keywords toys, toilets, environmental faecal contamination, Bangladesh, sanitation, coliforms

Transcription

1 Tropical Medicine and International Health doi: /tmi volume 19 no 5 pp may 2014 Toys and toilets: cross-sectional study using children s toys to evaluate environmental faecal contamination in rural Bangladeshi households with different sanitation facilities and practices Jelena Vujcic 1, Pavani K. Ram 1, Faruqe Hussain 2, Leanne Unicomb 2, Partha Sarathi Gope 2, Jaynal Abedin 2, Zahid Hayat Mahmud 2, M. Sirajul Islam 2 and Stephen P. Luby 2,3 1 State University of New York at Buffalo, New York, NY, USA 2 International Centre for Diarrhoeal Disease Research Bangladesh, Dhaka, Bangladesh 3 Stanford University, Stanford, CA, USA Abstract objective To evaluate household faecal contamination using children s toys among 100 rural Bangladeshi households categorised as cleaner (toilet that reliably separates faeces from the environment and no human faeces in/around living space) or less clean (no toilet or toilet that does not reliably separate faeces from the environment and human faeces in/around living space). methods We distributed toy balls to each household and rinsed each study toy and a toy already owned by the household in 200 ml of Ringer s solution. We enumerated faecal coliforms and faecal streptococci from each rinse using membrane filtration methods. results Study toys from 39 cleaner households had lower mean faecal coliform contamination than toys from 61 less clean households (2.4 log 10 colony-forming units (CFU)/200 ml vs. 3.2 log 10 CFU/ 200 ml, P = 0.03). However, wealth measures explained a portion of this relationship. Repeat measures were moderately variable [coefficient of variation (CV) = 6.5 between two toys in the household at the same time, CV = 37.6 between toys in the household at two different times 3 4 days apart]. Too few households owned a non-porous toy to compare groups without risk of bias. conclusion In rural Bangladesh, improved sanitation facilities and practices were associated with less environmental contamination. Whether this association is independent of household wealth and whether the difference in contamination improves child health merit further study. The variation found was typical for measures of environmental contamination, and requires large sample sizes to ascertain differences between groups with statistical significance. keywords toys, toilets, environmental faecal contamination, Bangladesh, sanitation, coliforms Introduction Sanitation interventions aim to decrease the burden of diarrhoeal disease a leading cause of death in young children in low- and middle-income countries (Liu et al. 2012). While some studies show improvements in sanitation reduce diarrhoea risk, in a recent meta-analysis, authors could not determine a collective effect of sanitary interventions on diarrhoea risk due to variable methodology and generally poor quality of data from these studies (Clasen et al. 2010). One way to better understand the potential contribution of improvements in sanitation is to assess whether they impact faecal contamination of the household and surrounding environment. Improvements in sanitation may reduce the risk of childhood diarrhoea by separating faeces from the domestic and peri-domestic environments of young children, presumably by decreasing microbial contamination of these environments, including household surfaces, objects and other fomites (Figure 1). In high-income settings, microbiological testing of kitchen or bathroom surfaces is used to assess household contamination. In low-income settings, where the burden of diarrhoeal disease is high, finding a standardised surface in the cooking or bathroom areas to make such an assessment is difficult; these structures or surfaces might not exist, or the type of material that composes these surfaces could vary substantially between households. Also, the extent to which young children John Wiley & Sons Ltd

