Ethics statement

This study was approved by the Office for Protection of Human Subjects at the University of California, Berkeley (IRB# 2019-02-11803), the Bioethics Committee at the Universidad San Francisco de Quito (#2017-178M), and the Ecuadorian Health Ministry (#MSPCURI000243-3). Formal written consent was obtained from each primary caregiver of children enrolled in the study prior to participation.

Study site

This study was carried out in semirural communities east of Quito, Ecuador by researchers at the Instituto de Microbiología at the Universidad San Francisco de Quito (USFQ) and the University of California, Berkeley School of Public Health. The study area was approximately 320 km² and included commercial food animal operations and small-scale or “backyard” production of food animals for both subsistence farming and trade. Irrigation canals fed by nearby rivers flow throughout communities in the study area and are used for subsistence farming and small-scale agricultural production.

Study design

This repeated measures observational study aimed to enroll 360 households through stratified random sampling across 7 semirural parishes. For power calculations and sample size determination, please see SI (Sample Size and Power Calculations in S1 Files). Random sampling was stratified at the neighborhood level (within parishes) based on the following strata, as determined by community-based fieldworkers: (1) backyard food animal production present in neighborhood; (2) backyard food animal production present and commercial food animal production within 1 km of neighborhood; and (3) no backyard food animal production and no commercial production within 1 km. Households were enrolled if they met the following inclusion criteria at the time of enrollment: (1) a primary child caretaker who was over 18 years of age was present; and (2) a child between the ages of 6 months and 5 years old was present in the household. Informed written consent was provided by the child’s caretaker prior to participation in the study. If there was more than 1 child at a given household, the youngest child was selected for participation. Due to high rates of migration out of the study site and loss to follow-up after the onset of the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) pandemic, we enrolled new households each cycle. Households were visited up to 5 times between August 2018 and August 2021 by trained field staff who conducted household surveys and collected household GPS coordinates and biological samples at each visit. This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (Supporting information (SI), Checklist in S1 Files). Analysis plans for this manuscript were developed beginning in May 2020 and were last revised in March 2021 after sequencing results became available; we amended our analysis plan to include descriptive results of antibiotic resistance genes (ARGs) identified from sequenced E. coli isolates. This research was supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health (NIH) under Award Number R01AI135118.

Exposure assessment

The primary exposures of interest were exposures to commercial food animal production in the community and household-level exposures to small-scale food animal production. Exposures and other household characteristics and practices were assessed at each household visit to capture time-varying exposures. Exposure to commercial food animal production was assessed in 3 ways: (1) distance to the nearest commercial food animal production facility; (2) density of commercial food animal production facilities; and (3) proximity to drainage paths of commercial food animal production facilities. Commercial poultry production facilities—vertically integrated operations marked by long barns with a metal roof that typically held approximately 20,000 birds or more—were located using satellite imagery in Google Maps and confirmed as active operations through site visits and ground-truthing. Other types of food animal production operations were identified through local knowledge and ground-truthing.

For the first measure of commercial food animal production exposure, Euclidean distance between each household and the nearest active commercial food animal operation was measured at each time point. For the second measure of exposure, the density of commercial food animal operations was assessed by summing the number of operations within a five-kilometer buffer of each household at each time point. We used the sf package in R to create these first 2 exposure variables [37,38]. While proximity may serve as a proxy for the likelihood of environmental contamination from nearby commercial operations, density may serve as a proxy for the extent of environmental contamination from nearby operations. Density of commercial poultry operations has been associated with cephalosporin-resistant E. coli in nearby stream water and sediment in the United States [39]. Finally, for the third exposure measure, drainage paths from each commercial food animal operation were identified using ArcGIS Online Trace Downstream tool, which uses a digital elevation model to identify downstream flow paths from elevation surfaces, drainage directions, river networks, and watershed boundaries obtained from the HydroSHEDS 90 m database [40]. We then created buffers around drainage paths, which ranged from 2.9 to 3.1 km in length, and spatially joined study households to buffer layers in QGIS to identify which households were within 100 or 500 meters of commercial operation drainage paths [41] (Fig 1). Proximity to these drainage paths may capture exposures to antibiotic-resistant bacteria transported from commercial food animal operations through waterways, even when households are located further from the food animal operation, itself.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Map of active commercial food animal production operations (n = 130) and their drainage flow paths in the study area, east of Quito, Ecuador.

