Exposure, hazard, and vulnerability all contribute to Schistosoma haematobium re-infection in northern Senegal.

BACKGROUND: Infectious disease risk is driven by three interrelated components: exposure, hazard, and vulnerability. For schistosomiasis, exposure occurs through contact with water, which is often tied to daily activities. Water contact, however, does not imply risk unless the environmental hazard of snails and parasites is also present in the water. By increasing reliance on hazardous activities and environments, socio-economic vulnerability can hinder reductions in exposure to a hazard. We aimed to quantify the contributions of exposure, hazard, and vulnerability to the presence and intensity of Schistosoma haematobium re-infection. METHODOLOGY/PRINCIPAL
FINDINGS: In 13 villages along the Senegal River, we collected parasitological data from 821 school-aged children, survey data from 411 households where those children resided, and ecological data from all 24 village water access sites. We fit mixed-effects logistic and negative binomial regressions with indices of exposure, hazard, and vulnerability as explanatory variables of Schistosoma haematobium presence and intensity, respectively, controlling for demographic variables. Using multi-model inference to calculate the relative importance of each component of risk, we found that hazard (Ʃwi = 0.95) was the most important component of S. haematobium presence, followed by vulnerability (Ʃwi = 0.91). Exposure (Ʃwi = 1.00) was the most important component of S. haematobium intensity, followed by hazard (Ʃwi = 0.77). Model averaging quantified associations between each infection outcome and indices of exposure, hazard, and vulnerability, revealing a positive association between hazard and infection presence (OR = 1.49, 95% CI 1.12, 1.97), and a positive association between exposure and infection intensity (RR 2.59-3.86, depending on the category; all 95% CIs above 1).
CONCLUSIONS/SIGNIFICANCE: Our findings underscore the linkages between social (exposure and vulnerability) and environmental (hazard) processes in the acquisition and accumulation of S. haematobium infection. This approach highlights the importance of implementing both social and environmental interventions to complement mass drug administration.

Introduction

When a natural hazard like a flood or hurricane is imminent, risk managers and first responders must account for the exposure and vulnerability of affected populations, conditions and processes that are dynamic in both space and time [1]. Precious resources will be misplaced if authorities help families whose homes have not flooded, because those resources will have been diverted away from families who have been exposed to the impacts of the hazard. Similarly, vulnerable families with a reduced capacity to anticipate, respond to, or recover from, the impacts of a natural hazard will require more assistance than less vulnerable families [1]. An effective response to natural disasters, as a result, requires evaluation of where a natural hazard overlaps with exposure, and how vulnerability mediates exposure and its impacts in time and space.

In contrast to the risk of natural disasters, the risk of infectious disease is often defined in terms of its environmental determinants [2-4]. Risk of insect-borne diseases like Lyme disease and malaria are often described in terms of ‘entomological risk,’ such as the density of infected nymphal ticks (for Lyme disease) [5] or the probability of an infective mosquito bite (for malaria) [6]. Similarly, wildlife biodiversity–representing the size of the natural reservoir of potentially zoonotic pathogens–is used as a proxy for spillover risk [7]. While these measures reflect important components of pathogen transmission, their focus on the source of the disease-causing agent overlooks the socio-ecological processes that connect those infectious agents with susceptible people, thereby facilitating transmission and determining risk. Environmental metrics represent a single component of infectious disease risk: hazard [6, 8]. Meanwhile, exposure and vulnerability reflect the roles of human behavior and social structures in transforming hazard into risk: exposure describes contact between a person and a pathogen (e.g., exposure to a hazard), while vulnerability determines the extent to which people can anticipate, adapt to, and potentially mitigate the impacts of infection [6, 9].

These three components of risk–exposure, hazard, and vulnerability–operate in concert rather than isolation. They comprise the breadth of socio-ecological conditions that put and keep people in contact with infectious agents of disease. Few studies address all three components simultaneously, but those that do show how human interaction with the environment transforms the presence of a hazard into risk for infection and disease. For example, accounting for both entomological risk (i.e., hazard) and exposure explained Lyme disease incidence in New York when entomological risk alone did not [5]. In East Africa, incorporating social vulnerability and entomological risk into a spatially-explicit risk assessment of malaria showed how different interventions could be targeted to different places with different needs [10, 11]. Across disease systems, insights about the relative contributions of exposure, hazard and vulnerability can inform which interventions to employ to mitigate risk and where, when, and how to implement them.

The risk for schistosomiasis–a disease caused by parasitic trematodes from the genus Schistosoma–can also be divided into these three distinct components. Human schistosomes complete a two-host transmission cycle between humans and specific species of freshwater snails. People become infected through contact with freshwater where schistosomes’ free-living infective larval stage (called cercariae) emerge from infected snails and burrow into human skin. Schistosome hazard is often described as ‘malacological risk,’ and quantified using ecological indicators of snail populations within a water source [12-14]. Similarly, exposure to schistosomes is often approximated by studying human water contact behavior because the process of cercariae burrowing into skin upon contact cannot be directly observed [15-17]. Even so, exposure to surface water on its own does not imply risk for infection [18, 19], as the hazard of infected snails in a water source, and the circulation of parasites through the snail population is what renders water contact harmful [6]. Lastly, vulnerability determines whether a person can avoid or mitigate the schistosome hazard in their surroundings, either by reducing exposure to or contamination of water bodies [20, 21], or seeking treatment for infection [22]. The relative disadvantage of certain groups may make risk mitigation difficult, obliging people to remain in contact with parasite-laden water even if they are aware of the risk [6].

Approximately 200 million people are infected with schistosomes worldwide, making schistosomiasis second only to malaria in the global burden of parasitic disease [23, 24]. Schistosome infections are treatable with the antiparasitic drug praziquantel. Since the 1980s, mass drug administration (MDA) campaigns employing praziquantel have been the central strategy of schistosomiasis control, displacing the snail control interventions of the preceding decades [25, 26]. These MDA programs have played an important role in reducing morbidity due to schistosomiasis, but because praziquantel does not prevent new infections, the benefits of MDA have not been realized universally across all contexts [27]. The identification of persistent hotspots–or settings whose prevalence and intensity of infection remains high despite repeated rounds of treatment–has underscored the need for interventions that control transmission and complement the morbidity control achieved by MDA [28-30]. Controlling transmission will require both ecological and social points of intervention, whose identification and implementation can be informed by understanding the relative contributions of hazard, exposure, and vulnerability in the risk for acquiring and accumulating schistosome infections.

In this mixed methods study, we combined parasitological, ecological, and socio-economic data from 13 villages in the lower basin of the Senegal River to better understand the relative contribution of exposure, hazard, and vulnerability to the burden of urogenital schistosomiasis (caused by Schistosoma haematobium) in an endemic region of northern Senegal. In this setting, the construction of a dam and irrigation infrastructure has reduced the region’s susceptibility to drought and famine, while simultaneously facilitating schistosome transmission. Prior to dam construction, the transmission of S. haematobium occurred at low levels on a seasonal basis, but the disease has since become hyperendemic with perennially high prevalence and intensity of infection [31-33]. Dam construction also coincided with the introduction and rapid spread of S. mansoni, an intestinal schistosome, which remains co-endemic with S. haematobium [34, 35]. Simultaneously, agricultural livelihoods have shifted from traditional flood recession and rainfed cultivation to more intensively irrigated crops [36]. The social and environmental changes in this setting have collectively affected exposure, hazard, and vulnerability to schistosome infection.