2 Faecal contamination of hands Defecation (human & animal) (animal faeces) (human faeces) POOR SANITATION (unimproved toilet, open defecation) (animal faeces) Unsafe disposal of faeces No disposal of faeces Faecal contamination of the peri-domestic environment Faecal contamination of vehicles and vectors (fomites, flies, tracking by people, objects and animals) Faecal contamination of the domestic environment (food, water, surfaces, objects) Ingestion of faecal microbes by children Increased risk of diarrheal disease Figure 1 Model of the mechanism for transmission of faecal microbes from human and domestic animal faeces. come into contact with kitchen or bathroom surfaces is unclear. A standardised fomite that is readily encountered by young children could serve as an appropriate indicator of the faecal contamination level of the household. Data from child care facilities in high-income settings indicate toys can contain high levels of faecal contamination (Ekanem et al. 1983; Van et al. 1991; Laborde et al. 1993; Jiang et al. 1998). Coinvestigators of this study conducted preliminary analyses among 50 rural Bangladeshi households to assess differences in toy contamination between households with improved and unimproved sanitation facilities. Toys in households with an improved latrine were less contaminated with faecal coliforms than toys from households with unimproved latrines (DiVita et al. 2008). However, the study was constrained by a small sample, and the association may have been due to chance. Recently, Pickering and colleagues tested surfaces and fomites in 20 peri-urban, Tanzanian households to understand the pattern of environmental faecal contamination in the domestic and peri-domestic environment. They reported high levels of E. coli and enterococci on fomites (including toys) and on wet/damp surfaces (Pickering et al. 2012). We chose toy balls as the standard fomite to investigate whether faecal contamination of children s toy was a marker of household contamination by evaluating the association between household cleanliness and faecal contamination of toys (objective 1). Cleaner households had a toilet that reliably separated faeces from the environment and had no visible human faeces in the living or adjacent space, whereas less clean households did not have a toilet that reliably separated faeces from the environment and had visible human faeces in the living or adjacent space. We also assessed the variability and internal validity of the toy contamination measure (objective 2). As rinsing toys already present in the household would require fewer visits and fewer resources than providing toys and then retrieving them at a later date, we compared levels of faecal contamination of toys provided by the study to the faecal contamination of toys already owned by the household (objective 3). Methods We conducted this cross-sectional study in June August 2011 in a densely populated rural area in northern Bangladesh. We recruited primary caregivers of children 3 years old who participated in the WASH Benefits pilot study ( specifically in the control arm, and two water treatment arms. The pilot study was completed prior to this study. Using baseline data from the WASH Benefits pilot study (collected 1 year prior), we identified households with a child 36 months old that met criteria for either comparison group: Cleaner households had a toilet that could reliably separate faeces from the environment (flush/septic/piped sewerage or pit latrine with slab and water seal), and no visible human faeces inside the household, within the compound (multiple households sharing a yard) and on the way to the compound. Less clean households did not have a toilet that could reliably separate faeces from the environment (practised open defecation, had a hanging toilet, open pit latrine, pit latrine with slab but no/broken water seal and no lid, or flush/pour to somewhere else) and had human faeces present at two or more of the observed places (way to compound, within compound, inside household). Field workers approached households starting with those closest to the field office. Prior to consent, field workers observed the latrine and household to confirm each household met the inclusion criteria and confirmed the age of the youngest child. When the list of candidate 2014 John Wiley & Sons Ltd 529

