INTRODUCTION
Campylobacter species are zoonotic bacteria, and the animal reservoirs include poultry, free-living birds, and ruminants, such as cattle and goats, as well as other warm-blooded animals.1,2 Common symptoms of Campylobacter infection include fever, diarrhea, and abdominal pain.3 In addition to symptomatic infections, asymptomatic Campylobacter cases are common, specifically among children and adults in low- and middle-income countries (LMICs).4,5 Evidence from studies in low-resource settings has shown that colonization of Campylobacter species acquired at the early stage of child development potentially contributes to environmental enteric dysfunction, malnutrition, and growth faltering in children,4,6,7 which draws increasing attention to the long-term health effects of Campylobacter infections among children younger than 5 years of age in LMICs.
Campylobacter jejuni and Campylobacter coli have been reported as the two most common species that cause illness in humans.3 Traditional methods selectively isolating these two species have resulted in a large body of knowledge specific to their disease burden and clinical manifestations.8 However, several Campylobacter species other than C. jejuni/coli (non-C. jejuni/coli species), such as Campylobacter concisus, Campylobacter lari, Campylobacter upsaliensis, and Campylobacter ureolyticus, have been recognized as “emerging species” and have shown increasing clinical importance.9 Increasing proportions of non-C. jejuni/coli species, including a new species “Candidatus Campylobacter infans,” were isolated from child stool samples collected in low-resource settings, and several studies have revealed their potential linkage with child stunting.7,10–13 This evidence shifts our attention toward the genus Campylobacter as a whole rather than focusing solely on C. jejuni/coli in the context of child health.
The classic “F diagram” (fluids, fields, flies, fingers, fomites, and food) from WHO indicates that adequate sanitation and hygiene can be effective in interrupting the fecal–oral transmission routes of diarrheic pathogens, including Campylobacter, in low-resource settings.14,15 Driven by this idea, water, sanitation, and hygiene (WaSH) interventions, including improved pit latrines, handwashing stations, liquid soap, and point-of-use water chlorination, have been developed and implemented in LMICs.16,17 Substantial evidence suggests that adequate WaSH contributes to reducing the risk of childhood diarrheal disease.18,19 However, several recent large-scale randomized, controlled trials have found that WaSH interventions did not significantly reduce the occurrence of diarrhea or other enteric infections.20–22 One possible explanation is that traditional WaSH interventions generally target routes of children’s exposure to human feces but fail to pay adequate attention to children’s exposure to animal feces.15,23 Risk factors for exposing children to animal feces might include livestock ownership, proximity to livestock, cohabitation of humans and animals, etc.
The first year of a child’s life is marked by remarkable developmental milestones, and the differentiation between the first and second 6 months brings about major changes in both biological risk and environmental interaction. In the initial 6 months, breastfeeding plays a crucial role in offering infants essential nutrients and immune protection against pathogens.24 However, as the infant transitions into the second 6 months, the introduction of solid foods broadens the child’s nutritional spectrum but also, amplifies exposure to contaminants and pathogens through the food pathway.25 Meanwhile, the infant becomes more mobile, gaining the ability to explore the surroundings independently. This newfound mobility empowers infants to explore their environment actively through touching and tasting,26 potentially increasing exposure to pathogens through the fecal–oral pathway. This developmental transition underscores the dynamic effects of age on infant’s biological risk and environmental interaction.
Moreover, other environmental factors could impact the nonfood transmission route of Campylobacter to humans directly or through animal reservoirs indirectly.27,28 Studies conducted in European countries, the United States, and Canada found campylobacteriosis incidence to be correlated with climatic variables, including temperature and precipitation.29–33 In addition to weather and climate factors, Sanderson et al.34 also examined the potential impacts of hydrology and landscape features, like soil type and land use, on the rates of human Campylobacter cases in the United Kingdom. This study showed that an increased risk of Campylobacter infections was associated with periods of high surface-water flow and catchment areas with cattle/sheep grazing on stagnogley soils. Another similar study conducted in New Zealand linked the risk of infection to a high dairy cattle density.35 However, these kinds of studies that investigate the influence of environmental factors have been predominantly conducted in developed countries with a focus on C. jejuni/coli. There is a critical knowledge gap in environmental effects on Campylobacter infections in LMICs.
Although multiple factors are involved in the transmission pathways of Campylobacter, few studies have assessed the combined effects of environmental covariates with other relevant factors from different domains, including humans and animals. Because environmental data often comprise spatial information, such as land cover, precipitation, and topography, geospatial analysis is needed to address the inherent differences in data types between environmental factors and human/animal factors by integrating them within a spatial framework (e.g., spatial regression). This capability to analyze data at different levels from individual households to broader environmental landscapes enhances the depth of risk factor identification. Coupled with the One Health perspective, this kind of analysis will help unravel the complex interplay between human, animal, and environmental factors involved in Campylobacter transmissions among children in a low-resource setting.
The longitudinal study of the Campylobacter Genomics and Environmental Enteric Dysfunction (CAGED) project conducted in rural eastern Ethiopia aimed to examine the association between Campylobacter infection, related reservoirs, and child health outcomes.36 In addition to child fecal samples, the CAGED project also collected environmental (e.g., soil and drinking water) and livestock samples for detection and quantification of Campylobacter. Results from a previous study showed that the prevalence of Campylobacter in all infant stool samples was 64%, and it could increase to as high as 89% as the infants grow older. Moreover, the prevalence in livestock feces ranged from 93% to 99%. In addition to these fecal samples, the household questionnaire used in this project included components on people (e.g., demographics as well as livelihood and wealth); livestock; and the interactions between people, livestock, and the environment. Combined with the child and environmental samples, these data provide an opportunity to investigate the potential combined effects of these risk factors on Campylobacter infection in infants.
MATERIALS AND METHODS
Study design and protocol.
A detailed description of the CAGED study design and protocol can be found elsewhere.36 Briefly, a total of 106 infants from 10 kebeles (the smallest administrative unit) of Haramaya woreda, Eastern Hararghe Zone, Oromia, Ethiopia (Figure 1) were enrolled at birth and followed until approximately 13 months of age. Written informed consent was obtained from both parents of each participating child in the local language (Afan Oromo). Household information, including demographics, livelihood and wealth, livestock ownership, and child health and nutrition status, was collected through household surveys at baseline and end line along with short surveys conducted monthly. During the study period, child stool samples were collected monthly and tested for Campylobacter species. In addition, fecal samples from mothers and siblings of the enrolled children, feces from livestock (i.e., chicken, cattle, goat, and sheep), and environmental samples (soil and drinking water) were collected biannually. DNA extraction, genus-specific TaqMan real-time polymerase chain reaction (PCR; Thermo Fisher Scientific, Waltham, MA), and species-specific Sybr Green real-time PCR were performed afterward to detect, quantify, and characterize Campylobacter spp. in these samples.37
Study area and enrolled households in the Campylobacter Genomics and Environmental Enteric Dysfunction project.
Citation: The American Journal of Tropical Medicine and Hygiene 112, 3; 10.4269/ajtmh.24-0401
Outcome variable.