The inset is an aerial photograph of commercial poultry production facilities in the study area (image credit: Jay P. Graham). Map created in QGIS; contains information from OpenStreetMap and OpenStreetMap Foundation, which is made available under the Open Database License.

https://doi.org/10.1371/journal.pmed.1004299.g001

Household-level exposure to small-scale food animal production was defined as household ownership of any food animals (i.e., chickens, pigs, cattle, sheep, goats, ducks, guinea pigs, rabbits, or quail). Household surveys captured information on caregiver-reported household ownership of food animals and other domestic animals, as well as household characteristics, caregiver and child demographics, health status, and other potential risk factors. Surveys were created using Open Data Kit (ODK) Build and trained enumerators used ODK Collect on Android devices for mobile data collection [42]. Encrypted survey forms were sent to ODK Aggregate on a secure server at USFQ upon completion and were subsequently downloaded for de-identification and analysis.

Outcome assessment

A single stool sample from the youngest eligible child per household was collected at each visit to determine enteric carriage of 3GCR-EC and ESBL production based on phenotypic testing, described below. Caregivers were given supplies to collect a child stool sample, which were double-bagged and stored in a fridge or on ice (4°C), and the enumerator returned to collect the sample the next day. In the case that a child had not defecated before the enumerator returned, the enumerator came back to the household the following day to collect the sample. If domestic or food animals were present at the household at the time of data collection, a single stool sample per animal species was collected from the environment where the animals defecate. Fecal samples were placed in sterile containers and stored on ice (4°C) during transportation to the microbiology lab at USFQ. Samples were processed at USFQ within 5 h of collection to identify 3GCR-EC and conduct antibiotic susceptibility testing.

Microbiological methods

To screen for 3GCR-EC and improve sensitivity for detecting ESBL production among both dominant and nondominant strains of E. coli, fecal samples were plated on MacConkey agar (Difco, Sparks, Maryland) with 2 mg/L of ceftriaxone and incubated for 18 h at 37°C [43]. Up to 5 ceftriaxone-resistant isolates phenotypically matching E. coli were selected from each fecal sample and preserved at −80°C in Trypticase Soy Broth medium (Difco, Sparks, Maryland) with 20% glycerol. 3GCR-EC isolates for each fecal sample were thawed and regrown on MacConkey agar at 37°C for 18 to 24 h for evaluation of antibiotic susceptibility by the disk diffusion method (Kirby Bauer test) on Mueller–Hinton agar (Difco, Sparks, Maryland). To confirm presumptive E. coli isolates, colonies were inoculated onto Chromocult coliform agar (Merck, Darmstadt, Germany).

Antibiotic susceptibility testing

Antibiotic susceptibility testing of all 3GCR-EC isolates was conducted for 10 antibiotics: ampicillin (AM; 10 μg), ceftazidime (CAZ; 30 μg), ciprofloxacin (CIP; 5 μg), cefotaxime (CTX; 30 μg), cefazolin (CZ; 30 μg), cefepime (FEP; 30 μg), gentamicin (GM; 10 μg), imipenem (IPM; 10 μg), trimethoprim/sulfamethoxazole (SXT; 1.25 per 23.75 μg), and tetracycline (TE; 30 μg). Isolates were identified as either susceptible or resistant to each antibiotic according to the resistance or susceptibility interpretation criteria from Clinical and Laboratory Standards Institute (CLSI) guidelines [44]. E. coli ATCC 25922 was used as the quality control strain. Multidrug resistance (resistant to 3 or more classes) was determined based on the number of macro-classes to which each isolate was resistant. Macro-classes were defined as cephalosporin/beta-lactamase inhibitors, penicillins, aminoglycosides, carbapenems, fluoroquinolones, tetracyclines, and folate pathway inhibitors.