3 households was exhausted, we had not met the target sample size. WASH Benefits pilot study baseline data was collected 1 year prior to this study, and it was possible that households we initially excluded (because they did not fit in either group) changed their latrine or have different environmental faecal conditions. Field workers approached such households and sought informed consent if a household met the criteria for either of the groups. Based on the preliminary analysis of toy contamination (DiVita et al. 2008), we assumed the variance of log scaled data was 0.93 for the cleaner group and 1.4 for the less clean group. Using a 2:1 ratio, we estimated 100 households (40 cleaner and 60 less clean households) would be sufficient to detect a difference of 0.03 log 10 in arithmetic means of log-transformed CFU/100 ml between cleaner and less clean groups, at the 95% significance level. At the initial visit, field workers collected information about sanitation, water treatment and handwashing practices and distributed two sterilised non-porous, rubber toy balls that were similar in size and colour/type (Figure 2). Three to four days later, field workers, who were trained to collect and handle samples, rinsed each toy ball for s in 200 ml of Ringer s solution, and distributed two new, similar toy balls. If the household owned a non-porous toy that would fit inside a 24 oz. Whirlpak TM bag, the field team washed one such toy using the same procedure as for the study-provided toys. Three to four days after the first follow-up visit, field workers repeated the same steps as in the first follow-up visit but did not wash a household-owned toy. Primary caregivers provided written informed consent. The study protocol was approved by the icddr,b Ethical Review Committee. Laboratory analysis of samples We maintained rinse samples at 4 10 C and transported them to the icddr,b laboratory within 24 h of collection. We used faecal coliforms and faecal streptococci as indicator bacteria for monitoring the faecal contamination of toys. Faecal coliforms and faecal streptococci are isolated from faeces of warm-blooded animals and are widely used indicators (Clescerl et al. 1999; Hurst & Crawford 2007). As measuring toy contamination is a relatively new approach to understanding household contamination, we measured faecal streptococci, which are more persistent than faecal coliforms in the environment. We did not use E. coli as an indicator as pilot work indicated that E. coli counts were low and not different between households with improved and unimproved latrines (DiVita et al. 2008). We used standard membrane filtration methods to quantify the number of colony-forming units (CFU) of faecal coliforms and faecal streptococci. For each sample, we filtered 100 and 5 ml through a 0.22 lm Millipore (Billerica, MA) membrane filter, then placed the filters on modified faecal coliform (mfc) plates and Kenner Fecal Streptococcus Agar (KFSA) plates. The remaining samples were stored at 4 C to prevent growth. Some toys were encrusted with mud/dirt, which created hazy samples. If a sample appeared moderately hazy, we filtered 20 ml instead of 100 ml to allow us to enumerate CFU within detection limits. Similarly, if a sample appeared very hazy or opaque, a 2-ml aliquot was also filtered and plated to enumerate each microbe. For all samples, we inoculated 100 ll of sample on to mfc and KFSA media following the drop plate technique in order to quantify samples where the colonies on the filter were numerous to count. We reported CFU from the drop plate method if we could not count the number of CFU on the filter. If the number of CFU were too numerous to count, we made 10-fold dilutions and plated following the same protocol 1 day later for faecal coliform enumerations, and 2 days later for faecal streptococci enumerations. We incubated mfc plates at 44 C for h then counted the characteristic blue colonies as faecal coliforms. We incubated KFSA media at 35 C for h then counted the characteristic pink/red colonies as faecal streptococci. Data analysis Figure 2 Rubber toy balls distributed to households. We compared respondent and household characteristics between cleaner and less clean households. To evaluate difference in toy contamination between cleaner and less clean households (objective 1), we compared means of log 10 -transformed CFU of faecal coliforms, and faecal John Wiley & Sons Ltd