Using these data from this setting, we compared the relative contributions of exposure, hazard and vulnerability as explanatory variables of both presence and intensity of S. haematobium infection. We then quantified the magnitude and precision of the association between infection outcomes and indices of each component of risk as well as demographic characteristics, whose importance has been demonstrated in this and other settings [37-39]. Using a mix of survey, interview, and ecological data to assess the relative contributions of exposure, hazard, and vulnerability to schistosomiasis risk, we identify the circumstances under which ecological, infrastructural, and behavioral interventions can complement mass drug administration in campaigns to control and eliminate schistosomiasis.

Methods

Ethics statement

This study received approval from the National Committee of Ethics for Health Research from the Republic of Senegal (Protocol #SEN14/33) as well as the Institutional Review Boards of the University of California, Santa Barbara (Protocol #19-17-0676) and Stanford University (Protocol #43130). Children were enrolled in the parasitology study after providing assent alongside written informed consent from the child’s parent or guardian. All interview and survey respondents provided verbal informed consent.

Study area and population

This study used cross-sectional, mixed methods data collected in 2017 and 2018 as part of a longitudinal study of schistosome infection in school-aged (5–15 years) children. Primary data collection included parasitological examinations of urine samples screening for S. haematobium infections, survey data from the residents and heads of households where those children reside, and interview and ecological data from water access sites in their corresponding villages. Sixteen villages along the lower Senegal River, its tributaries, and the Lac de Guiers in northwest Senegal were selected for participation in the longitudinal study; these villages were representative of the rural, high-transmission sites common in the region (see appendix of [33] for details on village selection; Fig 1). Briefly, the criteria used to select villages included proximity to freshwater, presence of water access sites, presence of a school with sufficient enrollment in target primary school grades, a non-zero prevalence of self-reported infection, and accessibility in the rainy season [33]. Of these 16 study villages, six were located along the Senegal River and its tributaries while ten were located on the shores of Lac de Guiers. Three of these 16 villages (one on the river and two on the lake) were excluded from this analysis because a vegetation removal intervention took place in these sites in 2017, potentially influencing hazard indices (Fig 1)

Fig 1

The lower basin of the Senegal River, where parasitological, socio-economic, and ecological data were collected from 821 school-aged children in 411 households and 24 water access points in 13 villages.

Three villages (shown in white circles) were excluded from this analysis because they were sites of a vegetation removal intervention in 2017 that may have affected hazard indices.

School-aged children were recruited from grades 1–3 in each village school, and each student provided urine samples for determining S. haematobium infection presence and intensity in January–March 2018. In July 2018, we attempted to reach all the households where students enrolled in the school-based parasitology study lived and administer a questionnaire about household demographics and water contact behavior.

Data collection

Parasitology data collection

Through schools in all 16 villages, a total of 1,480 school-aged children was enrolled at baseline in February–April 2016. Of those, 1,301 remained enrolled during the third year of follow-up in January–March 2018 and successfully produced one urine sample on each of two testing days. After excluding children living in the three villages where the vegetation removal intervention occurred in 2017, parasitological data were available for 821 children in 13 villages (S1 Fig). At baseline and the subsequent two years, urine samples were analyzed by urine filtration for S. haematobium infection [40]. Following sample collection, all children were treated at school (with parental consent) with two rounds of praziquantel administration at 40 mg/kg. Because the cross-sectional data used in this study were embedded in a large longitudinal study in the same villages, infection data collected in January–March 2018, which were used as outcome variables in this study, reflect re-infection following treatment over the course of the preceding year [41-43].

Household survey data collection

Household-level survey data were collected at the beginning of the rainy season in July 2018, following parasitological data collection for that year. In all 16 villages, we aimed to reach all households with children enrolled in the school-based parasitology study. Of the 1,301 students enrolled in the parasitology study, we reached the households (n = 524) of 1,216 students (93.5%; S1 Fig). The household survey instrument included six main modules (S1 Table). The modules used in this analysis included: (1) individual-level demographic information, (2) individual-level, two-week recall of contact with surface water for six primary domestic and occupational chores requiring contact with water, (3) average frequency of each water contact activity reported at the household level, (4) socio-economic indicators, such as reported access to improved water and sanitation infrastructure and (5) geographic coordinates of household locations. The collection of water contact data in the survey was designed to elicit information on water contact behavior over the course of a typical week.

Body surface area interview data collection

In 15 villages, brief interviews were conducted with village residents, usually at village water access sites in August 2016. In each interview, respondents were asked how much of the body typically comes into contact with water for each of six common water contact activities: (1) washing laundry, (2) washing dishes, (3) collecting water, (4) irrigating fields, (5) washing or watering livestock and (6) fishing from shore [39]. Responses for each activity were registered on a diagram used to measure burn size [44]. Two interviews were conducted in each village–one with an adult man and one with an adult woman–for a total of 30 interviews and 210 activity-specific observations (30 interviews x 7 water contact activities).

Ecological data collection

Snail surveys were performed at 32 water access sites in all 16 study villages, the details of which are described elsewhere [33]. Briefly, randomized snail sampling was stratified by three microhabitat types (emergent vegetation, non-emergent vegetation, and open water/mud bottom), such that sampling effort within a site was proportionate to the relative area of each microhabitat. Imagery from Google Earth and an unmanned aerial vehicle (drone) was used to identify site boundaries and designate fifteen random sampling points within each microhabitat. At each point, exhaustive sampling for snails occurred within a quadrat (76.2 cm x 48.26 cm x 48.26 cm; area = 0.3677 square meters). The snails collected from each quadrat were placed into labeled vials and returned to the laboratory for identification and infection screening via shedding and dissection assays. Molecular identification of schistosome cercariae shed from captured snails in the laboratory distinguished human-infecting schistosome species from species that affect non-human animals. Data from the 24 water access sites that were not part of the vegetation removal intervention were included in the analysis.

Data availability

A village-level summary of all variables used in this analysis is available online [45].

Data analysis

Infection metrics

Infections with S. haematobium were determined to be present when any of the samples collected from an individual student contained at least one S. haematobium egg. Infection intensity was quantified by taking the median egg count (per 10 mL urine) across the samples collected from each student at both parasitological sampling visits. Medians were taken because the distributions of infection intensity were strongly right-skewed and included a large number of zeroes, reflecting a negative binomial distribution.

Exposure indices

Exposure indices were calculated using data from three sources: (1) individual- and household-level survey data, (2) body surface area interviews and (3) a previous study of water contact behavior in a similar setting [39]. These indices aggregated reported water contact across the six water contact activities described above, including in the body surface area interview form and the water contact module in the household survey. Whether or not a person performed a given activity in the preceding two weeks was recorded as binary individual-level variables in the household survey. The average frequency of each activity was reported at the household level in the survey, while duration and extent of contact were drawn from literature-based estimates and body surface area interviews, respectively (S2 Table). Exposure indices based on these activity-specific data incorporated varying degrees of complexity, including weekly frequency-only estimates, weekly time exposed and weekly time exposed adjusted for mean body surface area exposure.