For phenotypic confirmation of ESBL production, the combination disk diffusion test was used with CAZ and CAZ/CLA (ceftazidime with clavulanic acid) as outlined in the CLSI guidelines [44]. In the first 4 cycles of data collection, up to 5 E. coli isolates per sample were selected and preserved for analysis. 3GCR-EC isolates from the same fecal sample with identical phenotypic resistance profiles were considered duplicates and were de-duplicated prior to analyses. Due to limited laboratory resources after the SARS-CoV-2 pandemic began in 2020 and the high rate of clonal relationships between E. coli isolated from the same sample (based on preliminary sequencing), only 1 E. coli isolate per sample was selected and preserved for analysis during the fifth cycle of data collection.

DNA sequencing and analysis

Genomic DNA was extracted from the isolates using Wizard Genomic DNA Purification (Promega) kits and QIAGEN DNEasy Blood & Tissue Kits according to the manufacturer’s instructions. We conducted quality control of extracted DNA prior to library creation at the University of Minnesota Genomics Center, including PicoGreen quantification, quantitative capillary electrophoretic sizing (Agilent), and functional quantification (KAPA Biosystems qPCR). Whole-genome sequencing was carried out at the University of Minnesota. In brief, we sequenced whole-genome E. coli isolates using either Illumina MiSeq or NovaSeq with Nextera XT libraries. Following sequencing, raw reads were quality-trimmed and adapter-trimmed using trimmomatic [45]. Assemblies of reads was performed using SPAdes [46], then ARGs were identified using ABRicate (version 0.8.13) and a curated version of the ResFinder database [47]. We also performed in silico multilocus sequence typing (MLST) based on 7 housekeeping genes (adk, fumC, gyrB, icd, mdh, purA, and recA), an additional 8 housekeeping genes (dinB, icdA, pabB, polB, putP, trpA, trpB, and uidA), and core genome (cgMLST) using MLST 2.0 [48] and cgMLSTFinder 1.1 [49]. Detailed methods are previously described [50]. Raw reads from isolates sequenced in this study are available at the NCBI Short Read Archive (SRA) under BioProject accession no. PRJNA861272.

Statistical analyses

Multivariable log-binomial regression models were used to estimate unadjusted and adjusted relative risks (RRs) for the associations between commercial or household food animal exposures and 3GCR-EC and ESBL-EC carriage (i.e., main effects). To control for confounding and strong predictors of the outcomes variables, all adjusted models included prespecified covariates (child age, child sex, child antibiotic use in the past 3 months, household asset score as a proxy for socioeconomic status, and caregiver education level) identified using a directed acyclic graph and existing literature (Fig A in S1 Files) [51]. Next, we included an interaction term in adjusted models, comparing combined effects of household and commercial food animal exposures to a single referent group (unexposed to both commercial and household food animals). We also assessed effect measure modification by estimating the effect of each measure of exposure to commercial food animals among those with versus without food animals (i.e., stratum-specific effects). Estimates and P-values from interaction models were used to determine multiplicative interaction based on an alpha level of <0.10. Finally, we ran additional log-binomial regression models to estimated associations between secondary risk factors of interest related to household animal ownership, animal contact and feces management, and presence of 3GCR-/ESBL-EC in animal stool collected at households. Exposures were treated as binary and categorical using cut-points selected based on the data distribution and interpretability. The outcomes, 3GCR-EC and ESBL-EC carriage, were treated as binary and modelled at the isolate level, adjusting for clustering at the individual/household level and estimating robust standard errors with generalized estimating equations. In a sensitivity analysis, we reproduced this main analysis using only the first E. coli isolate per fecal sample to assess the potential for bias due to the change in number of isolates and probability of detecting the outcomes at each time point. Individual observations with missing exposure, outcome, or covariate data were removed for this complete case analysis. Statistical analyses and visualizations were completed in R version 3.6.1 [38] using the dplyr [52], tableone [53], ggplot2 [54], geepack [55], multcomp [56], and car [57] packages. Exposure and outcome data used in epidemiological analyses are published and publicly available on Dryad (https://doi.org/10.5061/dryad.41ns1rnm7).