4 streptococci between cleanliness groups and used t-tests to determine whether the means were statistically different between groups. If the number of CFU detected was zero, we replaced the zero value with 0.5 (half the lower detection limit) to carry out the log 10 transformation. As each household received four toy balls over the study period, we made the following comparisons of the mean of log 10 -transformed contamination between cleanliness groups: (i) for each toy individually; (ii) at each time point (time point 1 where toys 1 and 2 were dispersed, and time point 2 where toys 3 and 4 were dispersed) where we first calculated the mean of log 10 -transformed CFU of two toys at each time point; and (iii) for all toys retrieved within a given household where we first calculated the mean of log 10 -transformed CFU of all retrieved toys per household. We used linear regression to model the relationship between household cleanliness and toy contamination (as calculated in comparison iii, described above) with and without adjustment for potential confounders. Confounders included household assets, such as an electricity connection, and education level, which served as a proxy for wealth. Due to our small sample size, some cell sizes had fewer than five observations when we included more than two variables in the model. Therefore, we tested one potential confounder at a time in the regression model. To assess variability of toy contamination (objective 2), we calculated the coefficient of variation (CV) between the two toy balls placed in each household at the same time and we calculated CV between the toy balls at two different times. We assessed the internal validity of toy contamination as a measure of environmental contamination by comparing toy contamination to the number of human and/or animal faeces piles observed in the domestic and peri-domestic environment of the household among the entire study population. We separately constructed tertiles of number of piles of human and nonhuman faeces based on the distribution of number of piles observed per compound. We calculated the mean log 10 -transformed faecal coliform contamination of toys for each tertile and used one-way ANOVA to test for differences between tertiles. To assess the correlation between faecal contamination found on the study-provided toys and toys already owned by the households (objective 3), we calculated the correlation coefficient. Results We enrolled 39 cleaner households and 61 less clean households. Households in the cleaner group had more members, fewer male children (younger than 15 years) living in the compound, more years of paternal education, higher income, and had a higher proportion of respondents reporting household assets and access to electricity (Table 1). Faecal contamination of toys between cleaner and less clean groups (objective 1) Thirteen samples were below the detection limit for faecal coliforms and none for faecal streptococci. No samples were above the detection limit when considering both methods. Our data appeared normally distributed after log-transformation. In bivariate analysis, toys from cleaner households were less contaminated with faecal coliforms (mean = 2.6 log 10 CFU/200 ml, SD = 1.4) than toys from less clean households (mean = 2 log 10 CFU/ 200 ml, SD = 1.0; P-value 0.03; Table 2). When we compared faecal coliform contamination between groups for each toy ball separately, toys were consistently less contaminated in the cleaner households than in the less clean households. However, the comparison between groups of faecal coliform contamination of toys 2 and 4 did not reach statistical significance (Table 2). Faecal streptococci contamination of toys was not different between cleaner and less clean households (Table 2). In linear regression, toys in less clean households had a 0.57 higher mean log 10 -transformed faecal coliform count compared to toys in cleaner households (beta-coefficient = 0.57, 95% CI = , P-value = 0.02). When we included electricity connection in the model, less clean households had a 0.49 higher mean log 10 -transformed faecal coliform count compared to cleaner households (beta-coefficient = 0.49, 95% CI = 0.01, 0.98, P-value = 0.05). When the model included father s education, less clean households had a 0.48 higher mean log 10 -transformed faecal coliform count than cleaner households (beta-coefficient = 0.48, 95% CI = 0.1, 0.97, P-value = 0.06). Variability and internal validity (objective 2) We found similar variability for faecal contamination of two study toys placed in the household at the same time (CV = 36.5) and for toys placed in the household at two different times three to four days apart (CV = 37.6) (Figures 3 and 4). Mean faecal coliform contamination of toys increased as the number of human faeces piles observed in the household and compound increased (2.7 log 10 CFU/200 ml (SD = 1.4, n = 40) for households with no human faeces in the living space, 3.1 log 10 CFU/ 200 ml (SD = 1.0, n = 23) for households with 1 pile of human faeces, and 3.2 log 10 CFU/200 ml (SD = 0.92, n = 37) for households with 2 or more piles of human 2014 John Wiley & Sons Ltd 531

