with different CDI diagnosis strategies was undertaken. This study was approved by the local committee for healthcare-associated infections.
All diarrhoeic stools (taking the shape of the container) from adult inpatients with a clinical suspicion of CDI or a healthcare-associated diarrhoea were included.
Stools samples from outpatients and from patients from daycare centres were excluded.
During P1 (5 December 2010 to 28 February 2011), diagnosis of CDI was performed according to conventional reference methods combining the stool cytotoxicity assay on MRC-5 cells and the toxigenic culture, as previously described . During P2 (1 March to 15 May) and P3 (16 May to 15 August), the diagnosis was performed using the Xpert C. difficile assay (Cepheid, Sunnyvale, CA, USA) and a two-step algorithm based on GDH detection (Quik Chek; Alere, Waltham, MA, USA) as a screening test, followed by the Illumigene C. difficile assay (Meridian Bioscience, Cincinnati, OH, USA) on GDH-positive samples, respectively. Stool samples were processed in a batch once a day from Monday to Friday during P1 whereas they were tested on demand from Monday to Saturday during P2 and P3. Xpert C. difficile, GDH detection and Illumigene C. difficile assays were performed according to the manufacturers' instructions. To assess sensitivity and specificity of the Xpert C. difficile assay and the two-step algorithm, toxigenic culture was systematically performed on every stool sample during P2 and P3. Positive and negative results of tests for C. difficile from inpatients were notified by telephone to the clinical service by medical resident microbiologists. In case of positive toxigenic culture but negative Xpert C. difficile or two-step algorithm assays, the result was transmitted to the clinician but excluded from the analysis of the impact of a rapid diagnosis strategy on patient management.
An information letter was sent to every clinical ward to inform clinicians about the changes of diagnosis strategy.
Data collection and analysis
For all patients tested for C. difficile, the following data were collected by medical microbiologists using a standardized questionnaire: age, gender, ward, origin of the diarrhoea, date of admission, date and hour of reception of stool sample at the laboratory, date and hour of result notification, date of initiation of specific treatment by metronidazole or vancomycin, date of implementation of isolation precautions, and date of discharge. For patients with a diagnosis of CDI, clinical parameters (number of stools per day, tenderness, abdominal pain) were recorded at the day of stool culture. Outcome at day 10 (clinical cure, death) and at day 30 (death or recurrent CDI) was assessed.
The following criteria were used to assess quality of patient management:
time to receipt of results defined as the time elapse between stool reception at the laboratory and notification of the result to the clinician,
frequency of repeat testing within 7 days,
for patients with CDI: time elapse between stool collection and beginning of treatment, length of hospital stay after CDI diagnosis, mortality at day 10 and day 30
for patients without CDI: frequency and length of empirical treatment for C. difficile frequency and length of pre-emptive contact precautions.
Patients who were on isolation precautions for another reason or those who received intravenous metronidazole for another infection were not considered to be specifically isolated or treated for their C. difficile infection.
Data were entered into Epi Info software (version 6.04dfr; Centers for Disease Control and Prevention, Atlanta, GA, USA). A chi-square analysis was used to compare categorical variable and analysis of variance tests or the non-parametric Mann–Whitney U test for variables with non-Gaussian distribution were used to compare continuous variables. The Bonferonni's correction was used for multiple comparisons (P1 vs. P2, P1 vs. P3 and P2 vs. P3). A two-sided p value <0.017 was considered statistically significant.
Assessment of prescriber satisfaction
The clarity and rapidity of reporting was assessed through a simple anonymous questionnaire sent to 30 senior and resident prescribers from nine different wards at the end of periods 2 and 3. They were asked to rate their satisfaction on a quantitative scale ranging from 0 (no satisfaction) to 10 (high satisfaction).
Patients were considered to have CDI if they had a positive test for C. difficile (positive cytotoxicity assay, or toxigenic culture or Xpert C. difficile or GDH/Illumigene assays) or evidence of pseudomembranes during lower endoscopy examination. Healthcare-associated infection surveillance was performed using the ECDC (Centers for Diseases Control and Prevention) definitions .