Household and child characteristics

In a majority of the 594 included households, most primary caregivers had at least a high school or college level education, ranging from 67.1% to 79.5% across the data collection cycles (Table 1). The average age of children participating in the study was 1.8 years during the first cycle of data collection (Table 2). Access to drinking water and sanitation was high; 98.8% of households had a flush toilet (to sewer or septic tank), 92.2% had piped drinking water inside their home, and 98.6% had 24-h access to drinking water (Table M in S1 Files). Over 93% of households had household handwashing stations with the presence of both soap and water, confirmed by observation (Table 1). However, caregiver-reported child handwashing frequency suggested that most children rarely (44.4% to 69.6%) washed their hands after contact with animals (Table 2). Over the course of the entire study period, 36.7% of caregivers reported that their child played near animal feces in the last 3 weeks, 36.0% reported that their child had contact with livestock at least once per week in the last 3 months, and 67.9% reported child contact with pets at least once per week in the last 3 months. Seven-day caregiver-reported child diarrhea prevalence was 20.3% at the beginning of the study period in 2018 and declined to 4.2% at the final data collection cycle in 2021 during the SARS-CoV-2 pandemic (Table 2). Similarly, treatment for infection in the last 3 months declined from 31.1% at the first cycle to 8.1% at the fifth cycle; child antibiotic use in the last 3 months also declined from 26.2% at cycle 1 to 5.9% at cycle 5 (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of study households in semirural Quito, Ecuador at each cycle of data collection between July 2018 and September 2021.

https://doi.org/10.1371/journal.pmed.1004299.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Characteristics and behaviors of children in study households at each cycle of data collection between July 2018 and September 2021.

https://doi.org/10.1371/journal.pmed.1004299.t002

Food animal production and domestic animals

We identified 130 active commercial food animal operations in our study site including 122 poultry facilities, 5 hog facilities, 2 horse facilities, and 1 milk production facility. Across the 7 parishes in the study site, 4 parishes (Tababela, Tumbaco, Checa (Chilpa), and Pifo) had low-intensity commercial production with fewer than 10 commercial food animal operations, each. Three parishes (El Quinche, Puembo, and Yaruqui) had high-intensity production, with more than 30 commercial food animal operations, each (Table F in S1 Files). More than half of study households were located <1.5 km from a commercial food animal operation (55.7%) or were within 500 m of a commercial operation’s drainage flow path (56.1%) throughout the study period. During the first 4 cycles of data collection, at least 67% of households were within 5 km of 6 or more commercial food animal operations; this dropped to 43% in the fifth cycle (Table 1).

We collected 1,871 animal fecal samples from 376 matched households and identified 1,060 3GCR-EC isolates to assess as a potential risk factor. Throughout the study period, 66.4% of caregivers reported owning any type of animal, with 29.2% owning food animals and an average of 5.7 to 9.2 food animals at each data collection cycle (Table 1). Among households that owned food animals, reported antibiotic use in food animals was low, declining from 26.4% at the beginning of the study period to 0% at the end of the study (Table 1). Thirty-two percent of households owned backyard chickens, 17% owned guinea pigs, 11% owned pigs, 7% owned cattle, and 3% owned goats or sheep. The average flock size for backyard chickens was 13.7 (standard deviation (SD): 29.9). There was an average of 16 guinea pigs (SD: 19.6), 4.3 pigs (SD: 6.6), 2.8 cows (SD: 2.6), 2.9 goats (SD: 3.2), and 4.0 sheep (SD: 6.3). Eighty percent of households owned dogs, with an average of 2.3 dogs per household (SD: 1.6).

Characterization of 3GCR-EC isolates

Sixty-one percent of children (n = 365/600) were carriers of 3GCR-EC and 25% (n = 149/600) were carriers of ESBL-EC at least once throughout the study period. 3GCR-EC were detected in 38% of child samples (n = 652/1,699) and 51% of animal samples (n = 959/1,871) from matched households. Of 910 3GCR-EC isolated from child samples, 36% were resistant to fourth-generation cephalosporin, cefepime, in phenotypic susceptibility testing (Table B in S1 Files). Eighty-six percent of child 3GCR-EC isolates were multidrug-resistant (3 or more classes), 37% were extensively drug-resistant (5 or more classes), and 22% were phenotypic ESBL producers (Table B in S1 Files). Notably, 5 children (<1%) from different households had 1 E. coli isolate that was phenotypically resistant to the carbapenem drug, imipenem (Table B in S1 Files). All of these imipenem-resistant isolates were multidrug-resistant to 4 or more drug classes, with each isolate also expressing resistance to ampicillin and cephalosporin antibiotics. Phenotypic resistance to antibiotics among 3GCR-EC isolates from animal fecal samples are described in Table C in S1 Files.