5 Table 1 Respondent/household demographic characteristics Mean (standard deviation) or number (%) for Household cleanliness category:* Characteristics All households (n = 100) Cleaner (n = 39) Less clean (n = 61) P-value Respondent s age 27.0 (5.9) 28.1 (6.6) 26.3 (5.4) 0.13 Age of child (months) 23.5 (8.5) 23.6 (8.7) 23.4 (8.5) 0.91 Male gender of child 60 (60) 23 (59) 37 (61) 0.87 Household composition # Household members 5.5 (2.2) 6.2 (2.6) 5.2 (1.7) 0.02 # Children under (0.7) 1.5 (0.8) 1.7 (0.7) 0.35 # Children under (1.5) 2.9 (1.6) 2.8 (1.4) 0.74 # Male children under (0.9) 1.2 (0.9) 1.5 (0.9) 0.07 Compound composition # Household in compound 5.9 (2.8) 5.8 (2.6) 5.9 (3.0) 0.90 # Compound members 28.1 (12.6) 29.9 (13.3) 27.0 (13.8) 0.30 # Children under (3.2) 5.3 (2.4) 5.8 (3.6) 0.49 # Children under (5.9) 12.2 (5.3) 12.0 (6.4) 0.88 # Male children under (3.5) 4.2 (2.6) 5.8 (3.9) 0.01 Income (in BDT per month) Up to 4000 BDT 23 (23) 3 (8) 20 (33) < BDT 39 (39) 10 (26) 29 (48) 6001 BDT and above 38 (38) 26 (66) 12 (19) Mother s education Up to/including primary school 73 (73) 24 (62) 49 (80) 0.04 Above primary 27 (27) 15 (38) 12 (20) Father s education Up to/including primary school 78 (78) 24 (62) 54 (89) <0.01 Above primary 22 (22) 15 (38) 7 (11) Assets Black and white television 12 (12) 5 (13) 7 (11) 0.84 Electricity 51 (51) 28 (72) 23 (38) <0.01 Mobile phone 49 (49) 27 (69) 22 (36) <0.01 Materials at handwashing place None 17 (17) 4 (10) 13 (21) 0.17 Water only 75 (75) 30 (77) 45 (74) Any type of soap and water 8 (8) 5 (13) 3 (5) *Cleaner households have a toilet that reliably separates faeces from the environment and have no visible human faeces in the household, in the compound and on the way to the compound at baseline. Less clean households do not have toilet that reliably separates faeces from the environment and had human faeces at 2 or more places. BDT, Bangladeshi Taka (1 USD was approximately 70 Taka at the time of the study). faeces). However, the number of faeces piles explained little of the variance, and the correlation may have been due to chance [F (2, 97) = 2.8, P-value=0.07, R 2 = 0.05]. Similarly, the mean faecal coliform contamination of toys increased as the number of non-human animal faeces piles observed in the household or within the compound increased (2.7 log 10 CFU/200 ml (SD = 1.2, n = 27) for households with <15 piles of animal faeces, 3.0 log 10 CFU/200 ml (SD = 1.1, n = 39) for households with piles of animal faeces and 3.2 log 10 CFU/ 200 ml (SD = 1.2, n = 34) for households with 19 or more piles of animal faeces), but the association explained little of the variance and was likely due to chance [F(2,97) = 1.0, R 2 = 0.02; P-value = 0.36]. Faecal contamination of toys owned by the household compared to study-provided toys (objective 3) Only 30 households owned toys (ex. plastic toy car, small plastic cup) that met our criteria (made of non-porous material and fits in a 24 oz. Whirlpak TM bag) for microbial contamination testing. Three samples were below the detection limit for faecal coliforms and none for faecal streptococci. No samples were above the detection limit John Wiley & Sons Ltd

6 Table 2 Means of log 10 -transformed counts of faecal coliform and faecal streptococci on toys distributed to cleaner and less clean households Toys (n = number of households) Household cleanliness* Cleaner (SD) Less clean (SD) P-value Study-provided toys Toy 1 (n = 100) (n = 39) (n = 61) Faecal coliform CFU 2.69 (1.40) 3.25 (1.33) 0.04 Faecal streptococci CFU 4.87 (0.66) 4.69 (0.80) 0.24 Toy 2 (n = 96) (n = 35) (n = 61) Faecal coliform CFU 2.61 (1.50) 3.09 (1.23) 0.09 Faecal streptococci CFU 4.86 (0.73) 4.63 (0.87) 0.18 Mean of toys at 1st (n = 39) (n = 61) follow-up visit (toys 1 & 2) (n = 100) Faecal coliform CFU 2.71 (1.36) 3.17 (1.11) 0.07 Faecal streptococci CFU 4.87 (0.64) 4.66 (0.72) 0.13 Toy 3 (n = 99) (n = 39) (n = 60) Faecal coliform CFU 2.49 (1.73) 3.33 (1.44) 0.01 Faecal streptococci CFU 4.72 (0.86) 4.78 (0.82) 0.72 Toy 4 (n = 93) (n = 36) (n = 57) Faecal coliform