Do domestic animals contribute to bacterial contamination of infant transmission pathways? Formative evidence from Ethiopia

Child stunting is associated with poor water, sanitation and hygiene (WASH), partly due to the effect of infection on intestinal nutrient absorption. WASH interventions, however, show little effect on growth. A hypothesis is that bacterial contamination of hands and ﬂ oors from domestic animals and their faeces, and subsequent ingestion via infant hand-to-mouth behaviours, may explain this. This formative study used microbial testing and survey and observational data from 20 households in Ethiopia to characterise principle bacterial transmission pathways to infants, considering WASH facilities and practices, infant behaviours and animal exposure. Microbial swabbing showed the contamination of hands and ﬂ oor surfaces from thermotolerant coliform bacteria. Animal husbandry practices, such as keeping animals inside, contributed signi ﬁ cantly (all p < 0.005). There was no evidence that latrine facilities mitigated contamination across infant ( p ¼ 0.76) or maternal ( p ¼ 0.86) hands or ﬂ oor surfaces ( p ¼ 0.36). This small study contributes to the evidence that animal faeces are an important source of domestic bacterial contamination. The results imply that interventions aiming to reduce pathogen transmission to infants should think beyond improving WASH and also consider the need to separate infants and animals in the home. Intervention studies will be required to determine whether this reduces infant infection and improves linear growth.

and hand-to-mouth contact appears to result in significantly higher faecal intake in IYC than consumption of stored water (Mattioli et al. ). While direct ingestion of faeces is more rare, IYC frequently place soil in their mouth or put their hands in their mouths after touching soil (Ngure et al.

Study sites and sampling frame
Formative research is intended as an initial part of the process of a larger study design and can use both qualitative and quantitative methods to provide data for research teams to plan interventions or further data collection. While formative research is critical to designing and delivering interventions or programmes which are efficient and effective, it is earlyphase data not powered to detect differences between groups (Gittelsohn et al. ). As such, this study must be interpreted in this contextit is intended to provide indicative evidence on a hypothesis but is not sufficiently powered to provide conclusive evidence in this area. This formative study was conducted in the Sidama zone, Southern Nations, Nationalities, and Peoples (SNNPR) region, Ethiopia as the geographical outreach area of the nongovernmental organisation People In Need. The study took part in the month of May, which in the region is a mostly dry period with some afternoon rainfall. From the six woredas (districts) in the zone, 16 kebeles (neighbourhoods) were grouped into peri-urban (n ¼ 6) or rural (n ¼ 10). A simple random sampling method was used, whereby kebeles within the People In Need intervention area were listed, given a number and using a lottery method, eight kebeles were  1 | Age-specific pathways and contributing sources by which infants are exposed to and ingest pathogenic bacteria within the home. Where the dashed lines constitute the original 'F diagram', the figure illustrates how the original diagram does not consider specific infant behaviours (boxes in bold), nor the contribution of animal faeces to domestic contamination, but the integral relationship with those three components and microbial ingestion. Adapted alongside the 'F diagram', as published by Wagner & Lanoix (1958).

Survey and infant observation period
A survey was designed to assess sanitation facilities, handwashing practices, animal presence, livestock husbandry and diarrhoea prevalence and duration of diarrhoeal episodes (where to avoid reporting bias, researchers asked caregivers for the frequency of loose or watery stools over the past 7 days and subsequently applied World Health Organization criteria) (Baqui et al. ). This was administered alongside a 1-hour observation period which noted infant hand-to-mouth behaviours and general sanitary conditions. During this, a pre-tested semi-structured survey tool was used to record every object that was either mouthed or touched by the infant, where mouthing was defined as an infant putting fomites or fingers into their mouths whether swallowed or not. This semi-structured tool was calibrated to capture some of the underemphasised key pathogen transmission pathways in infants. As illustrated in Figure 1, this is also well described by the 'modified At the end of data collection, only one blank sample was found to be contaminated during sample processing.