We calculated the cumulative burden of Campylobacter infection for each enrolled child as the outcome variable of interest derived from the Ct values of their fecal samples collected over time. A genus-specific standard curve for expected bacterial load (log genome copies per 50 ng DNA) against the Ct values was first generated using the 16S TaqMan approach.38 One milliliter normalized bacterial culture cocktail (including C. jejuni, C. coli, Campylobacter hyointestinalis, C. lari, and Campylobacter fetus) was used for DNA extraction, and 2 µL extracted DNA was used for quantitative PCR afterward. We used genus-specific primers targeting the Campylobacter 16S ribosomal RNA gene (forward primer: GATGACACTTTTCGGAGCGTAA; reverse primer: GCTTGCACCCTCCGTATTA; probe: CGTGCCAGCAGCC-MGB).38 The thermocycling conditions consisted of an initial cycle at 95°C for 10 minutes followed by 45 cycles of 95°C for 15 seconds and 55°C for 60 seconds. Nuclease free water and Salmonella genomic DNA were used as negative controls. We tested up to 10 Campylobacter DNA concentrations and repeated the experiment three times. Then, the Ct values of the tested stool samples were converted to the expected bacterial load using this standard curve. The cumulative Campylobacter burden was defined as the average of the expected bacterial loads from available stool samples for each child.37
Explanatory variables: Household surveys.
Human and animal data were derived from household surveys. Demographic data, including child’s sex, mother’s age, and mother’s education level, were selected from the baseline survey along with ownership of livestock (i.e., cattle, goat, sheep, and chicken) and assets (Table 1). To quantify all of the livestock kept by a household, a composite metric (tropical livestock unit) was calculated based on the number of each species of livestock recorded in the baseline household survey.39
Demographics and socioeconomic status of the households enrolled in this study
| Variable | N = 106* |
|---|---|
| Sex | |
| Female | 51 (48%) |
| Male | 55 (52%) |
| Mother’s age (years) | 27.0 (22.0–32.0) |
| Unknown | 1 |
| Mother’s education | |
| No primary education | 76 (72%) |
| Some primary education | 29 (28%) |
| Unknown | 1 |
| Livestock ownership | |
| Cattle | 52 (49%) |
| Goat | 60 (57%) |
| Sheep | 48 (45%) |
| Chicken | 53 (50%) |
| Tropical livestock unit | 0.62 (0.20–1.40) |
Values are presented as n (percentage) or median (interquartile range).
In the monthly short surveys, we selected variables that reflect interactions between the target child and animals or the environment (e.g., physical contact with animals, crawling in areas with animal droppings, and mouthing of soil or animal feces), diet and nutrition (e.g., consumption of animal source food), feeding practices (e.g., prelacteal feeding, introduction of complementary foods, and consumption of any solid food in the past 24 hours), and use of antibiotics (i.e., being treated with antibiotics in the past month) (Table 2). In addition, two composite variables, minimum dietary diversity (MDD) and household food insecurity access score, were generated according to their respective literatures.40,41
Variables used for regression modeling and univariate analysis results
| Variable | First Period | Second Period | ||
|---|---|---|---|---|
| OR (CI)* | P-Value | OR (CI) | P-Value | |
| Child’s sex | 0.8 (0.4–1.7) | 0.56 | 2.3 (1.1–5.1) | 0.03 |
| Mother’s age (years) | 1.3 (0.6–2.8) | 0.55 | 0.8 (0.4–1.7) | 0.55 |
| Mother’s education | 1.2 (0.5–2.9) | 0.49 | 1.1 (0.4–2.6) | 0.49 |
| Cattle ownership | 0.6 (0.3–1.4) | 0.24† | 2.1 (1.0–4.7) | 0.05 |
| Goat ownership | 1.4 (0.6–3.0) | 0.43 | 0.9 (0.4–1.9) | 0.70 |
| Sheep ownership | 0.6 (0.3–1.4) | 0.24 | 0.5 (0.2–1.2) | 0.12 |
| Chicken ownership | 0.8 (0.4–1.7) | 0.56 | 1.5 (0.7–3.2) | 0.33 |
| Tropical livestock unit | 1.1 (0.5–2.3) | 0.85 | 1.3 (0.6–2.7) | 0.56 |
| Asset | 1.3 (0.6–3.0) | 0.44 | 0.8 (0.4–1.8) | 0.70 |
| Prelacteal feeding | 2.3 (1.0–5.5) | 0.06 | NA | NA |
| Time of introduction of complementary foods (days) | 0.6 (0.3–1.3) | 0.17 | NA | NA |
| Drinking from a bottle with a nipple in the past 24 hours | 0.6 (0.3–1.3) | 0.17 | 1.0 (0.5–2.1) | 1 |
| Consumption of unpasteurized animal milk in the past 24 hours | 1.0 (0.2–4.4) | 1 | 1.4 (0.6–2.9) | 0.44 |
| Consumption of any solid food in the past 24 hours | 4.5 (1.1–31.1) | 0.04 | 1.8 (0.9–4.0) | 0.12 |
| Consumption of animal source food in the past 24 hours‡ | NA | NA | 0.9 (0.4–1.9) | 0.70 |
| Minimum dietary diversity‡ | NA | NA | 0.8 (0.4–1.7) | 0.56 |
| Household food insecurity access score | 1.3 (0.6–2.7) | 0.56 | 2.3 (1.1–5.1) | 0.03 |
| Vitamin A supplementation in the past month§ | NA | NA | 1.2 (0.5–2.6) | 0.69 |
| Being treated with antibiotics in the past month | 1.6 (0.7–3.7) | 0.22 | 0.8 (0.4–1.7) | 0.55 |
| Contact with animals | 3.2 (1.4–7.7) | 0.01 | 1.3 (0.6–2.7) | 0.56 |
| Crawling in areas with animal feces | 3.0 (0.8–14.2) | 0.11 | 1.5 (0.7–3.2) | 0.33 |
| Mouthing of soil or animal feces | 12.1 (2.2–226) | 0.00 | 1.5 (0.7–3.2) | 0.33 |
| Preventive actions (mother) for mouthing behavior | 1.1 (0.5–2.4) | 0.77 | 0.9 (0.4–2.0) | 0.84 |
| Elevation (m) | 0.6 (0.3–1.3) | 0.17 | 0.7 (0.3–1.5) | 0.33 |
| Maximum daily land surface temperature, 2021 (°C) | 1.3 (0.6–2.7) | 0.56 | 1.1 (0.5–2.3) | 0.85 |
| Mean daily land surface temperature, 2021 (°C) | 0.9 (0.4–2.0) | 0.85 | 0.8 (0.4–1.7) | 0.56 |
| Minimum daily land surface temperature, 2021 (°C) | 1.3 (0.6–2.7) | 0.56 | 1.5 (0.7–3.2) | 0.33 |
| Maximum 16-day NDVI, 2021 | 1.5 (0.7–3.2) | 0.33 | 0.9 (0.4–2.0) | 0.85 |
| Mean 16-day NDVI, 2021 | 1.3 (0.6–2.7) | 0.56 | 0.7 (0.3–1.5) | 0.33 |
| Minimum 16-day NDVI, 2021 | 0.8 (0.4–1.7) | 0.56 | 0.5 (0.2–1.1) | 0.08 |
| Population count 100 × 100 m, 2020 | 1.2 (0.5–2.5) | 0.70 | 0.3 (0.1–0.6) | 0.00 |
| Slope (degree) | 0.4 (0.2–0.9) | 0.03 | 1.3 (0.6–2.7) | 0.56 |
| Proportion of clay particles in the fine earth fraction (g/kg) | 1.3 (0.6–2.7) | 0.56 | 2.3 (1.1–5.1) | 0.03 |
| Soil organic carbon content in the fine earth fraction (0.1 g/kg) | 0.9 (0.4–1.8) | 0.70 | 1.2 (0.5–2.5) | 0.70 |
| Soil pH | 0.5 (0.2–1.3) | 0.16 | 0.8 (0.3–2.0) | 0.64 |
NA = not applicable; NDVI = normalized difference vegetation index; OR = odds ratio; CI = confidence interval.