CFU 2.74 (1.90) 3.18 (1.46) 0.20 Faecal streptococci CFU 4.79 (0.93) 4.68 (0.78) 0.57 Mean of toys at 2nd (n = 39) (n = 61) follow-up visit (toys 3 & 4) (n = 100) Faecal coliform CFU 2.57 (1.74) 3.24 (1.33) 0.03 Faecal streptococci CFU 4.75 (0.83) 4.71 (0.71) 0.81 All study-provided toys (n = 39) (n = 61) (n = 100) Faecal coliform CFU 2.64 (1.40) 3.21 (0.95) 0.03 Faecal streptococci CFU 4.81 (0.66) 4.69 (0.53) 0.35 Household-owned toys (n = 13) (n = 17) (n = 30) Faecal coliform CFU 1.95 (1.71) 3.36 (1.23) 0.01 Faecal streptococci CFU 4.45 (1.13) 4.53 (0.73) 0.82 *Clean households had a toilet that reliably separated faeces from the environment and had no visible human faeces in the household, in the compound and on the way to the compound at baseline. Less clean households did not have toilet that reliably separated faeces from the environment and had human faeces at 2 or more places of observation at baseline. Each household was given a 2 toy balls at the baseline visit and 2 toy balls at the 1st follow-up visit. Toys 1 and 2 were collected for culture at the 1st follow-up visit. Toys 3 and 4 were collected for culture at the 2nd follow-up visit. Some toys were lost before data collectors returned to rinse the toy. Number of retrieved/ rinsed toys are marked. At follow-up visit 1, each household was asked whether they owned toys other than those provided by the study. One nonporous toy that could fit in the collection bag was cultured per household of the households that possessed non-study toys. Log-transformed CFU was averaged between multiple toys per household. If values were missing, then value of the remaining toy(s) was used for the calculation. when considering either method. The mean faecal coliform contamination of household-owned toys was higher among less clean households (3.4 log 10 CFU/200 ml, SD = 1.2) compared to cleaner households (2.0 log 10 CFU/200 ml, SD = 1.7). Similarly to the toys provided by the study, faecal streptococci contamination of the toys already owned by the household was not different between cleaner and less clean households (Table 2). Contamination of household toys was correlated with contamination of study-provided toy balls (r = 0.47, P-value = 0.01 for faecal coliforms, and r = 0.42, P-value = 0.02 for faecal streptococci). Discussion In rural Bangladesh, improved sanitation facilities and absence of human faeces in the household compound were associated with less faecal coliform contamination of children s toys. These findings suggest that improvements in sanitation could reduce exposure to environmental faecal contamination in the domestic and peridomestic environments (Figure 1). Measuring fomite contamination may provide a more objective assessment of how well sanitation interventions separate faeces from the household environment than self-report or rapid observation indicators of latrine use and access. Latrine inspections and sanitary surveys for presence of human faeces are subjective (O Loughlin et al. 2006; Rodgers et al. 2007; Montgomery et al. 2010). However, microbiologic methods require expertise and are time and resource intensive. Whether we gain additional explanatory information using this method to assess environmental faecal contamination compared to less resource-intensive methods merits further study. We characterised household cleanliness by the sanitation attributes of a single household. However, we expect that the sanitation of nearby households in these communal living spaces could affect toy contamination or contribute to the variability in toy contamination observed in this study as nearby households may have different sanitation facilities. We do not expect that toy contamination perfectly describes the effect of sanitation on a household s microbial burden, but it could describe the contamination of the broader living environment encountered by the child. As we found no difference in faecal streptococci contamination of toys between different sanitary conditions, we do not consider faecal streptococci to be useful for similar work in the future. We can continue to assess whether other microbes are better suited to discern faecal environmental contamination and/or with less variability compared to faecal coliforms. Neither faecal 2014 John Wiley & Sons Ltd 533