Two indices measured frequency of contact with water. One index [F (raw)] was a categorical variable derived directly from a single survey question, where each child was asked to recall the number of visits made to a water access point in the preceding seven days. The second [F (sum)] was a numeric variable derived from activity-specific measures collected through the survey at the individual and household levels. Individual-level water contact data indicated whether each member of the household had performed each activity in surface water (e.g., lake, river, irrigation canals) in the preceding two weeks. Dichotomous responses were registered for all members of the household. Response categories for frequency of each water contact activity were collected at the household level (e.g., one value per household) and converted to numeric values representing times per week (e.g., a reported frequency of once per day was converted to a numeric value of 7 times per week). The numeric values of household-level activity frequency were multiplied by indicators of individual-level activity to generate estimated weekly contacts by activity for all children in the data set:

Total time exposed per week (FD) was calculated by multiplying the weekly frequency of an activity by the estimated duration of that activity, as calculated from direct observations in another study in a similar setting [39]. The product of frequency and duration was then summed across all activities for each child to generate an estimated total time (in minutes) of contact with water over the course of a week.

Total exposure time per week was then adjusted by estimated body surface area in contact with water for a given activity (FDB). We used data from body surface area interviews (S2 Table) to estimate how much of the body came into contact with water for each activity. For each observation, the percent of body surface area (%BSA) was summed for each body segment reported as exposed for that activity using published values for burn size estimation [44]. Mean %BSA for each activity was calculated across all interview respondents. Activity-specific estimates of body surface area exposed (in square meters) were calculated by multiplying mean %BSA by an estimated total body surface area for children under 14 years of age [46]. Estimated body surface area exposure required for each activity was multiplied by the total time exposed for that activity, giving a BSA-adjusted time per week.

Hazard indices

Hazard indices were calculated from ecological field data collected at the 24 water access points in the same 13 villages where parasitology and household survey data were collected. A previous study in this setting found that area of suitable snail habitat (e.g., area covered by non-emergent aquatic vegetation) was the best environmental predictor of presence and intensity of S. haematobium infection [33]. We developed six hazard indices based on area of snail habitat, which varied by (1) season, (2) spatial relationship and (3) level of data aggregation. Two hazard variables captured the area of suitable snail habitat (1) for the peak transmission season preceding parasitological data collection for each water access point (July–August 2017; peak) and (2) summed across the three ecological data collection periods in the year preceding parasitological data collection (May–June 2017, July–August 2017 and January 2018; year). Summing habitat area across the year represents an integrated hazard metric across the three dominant climatic seasons in the region. We primarily used data from the water access site nearest to each household, based on straight-line distance, but two additional hazard variables captured the spatial relationship between households and the nearest water access sites by scaling peak and year hazard indices by the distance between a child’s household and the nearest water access site. Two final hazard variables aggregated peak and year indices to the village level, where the area of suitable snail habitat during the transmission peak and the entire year were summed across all water points in a village.

Vulnerability indices

Vulnerability indices were calculated from the household survey data and focused on access (or lack thereof) to water and sanitation infrastructure at the household level (S1 Table). Two survey questions concerned water sources: (1) what is the principal source of water used by members of the household for drinking? and (2) what is the principal source of water used for doing laundry? Responses to these two questions were used to calculate the extent to which each household depended on surface water for household needs (e.g., surface water is used for neither drinking water nor laundry; surface water is used for either drinking water or laundry; or, surface water is used for both drinking water and laundry). This variable was also dichotomized (e.g., a household used surface water for any activity versus no activities).

Two additional survey questions concerned sanitation infrastructure: (1) what type of toilet is primarily used by members of this household? and (2) is this toilet shared with other households? Responses to these two questions were used to calculate a three-level variable describing sanitation access (e.g., private toilet, shared toilet, no toilet). A dichotomous version of this variable was also calculated (e.g., private toilet versus shared or no toilet). This dichotomous categorization was chosen because few households reported having no toilet.

A final index of vulnerability was an index of socio-economic status (SES) derived from data on durable assets and household conditions reported in the household survey [47, 48]. Principal components analysis was performed using the prcomp() function in the stats package (version 3.5.1) in R [49]. Loadings for individual variables in the first principal component were used to calculate a numeric socio-economic score for each household. Numeric scores were then divided into quintiles.

Statistical analysis

We used mixed-effects logistic and negative binomial regression models to assess exposure, hazard, and vulnerability indices as explanatory variables of S. haematobium infection presence (i.e., infected versus uninfected) and intensity (i.e., number of eggs detected per 10mL urine), respectively. In all models, we controlled for key social and demographic covariates: age and sex of the child, village location (i.e., river versus lake) and village population size. We also included random intercepts (households nested within villages) to account for the hierarchy in the data. Mixed effects logistic regression was performed using the lme4 package (version 1.1–21) [50], while mixed effects negative binomial regression was performed using the glmmTMB package (version 1.0.1) [51] in R (version 3.5.1) [49].

We used information-criterion (IC)-based model selection and multi-model inference to determine the relative importance of each component of risk [52]. For each outcome, we first fit a set of exposure-only, hazard-only, and vulnerability-only models. The Akaike Information Criterion (AIC) was used to discern the best-fitting index (or indices) of each component of risk. We retained all variables in exposure-only, hazard-only and vulnerability-only models whose ΔAIC was less than 2 across the models in a given group [52]. This subset of variables was then used to fit models that included various combinations of the three components of risk: (1) exposure and hazard (E-H), (2) exposure and vulnerability (E-V), (3) hazard and vulnerability (H-V) and (4) all three: exposure, hazard, and vulnerability (E-H-V). Because fewer indices of exposure, hazard and vulnerability were used in these combination models of infection presence (based on AIC for exposure-, hazard-, and vulnerability- only models), we fit a set of 28 candidate models with the presence of S. haematobium as the outcome and 47 candidate models with intensity of S. haematobium re-infection as the outcome. Regression diagnostics indicated no problem with multicollinearity in models combining indices of exposure, hazard, and vulnerability (S2 Fig).

For both sets of models, we calculated AIC, ΔAIC and Akaike weights (wi) [52]. The sum of Akaike weights across the exposure, hazard and vulnerability components of risk was used to quantitatively assess their relative importance across the full set of candidate models for both presence and intensity outcomes [52]. Additionally, for each index of exposure, hazard and vulnerability, we calculated model-averaged estimates of the magnitude and precision of the association with both infection outcomes using the AICcmodavg package in R [53]. Odds ratios (OR) and rate ratios (RR) with 95% confidence intervals (95% CIs) were estimated for presence and intensity outcomes, respectively.

Results

Characteristics of the study population

Of the 821 school-aged children with complete information across the parasitological, survey and ecological data sets, the mean age was 9.3 years and approximately half (414/821; 50.4%) were male (Table 1). On average, these children lived within 521 meters of a water access point on the Senegal River, a tributary, or the Lac de Guiers, which contained between zero and 149,775 square meters of suitable snail habitat in the peak transmission season. Approximately half of the households where these children lived relied on surface water collected from water access points for at least one household need, and most households (82.6%) reported primarily using private sanitation infrastructure (Table 1).

Table 1

Demographic characteristics of the overall study population, and stratified by sex.

