The study was designed as a cohort study. Following approval (PV4643) by the ethics committee of the Ärztekammer Hamburg (Hamburg’s General Medical Council), 262 patients without diarrhoea consecutively admitted to the geriatric unit of the Marienkrankenhaus from March to November 2014 were examined. The Katholisches Marienkrankenhaus gGmbH is a teaching hospital of the University of Hamburg with 550 inpatient beds in total in various different medical units. The geriatric department consists of 5 wards of 126 beds. Written consent was obtained after detailed information and explanation of the study procedures.
First, we tried to assess the association of clinical variables with Clostridium difficile colonization.
Secondly, we explored the association of C. difficile colonization with subsequent development of CDI. Patients were monitored throughout their hospital stay with regard to the development of a symptomatic CDI.
As described in 2.5, Statistical analyses, a required number of at least 250 patients had been calculated based on an examined rate of CDI of 4% in geriatric in-patients in 2012 in the Katholisches Marienkrankenhaus.
Only patients without diarrhoea were included with diarrhoea being defined as the occurrence of >3 unformed stools per day. Participation could be revoked at any time without stating any reasons. Stool samples at the first bowel movement after hospital admission were collected for testing. They were to be delivered within 6 days after admission obtained spontaneously; later stool samples were not evaluated.
Patients who had acute diarrhoea, or who declined participation, or who were incapable of giving consent and had no legal representative, or whose legal representative refused consent were excluded.
Patients who agreed to take part but failed to deliver a stool sample within the first 6 days after admission were excluded as well. The first stool sample of each participant was tested for toxigenic Clostridium difficile using PCR (see Lab analysis.).
Data collection
The following anamnestic data of the patients taking part were recorded: age, sex, date of admission, current duration of hospital stay, number and duration of previous hospital stays within the past 6 months, current or previous treatments with antibiotics within the past 6 months and the respective agents, medication with proton pump inhibitors (PPI) or immunosuppressants. In our patients the following immunosuppressant drugs were used: systemic corticosteroids, methotrexate, ciclosporine and leflunomide; two patients had just finished a cycle of cytostatic medication with R-CHOP (including rituximab, cyclophosphamide, vincristine, doxorubicine and prednisolone).
The immediately previous place of residence was categorized as transfer from 1) a stay at an external hospital, 2) a different department of the same hospital, 3) home, which also included a care facility. Further, the place of residence before admission to hospital was recorded: living independently, in sheltered accommodation, or within a care facility; in addition, known previous CDI episodes (medical history, referral letters) and one further category, “post-surgery” (surgery as the initial reason for admission to hospital), were included, assuming a higher risk of colonization in this patient group.
Geriatric assessment at admission
At the time of admission, all participating patients underwent the following geriatric assessment: Barthel Index [20], Mini Mental State Examination (MMSE) [21], measuring Handgrip Strength in kPa [22], and Timed up and Go [23]. Additionally, the Body Mass Index was determined and the presence of comorbidities was recorded using the Charlson Comorbidity Index [24].
Lab analysis
In all participating patients, the first stool sample after admission to the geriatric ward was tested for toxigenic C. difficile using PCR. A commercially available PCR test system was used: the Xpert C. difficile test on the GeneXpert by Cepheid (Germany, Frankfurt).
This test uses real-time PCR to determine the presence of toxigenic C.difficile strains by means of nucleic acid detection of gene sequences of the cytotoxin B (tcdB) and of the binary toxin (cdt), and also the deletion of the repressor gene tcdC. This deletion leads to an increased toxin production and is characteristic of the PCR-ribotype 027 (PFGE-type NAP1 and REA-type B1).
The result either confirms or rules out the existence of toxigenic C.difficile; if a positive result occurs, additional information is given on whether the pathogenic strain ribotype 027 (PFGE-type NAP1, REA-type B1) is likely to be present or not.
The performance data provided by the test manufacturer compared with the toxigenic culture result in a sensitivity of 100% at a specificity of 93% (Xpert C. difficile; GXCDIFFICILE-CE-10; 300-9291G Rev E, November 2012).
In patients who subsequently developed clinical symptoms of CDI, an EIA test for glutamate dehydrogenase (GDH) and, if positive, a follow-up toxin A/B EIA test were performed (C.DIFF CHEK–60 and C. DIFFICILE TOX A/B II; TECHLAB/ Alere USA).
Surveillance of the occurrence of a symptomatic CDI
For clinical follow-up, Clostridium difficile infection (CDI) was defined as diarrhoea with loose stools at least three times in 24 h, elevated C-reactive protein and blood leukocyte count, and positive results of GDH and toxin A/B EIA tests as described above.
Occurring diarrhoea was documented throughout the entire hospital stay. When diarrhoea occurred, the diagnostic lab tests mentioned above were carried out to either confirm or rule out a CDI.
At discharge the patient file was evaluated and the marked fact of a CDI during the stay in the geriatric ward was recorded for the study.
Statistical analysis
The sample size calculation was based on the follow-up part of the study: Preliminary examinations in 2012 showed that 4% of all in-patients of the geriatric department in our hospital suffered from clinical CDI. As this rate is in accordance with a literature report in a comparable institution [25], we based our sample size calculation on this assumption. In addition, we assumed that 20% of our patients would be colonized carriers at admission, but asymptomatic [2, 26]. We aimed to detect a tenfold increase of clinical infection rate in carriers vs. non-carriers by means of univariate analysis with a power of 80% at a level of significance (α) of 5%. Based on these assumptions, sample size calculation yielded a sample size of 250.
First, a descriptive characterization of the participants was made. As a next step, clinical variables were compared between the PCR positive and PCR negative groups by univariate analysis. Statistical significance was reached for a bilateral p value ≤ 0.05. Clostridium difficile colonization (PCR positive) was treated as dependent variable while clinical features, such as previous antibiotic treatment and CGA domains, were treated as independent variables.
In addition, the significant variables of this univariate analysis were entered into the binary logistic regression (“backwards: LR” method) in order to predict the PCR results. As level of significance for variable entry p ≤ 0.05 was used. The level used for the backward method was p(out) = 0.10, p(in) = 0.05.
The correlation matrix of the significant variables was calculated and used for variable entry into the model in order to minimize the risk of overfitting and multicollinearity.
Thus, only significant variables of the univariate analysis were chosen for the multivariate analysis. The calculation was carried out by an independent statistician using SPSS, version 22.0, according to Schendera [27].
Finally, we compared the initially colonized patients who developed CDI with the initially colonized patients who did not fall ill during hospital stay by cross tabulation and Fisher’s exact test (two- sided, p = 0.000).