ORs are shown with CIs in parentheses.
Bold script represents that the P-value is less than 0.25 and that the corresponding variable was included as a candidate for the multivariate analysis.
Data were not available for the first period as the minimum dietary diversity survey was administered after 6 months of age.
Only one child had vitamin A supplements for the first period. This variable was excluded from the univariate analysis for the first period.
Explanatory variables: Environmental covariates.
Thirteen environmental covariates used for ecological niche modeling of the genus Campylobacter in a previous analysis42 were also included in this study (Table 3). Corresponding values in grids within which the 106 enrolled households were located were extracted from the raster layer of each environmental covariate using the raster package43 in R (R Foundation, Vienna, Austria).
Environmental variables used in this study and their sources
| Variable | Median | Data Source |
|---|---|---|
| Elevation (m) | 2,083.0 | WorldPop (https://www.worldpop.org/) |
| Maximum daily land surface temperature, 2021 (°C) | 44.18 | MODIS Land Surface Temperature/Emissivity Daily (MOD11A1) v. 6.1 |
| Mean daily land surface temperature, 2021 (°C) | 31.17 | MODIS Land Surface Temperature/Emissivity Daily (MOD11A1) v. 6.1 |
| Minimum daily land surface temperature, 2021 (°C) | 16.58 | MODIS Land Surface Temperature/Emissivity Daily (MOD11A1) v. 6.1 |
| Maximum 16-day NDVI, 2021 | 0.68 | MODIS Vegetation Indices (MOD13Q1) v. 6.1 |
| Mean 16-day NDVI, 2021 | 0.48 | MODIS Vegetation Indices (MOD13Q1) v. 6.1 |
| Minimum 16-day NDVI, 2021 | 0.30 | MODIS Vegetation Indices (MOD13Q1) v. 6.1 |
| Population count 100 × 100 m, 2020 | 6 | WorldPop (https://www.worldpop.org/) |
| Slope (degree) | 5.0 | WorldPop (https://www.worldpop.org/) |
| Proportion of clay particles in the fine earth fraction (g/kg) | 421.5 | SoilGrids v. 2.0 (https://soilgrids.org/) |
| Soil organic carbon content in the fine earth fraction (0.1 g/kg) | 328 | SoilGrids v. 2.0 (https://soilgrids.org/) |
| Soil pH | 7.4 | SoilGrids v. 2.0 (https://soilgrids.org/) |
NDVI = normalized difference vegetation index.
STATISTICAL ANALYSES
Given that feeding practices (exclusive breastfeeding versus complementary feeding) and the motor ability of infants are most often quite different between the first and second halves of infants’ first year of life, related risk factors, such as probability of exposure to livestock and contaminated environments, are likely to dynamically change over time. Child age is also an important confounding factor for Campylobacter infections.4,39 To be consistent with our previous analysis about the prevalence of Campylobacter by age groups,42 we split the whole study period into two parts, with a cutoff of 177 days of child age, which reflects the boundary between the first two and second two age groups classified in that analysis.
The outcome variable, cumulative Campylobacter burden, and all explanatory variables derived from the short household surveys were calculated separately for the two periods. We took the average of the monthly short survey data for each selected short survey variable, which resulted in an average value for the numeric variables or a proportion of being “one” for the binary variables (ordinal variable MDD was dichotomized using a generally accepted cutoff of less than five and greater than or equal to five) during each period. Given the relatively small sample size, a median split approach was applied to dichotomize all continuous variables to improve the model robustness following a previous study.44
The purposeful selection of covariates approach was used in this study to choose the candidate variables and determine which to include in the final model.45,46 The likelihood ratio test from logistic regression was performed for each explanatory variable with the outcome variable. The univariate analysis was conducted for the two time periods separately with the same pool of explanatory variables. For the first period, two variables (consumption of animal source food in the past 24 hours and MDD) were excluded from the univariate analysis as these questions were not asked until a child reached 6 months of age. Given that 105 of 106 children did not have vitamin A supplements in the first period, the variable vitamin A supplementation was also excluded from the first-period univariate analysis. Any variables with a P-value less than 0.25 were selected as candidates for the following multivariate analysis.
We built a multivariate logistic regression model using an iterative process to select variables. In each iteration, the variable with the highest P-value was temporarily removed if it was not significant at the 0.05 alpha level. We checked for changes in the remaining coefficients to assess potential confounding effects. If removing the variable caused a change greater than 20% in any coefficient, indicating confounding, it was added back. This process was repeated until the final model included only significant covariates and potential confounders.
Nagelkerke R2,47 a so-called pseudo-R2 measure, and the Hosmer and Lemeshow goodness-of-fit test48 were used to test the model fit. For a sample size of less than 200 and percentage of success in the outcome variable ranging from 38% to 62%, benchmark values between 0.32 and 0.58 of Nagelkerke R2 indicate good fit of the model.49 The Hosmer and Lemeshow test is a hypothesis test and evaluates if the expected event frequencies from the logistic regression model match the observed event frequencies in subgroups. The area under the receiver operating characteristic curve (AUC) was used to assess the model performance in discriminating the positive results from the negatives.50 AUC values range from 0 to 1, and empirically, values between 0.7 and 0.9 are a sign of good predictive performance. Values greater than 0.9 represent an excellent performance.51
We first fitted a multivariate model using the data collected from the first period and tested if it was a good fit for the second-period data. If not, a separate model was built for the second period, and effects of the covariates between two periods were evaluated. All of the statistical analyses were performed using R v. 4.1.1.52
Spatial autocorrelation test.
Spatial autocorrelation refers to the correlation within variables across different spatial units.53 If the values of a particular variable in nearby locations tend to be similar, a positive spatial autocorrelation exists in this variable, whereas a negative spatial autocorrelation occurs when the values of a variable are more dissimilar than expected with their spatial neighbors. If spatial autocorrelation exists in the residuals of a regression model, it violates the assumption of independent errors and may result in an underestimation of the standard errors of the coefficient estimates of the model.54 To test the spatial autocorrelation in the residuals of the logistic models fit in this study, we performed the Moran I test on the residuals using the spdep package in R.55
RESULTS
Demographics and socioeconomic status.
Among the 106 enrolled children, 51.9% were male, and 48.1% were female (Table 1). The mother’s age at baseline ranged from 17 to 43 years, with a median age of 27 years. A high proportion of mothers reported not attending school at any level; only 28% had some primary education. For livestock ownership, 49%, 57%, 45%, and 50% of households owned cattle, goat, sheep, and chicken, respectively. Accordingly, the indicator of tropical livestock units kept by the household ranged from 0 to 5.54, with a median value of 0.62.