Variable	Overall	Males	Females
Observations [n]	821	414	407
Age (years) [mean (SD)]	9.3 (3.1)	9.2 (2.2)	9.3 (3.8)
Distance to water point (m) [mean (SD)]	521.6 (505.3)	522.2 (530.6)	521.0 (475.5)
Water point habitat area (m2) [mean (range)]	38,857 (0–149775)	38,898 (0–100,361)	38,815 (0–149,775)
Surface water dependence [n (%)]
Neither drinking nor laundry	400 (48.7)	211 (51.0)	189 (46.4)
Either drinking or laundry	265 (32.3)	129 (31.2)	136 (33.4)
Both drinking and laundry	156 (19.0)	74 (17.9)	82 (20.1)
Sanitation facility [n (%)]
Private toilet	679 (82.7)	342 (82.6)	337 (82.8)
Shared toilet	134 (16.3)	68 (16.4)	66 (16.2)
No toilet	8 (0.9)	4 (0.9)	4 (1.0)
Asset-based wealth index [n (%)]
1st quintile (Lowest)	133 (16.2)	76 (18.4)	57 (14.5)
2nd quintile	116 (14.1)	57 (13.7)	59 (14.5)
3rd quintile	150 (18.3)	82 (19.8)	68 (16.7)
4th quintile	201 (24.5)	100 (24.2)	101 (24.8)
5th quintile (highest)	221 (26.9)	99 (23.9)	122 (30.0)

The burden of S. haematobium among this study population was high. The overall prevalence was 66.7% and was slightly higher among boys (72.9%) than girls (60.4%) (Table 2). The geometric mean of S. haematobium infection intensity was 4.9 eggs per 10 mL urine (SD = 117.4) overall, 6.4 eggs per 10 mL urine (SD = 66.8) for boys and 3.7 eggs per 10 mL urine (SD = 152.0) for girls (Table 2).

Table 2

Descriptive analysis of outcome variables (presence and intensity of S. haematobium infection) in the overall study population and stratified by sex.

Variable	Overall	Males	Females
Observations [n]	821	414	407
Infection presence [infected n (%)]	548 (66.7)	302 (72.9)	246 (60.4)
Infection intensity [GM (SD)]	4.9 (117.4)	6.4 (66.8)	3.7 (152.0)

GM = geometric mean eggs/10mL urine

When asked about contact with surface water generally, more than one-third (39.1%) of school-aged children reported no visits to village water access sites in the preceding seven days. Approximately one-quarter (26.7%) visited once or twice a week and the remainder (34.3%) reported more frequent visits (Table 3). When prompted to recall activity-specific contact with surface water (e.g., doing laundry or washing livestock), the average sum of total weekly contacts across all activities was much higher (mean = 8.5 times per week, SD = 10.9). The weekly frequency of some of these activities (e.g., bathing and dish washing) exceeded that reported from the single survey question prompting seven-day recall of all visits to the water access point (Table 3).

Table 3

Descriptive analysis of exposure indices calculated from individual-level responses to water contact questions in the household survey, overall and stratified by sex.

Variable	Overall	Males	Females
Categorical frequency [n (%)]
No visits per week	321 (39.1)	145 (33.0)	176 (43.2)
1–2 visits per week	219 (26.7)	121 (29.2)	98 (24.1)
3–6 visits per week	68 (8.3)	33 (8.0)	35 (8.6)
7 visits per week	173 (21.1)	95 (22.9)	78 (19.2)
7+ visits per week	40 (4.9)	20 (4.8)	20 (4.9)
Weekly activity-specific frequency [mean (SD)]
Bathing	4.7 (5.2)	4.6 (5.1)	4.8 (5.3)
Water collection	1.4 (3.4)	1.0 (2.9)	1.8 (3.9)
Dishes	1.2 (3.3)	0.1 (1.0)	2.2 (4.3)
Laundry	0.7 (1.9)	0.1 (0.7)	1.3 (2.5)
Livestock	0.3 (1.5)	0.5 (1.9)	0.1 (0.9)
Irrigation	0.2 (1.2)	0.2 (1.2)	0.1 (1.3)
Fishing	0.1 (0.8)	0.1 (0.8)	0.03 (0.7)
Numeric frequency [mean (SD)]	8.5 (10.9)	6.6 (7.5)	10.4 (13.2)
Total weekly time exposed [mean (SD)]	87.3 (111.4)	66.0 (74.6)	109.0 (135.9)
BSA-adjusted weekly time exposed [mean (SD)]	74.5 (83.4)	65.6 (71.7)	83.5 (93.0)

Bathing was assumed to involve contact with the entire body surface, while shore fishing (47.6%) and water collection (45.3%) involved the next greatest body surface area in contact with water (S2 Table). Irrigation (22.5%) was the activity with the least body surface area exposed (S2 Table). Using the reported duration of activity-specific water contact, we estimated that the school-aged children in this setting spend, on average, 87.3 minutes (SD = 111.4) in contact with surface water per week, an estimate that is higher for girls (mean = 109.0 minutes, SD = 135.9) than boys (mean = 66.0 minutes, SD = 74.6). Adjusting weekly total duration of contact by total body surface exposed during a specific activity reduced this overall estimate slightly to 74.5 m2-minutes, with similar variation between boys (mean = 65.6 m2-minutes, SD = 71.7) and girls (mean = 83.5 m2-minutes, SD = 93.0).

While survey data indicate that dependence on surface water is not evenly distributed across the quintiles of the asset-based SES index (Kruskal-Wallis test, Χ2 = 14.0, df = 4, P-value = 0.007), we find that those who do depend on surface water for at least one household task are relatively evenly distributed across SES quintiles (S3 Table). For example, almost a third of the households (29.8%) that use surface water for either drinking or laundry are in the highest SES quintile. We find a similar pattern for dependence on surface water for both drinking and laundry: almost a quarter of households (21.8%) in the highest SES quintile rely on surface water for both chores, which is identical to the proportion of households in the lowest SES quintile with the same reliance on surface water (S3 Table).

Best-fitting indices of exposure, hazard, and vulnerability

Four exposure indices, six hazard indices, and five vulnerability indices were considered in a set of mixed-effects logistic regression models of S. haematobium infection presence (n = 28) and a set of mixed-effects negative binomial regression models of S. haematobium infection intensity (n = 47; Table 4). The full comparison of model fit, including the Akaike weight (wi) for each individual model, is summarized in S4 Table for presence models and in S5 Table for intensity models. Data in Table 4 summarize each variable used to approximate exposure, hazard, or vulnerability as well as the number of models that variable appeared in for each outcome. For infection presence, all four exposure-only models were within a ΔAIC of 2. For infection intensity, the categorical frequency of water point visits (F (raw)) outperformed all three numeric indices of exposure (numeric frequency [F (sum)], total time exposed [FD] and BSA-adjusted time exposed [FDB]). The exposure-only model of infection intensity containing the F (raw) index had the lowest AIC of the four exposure-only models, with the ΔAIC for the next best-fitting model of infection intensity far exceeding the threshold of 2 (ΔAIC = 16.52; S5 Table). As a result, we retained all four indices of exposure for models of infection presence, each of which appeared in a total of four of the 28 (14.2%) presence models. In contrast, the categorical frequency of water access site visits was the only index of exposure retained in subsequent models of infection intensity, appearing in a total of 20 of 47 (43.5%) models in that set (Table 4).

Table 4

Summary of the indices of exposure, hazard and vulnerability that were compared in a set of 28 mixed effects logistic regression models of S. haematobium infection presence and a set of 47 mixed effects negative binomial regression models of S. haematobium infection intensity and the number (%) of models in each set in which each variable appeared.

