This study was nested within the MAL-ED cohort in three peri-urban communities of Iquitos, Peru: Santa Clara de Nanay, Santo Tomas, and La Union (3°47’S, 73°20’W). In order to be eligible for the floor sampling study, a household had to have a child less than 48 months of age who was still enrolled in the MAL-ED study at the time of sampling. Households were enrolled in the MAL-ED study if they were located within the catchment area and had a healthy infant born during the two years of enrollment.²³

Prior work has shown that these communities lag behind the rest of the Peru in terms of development indicators.²⁴ Only 20.2 percent of the population had access to an improved sanitation facility while 58.4 percent of the overall Peru population had access. Similarly, 46.7 percent of households in the study communities had access to clean water versus 77.1 percent in all of Peru. Child growth also lagged behind with 46.3 percent of children under 5 years old being stunted versus 19.5 percent in Peru. Children under 5 years old who were reported to have diarrhea in the past week was 35.4 percent versus 13.9 percent in Peru.²⁴ The households were low-income with the mean monthly per-capita income at $28 US dollars.²⁵ The temperature ranges between 21.9 and 32.4 degrees Celsius with an average of 25.8 degrees Celsius.²⁴ Rainfall is frequent and occurs throughout the year on about half of all days with the heaviest rainfall in January.²⁴

The communities are located proximal to the Nanay River, which is a major branch of the Amazon river system. The river levels rise until March and, at the time of initiation of the study, the Nanay River was receding and no flooding was apparent within any of the households visited. There is no centralized sewerage infrastructure in the community and hence open ditches are used to drain storm and gray water away from the home. The frequent flooding in this riverine community also leads to fecal matter from latrines being released into the environment.

During each household visit, a household questionnaire was administered in Spanish prior to floor sampling. The questionnaire was based on the Demographic and Health Surveys³ and was a shortened version of the standardized questionnaire. In addition, study staff conducted a standardized visual inspection of floor type by room within households and noted the materials used as either dirt, wood, cement or tile.

The questionnaire assessed the primary exposure variable of the type of sanitation facility used by household members and whether or not this facility was shared. The options for type of sanitation facility were: i) no facility/open field; ii) pit latrine; iii) pour flush toilet to a septic; iv) flush to somewhere else; or v) ventilated improved pit latrine. Responses to water and hygiene questions provided covariate data on the household’s primary water source, mode of water treatment, time it takes to fetch water, hygiene behavior and crowding. Information was also collected on socio-economic factors such as housing construction materials, length of tenancy, electricity access, maternal education, and monthly income. Given the propensity for households to keep free-ranging or corralled chickens in this community, participants also were interviewed regarding the presence of chickens in the home to evaluate the influence of chicken feces on the bacterial contamination of household floors.

From August to September 2015 household floors were sampled for E. coli bacteria using a modified dry electrostatic cloth method based on one designed for household settings.²⁶ Samples were collected from highly trafficked areas, namely the household entrance and the kitchen, which has been shown to have higher levels of fecal bacteria than the bathroom areas.^{18, 27} These areas were also selected for high likelihood of fecal pathogen exposure for children under five years of age who spend large amounts of time in play near the entrance and near the primary caregiver engaged in cooking activities. The first area sampled was the entryway floor, typically located at the front of the house and closest to the open drains that conveyed untreated wastewater and had a tendency to overflow during periods of rainfall. The second area sampled was the kitchen floor area, typically located at the back of the house (See Figure S1, Supporting Information). If there was a latrine or toilet, it was most commonly in the back of the house, closest to the kitchen area. The kitchen area was also observed to be the area of the house where most water use and storage activities took place, creating a potentially favorable environment for bacteria. Homes were visually inspected for presence of animals inside the home at the time of sampling. Duplicate samples were taken side by side at the entryway location to investigate the heterogeneity of fecal contamination across floors. To assess the potential influence of different floor material types (e.g. dirt, wood, cement) on fecal contamination, we recorded information about the floor material types at the household entrance and in the kitchen area at the time of sampling.

Prior to field collection, sterile packets of dry electrostatic cloths (Swiffer™; Proctor & Gamble, Cincinnati, OH) were separated, quartered, and individually wrapped in autoclave paper (Fisher Scientific™, Pittsburgh, PA). Wrapped packets were then sterilized by autoclaving. For each collection an adapted protocol from Davis et al. (2012) was used where a prepared cloth was passed over a 30 cm by 30 cm floor surface with medium pressure to maximize the amount of pick-up from the surface.²⁶ The cloth was then placed into a sterile 700 mL Whirlpak bag (Nasco, Fork Atkinson, WI) and 5 mL of sterilized Milli-Q ultra-pure water to guard against microbial desiccation during transport. Samples were stored in a cooler on ice at 4°C during field collection and transported to the fully equipped microbiologic, immunologic, and PCR based diagnostic laboratory of the Asociación Benéfica PRISMA, approximately 15km from the study site in the city center of Iquitos. Samples were processed within six hours of collection.