A severe CDI case was defined as a patient who meets any of the following criteria: (i) history of admission to an intensive care unit for complications associated with CDI (e.g. for shock that requires vasopressor therapy); (ii) history of surgery (e.g. colectomy) for toxic megacolon, perforation or refractory colitis; or (iii) death caused by CDI within 30 days after symptom onset .
Between 5 December 2010 to 15 August 2011, 1122 stool samples were tested for C. difficile (Fig. 1) of which 130 (11.58%) were toxigenic culture-positive. The overall CDI incidence during the entire study period was 8.18 cases CDI per 10 000 patient-days. The CDI incidence did not differ significantly across the different periods but a slight increase in incidence was observed after implementing rapid methods for diagnosis (Table 1).
Table 1. Incidence of Clostridium difficile infection and density testing during each study period
p (P1 vs. P2)
p (P1 vs. P3)
p (P2 vs. P3)
P1, period 1; P2, period 2; P3, period 3.
Total length of hospital stay (patient-days)
No. of stool samples tested
Density of C. difficile testing (/1000 patient-days)
No. of CDI
Incidence of CDI (per 10 000 patient-days)
No. of positive toxigenic cultures
Frequency of positive toxigenic cultures (/100 stool samples)
Compared with toxigenic culture as a reference method, nine, two and three false-negative results were found with cytotoxicity assay, Xpert C. difficile and GDH/Illumigene algorithm, respectively. In addition, we recorded one false-positive result with the Xpert C. difficile assay and none by cytotoxicity assay or GDH/Illumigene algorithm. Sensitivity of cytotoxicity assay, Xpert C. difficile and two-step algorithm was 75% (95% CI 57.46–87.27), 95.65% (95% CI 83.76–99.24) and 93.75% (95% CI 81.79–98.37), respectively. Specificity was 100% (95% CI 98.53–100), 99.69% (95% CI 98.84–99.98) and 100% (95% CI 98.60–100), respectively.
Demographic data of patients (age, gender, origin of the diarrhoea, wards) are described in Table 2 and were similar across the three study periods. Changes in the management of patients with a positive and a negative result are described in Tables 3 and 4, respectively. Reporting of a positive or a negative result was significantly shorter in P2 and P3 compared with P1. Frequency of redundant stool samples within 7 days was lower during P2 and P3 compared with P1, whatever the initial result.
Table 2. Demographic features of patients included in each study period
Time to stop an empiric treatment (days) mean ± SD (median)
5.5 ± 3.3 (4)
3.6 ± 4.2 (2)
3.8 ± 4.6 (2)
Number of unjustified treatment days
Contact precautions, n (%)
Length of contact precautions (days) mean ± SD (median)
4.5 ± 1.8 (4)
4.7 ± 6.8 (1.5)
3.7 + 3.6 (2)
Number of unjustified contact precautions days
Table 4. Management of patients with Clostridium difficile infection
Period 1 (CTA+TC) (n = 36)
Period 2 (Xpert) (n = 45)
Period 3 (GDH+Illumig.) (n = 45)
p (P1 vs. P2)
p (P1 vs. P3)
p (P2 vs. P3)
CTA, cytotoxicity assay; GDH, glutamate dehydrogenase; LOS, length of stay; MTZ, metronidazole; TC, toxigenic culture; VA, vancomycin.
P1, period 1; P2, period 2; P3, period 3.
Patients with ≥10 stools per day, n (%)
Abdominal pain, n (%)
Tenderness, n (%)
Time for return of results
Days, mean ± SD (median)
3.1 + 2.58 (2)
0.53 + 0.66 (0)
1.20 + 1.64 (1)
Hours, mean ± SD (median)
75.7 + 61.9 (51)
15.4 + 15.4 (5)
31.4 + 38.7 (27)
Redundant stool samples (<7 days), n (%)
Specific treatment by VA or MTZ, n (%)
Time (days) elapse between C. difficile testing and specific treatment mean ± SD (median)
2.00 + 1.68(2)
0.49 + 0.56 (0)
1.03 + 1.80 (0)
Contact precautions, n (%)
Clinical cure at day 10, n (%)
Recurrence (within 30 days), n (%)
Severity, n (%)
Mortality at day 10, n (%)
Mortality at day 30, n (%)
LOS (days) mean ± SD (median)
30.3 ± 36.3 (19.5)
23.2 ± 25.4 (15)
26.9 ± 28.9 (20)
LOS after stool culture mean ± SD (median)
15.8 ± 14.0 (10.5)
12.3 ± 19.7 (8)
12.5 ± 12.5 (9)
Treatment of patients with CDI with vancomycin or metronidazole started earlier during P2 and P3 than during P1 (0.5 ± 0.5 days, 1.0 ± 1.8 days vs. 2.0 ± 1.7 days). Length of hospital stay following the diagnosis of CDI was shorter in patients from P2 to P3 compared with patients from P1 but the difference was not statistically significant (mean: 12.3 days and 12.5 days vs. 15.8 days, respectively). Empirical therapy among patients without CDI decreased significantly from 13.6% during P1 to 6.4% during P2 and 5.6% during P3. Numbers of unnecessary treatment-days were 243, 75 and 73 during P1, P2 and P3, respectively.
