This study is the first to investigate the impact of HH improvement and reduction of antibiotic use on the dynamics of within-household transmission of ESBL-EC. Our results underline the importance of HH and the little impact that antibiotic reduction has on the dynamics of ESBL-EC transmission. Improving HH compliance by 50% reduced the probability of ESBL-EC acquisition by 30 to 60% according to the household composition and the category of index carrier. One reason that HH was the most effective control measure was that it acts in three ways, i.e. preventing contamination after contact with potentially ESBL-positive faecal matter; accelerating the spontaneous decontamination of hands and preventing cross-transmission of contamination; and preventing gut colonisation from contaminated hands.
HH with soap is a highly effective means of reducing infectious disease transmission; a systematic review showed that HH reduces the risk of diarrhoea episodes by 42–47% [22] and reduces the rate of respiratory infections by 5–34% [23]. Although the importance of HH in preventing infections is obvious, compliance in the community remains low. A systematic review showed that approximately 19% of the world population washes their hands with soap after contact with faeces (13–17% in low- and middle-income regions, and 46–49% in high-income regions) [24].
In our study, we used detailed data on HH based on a single study [17]. To our knowledge, it was the only study describing HH behaviour in households, including key moments for ESBL-EC transmission, and the category of household members. Further research is urgently needed to assess current household levels of HH and their impact on ESBL-EC transmission in the community.
Our modelling study showed that the probability of ESBL-EC acquisition was higher in households with children and especially those with a baby. Other studies also indicated a unique place of children in the transmission dynamics of ESBL-PE [26]. The higher probability of acquisition and transmission could be explained by the intensity of contacts between children and other household members, frequent contacts with contaminated environment and limited HH. Islam et al. found that the prevalence of intestinal carriage of ESBL-PE in U.S. children was the highest in 1 to < 2-year-olds and < 5-year-olds (6.5% and 5.2% vs.1.7% in children over 5 years old). Another study reported the transmission between a child carrying ESBL-EC and their family members in 23% of cases.[27] In our study, the probability of acquisition from a child or a baby was estimated at 31.2 or 52.8%, respectively. This could be explained by the long duration of colonisation considered in our model (111 days vs. 36 days (4–60), observed in the cited study). In an additional analysis, we fixed that the duration of gut colonisation at 36 days; this reduced the probability of transmission in a family composed of three persons and was 11.4 or 22.1%, when an index carrier was a child or a baby (Supplementary Text S2).
Human exposure to ESBL-EC may occur via raw meat, vegetables, animals, the environment, and human-to-human transmission. In particular, a high prevalence of ESBL-EC has been reported in retail chicken meat; however their role as a main cause of human EC infections remains controversial [8]. In a recent study, Mughini-Gras et al. quantified the significance of different sources of community-acquired ESCB-EC colonisation [9]. They indicated humans as the most important cause; however, the other sources also represented a large reservoir of ESBLs. In our model, we included the daily probability of ESBL-EC background colonisation based on a recent study [3]. This value was much lower that the probability of colonisation due to cross-transmission considered here. Thus, in an additional analysis, we investigated the impact of an increased probability of background colonisation on model predictions. We found that if the probability of background colonisation increased, it had a major impact on the persistence of colonisation in households, limited the impact of HH, and thus may subsequently contribute to community transmission.
Antibiotic use and misuse are the major forces associated with selection of resistant bacteria. However, reducing antibiotic use in the community gave divergent results on the reversion of antibiotic resistance. Indeed, studies examining the impact of antibiotic restriction on resistance were mostly performed in hospital settings and extrapolation from the hospital to the community is not straightforward [28]. Few studies have investigated the impact of antibiotic reduction on the resistance of E. coli in the community. One showed that a 28% reduction in the overall use of quinolones resulted in a significant increase in the susceptibility of E. coli to quinolones [28]. Another showed that antibiotic stewardship led to reduction of ciprofloxacin and cephalosporins and decreased the incidence of infections caused by ESBL-EC in the community [29]. One other study investigated the impact of restriction of sulphonamide prescription in the UK on the prevalence of resistance in E. coli.[30] Although the number of prescriptions decreased by 98% from 1991 to 1999, the frequency of E. coli resistance to sulphamethoxazole increased from 39.7% and 46%.
In our study, we compared the effectiveness of antibiotic reduction with improvement of HH in the community. Our results showed that an optimistic scenario with 50% restriction in antibiotic use reduced the probability of transmission modestly, by 2–6%, and that even a 10% improvement of HH compliance was more effective than a 62% reduction in antibiotic use in the community. Although antibiotic stewardship programmes may be important [29], our results show that improvement of HH was more effective in controlling the transmission of ESBL-EC in the community.
Our study has several limitations. Firstly, there are several uncertainties surrounding input parameters, in particular the probability of hand contamination with ESBL-EC after contact with faeces, the probability of cross-transmission and the probability of gut colonisation. Further research is needed to provide better estimates of these important inputs of the model. Secondly, we did not model households with elderly people. The elderly population may be important in the transmission dynamics of ESBL-PE by higher care needs and frequent contact with the healthcare system. Finally, we based our predictions on data from developed countries. Further research is needed to study the impact of HH on the transmission of ESBL-EC in developing countries where access to sanitation is limited and the probability of direct contamination from the environment could be very high.
The major strength of our study is the use of an individual-based model that incorporates key but still rare elements of the transmission dynamics of ESBL-EC, such as the frequency and nature of contacts among household members, impact of antibiotic treatment and HH. Furthermore, it was calibrated on actual data. Secondly, we quantified the effectiveness of antibiotic reduction and the improvement of HH in the community which would be very difficult to implement and compare in an interventional study. Finally, we used sensitivity analyses to assess the impact of uncertain input parameters on the outcomes of interest and to identify parameters to prioritise in future research. These parameters should be carefully documented if modelling studies are to guide policies regarding infection control measures.
In conclusion, our model findings suggest that the probability of ESBL-EC is high in households and especially those with a baby. Improving HH was the most effective intervention to reduce the spread of ESBL-EC in the community, as compared to antibiotic reduction. Major efforts should be directed towards improving hygiene in the community in order to limit the spread of ESBL-EC.