For elution, 100 ml of sterile 0.1% Peptone buffer was added into the Whirlpak bag containing the cloth and vigorously shaken for one minute. The cloth was aseptically removed and E. coli in the buffer were enumerated following USEPA Method 1604²⁸ using m-coliblue24 commercial media (HACH, Loveland, U.S.A.). Positive E. coli were identified as blue colonies. Pre, intermittent and post blanks were run to confirm the absence of cross contamination of samples. To obtain a countable number of colonies (i.e. 20–200), undilute, 10-fold, 100-fold and 1000-fold dilutions of eluate for samples collected on dirt floors and undilute, 10-fold and 100-fold for samples collected on wood and cement floors were processed, enabling a detection range of 0 to 200,000 colony forming units (CFU) per 900 cm²of floor area to be enumerated.

The primary independent variable of sanitation facility was categorized into “improved” and “unimproved” sanitation facilities as defined by the Joint Monitoring Program (JMP) for Water Supply and Sanitation.²⁹ The JMP classifies improved facilities as those that ensure hygienic separation of human excreta from human contact and include facilities that flush or pour flush to a piped sewer system, septic tank or pit latrine. Unimproved facilities on the other hand, do not ensure hygienic separation of human excreta from human contact and include pit latrines without a slab. For the purposes of this study, those households that did not have a toilet facility were also categorized as “unimproved”.

Water source, water treatment and floor type covariates were analyzed as categorical variables. A hygiene index variable score was calculated as a cumulative score from the following four questions: i) Do you wash your hands after helping your child defecate? ii) Do you wash your hands before preparing food? iii) Do you wash your hands after going to the bathroom? and iv) Do you use toilet paper?. The hygiene index score had three levels with good indicating the interviewee answered all questions as always practicing the hygienic behaviors; average indicated that for one of the four questions the interviewee only sometimes practiced the hygienic behavior; and poor indicated that for two or more questions the interviewee only sometimes practiced the hygienic behavior. E. coli concentrations were log₁₀-transformed and reported as log₁₀ E. coli CFU/900 cm².

Two-sample t-tests with equal variances and Pearson Chi-squared analysis were used to compare household characteristics across improved and unimproved sanitation facility types. Unadjusted linear regression analyses were conducted to evaluate associations between water, sanitation, hygiene (WASH) and household characteristics with log₁₀ E. coli CFU/900 cm² in the entrance and kitchen areas. Using generalized linear models we conducted a stratified analysis by sanitation type (improved versus unimproved). For this analysis of the relation between floor types and the levels of log₁₀ E. coli CFU in strata of household sanitation type (unimproved and improved), observations in the entrance and kitchen of each house were combined. We adjusted for potential confounding covariates in linear regression models using a backward elimination approach. A final parsimonious covariate adjustment set was selected based on considerations of sample size and the minimization of the Akaike Information Criterion (AIC).³⁰ Interaction terms between sanitation type and floor type were included to determine if the association between sanitation and log₁₀ E. coli CFU/900 cm² was modified by floor type. Beta coefficients and 95% confidence intervals were estimated and represent the log₁₀-unit change in E. coli CFU/900 cm² per unit of each of the independent variables (household sanitation type, floor type, etc). Pearson correlation coefficients and 95% confidence intervals were calculated to estimate the variability between duplicate floor samples within the same household.

The data processing and visualization were performed in R 3.0.3³¹ using the ggplot2 package³² and subsequent statistical analyses were conducted using Stata version 12.1 (College Station, TX).

The study protocol and questionnaires were approved by the institutional review boards from Johns Hopkins Bloomberg School of Public Health (Baltimore, MD) and Asociación Benéfica Proyectos de Informática, Salud, Medicina, y Agricultura (A.B. PRISMA), Iquitos, Peru. All participants gave written consent prior to household sampling.

For households with unimproved versus improved sanitation facilities, there were significant differences in household characteristics (Table 1). Households with unimproved versus improved sanitation had a higher percentage of dirt floors in both the entrance (77.4 vs 37.5, p<0.01) and kitchen (87.1 vs 59.4, p<0.05 level) and a more frequent reporting of chickens in the home (45.2 vs 12.5, p<0.01). Households with improved versus unimproved sanitation had a higher percentage of cement floors in both the entrance (56.3 vs 19.4, p<0.01) and kitchen (31.3 vs 9.7, p<0.05). There were no significant differences across sanitation type for other household WASH characteristics such as sharing sanitation facilities, type water connection, time to fetch water, household chlorine use to treat drinking water, crowding, income, education, electricity connection, wall and roof type and tenancy in the house.