When comparing the two periods with a rapid diagnosis (P2 and P3), the only statistically significant difference was the time restituer a positive result to the physician, which was shorter for Xpert C. difficile than for the two-step algorithm (15.4 ± 15.4 h vs. 31.4 ± 38.7 h, respectively, p 0.004).
Clinicians reported being very satisfied with the rapidity of the result return, whatever the strategy used (Xpert C. difficile or the two-step algorithm) (Fig. 2). Nevertheless they found that clarity of reporting was better with Xpert C. difficile than with the two-step-algorithm (mean satisfaction rating 8.27 vs. 9.20, p 0.06).
Accurate and early identification of C. difficile infections is a crucial step for successfully treating patients and reducing severe outcomes and complications. In addition, a rapid diagnosis is also essential for implementing contact precautions and preventing transmission of the organism from patient to patient. Toxigenic culture and the stool cytotoxicity neutralization assays are considered the reference methods but paradoxically they are not frequently used in clinical laboratories in Europe. Two reasons for their low utilization are the technical difficulty of performing the assays and their long turnaround times. Enzyme immunoassays for toxins A and B lack sensitivity and they should not be used as stand-alone tests for the diagnosis of CDI according to recent European and American guidelines [6, 7]. The changes in the epidemiology of CDI encouraged companies to find new rapid and sensitive methods for the diagnosis of CDI. Starting in 2009, nucleic acid amplification tests based on the detection of tcdA or tcdB genes became commercially available. In clinical practice, these methods are used either on every stool sample or as a confirmation method for stool samples that are positive by GDH assays. However, these different strategies based on nucleic acid amplification tests are costly and their performance and their impact on patient management should be assessed before implementing them in a laboratory.
Our findings confirm that PCR on every stool sample and a two-step algorithm based on GDH detection followed by confirmation of positive results by Illumigene are both sensitive methods for the diagnosis of CDI. These results confirm recent data that reported mean sensitivity and specificity of PCR assays of 90% and 96%, respectively  compared with toxigenic culture. The sensitivity of a GDH-based algorithm is more controversial. On one hand, these tests have displayed good correlation with culture on selective medium. A recent meta-analysis of GDH tests demonstrated a high diagnostic accuracy for the presence of C. difficile in faeces; when compared with culture, GDH tests achieved a sensitivity of >90% . On the other hand, a recent report from Tenover et al.  suggested that the performance of GDH assays depends upon the strain genotype, with a reduced sensitivity of the GDH assay for detecting PCR ribotypes other than 027. However, this result was challenged by a more recent study . The objective of the two-step algorithm based on GDH detection followed by Illumigene is to save money compared with PCR performed for every stool sample and to reduce the workload of cell culture, PCR or toxigenic culture, which are often used as confirmatory tests .