Cumulative Campylobacter burden.
The calculated cumulative Campylobacter burden for infants in the first period ranged from 0.77 to 3.75 log genome copies per 50 ng DNA, with a median value of 2.10, whereas the minimum and maximum of cumulative Campylobacter burden for the second period were 1.95 and 5.06 log genome copies per 50 ng DNA, respectively. The Campylobacter burden in the second half year of life (M = 3.52, SD = 0.75) was significantly higher than in the first half year of life (M = 2.14, SD = 0.62) with t(210) = 14.7 and P <0.01 (Figure 2).
Frequency distribution of the cumulative Campylobacter burden for the first and second periods in this study.
Citation: The American Journal of Tropical Medicine and Hygiene 112, 3; 10.4269/ajtmh.24-0401
Univariate analysis.
Based on the P-value of the likelihood ratio test with a cutoff of 0.25, 13 and 8 variables were selected as candidate variables for the first period and the second period, respectively (Table 2). Among the candidates, three variables, namely cattle ownership, sheep ownership, and consumption of any solid food in the past 24 hours, were included in the multivariate analysis for both periods. Although child’s sex was not significant at the alpha level of 0.25 in the first period, we still included it in the multivariate analysis given the confounding effect of child’s sex in Campylobacter infection reported in previous studies.4,7,39,56
Multivariate analysis.
Period 1.
The final logistic regression model for the first period showed that contact with animals, mouthing of soil or animal feces, and drinking from a bottle with a nipple in the past 24 hours were statistically significant at the 0.05 level (Table 4). The direction of the estimated coefficients indicated that children in the first period who had more physical contact with animals and more mouthing of soil or animal feces had greater odds of high cumulative Campylobacter burden, whereas drinking from a bottle with a nipple was found to be a protective factor for high Campylobacter burden. Sheep ownership was marginally significant in the model and had a negative coefficient estimate. Through the purposeful selection of the covariates process, child’s sex, prelacteal feeding, elevation, and soil pH were identified as potential confounding factors for Campylobacter burden.
Logistic regression model for the cumulative Campylobacter burden in the first period
| Characteristic | OR | 95% CI | P-Value |
|---|---|---|---|
| Child’s sex | 0.59 | 0.22–1.51 | 0.3 |
| Sheep ownership | 0.40 | 0.15–1.01 | 0.06 |
| Prelacteal feeding | 1.77 | 0.66–4.87 | 0.3 |
| Contact with animals | 3.13 | 1.11–9.6 | 0.04* |
| Mouthing of soil or animal feces | 12.8 | 1.80–272 | 0.03 |
| Drinking from a bottle with a nipple in the past 24 hours | 0.35 | 0.12–0.94 | 0.04 |
| Elevation | 0.61 | 0.23–1.60 | 0.30 |
| Soil pH | 0.43 | 0.14–1.29 | 0.14 |
| Nagelkerke R2 | 0.40 | – | – |
| Hosmer and Lemeshow test | – | – | 0.42 |
| AUC | 0.79 | 0.70–0.88 | – |
| Moran I statistic | −0.06 | – | 0.75 |
AUC = area under the receiver operating characteristic curve; OR = odds ratio; CI = confidence interval.
Bold represents that the P-value is less than 0.05.
The Nagelkerke R2 of the logistic regression model was 0.40, which fell into the benchmark values (0.32–0.58) for good model fit (Table 4). The Hosmer and Lemeshow test also showed that there was no evidence of poor model fit. The model also had a relatively good predictive performance with an AUC of 0.79 (95% CI: 0.70–0.88) (Figure 3A). Moran I statistics showed that no spatial autocorrelation existed in the model residuals, suggesting that there was no need to consider a spatial regression model to account for the spatial autocorrelation.
Receiver operating curves for logistic regression models fit in this study. (A) Model fit with the first-period data. (B) Test of the fit of the first-period model to the second-period data. (C) Model fit with the second-period data. AUC = area under the receiver operating characteristic curve.
Citation: The American Journal of Tropical Medicine and Hygiene 112, 3; 10.4269/ajtmh.24-0401
To test if one single model can be obtained for both periods, the second-period data were plugged into the model for the first period as a testing dataset. The Hosmer and Lemeshow test checked the difference between observed data and predicted values, and it showed that this difference was significant (P = 0.00), suggesting a poor fit of the model to the second-period data. It was also supported by a low AUC of 0.57 (95% CI: 0.46–0.69) (Figure 3B).
Period 2.
Logistic regression models were then fitted using the candidate variables selected from the univariate analysis for the second period, and the final model included eight variables, with child’s sex, cattle ownership, and population density (per 100 m2) being statistically significant at the 0.05 level (Table 5). Children being female and living in households that kept cattle had greater odds of high cumulative Campylobacter burden, whereas population density showed protective effects on Campylobacter burden. Household food insecurity access score, proportion of clay particles in soil, consumption of any solid food in the past 24 hours, and sheep ownership were identified as potential confounding factors for Campylobacter burden in the second period.
Logistic regression model for the cumulative Campylobacter burden in the second period
| Characteristic | OR | 95% CI | P-Value |
|---|---|---|---|
| Child’s sex | 2.85 | 1.15–7.47 | 0.027* |
| Cattle ownership | 2.87 | 1.14–7.70 | 0.029 |
| Sheep ownership | 0.49 | 0.91–1.21 | 0.12 |
| Household food insecurity access score | 1.78 | 0.69–4.59 | 0.2 |
| Consumption of any solid food in the past 24 hours | 2.01 | 0.79–5.30 | 0.15 |
| Minimum 16-day NDVI, 2021 | 0.43 | 0.17–1.05 | 0.069 |
| Population count 100 × 100 m, 2020 | 0.37 | 0.15–0.92 | 0.033 |
| Proportion of clay particles in the fine earth fraction | 2.12 | 0.84–5.54 | 0.12 |
| Nagelkerke R2 | 0.34 | – | – |
| Hosmer and Lemeshow test | – | – | 0.47 |
| AUC | 0.80 | 0.71–0.89 | |
| Moran I statistic | 0.014 | – | 0.37 |
AUC = area under the receiver operating characteristic curve; NDVI = normalized difference vegetation index; OR = odds ratio; CI = confidence interval.
Bold represents that the P-value is less than 0.05.
The Nagelkerke R2 for the second model was 0.34, and the Hosmer and Lemeshow test showed that there was no evidence of poor model fit (Table 5). This model had a good predictive performance with an AUC of 0.80 (95% CI: 0.71–0.89) (Figure 3C). Again, there was no spatial autocorrelation existing in the model residuals.