We analyzed whole-genome sequencing data for a subset of 571 3GCR-EC isolated from fecal samples of the 365 children that were 3GCR-EC carriers. We first selected 1 isolate per child for sequencing, then aimed to randomly select an additional 200 isolates to include more ESBL-EC in sequencing analyses. The most prevalent sequence type (ST) among these isolates was ST 10 (7%, 40/571), while other clinically important STs like ST 131 and ST 117 accounted for about 3% of sequenced isolates (Table D in S1 Files). 3GCR-EC isolates had a mean of 9.7 total ARGs (SD: 4.3). Overall, the proportion of 3GCR-EC isolates with bla genes was 65%, 52%, 8%, 5%, and 5% for bla_CTX-M-encoding, bla_TEM, bla_CMY, bla_OXA, and bla_SHV genes, respectively. Ten (2%) 3GCR-EC isolates had an mcr-1 gene, indicating resistance to the last-line antibiotic colistin. Eighty-two percent (n = 119/145) of phenotypic ESBL-EC and 77% (n = 330/426) of non-ESBL-producing 3GCR-EC carried at least 1 bla gene. Among phenotypic ESBL-EC, the most prevalent bla genes were bla_CTX-M-55, bla_TEM-141, bla_TEM-1B, bla_CTX-M-15, and bla_OXA-1 (Table E in S1 Files). Among the most prevalent CTX-M-encoding genes, the most prevalent genes in parishes with high-intensity commercial food animal production were bla_CTX-M-55, bla_CTX-M-65, bla_CTX-M-15, and bla_CTX-M-14 (Fig 2 and Table H in S1 Files). Parishes with lower intensity commercial food animal production had a higher prevalence of bla_CTX-M-3, though bla_CTX-M-55, bla_CTX-M-65, and bla_CTX-M-15 were also detected in these parishes (Fig 2 and Table H in S1 Files).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Map of distribution of CTX-M-type genes from a subset of 571 3GCR-EC isolated from children in 7 parishes east of Quito, Ecuador.

Pie chart sections represent the proportion of isolates with a specific CTX-M-type gene detected, among those represented in the plot legend. Shaded areas represent parishes, where darker shading indicates a higher number of commercial food animal operations in a given parish. Map created in RStudio; Ecuador parish boundary shapefiles are made publicly available by the INEC and the United Nations OCHA. Data and license information are available at https://data.humdata.org/dataset/cod-ab-ecu?. 3GCR-EC, third-generation cephalosporin-resistant E. coli; INEC, Instituto Nacional de Estadística y Censos; OCHA, Office for the Coordination of Humanitarian Affairs.

https://doi.org/10.1371/journal.pmed.1004299.g002

Risk factors for 3GCR-EC

The prevalence of 3GCR-EC was higher in all food animal exposure groups compared to the unexposed, where the denominators include all fecal samples and isolates within each exposure group (Fig 3). In models assessing main effects of household and commercial food animal exposures, the density of commercial operations (>5 in 5 km radius) was the only exposure associated with an increased risk of 3GCR-EC (RR: 1.27; 95% CI: 1.11, 1.46) (Table 3). In models adjusting for interaction between household and commercial food animal exposures, children with >5 commercial food animal operations in a 5-km radius of their household had 1.36 times the risk (95% confidence interval (CI): 1.16, 1.59) of 3GCR-EC carriage than those with ≤5 operations within 5 km when controlling for household food animal ownership (Table 4). However, among those with household food animals, there was no effect of commercial operation density on 3GCR-EC carriage (RR: 1.05; 95% CI: 0.83, 1.33) (Table 4). There was not an excess risk of being exposed to both household food animals and >5 commercial operations (relative excess risk due to interaction (RERI): −0.29, 95% CI: −0.64, 0.06) (Table 4). However, the combination of owning household food animals and living <1.5 km from the nearest commercial operation was associated—with borderline significance—with an increased risk in 3GCR-EC carriage (RR: 1.15; 95% CI: 0.98, 1.36) compared to those without food animals who lived further from commercial food animal operations (Table 4). Proximity to a drainage flow path from a commercial food animal operation was not associated with 3GCR-EC carriage, regardless of household food animal ownership (Table 4). These results were largely robust to sensitivity analyses (Table I in S1 Files). Other risk factors for 3GCR-EC among children included child pet contact in the last 3 months (RR: 1.23; 95% CI: 1.09, 1.39) and pig ownership (RR: 1.23; 95% CI: 1.02, 1.48) (Fig 4) (Table J in S1 Files).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Bar plot of distribution of 3GCR-EC and ESBL-EC carriage stratified by exposures to HH food animals and CFOs.