During this study, we observed a slight, non-significant increase of CDI incidence following implementation of a rapid diagnosis by PCR or the two-step algorithm. Recent studies have shown that switching diagnostic strategy to nucleic acid amplification tests may lead to an excess rate of CDI [15, 16]. This is particularly true when you move from a strategy that lacks sensitivity (enzyme immunoassay for example) to nucleic acid amplification tests. Nevertheless, this is not the case in the present study, where we showed that the sensitivity of PCR or the two-step algorithm (GDH + Illumigene) compared with the toxigenic culture (P1) was 95.6% and 93.7%, respectively. Moreover, the increased CDI incidence observed during P2 and P3 was paradoxical because the incidence is usually higher during winter months, which correspond to period P1 . This trend may be explained by a higher density of C. difficile testing, indicating that the more you test, the more you find, as recently suggested . Testing changes from 5 days a week during P1 to 6 days a week during P2 and P3 may have influenced ward practice and may have promoted C. difficile testing. This finding is also supported by the fact that patients with CDI in P2 and P3 presented with less severe clinical symptoms than in P1. Number of stools per day and frequencies of tenderness and abdominal pain were lower in CDI patients identified during P2 and P3, compared with patients from P1. This observation suggests that a rapid diagnosis may increase physicians' trust in the results and can potentially promote C. difficile prescribing, even for patients with very mild diarrhoea.
A key finding of this study is the positive impact of a rapid diagnosis (PCR or two-step algorithm) on patient care compared with conventional diagnosis based on stool cytotoxicity assay and toxigenic culture. Interestingly, there was no significant difference between the two rapid diagnostic strategies, except for the time for return of a positive result.
First, the delay of result reporting (defined as the interval time between stool reception at the laboratory and the notification of the result) was shorter, both for positive and negative results. This result was expected for two major reasons: (i) PCR or GDH results can be obtained within a working day, which is shorter than for the stool cytotoxicity assay (24 h) or the toxigenic culture (48–72 h); in addition, 25% of CDI cases during P1 were only identified by toxigenic culture, which accounts for a long delay in result reporting; (ii) it has been possible to perform C. difficile testing on demand every day including Saturday during P2 and P3, because of the simplicity of the techniques, which was not the case with the cytotoxicity assay. The difference in the return of positive results between P2 and P3 may be explained by the fact that only the GDH test was reported to physicians on Friday afternoon, confirmation of GDH-positive results being performed on the following Monday morning.
Second, the frequency of repeat testing decreased in P2 and P3 compared with P1. This finding had no impact on cost saving in our hospital, because redundant stool samples within a 7-day period are recorded but not processed. Although repeat testing is not recommended by different guidelines, it remains a frequent practice in many healthcare facilities . In the case of an initial negative test, repeat testing will also increase the likelihood of a false-positive result because of the lack of specificity of the methods. In the case of an initial positive test, repeat testing as a ‘test of cure’ may remain positive for toxin or culture in 56% of patients despite resolution of diarrhoea  and may lead to additional unnecessary treatment. This practice may be encouraged by delay in reporting results to the ward and induces extra costs.
Third, we found that the rate of patients specifically treated by metronidazole or vancomycin was higher during P2 and P3 compared with P1 (93.3%, 84.4% vs. 80.6%), suggesting that physicians considered that patients with a positive result by PCR were truly infected. Fewer patients were treated with vancomycin or metronidazole during P1, probably because the results of toxigenic culture were returned, in some cases, after the patients' discharge. Moreover, patients with CDI were treated more rapidly with metronidazole or vancomycin during P2 and P3. Rapid awareness of positive microbiology results on the part of healthcare professionals is essential for quality patient care. Our findings strengthen the results from Verdoorn et al.  who showed that telephone notification of test results for patients with CDI is associated with a decreased time to the ordering of antimicrobial therapy. They also showed that prolonged time to the ordering of antimicrobial therapy was associated with prolonged hospitalization. We found that length of stay after a positive C. difficile diagnosis was shorter by 3.5 days when a PCR is used (p 0.05) as a stand alone test and by 3.3 days (p 0.27) when a two-step algorithm is used, compared with conventional diagnosis. In addition, we also noticed that quicker treatment was associated with an overall decrease in recurrence rate, mortality at day 10, and less severe outcome, but the difference did not reach significance, probably due to lack of statistical power.