DISCUSSION
From a One Health perspective, this study identified potential factors involved in the Campylobacter transmission pathways between humans, animals, and the environment in a low-resource setting. Child age-specific behaviors and other factors that may increase the risk of children’s exposure to animal feces are largely missing from previous interventions that aimed to mitigate the burden of Campylobacter infection or other enteric illnesses among infants and young children in LMICs.57 Risk factors of Campylobacter infections that have been commonly identified in previous studies conducted among children younger than 5 years of age in low-resource settings are primarily limited to maternal education level, feeding practices, WaSH indicators, and ownership of domestic animals.4,7,58,59 However, fewer studies have delved into infant-specific behaviors and the interactions between infants and livestock, which could also contribute to the risk of Campylobacter infections.57 Considering the dynamic effects of age, particularly distinguishing between infants ages younger than 6 months and those ages 6–12 months, on biological risk and environmental interaction, our study introduces a novel approach. By fitting models and identifying risk factors separately for the first and second 6 months of infancy, we aim to capture the nuanced variations in infection dynamics during these critical developmental stages. This innovative methodology coupled with the exploration of previously overlooked factors using longitudinal data sets our study apart from existing research on Campylobacter infection risk factors.
At an early stage of life, infants and young children put nonfood objects in their mouths as one way to explore the surrounding environment.60 However, this exploratory behavior could put children at higher risk of contracting zoonotic pathogens if they live in an environment contaminated by livestock and poultry feces. Our results identify mouthing of soil or animal feces as a significant risk factor contributing to a higher cumulative Campylobacter burden during the first half of the first year of life. This association can be attributed to the high prevalence of Campylobacter spp. in the feces of animals, including domestic livestock and poultry. Results from the laboratory showed that the prevalences of Campylobacter in the fecal samples of cattle, sheep, goats, and chicken collected from the CAGED longitudinal study were 99%, 98%, 99%, and 93%, respectively.37 In rural areas of LMICs, infants and young children are frequently placed on the ground, sharing space with free-ranging livestock.57 As animals defecate on the homestead or sometimes, even inside the homes, enteric pathogens harbored in livestock feces or the soil contaminated by the feces may be ingested by infants through their routine mouthing behaviors while they are on the floor. Ingestion of soil and livestock feces is, therefore, a point of exposure specific to infant and young child behavior. This infant-specific transmission pathway needs to be considered in the future design of intervention strategies for preventing children from having fecal exposure. This finding echoes the collective conclusion from the WaSH Benefits and the Sanitation, Hygiene, Infant Nutrition Efficacy trials: that more effective interventions are needed to reduce the exposure to fecal contaminations in the domestic environments other than the traditional WaSH interventions.61
Similarly, physical contact with animals was identified as another risk factor existing in the child–livestock interaction in the first period (before 6 months of age) that increases the exposure of children to animal feces. It is very common for people in rural areas to share living and sleeping quarters with their livestock, which frequently exposes young children in the household to direct contact with livestock. Keeping animals inside human living spaces has been associated with Campylobacter-positive child stools in another study in rural Ethiopia,57 highlighting the need to design interventions that create adequate separation of domestic animals from the human living spaces to block the transmission pathway through direct contact.
The risks and benefits of raising livestock in smallholder families on child health are intertwined.23 On one hand, livestock raised in the household can provide animal source foods, which are seen as the best source of nutrition-rich food for infants and young children. Livestock production is also a source of household income. Increased income from livestock sales can grant families more purchasing power for food, which further helps improve child nutritional status. However, livestock ownership can potentially increase the risk of children’s exposure to Campylobacter species. Specifically, free-ranging chickens, which are very common in rural and periurban communities in LMICs, are considered a major source of Campylobacter infection.62 Their feces can often be found in the home, which potentially increases the risk of Campylobacter exposure. According to the responses to our household surveys, chickens were usually kept inside the house (living quarters), on the homestead, or outside the homestead during the day. The frequencies were similar among the three scenarios. However, almost all households owning chickens did not confine their chickens regardless of whether they were kept inside the house or outside. In another analysis of the same population,44 we found that keeping chickens unconfined inside the house was associated with a higher Campylobacter load over the whole study period. However, that signal was not picked up by the models in the current study after the study period was split into two. Although free-ranging chickens might remain a potential risk factor for Campylobacter colonization in rural areas as it is in urban/periurban settings, simple relationships between animal ownership/management and Campylobacter colonization were often counterintuitive in our analyses,44 suggesting complicated relationships that require more detailed approaches.
In this study, we did not find a significant association between consumption of animal source food and Campylobacter burden, but cattle ownership was identified as a significant factor associated with increased odds of having higher cumulative Campylobacter burden during the second half of the first year of life. In our study population, when cattle were kept either inside the home or on the homestead but outside the home during the day, 100% (n = 27/27) and 93.3% (n = 28/30) of households, respectively, confined those cattle. During the night, almost all cattle were kept and tied inside the home. Regarding the purposes of raising cattle, households reported selling livestock for income (44.2%; n = 23/52) and consumption of milk (51.9%; n = 27/52) as the two major purposes. These data support other findings indicating a strong cultural focus on milk consumption in this area, which may correspond with an underlying risk in the consumption of raw/unpasteurized milk that has been associated with reported outbreaks of campylobacteriosis in high-income countries.63,64 A recent Ethiopian study showed a higher prevalence of C. jejuni (16%) in raw milk compared with other dairy products.65 Although consumption of raw milk was not identified in our models as a significant risk factor of Campylobacter burden in our study population, further work is still needed to unpack this association among children in low-resource settings.
A previous study showed that Campylobacter was highly prevalent in livestock feces collected from these households, with prevalences of 99%, 98%, and 93% for cattle, sheep, and chickens, respectively.37 It is worth noting that soil samples also had a high Campylobacter prevalence of 58%. Even if children living in the household do not have direct contact with livestock, they are still likely to get exposed through other indirect pathways, such as putting soil in their mouth. However, this kind of behavioral data is usually hard to collect through surveys. The Exposure Assessment of Campylobacter Infection in Rural Ethiopia project, a sister project of CAGED, was designed to conduct organized child observations for the same cohort as well as collect environmental samples in the households. Results generated from that project will help dissect these potential pathways of children’s exposure to Campylobacter species in the future.66,67
Appropriate infant and young child feeding practices could improve child nutritional status, growth, and development.68 For infants ages 0–5 months, exclusive breastfeeding is strongly recommended by WHO, whereas bottle-feeding using a bottle with a nipple/teat to feed any liquid or semisolid food is discouraged at this early stage of life because of global concerns that include excessive weight gain, iron depletion, etc.69 In addition, a bottle with a nipple is more likely to be contaminated in low-resource settings where inadequate cleaning and disinfection of bottles are more common, which increases the risk of enteric infections.70 Lengerh et al.59 showed that bottle-feeding was associated with increased odds of Campylobacter infection among diarrheic children in northwest Ethiopia. However, in this study, drinking from a bottle with a nipple was shown to be a protective factor for higher cumulative Campylobacter burden in the first time period. Our results should not be interpreted as an encouragement of bottle-feeding. One potential explanation could be that this practice is linked to other socioeconomic factors that contribute to reducing the risk of Campylobacter infections. A previous study using Ethiopian Demographic and Health Surveys data to examine the determinants of bottle-feeding suggested that women who had a higher education, came from a richer household, and lived in urban areas were more likely to bottle-feed.69 The role of the mother’s education level and household wealth status in Campylobacter burden was not clear in our analysis with a relatively small sample size, and further work is needed to unpack these potential links.