Number of observations includes fecal samples that were negative for 3GCR-EC (light gray bars for 3GCR-EC), isolates from fecal samples that were 3GCR-EC (orange bars), or ESBL-EC (purple bars) based on phenotypic susceptibility testing, and isolates plus fecal samples negative for both 3GCR-EC and ESBL-EC (light gray bars for ESBL-EC). Percent of observations positive for each outcome are listed above each bar. 3GCR-EC, third-generation cephalosporin-resistant E. coli; CFO, commercial food animal operation; ESBL-EC, extended-spectrum beta-lactamase E. coli; HH, household.

https://doi.org/10.1371/journal.pmed.1004299.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Additional risk factors for 3GCR-EC and ESBL-EC carriage among children in semirural parishes of Quito, Ecuador.

Points are adjusted RR and error bars are 95% CIs; asterisks (*) indicate significance given alpha = 0.05 (corresponding data including sample sizes and P-values in Tables K and J in S1 Files). RR are adjusted for repeated measures and controlled for the following covariates: caregiver education, asset score, child age and sex, and child antibiotic use in the last 3 months. 3GCR-EC, third-generation cephalosporin-resistant E. coli; CI, confidence interval; ESBL-EC, extended-spectrum beta-lactamase E. coli; RR, relative risk.

https://doi.org/10.1371/journal.pmed.1004299.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Unadjusted and adjusted RRs of 3GCR-EC and ESBL-EC for main effects of household food animal ownership and exposures to commercial food animal production.

https://doi.org/10.1371/journal.pmed.1004299.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Adjusted RRs of 3GCR-EC carriage among children given combined exposures to commercial food animal production and effect measure modification by household food animal ownership.

https://doi.org/10.1371/journal.pmed.1004299.t004

Risk factors for ESBL-EC

We did not detect significant associations or interactions between the density of or proximity to commercial food animal operations, household food animal ownership, and ESBL-EC carriage. In models adjusting for interactions between exposures, household food animal ownership and increased proximity to a drainage path were not significantly associated with an increased risk of ESBL-EC carriage among children. However, the effect sizes were larger for combined and increasing intensity of exposures. When controlling for household food animal ownership, the RR for ESBL-EC was 1.10 (95% CI: 0.76, 1.60) for those living 101 to 500 m from a drainage path and was 1.32 (95% CI: 0.71, 2.46) for those 101 to 500 m from a drainage path compared to those >500 m (Table 5). The RR further increased to 1.80 (95% CI: 0.94, 3.45) for those living within 100 m of a drainage path and with household food animals compared to those >500 m from a drainage path and without household food animals; this interaction was near-significant with an interaction P-value of 0.108 (RERI: 0.85; 95% CI: −0.10, 1.81) (Table 5). These findings were also robust to sensitivity analyses (Table I in S1 Files). Additional risk factors of ESBL-EC carriage among children in our study site included placing animal feces on household land/crops (RR: 1.63; 95% CI: 1.09, 2.46), household dog ownership (RR: 1.35; 95% CI: 1.00, 1.83), and child pet contact in the last 3 months (RR: 1.54; 95% CI: 1.10, 2.16) (Fig 4 and Table K in S1 Files).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Adjusted RRs of ESBL-EC carriage among children given combined exposures to commercial food animal production and household food animals, including both interaction effects (effects of individual and combined exposures vs. no exposures) and stratum-specific effects (effects of commercial food animal exposures within strata of household food animal ownership vs. no exposures).

https://doi.org/10.1371/journal.pmed.1004299.t005