A rapid diagnosis also had major positive consequences on management of patients with negative results. We showed that the rate of empiric treatment with metronidazole or vancomycin decreased from P1 (13.4%) to P2 (6.4%) and P3 (5.6%), as well as the number of unjustified contact precaution days. All of these findings indicate that a rapid diagnosis improves patient care and outcome and may result in considerable savings to the hospital. These results are in agreement with previous reports. Sewell et al. reported that a rapid CDI diagnosis based on PCR impacts positively on patient care compared with conventional diagnosis by the cytotoxicity assay: patients were discharged earlier from hospital (−4.88 days for PCR-positive patients and −7.03 days for PCR-negative patients) (Sewell B. et al., ESCMID 2012, P2274). After implementing a PCR assay for C. difficile, Loo et al. showed that isolation days decreased by an average of 3–4 days per patient, depending on the hospital (Loo V. et al., ICAAC 2011, D1273). Similarly, empirical therapy among patients decreased from 29.1% to 10.9%. The high rate of empirical therapy reported in their study may be explained by the location of the facilities (Québec) where the large outbreaks of severe CDI due to the epidemic 027/BI/NAP1 strain were initially described in 2006 . Cantazaro et al.  showed that the switch from a toxin A/B enzyme immunoassay to a PCR method for the diagnosis of CDI led to a decrease of hospital onset healthcare-associated CDI (from 4.4 per 10 000 patient-days during enzyme immunoassay testing to 0.9 per 10 000 patient-days during PCR testing). They also found a significant decrease in patient isolation days, test ordered, and metronidazole usage for patients with negative C. difficile tests. However, these results were partly explained by the changes in their infection control practices: during the EIA period, patients with diarrhoea were placed in pre-emptive isolation until three stool samples tested negative for the presence of C. difficile, whereas during the PCR period, isolation was removed if a single negative PCR result for toxigenic C. difficile was obtained.
Although our main objective was not to perform a health-economic study of a rapid CDI diagnosis, our results represent valuable data that should be considered for further cost-effectiveness studies or for financial discussion with the administrative decision makers. Larson et al.  conducted a cost analysis study and showed that the increased cost of either direct PCR testing or PCR testing as part of a three-step algorithm was justified by the earlier detection of CDI cases, which would prevent additional cases of nosocomial CDI and shorten the length of stay of patients with CDI.
To our knowledge, this is the first report showing an increase of clinician satisfaction when a rapid diagnosis of CDI is available. Clarity of reporting was considered higher with PCR (binary answer: positive or negative) than with a two-step algorithm where positive GDH result reporting may be misinterpreted as a final positive diagnosis. This finding supports the idea that a rapid and reliable diagnostic test increases clinicians' trust in the results and can potentially influence empirical CDI therapy and decrease unnecessary antibiotic use, which represents a potential cost saving. It can also help to optimize the use of private hospital rooms by reliably identifying patients who do not have the disease.
The study has some limitations. First, it has been performed in a single centre, and was neither blinded nor randomized. It is therefore difficult to extrapolate the results to other healthcare facilities with different modalities of management of patients with CDI. Second, the frequency of patients with pre-emptive isolation precautions or empirical treatment was probably underestimated because patients who were isolated for another reason or those who were treated with intravenous metronidazole for another infection were not taken into account in the analysis. Third, we did not attempt to perform a medico-economic analysis to estimate whether the decrease of length of stay or repeat testing following rapid diagnosis may compensate for the higher cost of PCR or a two-step algorithm. However, with a daily cost of hospitalization attributable to C. difficile estimated to be between $1300 and $2000 [25-27] and a number of CDI cases in our healthcare facility of 150 per year, the PCR or two-step algorithm could save more than $585 000 annually simply by reducing by 3 days the length of hospitalization of each patient with CDI. Savings may be even more important if cost estimates take into account technician time, which is reduced when using PCR or two-step algorithm.
This study demonstrated that rapid C. difficile testing by PCR or a two-step algorithm impacts positively on patient management by reducing empirical treatment, length of hospitalization, and unnecessary repeat testing. It may also increase clinicians' satisfaction and their trust in laboratory results. All these changes can potentially translate into cost savings for the hospital.
We are deeply indebted to Professor Fred Tenover for his valuable comments and suggestions.
Cepheid provided Xpert C. difficile tests to perform the study. Cepheid reviewed the manuscript before submission but had no role in the data analysis and did not have control over the content of the manuscript.
FB has received research grants from Biomérieux, Biosynex, Diasorin, Cepheid, Alere, Meridian and R-biopharm. All other authors report no conflicts of interest relevant to this article.