7 Log-transformed faecal coliform contamination of the first toy Second follow-up visit (n = 92) CV = 33.7 R 2 = 0.32, P-value <0.01 First follow-up visit (n = 96) CV = 36.5 R 2 = 0.37, P-value <0.01 First follow-up visit Second follow-up visit Log-transformed faecal coliform contamination of the second toy Identity line, y = x Figure 3 Log-transformed faecal coliform contamination of identical toys present in a household at the same time at 2 independent time points. Mean of log-transformed faecal coliform contamination at first follow-up visit Identity line, y = x CV = 37.6 R 2 = 0.18, P-value < Mean of log-transformed faecal coliform contamination of toys at second follow-up visit 8 Figure 4 Log-transformed faecal coliform contamination of identical toys present in a household at different times, n = 100. coliforms nor faecal streptococci are specific to human faeces. Animal faeces are abundant in rural environments and may contribute to faecal contamination of toys. We suspect the faecal contamination we detected on toys is from both human and animal faeces. Faecal contamination from humans and animals could present different health risks of childhood illness. Using these microbial indicators, we risk imprecision in our estimates of faecal contamination levels because improvements in sanitation largely target disposal of human excrement as opposed to animal excrement. Microbial source tracking could determine the extent to which faecal microbes that are specific to human excrement are contributing to fomite contamination, and whether improvements in household sanitation facilities and practices are decreasing human faecal contamination of household objects. Our observations were not entirely independent of confounding by socioeconomic factors as our multivariate analyses suggest that proxies for wealth explain a portion of the lower toy contamination observed in cleaner households compared to less clean households. However, this seems unlikely as the beta-coefficient comparing faecal coliform contamination of toys and household cleanliness changed by <20% when a measure of wealth was included John Wiley & Sons Ltd

8 Our sample size limited the number of variables that can be included in the multivariate model, which limited our ability to adequately evaluate the effect of household wealth. Confounding has been a central issue in cross-sectional studies that assess the impact of sanitation interventions on diarrhoeal disease (Blum & Feachem 1983). Wealthier or more educated households may have improved hand hygiene behaviour or household hygiene practices, which could disrupt the contamination cycle, independent of type of sanitation facility (Figure 1), and result in reduced contamination of children s toys. In future research, exploring whether and to what extent the association between household cleanliness and toy contamination is independent of household wealth or education, and whether toy contamination is independently associated with childhood diarrhoea could clarify whether this is a meaningful method to measure faecal environmental contamination. Several of our comparisons indicated that faecal coliform contamination of toys varies. Standard deviations within a group were relatively large. This is similar to other measures of faecal environmental contamination in the household such as contamination of water and of hands (Levy et al. 2009; Pickering et al. 2011; Ram et al. 2011). However, toy contamination may be more representative of a child s exposure to faecal contamination from the household environment than water or hands, as toys are likely cleaned less frequently than hands, may be exposed more directly to the household environment than water and may be more readily encountered by children than other fomites, such as kitchen and bathroom surfaces. We attempted to test household-owned toys for faecal contamination but only a minority of households owned non-porous toys that met our pre-specified criteria. While the mean faecal coliform contamination of household-owned toys was lower among cleaner households than less clean households, a limitation of this comparison is the risk of bias due to the small numbers of households with such a toy. Washing householdowned toys may reduce the number of visits and resources needed to collect this measure, but most households (70%) included in our study did not own a toy suitable for washing. We found good correlation between household toy contamination with faecal coliforms and