Data analysis
Means for bacteria counts from swabs were calculated as

Survey results
The WASH survey asked questions regarding infant characteristics, diarrhoea prevalence and episode duration (as described), latrine ownership and use, handwashing and animal husbandry practices. Briefly, most houses had a pit latrine either with a slab (40%) or without (30%). The remainder (20%) had no toilet at all (assumed to openly defecate) or used the toilet of a neighbour (10%). Only five (25%) households had a specific place to wash their hands; of those, all households had water available, but only three (15%) had soap (in two instances visual inspection indicated the soap was likely not used). Seventy per cent of households raised animals of some kind, with the most common chickens (93%) and cattle (71%). When asked where animals lived during the day, only one household reported that their animals lived outside enclosed in an area, with the rest kept outside either unenclosed or living inside with the family, suggesting that animals were mostly uncontained. One hundred per cent of households reported that during the night animals lived inside with the family. Regarding diarrhoea prevalence, three infants (15%) were reported to have experienced three or more loose stools within a 24-h period; across these infants, the reported mean duration of a diarrhoeal episode was 3.3 days. Table 1 illustrates these findings along with general hygiene characteristics of the infant's environment.

Infant observation period results
Nineteen infant-caregiver pairs were observed for 20 h during the infant observation period. Infants were frequently observed to mouth their own hands (a mean of 31 times over 1 h), or to mouth those of their caregiver (mean of 21 times over 1 h), which in the majority of instances were both visibly dirty (90% and 86%, respectively). Throughout the observation, infants would typically have nothing to play with other than a plastic bottle, which may explain why infants were observed to frequently suck their hands or those of their caregiver. Animal faeces were directly ingested by two infants, and the floor surface material was also picked up and directly entered into the infants' mouths on seven occasions.

Microbiological data
Samples from the inside floor surface showed the highest bacterial count with a mean TTC CFU/dry g of 76.5 (1.88 log 10 ). The higher count in the floor sample may reflect the typical presence of animals inside as well as the common occurrence of animal faeces. High counts in the domestic floor sample are significant, given the sample was collected within crawling reach of the infant and was observed to enter the infants' mouths and contaminate their hands. Infant hand contamination showed a slightly higher mean count than those of their caregiver (mean TTC CFU/hand 33.3 (1.52 log 10 ) versus 23.6 (1.37 log 10 ), respectively) (p < 0.005, data not shown). This is unsurprising in a context where infants were frequently crawling on the floor and touching and mouthing objects, which were usually visibly dirty (Table 1). The relationship between key vectors and transmission pathways, as measured by microbial testing, is presented in Figure 2.

Study limitations
This study presents some limitations. Firstly, the small sample size in this study of 20 infants/households would not have comprehensively captured variability in TTC contamination across pathways, which likely varies considerably. As is noted elsewhere (Navab-Daneshmand

Raises animals 14 70
Hygiene characteristics (observed) In this study, even in households with a latrine (improved or other) contamination was still common, suggesting that even with sufficient sanitation infrastructure the presence of animals within the home may propagate contamination.
In one study in Bangladesh, while households with fewer contaminated toys and objects were those with high latrine coverage and WASH infrastructure (Torondel et al. ), the absence of animals was highlighted as a possible noteworthy factor to low levels of contamination (

CONCLUSION
Although the evidence presented here is of a small sample, results support the growing body of evidence which suggests that WASH interventions must address animal faecal contamination across the domestic environment: a 'total environmental hygiene' approach, which fully addresses multiple sources of contamination. Increased attention should be placed not only on WASH infrastructure and quality but also on addressing barriers from widespread faecal contamination to overall improved hygiene on a much wider level.
Similarly to another recent study (Kwong et al. ), this study did not observe human faeces within the home and suggests that faecal contamination from animals may be a primary limitation in WASH interventions, which tend to focus on that of humans. Further interventions, which aim to improve infant growth by addressing contamination, are likely to benefit from considering certain common animal husbandry practices, such as keeping animals indoors during the day and night, and the need to separate IYC from animals. If not considered, in this setting and other simi-