	Variable	Description	Models (n (%))
			Presence	Intensity
Exposure	F (raw)	Categorical frequency of visits to water access point in previous seven days, derived from single survey question	4 (14.2)	20 (43.5)
	F (sum)	Numeric frequency of weekly contacts with water, based on individual-level binary indicators for seven common water contact activities and average household-level frequencies, both of which were collected through the household survey	4 (14.2)	1 (2.2)
	FD	Weekly time exposed (in minutes) derived from activity-specific survey data, as in F (sum), combined with literature-derived data on duration of exposure [39]	4 (14.2)	1 (2.2)
	FDB	Weekly time exposed adjusted for mean body surface area exposed for each of seven common water contact activities, data estimated from body surface area interviews	4 (14.2)	1 (2.2)
Hazard	areaPeak	Area of snail habitat at the water access site nearest to household during peak season	1 (3.6)	8 (17.4)
	areaYear	Area of snail habitat at the water access site nearest to household summed over year	1 (3.6)	8 (17.4)
	areaPeak_d	Area of snail habitat at the water access site nearest to household during the peak season, adjusted by the distance between the site and the household	1 (3.6)	8 (17.4)
	areaYear_d	Area of snail habitat at the water access site nearest to child’s household summed over year, adjusted by the distance between the site and the household	10 (35.7)	8 (17.4)
	areaPeakV	Area of snail habitat summed across all water points in a village during the peak season	1 (3.6)	1 (2.2)
	areaYearV	Area of snail habitat summed across all water points in a village and across year	1 (3.6)	1 (2.2)
Vulnerability	surface	Dependence on surface water for household needs derived from survey data: (0) neither laundry or drinking, (1) either laundry or drinking or (2) both laundry and drinking	10 (35.7)	1 (2.2)
	surfaceYN	Dichotomous version of surface water dependence for any versus no activities	1 (3.6)	10 (21.7)
	sanitation	Primary sanitation infrastructure used by members of a household: (0) none, (1) shared toilet, (2) private toilet	1 (3.6)	10 (21.7)
	privateSan	Dichotomous version of sanitation, where members of a household use either a private toilet or shared/no toilet	1 (14.2)	10 (21.7)
	assetIndex	Quintiles of an asset-based SES index created using principal components analysis	1 (3.6)	1 (2.2)

A single index of hazard–the area of snail habitat at the water access site nearest to a child’s household summed over the three climatic seasons of ecological data collection, adjusted by the distance between the site and the household (areaYear_d)–had a ΔAIC ≤ 2 compared to other hazard-only models of S. haematobium presence (S4 Table). As a result, areaYear_d was the only index of hazard to appear in more than one model of S. haematobium presence (Table 4). In contrast, five indices of hazard had a ΔAIC ≤ 2 compared to other hazard-only models of S. haematobium intensity (S5 Table). These variables included the area of snail habitat at the water access site nearest to a child’s household during the peak transmission season (areaPeak) as well as area summed over the entire year (areaYear). The distance-adjusted analogs of these variables (areaPeak_d and areaYear_d) were also retained as was the area of snail habitat summed across all water points in a village and across the full year (areaYearV). These five indices of hazard were retained in subsequent models that combined hazard with other components of risk, such that all five appear in multiple models of S. haematobium intensity (Table 4).

One index of vulnerability–categorical dependence on surface water (surface)–best fit for the S. haematobium presence data (e.g., ΔAIC ≤ 2 among vulnerability-only models; S4 Table). This variable was retained for subsequent models that combined components of risk and appeared in ten (35.7%) of the 28 presence models (Table 4). In contrast, binary dependence on surface water (surfaceYN) as well as categorical and binary variables for sanitation infrastructure (sanitation and privateSan, respectively) all had ΔAIC ≤ 2 among vulnerability-only models of S.haematobium intensity (S4 Table). Each appeared in ten (21.7%) of the 47 intensity models (Table 4).

Relative importance of exposure, hazard, and vulnerability

Across these sets of candidate models, no single model emerged as the best-fitting model of presence or intensity of S. haematobium re-infection (S4 and S5 Tables). The largest wi for mixed effects logistic regression models of S. haematobium presence was 0.28. A total of five models were classified as having substantial support (ΔAIC < 2) [52]. Four of these five models included all three components of risk while the fifth (and best-fitting model) contained indices of hazard and vulnerability (Table 5)

Table 5

Inclusion of indices of exposure (E), hazard (H) and vulnerability (V) in models of S. haematobium presence and intensity models with substantial support (ΔAIC <2).

No.	E	H	V	AIC	ΔAIC	wi	Ʃ wi
Presence models with substantial support (from S4 Table)
24		X	X	837.37	0	0.28	0.28
26	X	X	X	838.32	0.96	0.18	0.46
27	X	X	X	838.57	1.21	0.16	0.62
28	X	X	X	838.89	1.52	0.13	0.75
25	X	X	X	839.21	1.84	0.11	0.86
Intensity models with substantial support (from S5 Table)
1	X			5494.29	0	0.08	0.08
16	X	X		5494.41	0.12	0.08	0.15
19	X	X		5494.46	0.17	0.08	0.22
42	X	X	X	5494.46	0.17	0.07	0.30
22	X		X	5494.49	0.20	0.07	0.37
18	X	X		5494.86	0.57	0.06	0.42
43	X	X	X	5494.86	0.57	0.06	0.48
40	X	X	X	5494.95	0.66	0.06	0.54
36	X	X	X	5494.98	0.68	0.06	0.59
21	X		X	5495.21	0.92	0.05	0.64
17	X	X		5495.30	1.01	0.05	0.69
20	X	X		5495.50	1.21	0.04	0.73
41	X	X	X	5495.70	1.41	0.04	0.77
38	X	X	X	5495.81	1.52	0.04	0.81
37	X	X	X	5495.95	1.66	0.03	0.84
39	X	X	X	5496.16	1.87	0.03	0.87
46	X	X	X	5496.19	1.90	0.03	0.90
23	X		X	5496.23	1.94	0.03	0.93

X indicates whether a variable from a given component of risk was included in the model; AIC = Akaike information criterion; wi = Akaike weight; Ʃ wi = sum of Akaike weights across models in a set

The largest wi for mixed-effects negative binomial regression models of S. haematobium intensity was 0.08, with 18 models classified as having substantial support [52]. Half (9 /18) of the intensity models with substantial support included indices of all three components of risk. An index of exposure was included in all the best models of re-infection intensity, including an exposure-only model which had the lowest AIC of all models of re-infection intensity (Table 5).

We used the sum of Akaike weights (Ʃwi) across each set of candidate models for each outcome to quantify the relative importance of each component of risk (Table 6). We found that hazard was the most important component of risk for re-infection presence in school-aged children (Ʃwi = 0.95). The relative importance of hazard for S. haematobium presence was followed by vulnerability (Ʃwi = 0.91) and then exposure (Ʃwi = 0.66; Table 6). The relative importance of each component of risk differed for models of S. haematobium re-infection intensity, where exposure (measured as categorical frequency of contact with surface water) was the most important metric (Ʃwi = 1.00), followed by hazard (Ʃwi = 0.77) and then vulnerability (Ʃwi = 0.63).

Table 6

Sum of Akaike weights (Ʃwi) as a measure of relative importance of the three components of risk for S. haematobium infection presence and intensity.

