Modelling diverse sources of Clostridium difficile in the community: importance of animals, infants and asymptomatic carriers

Abstract

Clostridium difficile infections (CDIs) affect patients in hospitals and in the community, but the relative importance of transmission in each setting is unknown. We developed a mathematical model of C. difficile transmission in a hospital and surrounding community that included infants, adults and transmission from animal reservoirs. We assessed the role of these transmission routes in maintaining disease and evaluated the recommended classification system for hospital- and community-acquired CDIs. The reproduction number in the hospital was 1 for nearly all scenarios without transmission from animal reservoirs (range: 1.0-1.34). However, the reproduction number for the human population was 3.5-26.0%) of human exposures originated from animal reservoirs. Symptomatic adults accounted for <10% transmission in the community. Under conservative assumptions, infants accounted for 17% of community transmission. An estimated 33-40% of community-acquired cases were reported but 28-39% of these reported cases were misclassified as hospital-acquired by recommended definitions. Transmission could be plausibly sustained by asymptomatically colonised adults and infants in the community or exposure to animal reservoirs, but not hospital transmission alone. Under-reporting of community-onset cases and systematic misclassification underplays the role of community transmission.

Keywords: Clostridium difficile; community-acquired infection; hospital-acquired infection; mathematical disease model; zoonotic infection.

Fig. 1.

Model structure showing including colonisation, gut flora status, symptoms and treatment. Adults in the immune classes do not have symptoms and therefore not all individuals with overgrowth seek or receive treatment (dashed arrows and box). The details for infants, immunity, demographics and hospital–community structure are summarised in Figure S1. The definitions and values of the parameters associated with each transition can be found in Table S1. †The force of colonisation depends in the number and type of infectious individuals in the same setting (hospital or community).

Fig. 2.

The classification of CDI cases based on IDSA and SHEA surveillance recommendations that we assessed with our model. Lessa et al. used a similar classification scheme to estimate incidence in the USA. *Lessa et al. used a 12-week cut-off and therefore do not classify any cases as ‘indeterminate’. ‡Lessa et al. used a 3-day cut-off. †We used symptom onset or hospital admission as reference points in our simulations. However, the classification system recommended by IDSA and SHEA uses onset of symptoms as the reference point for all cut-offs. Our classification is otherwise identical. Lessa et al. used date of positive faecal sample as reference point.

Fig. 3.

The reproduction number at the disease-free equilibrium for various plausible assumptions for the colonisation prevalence in adults and relative infectiousness of infants for (a) the whole population, (b) the community only and (c) the hospital only. The model had poorer model fit for the combination of high infant infectiousness and low adult colonisation prevalence, so these combinations are omitted from the figures.

Fig. 4.

The animal-driven threshold under various plausible assumptions for the C. difficile colonisation prevalence in adults, and the relative infectiousness of infants as (a) a proportion of all transmission in the community and (b) as rate of exposure to adults in the community. The reproduction number is less than one in the community if transmission from animals exceeds the animal-driven threshold. The model had poorer model fit at the animal-driven threshold for the combination of high infant infectiousness and low adult colonisation prevalence, so these combinations are omitted from the figures.

Fig. 5.

Classification of the origin of reported CDIs by time since hospital discharge or admission, comparing the actual incidence of reported hospital-acquired (HA) and community-acquired (CA) CDIs vs. the classification recommended by IDSA and SHEA and three variants. Lessa et al. use a 3-day cut-off for recent hospital admission and a 12-week cut-off for recent hospital discharge. The optimal and balanced classifications we have identified use 7.4- and 6.6-day cut-offs, respectively, for recent hospital admission and 2.1- and 12.5-day cut-offs, respectively, for recent hospital discharge.