Linear regression models comparing individual household WASH characteristics and the levels of log₁₀ E. coli CFU demonstrated significant associations in both the entrance and kitchen areas (Table 2). Households with improved sanitation had lower levels of log₁₀ E. coli CFU/900 cm² on floors when compared to homes with unimproved sanitation in both the entrance and kitchen (Beta: −0.63 (95% CI: −1.12, −0.15); and Beta: −0.80 (95% CI: −1.27, −0.33) respectively). For shared sanitation, households that reported not sharing their sanitation facility versus those did share had lower levels of log₁₀ E. coli CFU/900 cm² (Beta: −0.70; 95% CI: −1.27, −0.13) in the kitchen area. Household entrance and kitchen areas with cement floors also had lower levels of log₁₀ E. coli CFU/900 cm² when compared to household entrance and kitchen areas with dirt floors (Beta: −0.89 (95% CI: −1.38, −0.40); and Beta: −1.31 (95% CI: −1.83, −0.79) respectively). For every additional minute that interviewees reported needing to fetch water, corresponding increases in the concentrations of log₁₀ E. coli CFU/900 cm² on entrance floors (Beta: 0.06; 95% CI: 0.02, 0.10) and kitchen floors (Beta: 0.05; 95% CI: 0.01, 0.09) were observed. Table 2 illustrates that wall type, crowding, electricity access, maternal education and housing tenancy were independently associated with increases in log₁₀ E. coli CFU/900 cm².

To further understand the relationship between floor type and sanitation type, the stratified data by sanitation type (improved versus unimproved) are shown in Figure 2. The lowest log₁₀ E. coli CFU/900 cm² were found in the homes with both improved sanitation and improved floor types (defined by their ability to be disinfected such that wood and cement floors are combined into the improved category and dirt as unimproved). The reduction in log₁₀ E. coli CFU/900 cm² among households with unimproved sanitation was −0.60 (95% CI: −1.03, −0.17) when comparing wood or cement (improved) floors to dirt floors (unimproved). An even greater reduction of −1.17 log₁₀ E. coli CFU/900 cm² (95% CI: −1.68, - 0.66) was observed among households with improved sanitation when comparing wood or cement floors to dirt floors (Table 3).

Two multivariate linear regression models were run for the entrance and kitchen floor areas with predictor variables that included both the sanitation type (improved or unimproved) as an interaction term with floor type and the variable for whether the sanitation facility was shared (Table 4). The models adjusted for time to fetch water, presence of chickens in the household, crowding, maternal education and wall type. For the entrance floor area, households with improved sanitation and cement floors had lower log₁₀ E. coli CFU/900 cm² on their floors when compared to households with unimproved sanitation and dirt floors (Beta: - 0.43; 95% CI: −1.08, 0.21). For the kitchen floor area, households with unimproved sanitation and wood floors and households with improved sanitation and cement floors both had statistically significantly lower log₁₀ E. coli CFU/900 cm² on their floors when compared to households with unimproved sanitation and dirt floors (Beta: −2.36 (95% CI: −3.86, −0.86) and (Beta: −1.18 (95% CI: −1.77, −0.60) respectively). Households that did not share their sanitation facility also had significantly reduced log₁₀ E. coli CFU/900 cm² on their kitchen floors (Beta: - 0.65; 95% CI: −1.15, −0.16) when compared to kitchen floors in households that did share their sanitation facility. The significant covariates in the adjusted model for the kitchen area included lack of chickens in the household (Beta: −0.63; 95% CI: - 1.12, −0.15; indicating lower log₁₀ E. coli CFU/900 cm² for those without versus with a presence of chickens) and maternal education (Beta: −0.08; 95% CI: (−0.15, - 0.004; indicating lower log₁₀ E. coli CFU/900 cm² in homes for every year increase in of education). The significant covariates in the adjusted model for the entrance area, were time to fetch water (Beta: 0.05; 95% CI: 0.003, 0.09; indicating higher log₁₀ E. coli CFU/900 cm² for every minute increase in time to fetch water) and maternal education (Beta: −0.10; 95% CI: −0.19, 0.00; indicating lower log₁₀ E. coli CFU/900 cm² for every year increase in of education).

For the entrance area where side-by-side samples were collected to understand the distribution of E. coli bacteria across floor surfaces, the Pearson correlation coefficient between the initial and duplicate samples was 0.83 with a p-value < 0.001 (n=63) (See Figure S2, Supporting Information). The 95% confidence interval for the Pearson correlation coefficient ranged from 0.73 to 0.89 indicating a homogenous spread of bacteria across the sampling area.