	Ʃwi
	Exposure	Hazard	Vulnerability
Infection presence	0.66	0.95	0.91
Infection intensity	1.00	0.77	0.63

Model-averaged estimates

We used model averaging to calculate the magnitude and precision of the association between both infection outcomes and all indices of exposure, hazard and vulnerability that appeared in more than one model across the full set of candidate models and the demographic control variables used in each model (Fig 2). For the infection presence outcome, we found that the odds of infection increased with distance-adjusted area of snail habitat at the nearest water point to a child’s household summed over the year preceding infection data collection. The association corresponds with a 45% increase in the odds of infection with each square meter of snail habitat over the course of an entire year, with a confidence interval contained entirely above the null value of 1 (Distance-adjusted habitat area over year; OR = 1.49, 95% CI 1.12, 1.97).

Fig 2

Model-averaged point estimates and 95% confidence intervals for association between infection outcomes and each component of risk.

Top: Associations between S. haematobium infection presence and components of risk are estimated by logistic regression and measured with odds ratios. Bottom: Association between S. haematobium infection intensity and components of risk are estimated by negative binomial regression and measured with rate ratios. In both panels, indices of exposure (E) are represented by black triangles, indices of hazard (H) by turquoise circles, indices of vulnerability (V) by blue squares and demographic control variables (D) by orange diamonds.

Associations between infection presence and indices of vulnerability were less precise, with many confidence intervals including the null value of 1 (Fig 2). When comparing children whose households relied on surface water for multiple household tasks to those who relied on surface water for no household tasks, we found an almost nine-fold increase in the odds of infection with a confidence interval entirely above the null value of 1 (Categorical surface water use [2 activities vs 0]; OR = 8.67, 95% CI 1.67, 45.60). A similar association comparing households’ use of surface water for a single activity was also positive, but its confidence interval crossed the null (Categorical surface water use [1 activity vs 0]; OR = 1.34, 95% CI 0.84, 2.14).

All levels of the categorical exposure index suggest a positive association with infection presence (categorical frequency [F (raw)]), but only one level (Categorical frequency [3–6 times/week vs none]) had a confidence interval that did not include the null value of 1 (OR = 2.18, 95% CI 1.03, 4.66). Remaining levels of categorical exposure (1–2 times/week vs none, 7 times/week vs none and >7 times/week versus none) cannot be considered statistically different from the null (Fig 2). All three numeric indices of exposure had point estimates slightly below 1 (e.g., 0.9) with confidence intervals that included in the null value (Fig 2).

Two of the four demographic control variables had significant associations with infection presence (Fig 2). Children living in lake villages were five times more likely to be infected compared to children living in river villages (Village location; OR = 5.37, 95% CI 1.14, 25.28). The odds of infection were 54% lower among girls compared to boys (Child sex; OR = 0.46, 95% CI 0.33, 0.63) and younger children were less likely to be infected than older children (Child age; OR = 0.78, 95% CI 0.63, 0.96). Village population size was not associated with infection presence (Village population; OR = 1.16, 95% CI 0.64, 2.10).

For the infection intensity outcome, we found that the median number of eggs detected in urine samples increased for all levels of exposure (Categorical frequency [F (raw)]; Fig 2). These point estimates ranged from a 2.59- to 3.86-fold increase in egg burden among children reporting any exposure compared to those reporting no exposure. Confidence intervals for all estimates were entirely above the null value of 1 (95% CIs: [1-2x/week vs none] 1.52, 4.44, [3-6x/week vs none] 1.88, 8.00, [7x/week vs none] 1.55, 5.70, [>7x/week vs none 1.22, 7.77; Fig 2). The directionality of the associations between infection intensity and indices of vulnerability are like those for infection presence (e.g., above 1 for surface water use, below 1 for sanitation infrastructure), but with point estimates closer to the null and wider confidence intervals that include the null value of 1 (Fig 2).

The measures of association between infection intensity and hazard are mixed: distance-scaled and village-level indices have point estimates less than 1 while raw values of habitat area (i.e., not scaled by distance) for the nearest water point to a child’s home are greater than 1. The confidence intervals for all associations between S. haematobium infection intensity and hazard indices include the null value of 1. Associations between infection intensity and demographic control variables were similar to the associations described for these variables with presence outcomes (Fig 2).

Discussion

We found that components of risk describing exposure, hazard, and vulnerability all had some importance in explaining the occurrence of S. haematobium in school-aged children in this endemic setting in the lower Senegal River Basin. Based on the sums of Akaike weights calculated across both sets of models, however, hazard and vulnerability were the most important components of S. haematobium infection presence, while exposure and hazard were the most important components of S. haematobium infection intensity. These differences highlight the distinct factors that influence whether someone becomes infected versus whether they accumulate a large worm burden, with implications for the design of schistosomiasis control and elimination programs and guidelines.

The importance of hazard and vulnerability to the infection presence outcome reflects the conditions that influence whether a person acquires an infection or not. Any exposure may lead to infection, making additional exposure irrelevant to this binary outcome once a person becomes infected. But hazard in the environment, or the number of infected snails emitting cercariae from a certain amount of available habitat, is more likely to increase the risk of infection so long as some exposure is happening. Vulnerability through the household activities that involve regular water contact, like dependence on surface water, for example, ensures some baseline level of exposure.

Similarly, the importance of exposure and hazard to infection intensity reflects the conditions that influence the accumulation of worms. The intensity of infection is theoretically proportional to exposure, with each new adult worm the result of a separate infection event [54]. While this has not always borne out empirically [18, 39], our data agree with the theoretical expectation that more exposure leads to higher intensity infection. Such accumulation of worms is important because it can lead to severe morbidity. A study in Uganda found the prevalence of liver fibrosis to be five times greater in people who had lived in the S. mansoni-endemic area since birth compared to those with just a 10-year residency, leading the authors to conclude that the duration of exposure over time is an important risk factor for developing the severe morbidity resulting from prolonged high-intensity infections [55]. Another study in China found that water contact itself was not associated with infection intensity but, similar to our findings, accounting for both water contact and a measure of cercarial risk did [19].

While the recognition that schistosomiasis risk is determined by the convergence of social (exposure and vulnerability) and environmental factors (hazard) is not new [56-59], few studies comprehensively address all three components of risk for schistosomiasis specifically or for infectious diseases generally. Studies employing mathematical models have jointly addressed social and environmental determinants of schistosomiasis [60-62], while empirical studies often focus on one [12, 13, 63] or the other [16, 64, 65]. Our results allow us to think more concretely about how to choose interventions and thereby target schistosomiasis control. Reducing exposure through behavior change interventions that minimize contact with water is likely to contribute more to morbidity control (e.g., reductions in infection intensity) than transmission control (e.g., reductions in infection presence). Reducing vulnerability through the provision and use of piped water infrastructure will give people an alternative to surface water, allowing them to avoid risky exposures and thus, lower infection intensity. Reducing hazard in the environment through snail control interventions will be essential to reducing transmission and the likelihood of acquiring and accumulating infections from the environment. Understanding the interplay between different components of risk in a specific setting can facilitate the design of effective and efficient context-specific control strategies.

In this study, we used primary data sets collected by three distinct teams under the same project in the same study villages: (1) collection, processing and diagnosis of S. haematobium infection from urine samples collected from school-aged children on a yearly basis with associated treatment with praziquantel, (2) exhaustive sampling of snail populations and their habitats in the water access sites in all villages, and (3) socio-economic surveys in all the households where school-aged children in the parasitology study lived. All three data streams, combined with data on the spatial relationship between water access sites and surveyed households, were important for explaining patterns of infection in children. Our results underscore the importance of interdisciplinary and mixed methods research when addressing complex socio-environmental problems like the transmission of schistosomiasis and other infectious diseases.