study-provided toy contamination, suggesting that faecal contamination of newly introduced toys is comparable to that on toys that already existed in the household. In this study, children s toys were less contaminated with faecal coliforms in households with better sanitation facilities and less visibly apparent human faeces than in less clean households. To account for the high variability of this measure, fairly large samples sizes may be needed to reliably distinguish differences in contamination between groups. Whether the level of difference in environmental contamination detected by toy contamination between households with improved and unimproved sanitation facilities and practices improves child health merits further study. Acknowledgements We thank the respondents and their families, the icddr,b field staff, Dr. Amy Pickering and Dr. Amal K. Halder. This study was funded by the United States Agency for International Development. References Blum D & Feachem RG (1983) Measuring the impact of water supply and sanitation investments on diarrhoeal diseases: problems of methodology. International Journal of Epidemiology 12, Clasen TF, Bostoen K, Schmidt WP et al. (2010) Interventions to improve disposal of human excreta for preventing diarrhoea. Cochrane Database Systematic Review, CD Clescerl LS, Greenberg AE & Eaton AE (1999). Standard Methods for Examination of Water & Wastewater, American Public Health Association. DiVita MA, Halder AK, Jahid IK et al. (2008) The Utility of Common Household Objects as Markers of Home Hygiene in the Context of Access to Improved Sanitation. Conference presentation: ISEE 20th Annual Conference, Pasadena, California, October 12 16, Epidemiology 19(6), S / 1001.ede d Ekanem EE, DuPont HL, Pickering LK, Selwyn BJ & Hawkins CM (1983) Transmission dynamics of enteric bacteria in day-care centers. American Journal of Epidemiology 118, Hurst CJ & Crawford RL (2007). Manual of Environmental Microbiology. ASM Press, Washington D.C. Jiang X, Dai X, Goldblatt S et al. (1998) Pathogen transmission in child care settings studied by using a cauliflower virus DNA as a surrogate marker. Journal of Infectious Diseases 177, Laborde DJ, Weigle KA, Weber DJ & Kotch JB (1993) Effect of fecal contamination on diarrheal illness rates in day-care centers. American Journal of Epidemiology 138, Levy K, Hubbard AE, Nelson KL & Eisenberg JN (2009) Drivers of water quality variability in northern coastal Ecuador. Environmental Science and Technology 43, Liu L, Johnson HL, Cousens S et al. (2012) Global, regional, and national causes of child mortality: an updated systematic analysis for 2010 with time trends since Lancet 379, John Wiley & Sons Ltd 535

9 Montgomery MA, Desai MM & Elimelech M (2010) Assessment of latrine use and quality and association with risk of trachoma in rural Tanzania. Transactions of the Royal Society of Tropical Medicine and Hygiene 104, O Loughlin R, Fentie G, Flannery B & Emerson PM (2006) Follow-up of a low cost latrine promotion programme in one district of Amhara, Ethiopia: characteristics of early adopters and non-adopters. Tropical Medicine and International Health 11, Pickering AJ, Julian TR, Mamuya S, Boehm AB & Davis J (2011) Bacterial hand contamination among Tanzanian mothers varies temporally and following household activities. Tropical Medicine and International Health 16, Pickering AJ, Julian TR, Marks SJ et al. (2012) Fecal contamination and diarrheal pathogens on surfaces and in soils among Tanzanian households with and without improved sanitation. Environmental Science and Technology 46, Ram PK, Jahid I, Halder AK et al. (2011) Variability in hand contamination based on serial measurements: implications for assessment of hand-cleansing behavior and disease risk. American Journal of Tropical Medicine and Hygiene 84, Rodgers AF, Ajono LA, Gyapong JO, Hagan M & Emerson PM (2007) Characteristics of latrine promotion participants and non-participants; inspection of latrines; and perceptions of household latrines in Northern Ghana. Tropical Medicine and International Health 12, Van R, Morrow AL, Reves RR & Pickering LK (1991) Environmental contamination in child day-care centers. American Journal of Epidemiology 133, Corresponding Author Jelena Vujcic, State University of New York at Buffalo, 3435 Main Street, 270 Farber Hall, Buffalo, NY 14214, USA. jelenavu@buffalo.edu John Wiley & Sons Ltd