Influences of environmental and climatic factors have been rarely reported in studies investigating risk factors of Campylobacter infections among children in LMICs. Here, we included several environmental covariates previously used to model the prevalence of enteric diseases globally and regionally. Because of the small geography of our study area, most environmental covariates have less heterogeneity to show signals in their association with Campylobacter infections. This is one of the limitations in this study. Evaluating these environmental effects on Campylobacter infections at a larger scale in LMICs would be a future direction.
We tested the spatial autocorrelation in the model residuals during the model-building process, a step that has seldom been taken in previous work on identifying risk factors of Campylobacter infections. We did so not only because we included environmental covariates in the models but also, to consider the potential spatial effects (i.e., spatial dependence and spatial heterogeneity) introduced by georeferenced data.71 Neglecting the spatial effects could lead to an inflation of variance in regression estimates and consequently, a less reliable regression model.72 Therefore, it would be appropriate for future studies to consider including the spatial autocorrelation test as a core component of the modeling-building process to ensure the reliability and accuracy of the regression model.
Another innovative aspect of this study involves the adoption of bacterial load-based cumulative burden, departing from traditional prevalence measures when dealing with longitudinal data. By calculating the cumulative burden, we gain insight into the persistent impact of Campylobacter infections over time, offering a nuanced perspective on disease dynamics. This method enables a comprehensive assessment of Campylobacter burden over a certain period and holds promise for advancing longitudinal studies on Campylobacter infections. Our previous study indicated that higher Campylobacter loads are associated with an increased frequency of diarrhea among children in eastern Ethiopia,44 suggesting a potential link between bacterial load and disease occurrence. In addition, results from the Etiology, Risk Factors, and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development Project (MAL-ED) study suggested that higher Campylobacter burden has negatively affected the linear growth (measured by length-for-age Z score) of children in low-resource settings.4,73 It is necessary to understand the clinical significance of Campylobacter burden, and further investigation is essential to better understand how bacterial burden impacts disease outcomes and to determine whether it could serve as a reliable marker for predicting or managing Campylobacter-related illnesses.73–75
CONCLUSION
In conclusion, our study reveals that factors involved in the interactions between humans (infants), livestock, and home environment impacted the presence of higher cumulative Campylobacter burden among infants in eastern Ethiopia. For infants younger than 6 months, being reported to have physical contact with animals and have mouthing of soil or animal feces were identified as risk factors of higher Campylobacter burden. Additionally, drinking from a bottle with a nipple was shown to be a protective factor. This result requires additional research to understand and should not be interpreted to encourage bottle-feeding; additional information is required to understand whether there is a direct causal mechanism or whether underlying factors or confounders, such as socioeconomic status or overall household hygiene, might explain the finding. In older infants (ages between 6 and 12 months), being female and living in households with cattle had increased odds of higher Campylobacter burden. High population density (potentially linked to urban residency) was identified as a protective factor for this age group. Future interventions should pay more attention to the infant-specific transmission pathway and create adequate separation of domestic animals from humans to prevent infants and young children from potential fecal exposure.
ACKNOWLEDGMENTS
This work is a result of the Campylobacter Genomics and Environmental Enteric Dysfunction Research Team, whose members include Amanda Evelyn Ojeda, Arie H. Havelaar, Abadir Jemal Seran, Abdulmuen Mohammed Ibrahim, Bahar Mummed Hassen, Belisa Usmael Ahmedo, Cyrus Saleem, Dehao Chen, Efrah Ali Yusuf, Gireesh Rajashekara, Getnet Yimer, Ibsa A. Ahmed, Ibsa Aliyi Usmane, Jafer Kedir Amin, Jason K. Blackburn, Jemal Y. Hassen, Kedir A. Hassen, Kunuza Adem Umer, Karah Mechlowitz, Kedir Teji Roba, Loic Deblais, Mussie Bhrane, Mark J. Manary, Mawardi M. Dawid, Mahammad Mahammad Usmail, Nigel P. French, Nur Shaikh, Nitya Singh, Sarah L. McKune, Wondwossen A. Gebreyes, Xiaolong Li, Yenenesh Demisie Weldesenbet, Yang Yang, and Zelalem Hailu Mekuria.
REFERENCES
- 1.↑
Sahin O, Fitzgerald C, Stroika S, Zhao S, Sippy RJ, Kwan P, Plummer PJ, Han J, Yaeger MJ, Zhang Q, 2012. Molecular evidence for zoonotic transmission of an emergent, highly pathogenic Campylobacter jejuni clone in the United States. J Clin Microbiol 50: 680–687.
- 2.↑
Chlebicz A, Śliżewska K, 2018. Campylobacteriosis, Salmonellosis, Yersiniosis, and Listeriosis as zoonotic foodborne diseases: A review. Int J Environ Res Public Health 15: 863.
- 3.↑
Kaakoush NO, Castaño-Rodríguez N, Mitchell HM, Man SM, 2015. Global epidemiology of campylobacter infection. Clin Microbiol Rev 28: 687–720.
- 4.↑
Amour C et al.; Etiology, Risk Factors, and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development Project (MAL-ED) Network Investigators, 2016. Epidemiology and impact of Campylobacter infection in children in 8 low-resource settings: Results from the MAL-ED study. Clin Infect Dis 63: 1171–1179.
- 5.↑
Platts-Mills JA, Kosek M, 2014. Update on the burden of Campylobacter in developing countries. Curr Opin Infect Dis 27: 444–450.
- 6.↑
Lee G, Pan W, Peñataro Yori P, Paredes Olortegui M, Tilley D, Gregory M, Oberhelman R, Burga R, Chavez CB, Kosek M, 2013. Symptomatic and asymptomatic Campylobacter infections associated with reduced growth in Peruvian children. PLoS Negl Trop Dis 7: e2036.
- 7.↑
Haque MA et al., 2019. Determinants of Campylobacter infection and association with growth and enteric inflammation in children under 2 years of age in low-resource settings. Sci Rep 9: 17124.
- 8.↑
François R et al., 2018. The other Campylobacters: Not innocent bystanders in endemic diarrhea and dysentery in children in low-income settings. PLoS Negl Trop Dis 12: e0006200.
- 9.↑
Man SM, 2011. The clinical importance of emerging Campylobacter species. Nat Rev Gastroenterol Hepatol 8: 669–685.
- 10.↑
Terefe Y et al., 2020. Co-occurrence of Campylobacter species in children from eastern Ethiopia, and their association with environmental enteric dysfunction, diarrhea, and host microbiome. Front Public Health 8: 99.
- 11.↑
Bian X et al., 2020. Campylobacter abundance in breastfed infants and identification of a new species in the Global Enterics multicenter study. mSphere 5: e00735-19.
- 12.↑
Parker CT et al., 2022. Shotgun metagenomics of fecal samples from children in Peru reveals frequent complex co-infections with multiple Campylobacter species. PLoS Negl Trop Dis 16: e0010815.
- 13.↑
Garcia Bardales PF et al., 2022. “Candidatus Campylobacter infans” detection is not associated with diarrhea in children under the age of 2 in Peru. PLoS Negl Trop Dis 16: e0010869.
- 14.↑
Wagner EG, Lanoix JN, 1958. Excreta disposal for rural areas and small communities. Monogr Ser World Health Organ 39: 1–182.