Measures of exposure, hazard, and vulnerability

We learned how different measures of exposure, hazard and vulnerability fit the infection data. The simplest exposure index, which used seven-day recall of visits to a surface freshwater access point, fit infection intensity data better than did more complex exposure metrics that included information on the reported frequency, duration and extent of contact experienced for specific activities. There are several potential explanations for this finding. This simple index was derived from a single survey question that was posed specifically for the children enrolled in the parasitology study. In contrast, the activity-specific estimates were derived from a survey module that prompted individual-level information from all members of that household. Because of the volume of information that respondents were asked to provide (e.g., whether a person performed each of six water contact activities for all members of the household), the accuracy and precision of the activity-specific responses may have suffered.

Other studies have found weak or no association with exposure measures that account for multiple aspects of exposure. One study in a similar setting in northern Senegal found all of their exposure indices to be correlated with each other but not with the presence or intensity of S. mansoni infection [18]. In the study from which we drew data on activity-specific duration of water contact, a relationship between water contact and intensity of S. mansoni infection was only detected after accounting for body surface area and the time of day of exposure [39]. In a study of S. haematobium infection in Zanzibar, a set of exposure indices that incorporated varying combinations of frequency, duration and body surface area were found to be positively associated with infection, but the strength of the association decreased with the complexity of the exposure index used [66]. Moreover, the distinct ecologies of schistosome species and their intermediate snail hosts and the methods used to measure water contact may confound the true relationship between water contact and infection.

Similarly, malacological hazard data often have weak or no association with infection. In one study in China, a single infected snail was found among 7,000 snails collected in a setting where prevalence of S. japonicum infection in people exceeded 25% [67]. Prevalence of human infection of 30–40% has been observed in lake regions of East Africa where snails have been found to occur in very low numbers [68, 69]. An ecological study of snail ecology in Senegal–the one from which ecological data in this study is drawn–found the area of snail habitat to be a superior predictor of human infection compared to seemingly more proximate predictors including density of infected snails [33]. The current study builds on these findings by matching individual-level infection to the water access site nearest to the household (in addition to village-level metrics used in the previous study) and accounting for the spatial relationship between the household and the water access site.

Given the shortcomings of previous studies on water contact and snail ecology, we found that accounting for the spatial relationship between people and the environmental hazard improved the fit of S. haematobium presence models. Hazard indices scaled by the distance to the nearest water point fitted the infection presence data better than the other hazard indices, though this was not the case for infection intensity data. Previous studies have found the proximity to water access points to be an important explanatory variable of schistosome infection in endemic settings [66, 70–73]. Another found that water contact was only related to re-infection after accounting for the spatial distance between people and the hazard in their environment [19]. Such proximity may be an important aspect of vulnerability that makes it difficult to avoid exposure to hazardous water.

Similarly, hazard indices that dealt with the site nearest to the child’s household fit the infection presence data better than village-level aggregations of snail habitat. This finding contradicts studies that found schistosomiasis to be associated with human behavior and snail ecology primarily at the community level [74-76]. However, because water contact activity for most people tends to occur at just one or two sites [75, 76] and since distance to water is commonly associated with infection [71, 72], it is possible that the distance-adjusted hazard indices capture the hazard where exposure is most likely to occur for people living in a given household.

Measures of association between components of risk and infection

We found that no single model best approximated either the presence or intensity of S. haematobium infection. Rather, a set of models comprised the evidence for our conclusions. Without a single best-approximating model, we used model averaging to generate point estimates and confidence intervals of the association between our two infection outcomes and all explanatory variables appearing in both model sets.

Associations between demographic control variables and infection were consistent across the two outcomes. These variables–particularly child sex and village location–had the strongest relationship with infection of all variables included in all models. This finding speaks to the importance of demographics in the dynamics of S. haematobium infection [77, 78] and the need for analyses that examine heterogeneity across sex and geographic strata [76, 79].

We found that exposure was associated with an increase in both infection presence and intensity, although the confidence interval for the association with infection presence included the null value. Consistent with our relative importance findings, these estimates support the notion that the amount of exposure influences the accumulation of worm burden more than the acquisition of infection. The substantial support we found for vulnerability as an explanatory variable of infection presence may approximate the presence (rather than amount) of exposure. Given the lifespan of adult schistosomes in the human host (3–10 years) [77, 80, 81], a single infection resulting from a single exposure may be present in the body for years until a person is treated.

Even though model-averaging found indices of vulnerability to have wide confidence intervals, there was suggestive evidence of a dose-response between the odds of infection and the number of activities for which a household uses surface water. Children living in households that relied on surface water for multiple activities were more likely to be infected than those whose households relied on surface water for a single activity. While this may be the result of a lack of alternative sources of water (such as wells or piped water), it is also important to note that water contact activities often have social value [82], making the use of surface water for social activities like laundry a matter of preference. It is also possible that available alternative sources of water may not be suitable for particular activities, making surface water preferable in some cases [65, 83].

The asset-based SES index was the poorest index of vulnerability (ΔAIC = 9.48 and 6.40 among vulnerability-only presence and intensity models, respectively), outperformed by indices that represented access to, and use of, water and sanitation infrastructure. It is possible that SES is a relatively poor predictor of infection because there is not much variability in ownership of durable assets across our study population. Similarly, while the household dependence on surface water is not evenly distributed across SES quintiles, we found similar proportions of households in the highest and lowest SES quintiles that use surface water for at least some household needs. It is easy to assume that dependence on surface water will be highly correlated with SES, but that may not be the case when decisions to continue using surface water are made for other reasons. Similarly, because our study focuses on children who go to school and play with each other, the influence of SES on infection outcomes may not be as pronounced as they would be among adults, for whom occupation and household factors are stronger determinants of exposure. Consequently, SES, as we measured it, may not be a strong determinant of water contact or S. haematobium infection.

Finally, our measures of hazard were improved by accounting for snail habitat in the water access site nearest to the household, and in the case of models of S. haematobium presence, the best metric accounted for distance between households and the nearest water point. These methodological insights highlight the important role of spatial relationships in schistosomiasis risk and the value of incorporating spatial data and analysis into future research.

Limitations

The limitations of this study are grounded in the challenge of measuring behavioral and ecological phenomena that are dynamic in both space and time. In using multiple methods to measure water contact behavior, specifically, we struck a balance between collecting complete information across the many facets of water contact behavior and collecting data from enough participants to fit statistical models. To do this, we supplemented individual-level information on regular water contact activities for all participants with household-level survey data on frequency of contact, literature-based estimates on duration of contact and interview data on body surface area exposure.

While incorporating frequency, duration and body surface area into exposure indices was meant to triangulate around the different dimensions of water contact, we note that the complex metrics were often outperformed by a simple metric of exposure derived from a single survey question. The poor fit of the complex exposure indices may reflect our use of aggregated data on frequency, duration and extent of exposure to the household and village levels, as they may obscure important individual-level heterogeneities [54, 84]. Ultimately, logistics limited our ability to capture water contact behavior comprehensively and precisely for hundreds of children across more than a dozen villages, each with its own complex landscape of water contact. Because we did not record individual episodes of water contact for each participant, we were also not able to determine the time of day when water contact occurred, a key criterion for exposure based on the chronobiology of cercarial release from snails [85, 86].

