Clostridium: Transmission difficile?

Stephan Harbarth and Matthew Samore discuss the implications, and the limitations, of new research that might indicate that most Clostridium difficile cases are imported into hospitals.

In a new study published in this issue of PLoS Medicine, Sarah Walker and colleagues examine the epidemiology of C. difficile infection, focusing on the role of within-hospital transmission among ward patients. The investigators used a simpli-fied model that was populated with observational data from one National Health Service (NHS) Hospital Trust in the United Kingdom, a country that introduced compulsory surveillance with mandatory C. difficile testing of all elderly inpatients with diarrhea in 2008 [10]. Surprisingly, based on the results of their network analysis combined with molecular strain typing, up to three-quarters of patients with C. difficile infection did not acquire their infecting C. difficile strain during their hospital stay. Using time intervals, strain types, and patient location as plausibility checks, the authors propose that within-hospital transmission accounted for a relatively small number of the overall C. difficile cases detected. However, the rates of transmission varied in different specialty wards, with renal and transplant wards having the highest documented rates. Most of the cases of C. difficile that were attributed to within-hospital transmission occurred shortly after the onset of symptoms of the index case, suggesting that the hospital environment was not, as has previously been claimed, a long-lasting reservoir for this pathogen [7]. Overall, this study suggests that alternative explanations need to be sought for the origin of most of the new onset cases of C. difficile infection.
Moving On-What Do We Need to Know Next about C. difficile Transmission?
This impressive study addresses an important question-to what extent can we control C. difficile infection by prevention of transmission from symptomatic C. difficile infection cases in hospitals? However, there are limitations to the approach chosen in this study, including several possible sources of bias already mentioned by the authors (e.g., selection, misclassification, and information biases).
Other potential limitations not considered in this study include the possibility of inter-ward transmission. Patients from different wards might, for instance, be transported to common sectors of the hospital for procedures and diagnostic  tests, e.g., X-rays. Potential vectors of transmission, including equipment and health care workers who might care for patients on different wards, could be similarly mobile. In this study, wards were small relative to the hospital size. It is likely that, on average, many more symptomatic C. difficile cases were housed on ''other'' wards than on the ''same'' ward. Even though rates of intra-ward transmission per infected case were probably significantly higher than rates of inter-ward transmission per infected case, the absolute number of inter-ward transmissions may have, in fact, exceeded the number of intra-ward transmissions. Second, the poor sensitivity of the Enzyme Immuno-Assay (EIA) testing method for C. difficile diagnosis may have ignored a potentially significant pool of undiagnosed C. difficile patients (which could have been selected as controls, introducing misclassification bias). Third, antibiotic exposure data were not recorded, which could have biased the dates of onset of symptoms and cross-transmission. Finally, transmission events linked to asymptomatic carriers were not routinely detected [11].

Practical Implications
The two key practical questions related to this study are 1) how much benefit is accrued by blocking transmission from symptomatic C. difficile infection cases; 2) what proportion of the C. difficile infections that are attributed to within-hospital transmission instead represent alreadyinfected individuals who come into the hospital carrying toxigenic C. difficile strains in their gut flora. The study by Sarah Walker and colleagues cannot provide definitive answers to these questions because it has significant limitations with respect to both issues. The study cannot answer question 1, about benefit accrued by blocking C. difficile transmission, because it did not examine interward transmission. Further, it cannot tell us how many patients came in already colonized or infected because it did not examine asymptomatic C. difficile carriage upon admission and discharge. Attempting to interpret the results of this study with respect to these practical issues highlights the need to utilise models that account for the non-linear dynamics of spread of C. difficile.

Future Studies
Further studies are needed to elucidate answers to the two key questions we have identified above. Investigations should examine the possibility of transmission from falsely EIA-negative symptomatic patients, asymptomatic carriers (patients or health care workers), and community acquisition with importation of C. difficile into the hospital setting [12], and this might require both more data and the use of more advanced transmission models such as hidden Markov models.
More detailed screening data, such as a study that reported screening of asymptomatic C. difficile carriers in a large prospective cohort [13], and new models will help to answer the question of whether C. difficile is less of an institutional and more of a community problem than has previously been thought. Proving that the majority of nosocomial C. difficile infections are actually imported into hospitals (with toxigenic C. difficile strains being already present on admission) would be ''revolutionary''-however, we believe that the evidence generated by this study, albeit tantalizing, is not yet sufficient to prove this hypothesis.

Author Contributions
Wrote the first draft of the manuscript: SH. Contributed to the writing of the manuscript: MS. ICMJE criteria for authorship read and met: SH MS. Agree with manuscript results and conclusions: SH MS.