Modelling studies of the transmission-dynamics andhospital burden of Clostridium difficile

Abstract

Clostridium difficile, a Gram-positive spore-forming bacterium, is a source of considerable morbidity and mortality for patients treated in hospitals and other healthcare settings. Intestinal colonisation by C. difficile can cause infection (CDI) if the normal flora is disrupted, e.g. by the use of antimicrobials and some other drugs. Vaccines targeting C difficile main virulence factors, toxins A and B are currently undergoing clinical trials, however, their potential population impact is largely unknown. The work presented in this thesis aims to quantify the effectiveness of C. difficile vaccination in preventing hospitalonset CDI, including both its direct effects (reduction in individual patient morbidity and mortality) and indirect effects (prevention of onward transmission of the bacteria) using a mathematical dynamic transmission model framework. Based on a systematic literature review, it was shown that mathematical dynamic-transmission models have become an increasingly popular tool to help understand the patient-to-patient spread of nosocomial pathogens and predict the impact of healthcare prevention and control strategies. Methods have generally improved, with an increased use of stochastic models, and more advanced methods for formal model fitting and sensitivity analyses. Nonetheless, in contrast to methicillin-resistant Staphylococcus aureus – another bacterium commonly found in the healthcare setting – the transmission of C. difficile has rarely been considered within a dynamic modelling framework. Using national English CDI hospital surveillance data to fit a generalised additive mixed-effects model, this thesis revealed that, in line with recent evidence based on highly discriminatory genetic typingmethods, whilst transmission between symptomatic carriers was significant, this did not account for the majority of CDI cases in English hospitals. Asymptomatic carriers have been suggested as cocontributors, but their role in transmission remains uncertain to date. Previous estimates of additional excess bed days attributable to healthcare-acquired-CDI have varied widely, partly due to methodological weaknesses, and no robust estimates from a European setting are available. Both form key determinants to help quantify the health and economic burden of CDI, and are also likely to have an impact on the transmission-dynamics of the infection. Therefore, this thesis quantified the hospital burden of CDI, expressed in excess length of stay and mortality. A Cox proportional hazard model revealed that CDI was associated with a significantly decreased daily risk of discharge and increased risk of mortality, where the former was even further reduced for severe CDI patients. Using a multi-state model more intuitive estimates, i.e. the excess length of stay associated with mild (5 days [1.1-9.5]) and severe CDI (11.6 days [95% CI = 3.6-19.6]) were obtained. Finally, the results of an individual-based “state-of-the-art” dynamic transmission model in an English ICU (with epidemiological parameters informed by the findings of the statistical models mentioned, and with data-driven patient movement between the community, LTCF and ICU) showed that in settings with in-hospital acquisition rates comparable to the national average in English ICUs, immunising three patient groups: LTCF residents, elective patients and patients with a history of CDI in the ICU, resulted in a 43%, reduction of ICU-onset CDI. This required a relatively high number of vaccine doses, and a targeted strategy involving patients at high risk of colonisation on admission, such as LTCF residents proved more efficient. As these results proved highly sensitive to the level of antimicrobial use and in-ward acquisition rates, it was concluded that vaccination might be most efficient when targeting patient risk groups or settings where implementation of antimicrobial stewardship proves challenging.