- 15.↑
Penakalapati G, Swarthout J, Delahoy MJ, McAliley L, Wodnik B, Levy K, Freeman MC, 2017. Exposure to animal feces and human health: A systematic review and proposed research priorities. Environ Sci Technol 51: 11537–11552.
- 16.↑
Sclar GD, Penakalapati G, Amato HK, Garn JV, Alexander K, Freeman MC, Boisson S, Medlicott KO, Clasen T, 2016. Assessing the impact of sanitation on indicators of fecal exposure along principal transmission pathways: A systematic review. Int J Hyg Environ Health 219: 709–723.
- 17.↑
Dangour AD, Watson L, Cumming O, Boisson S, Che Y, Velleman Y, Cavill S, Allen E, Uauy R, 2013. Interventions to improve water quality and supply, sanitation and hygiene practices, and their effects on the nutritional status of children. Cochrane Database Syst Rev 2013: CD009382.
- 18.↑
Wolf J et al., 2018. Impact of drinking water, sanitation and handwashing with soap on childhood diarrhoeal disease: Updated meta-analysis and meta-regression. Trop Med Int Health 23: 508–525.
- 19.↑
Freeman MC et al., 2017. The impact of sanitation on infectious disease and nutritional status: A systematic review and meta-analysis. Int J Hyg Environ Health 220: 928–949.
- 20.↑
Clasen T et al., 2014. Effectiveness of a rural sanitation programme on diarrhoea, soil-transmitted helminth infection, and child malnutrition in Odisha, India: A cluster-randomised trial. Lancet GlobHealth 2: e645–e653.
- 21.↑
Null C et al., 2018. Effects of water quality, sanitation, handwashing, and nutritional interventions on diarrhoea and child growth in rural Kenya: A cluster-randomised controlled trial. Lancet Glob Health 6: e316–e329.
- 22.↑
Pickering AJ, Djebbari H, Lopez C, Coulibaly M, Alzua ML, 2015. Effect of a community-led sanitation intervention on child diarrhoea and child growth in rural Mali: A cluster-randomised controlled trial. Lancet Glob Heal 3: e701–e711.
- 23.↑
Chen D, Mechlowitz K, Li X, Schaefer N, Havelaar AH, McKune SL, 2021. Benefits and risks of smallholder livestock production on child nutrition in low- and middle-income countries. Front Nutr 8: 751686.
- 24.↑
Hanson LA, Wiedermann U, Ashraf R, Zaman S, Adlerberth I, Dahlgren U, Wold A, Jalil F, 1996. Effects of breastfeeding on the baby and on its immune system. Food Nutr Bull 17: 1–5.
- 25.↑
Oriá RB, Murray-Kolb LE, Scharf RJ, Pendergast LL, Lang DR, Kolling GL, Guerrant RL, 2016. Early-life enteric infections: Relation between chronic systemic inflammation and poor cognition in children. Nutr Rev 74: 374–386.
- 26.↑
Gibson E, 1988. Exploratory behavior in the development of perceiving, acting, and the acquiring of knowledge. Annu Rev Psychol 39: 1–42.
- 27.↑
Bronowski C, James CE, Winstanley C, 2014. Role of environmental survival in transmission of Campylobacter jejuni. FEMS Microbiol Lett 356: 8–19.
- 28.↑
Whiley H, van den Akker B, Giglio S, Bentham R, 2013. The role of environmental reservoirs in human campylobacteriosis. Int J Environ Res Public Health 10: 5886–5907.
- 29.↑
Arsenault J, Berke O, Michel P, Ravel A, Gosselin P, 2012. Environmental and demographic risk factors for campylobacteriosis: Do various geographical scales tell the same story? BMC Infect Dis 12: 318.
- 30.↑
Louis VR, Gillespie IA, O’Brien SJ, Russek-Cohen E, Pearson AD, Colwell RR, 2005. Temperature-driven Campylobacter seasonality in England and Wales. Appl Environ Microbiol 71: 85–92.
- 31.↑
Kuhn KG, Nygård KM, Guzman-Herrador B, Sunde LS, Rimhanen-Finne R, Trönnberg L, Jepsen MR, Ruuhela R, Wong WK, Ethelberg S, 2020. Campylobacter infections expected to increase due to climate change in northern Europe. Sci Rep 10: 13874–13879.
- 32.↑
Soneja S, Jiang C, Romeo Upperman C, Murtugudde R, Mitchell CS, Blythe D, Sapkota AR, Sapkota A, 2016. Extreme precipitation events and increased risk of campylobacteriosis in Maryland, USA. Environ Res 149: 216–221.
- 33.↑
Weisent J, Seaver W, Odoi A, Rohrbach B, 2014. The importance of climatic factors and outliers in predicting regional monthly campylobacteriosis risk in Georgia, USA. Int J Biometeorol 58: 1865–1878.
- 34.↑
Sanderson RA, Maas JA, Blain AP, Gorton R, Ward J, O’Brien SJ, Hunter PR, Rushton SP, 2018. Spatio-temporal models to determine association between Campylobacter cases and environment. Int J Epidemiol 47: 202–216.
- 35.↑
Spencer SEF, Marshall J, Pirie R, Campbell D, Baker MG, French NP, 2012. The spatial and temporal determinants of campylobacteriosis notifications in New Zealand, 2001–2007. Epidemiol Infect 140: 1663–1677.
- 36.↑
Havelaar AH et al., 2022. Unravelling the reservoirs for colonisation of infants with Campylobacter spp. in rural Ethiopia: Protocol for a longitudinal study during a global pandemic and political tensions. BMJ Open 12: e061311.
- 37.↑
Deblais L et al., 2023. Prevalence and load of the Campylobacter genus in infants and associated household contacts in rural eastern Ethiopia: A longitudinal study from the Campylobacter Genomics and Environmental Enteric Dysfunction (CAGED) project. Appl Environ Microbiol 89: e0042423.
- 38.↑
Platts-Mills JA et al., 2014. Detection of Campylobacter in stool and determination of significance by culture, enzyme immunoassay, and PCR in developing countries. J Clin Microbiol 52: 1074–1080.
- 39.↑
Chen D et al., 2021. Campylobacter colonization, environmental enteric dysfunction, stunting, and associated risk factors among young children in rural Ethiopia: A cross-sectional study from the Campylobacter Genomics and Environmental Enteric Dysfunction (CAGED) project. Front Public Heal 8: 615793.
- 40.↑
World Health Organization, 2017. Global Nutrition Monitoring Framework: Operational Guidance for Tracking Progress in Meeting Targets for 2025. Available at: https://apps.who.int/iris/bitstream/handle/10665/259904/9789241513609-eng.pdf; jsessionid=5B7CD35139464EA9E9214B4F68A81B5E?sequence=1. Accessed December 1, 2022.
- 41.↑
Coates J, Swindale A, Bilinsky P, 2007. Household Food Insecurity Access Scale (HFIAS) for Measurement of Food Access: Indicator Guide. Washington, DC: FHI 360 Food and Nutrition Technical Assistance.
- 42.↑
Li X, 2023. Applying Geospatial Approaches to Studying the Epidemiology of Enteric Illnesses at Different Scales. Gainesville, FL: University of Florida.
- 43.↑
Hijmans RJ, 2024. Raster: Geographic Analysis and Modeling with Raster Data. Available at: https://rspatial.org/raster. Accessed May 1, 2024.
- 44.↑
Chen D et al., 2024. Campylobacter Colonization and Undernutrition in Infants in Rural Eastern Ethiopia: A Longitudinal Community-Based Birth Cohort Study. Available at: https://doi.org/10.1101/2024.05.21.24307707. Accessed May 30, 2024.
- 45.↑
Bursac Z, Gauss CH, Williams DK, Hosmer DW, 2008. Purposeful selection of variables in logistic regression. Source Code Biol Med 3: 17.
- 46.↑
Hosmer DW Jr., Lemeshow S, Sturdivant RX, 2013. Applied Logistic Regression: Third Edition. Hoboken, NJ: Wiley.
- 47.↑
Nagelkerke NJD, 1991. A note on a general definition of the coefficient of determination. Biometrika 78: 691–692.
- 48.↑
Hosmer DW, Lemeshow S, 1980. Goodness of fit tests for the multiple logistic regression model. Comm Stats Theory Methods 9: 1043–1069.
- 49.↑
Hemmert GAJ, Schons LM, Wieseke J, Schimmelpfennig H, 2018. Log-likelihood-based pseudo-R2 in logistic regression: Deriving sample-sensitive benchmarks. Sociol Methods Res 47: 507–531.
- 50.↑
Hanley JA, McNeil BJ, 1982. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 143: 29–36.
- 51.↑
Li H, Chen Y, Deng S, Chen M, Fang T, Tan H, 2019. Eigenvector spatial filtering-based logistic regression for landslide susceptibility assessment. ISPRS Int J Geo-Inf 8: 332.
- 52.↑
R Core Team, 2021. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing.
- 54.↑
Lichstein JW, Simons TR, Shriner SA, Franzreb KE, 2002. Spatial autocorrelation and autoregressive models in ecology. Ecol Monogr 72: 445–463.
- 55.↑
Bivand R, 2022. R packages for analyzing spatial data: A comparative case study with areal data. Geogr Anal 54: 488–518.
- 56.↑
Strachan NJC, Watson RO, Novik V, Hofreuter D, Ogden ID, Galán JE, 2008. Sexual dimorphism in campylobacteriosis. Epidemiol Infect 136: 1492–1495.
- 57.↑
Budge S, Barnett M, Hutchings P, Parker A, Tyrrel S, Hassard F, Garbutt C, Moges M, Woldemedhin F, Jemal M, 2020. Risk factors and transmission pathways associated with infant Campylobacter spp. prevalence and malnutrition: A formative study in rural Ethiopia. PLoS One 15: e0232541.
- 58.↑
Diriba K, Awulachew E, Anja A, 2021. Prevalence and associated factor of Campylobacter species among less than 5-year-old children in Ethiopia: A systematic review and meta-analysis. Eur J Med Res 26: 2.
- 59.↑
Lengerh A, Moges F, Unakal C, Anagaw B, 2013. Prevalence, associated risk factors and antimicrobial susceptibility pattern of Campylobacter species among under five diarrheic children at Gondar University Hospital, northwest Ethiopia. BMC Pediatr 13: 82.
- 60.↑
Ruff HA, Saltarelli LM, Capozzoli M, Dubiner K, 1992. The differentiation of activity in infants’ exploration of objects. Dev Psychol 28: 851–861.
- 61.↑
Pickering AJ et al., 2019. The WASH benefits and SHINE trials: Interpretation of WASH intervention effects on linear growth and diarrhoea. Lancet Glob Health 7: e1139–e1146.
- 62.↑
Schiaffino F, Trigoso DR, Colston JM, Olortegui MP, Shapiama Lopez WV, Garcia Bardales PF, Pisanic N, Davis MF, Yori PP, Kosek MN, 2021. Associations among household animal ownership, infrastructure, and hygiene characteristics with source attribution of household fecal contamination in peri-urban communities of Iquitos, Peru. Am J Trop Med Hyg 104: 372–381.
- 63.↑
Wulsten IF, Galeev A, Stingl K, 2020. Underestimated survival of Campylobacter in raw milk highlighted by viability real-time PCR and growth recovery. Front Microbiol 11: 1107.
- 64.↑
Davys G, Marshall JC, Fayaz A, Weir RP, Benschop J, 2020. Campylobacteriosis associated with the consumption of unpasteurised milk: Findings from a sentinel surveillance site. Epidemiol Infect 148: e16.
- 65.↑
Admasie A, Eshetu A, Tessema TS, Vipham J, Kovac J, Zewdu A, 2023. Prevalence of Campylobacter species and associated risk factors for contamination of dairy products collected in a dry season from major milk sheds in Ethiopia. Food Microbiol 109: 104145.
- 66.↑
Deblais L et al., 2024. Assessing Fecal Contamination from Human and Environmental Sources Using Escherichia coli as an Indicator in Rural Ethiopian Households—A Study from the EXCAM Project. Available at: https://doi.org/10.1101/2024.08.21.24312392. Accessed August 30, 2024.
- 67.↑
Wang Y et al., 2024. Quantitative Multi-Pathway Assessment of Exposure to Fecal Contamination for Infants in Rural Ethiopia. Available at: https://doi.org/10.1101/2024.08.29.24312786. Accessed September 1, 2024.
- 68.↑
World Health Organization, 2008. Indicators for Assessing Infant and Young Child Feeding Practices: Conclusions of a Consensus Meeting Held 6–8 November 2007 in Washington, DC. Geneva, Switzerland: WHO.
- 69.↑
Belay DG, Getnet M, Akalu Y, Diress M, Gela YY, Getahun AB, Bitew DA, Terefe B, Belsti Y, 2022. Spatial distribution and determinants of bottle feeding among children 0–23 months in Ethiopia: Spatial and multi-level analysis based on 2016 EDHS. BMC Pediatr 22: 120.
- 70.↑
Rothstein JD, Mendoza AL, Cabrera LZ, Pachas J, Calderón M, Pajuelo MJ, Caulfield LE, Winch PJ, Gilman RH, 2019. Household contamination of baby bottles and opportunities to improve bottle hygiene in peri-urban Lima, Peru. Am J Trop Med Hyg 100: 988–997.
- 71.↑
Anselin L, Getis A, 1992. Spatial statistical analysis and geographic information systems. Ann Reg Sci 26: 19–33.
- 72.↑
Bertazzon S, Olson S, Knudtson M, 2010. A spatial analysis of the demographic and socio-economic variables associated with cardiovascular disease in Calgary (Canada). Appl Spatial Anal 3: 1–23.
- 73.↑
Rogawski ET et al., 2018. Use of quantitative molecular diagnostic methods to investigate the effect of enteropathogen infections on linear growth in children in low-resource settings: Longitudinal analysis of results from the MAL-ED cohort study. Lancet Glob Health 6: e1319–e1328.
- 74.↑
Liu J et al., 2016. Use of quantitative molecular diagnostic methods to identify causes of diarrhoea in children: A reanalysis of the GEMS case-control study. Lancet 388: 1291–1301.
- 75.↑
Platts-Mills JA, McQuade ETR, 2023. Assigning pathogen etiology for childhood diarrhea in high-burden settings: A call for innovative approaches. J Infect Dis 228: 814–817.