The issues implicit in the choice of methods for measuring water contact (e.g., recall and/or social desirability bias for questionnaire-based studies and failure to capture contact at other sites for studies based on direct observations at a single site) may be resolved in future studies by combining questionnaire-based or activity diary methods with the use of wearable GPS data loggers to collect more accurate and precise data on the spatial and temporal dynamics of water contact behavior. Such methods have been used in China [87], Uganda [82] and Cameroon [88] and have been found to be widely acceptable [89].

Finally, this study examines the convergence of the processes of exposure, hazard, and vulnerability for just one of the two schistosome species that circulate and cause disease in this setting. S. mansoni is co-endemic, circulating within different species of snail intermediate hosts with distinct habitat associations. The human behaviors associated with contamination of the environment also differ for S. mansoni compared to S. haematobium. For these reasons, the particularities of S. mansoni transmission may not be governed by the same relationships between exposure, hazard, and vulnerability.

Conclusion

In summary, we find that hazard, as measured by the area of snail habitat in the water access point nearest a household, and vulnerability, in terms of surface water use and access to sanitation at the household level, contribute most to the acquisition of S. haematobium infections. In contrast, exposure, approximated by a self-reported categorical frequency of contact with surface water per week, and hazard contribute most to the accumulation of S. haematobium parasites in school-age children, and thus to infection intensity. Together, our findings underscore the importance of all three components of risk, which act together across time and space to facilitate the acquisition and accumulation of S. haematobium infections in an endemic setting with high incidence of post-treatment re-infection. They highlight how interventions to complement MDA may be strategically deployed to reduce the risk of post-treatment re-infection.

While our data are specific to the context in which we collected them, the insights about the interaction among different components of risk and their impact on different metrics of infection may apply in other endemic settings and generate new knowledge about the processes that lead to the occurrence of persistent hotspots. Such an approach is critical to both controlling the severe morbidity of schistosomiasis, reducing the transmission of parasites in the environment and achieving elimination in endemic regions. Moreover, building a more holistic understanding of infectious disease risk will improve our ability to intervene at the point where a source of infection and a susceptible person meet.

STROBE checklist for cross-sectional studies.

(DOC)

Click here for additional data file.

Data flow diagram.

Sample sizes resulting from the merging and cleaning of human data from parasitological and household surveys and removing data from manipulated sites.

(TIF)

Click here for additional data file.

Correlation plot.

Correlations between all variables used in all models.

(TIFF)

Click here for additional data file.

Household survey items.

Household survey items used in in this analysis, including the module to which an item belonged, the level at which it was measured as well as the wording of and response categories provided for each question.

(PDF)

Click here for additional data file.

Duration and extent of water contact.

Activity-specific values used to estimate duration and extent of water contact. Duration estimates were drawn from published research that took place in a similar setting and extent was estimated from body surface area interviews.

(PDF)

Click here for additional data file.

Household surface water use by socio-economic status.

Frequencies (n (row %)) of household surface water use by quintiles of asset-based index of socio-economic status (SES).

(PDF)

Click here for additional data file.

Presence models.

Full set of candidate models used for multi-model inference of the relative contributions of exposure, hazard, and vulnerability to S. haematobium infection presence.

(XLSX)

Click here for additional data file.

Intensity models.

Full set of candidate models used for multi-model inference of the relative contributions of exposure, hazard, and vulnerability to S. haematobium infection intensity.

(XLSX)

Click here for additional data file.

9 Apr 2021

Dear Ms. Lund,

Thank you very much for submitting your manuscript "Exposure, hazard and vulnerability all contribute to Schistosoma haematobium re-infection in northern Senegal" for consideration at PLOS Neglected Tropical Diseases. As with all papers reviewed by the journal, your manuscript was reviewed by members of the editorial board and by several independent reviewers. In light of the reviews (below this email), we would like to invite the resubmission of a significantly-revised version that takes into account the reviewers' comments. To me this reads as a mixed methods paper and could be sold more so on this aspect and the benefits that this provides, also helping to address the issues with smaller sample sizes, that would not be satisfactory for a purely quantitative study.

We cannot make any decision about publication until we have seen the revised manuscript and your response to the reviewers' comments. Your revised manuscript is also likely to be sent to reviewers for further evaluation.

When you are ready to resubmit, please upload the following:

[1] A letter containing a detailed list of your responses to the review comments and a description of the changes you have made in the manuscript. Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

[2] Two versions of the revised manuscript: one with either highlights or tracked changes denoting where the text has been changed; the other a clean version (uploaded as the manuscript file).

Important additional instructions are given below your reviewer comments.

Please prepare and submit your revised manuscript within 60 days. If you anticipate any delay, please let us know the expected resubmission date by replying to this email. Please note that revised manuscripts received after the 60-day due date may require evaluation and peer review similar to newly submitted manuscripts.

Thank you again for your submission. We hope that our editorial process has been constructive so far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.

Sincerely,

Poppy H L Lamberton

Deputy Editor

PLOS Neglected Tropical Diseases

Anna Last

Associate Editor

PLOS Neglected Tropical Diseases

***********************

Thanks for this interesting paper which we would be happy to see resubmitted after undertaking the recommended changes. To me this reads as a mixed methods paper and could be sold more so on this aspect and the benefits that this provides, also helping to address the issues with smaller sample sizes, that would not be satisfactory for a purely quantitative study.

Reviewer's Responses to Questions

Key Review Criteria Required for Acceptance?

As you describe the new analyses required for acceptance, please consider the following:

Methods

-Are the objectives of the study clearly articulated with a clear testable hypothesis stated?

-Is the study design appropriate to address the stated objectives?

-Is the population clearly described and appropriate for the hypothesis being tested?

-Is the sample size sufficient to ensure adequate power to address the hypothesis being tested?

-Were correct statistical analysis used to support conclusions?

-Are there concerns about ethical or regulatory requirements being met?

Reviewer #1: yes

Reviewer #2: (No Response)

Reviewer #3: Overall, the methods are appropriately chosen in response to the study questions/objectives. The only minor issues is related to interview desribed in the study. Only two interviews/village with a total of 30 interview were conducted – what are potential implications with the analysis with such a few samples? Also unclear how 210 activity-specific observations were conducted?

Reviewer #4: Methods are explained in detail. Some limitations with the use of existing data collected from the communities is discussed.

--------------------

Results

-Does the analysis presented match the analysis plan?

-Are the results clearly and completely presented?

-Are the figures (Tables, Images) of sufficient quality for clarity?

Reviewer #1: The authors ignore many relevant papers that have been published, which are given in a sticky note on the returned ms. They completely ignore the fact that very few snails have to be infected to result in a high (>87%) of school age children. They record the number of times that the children were in the water but they do not record the time of day they had water contact, which is extremely important.

Reviewer #2: (No Response)

Reviewer #3: Well done!

Reviewer #4: The results are clearly outlined and address the aims (integrated consideration of hazards, exposure and vulnerability).

SOme of the tables and figures could/should have acronyms described more clearly in headers and/or legends.

--------------------

Conclusions

-Are the conclusions supported by the data presented?

-Are the limitations of analysis clearly described?

-Do the authors discuss how these data can be helpful to advance our understanding of the topic under study?

-Is public health relevance addressed?

Reviewer #1: There conclusions ignore important literature and thus they make sweeping statements that are not supported by other studies. Over all they have an excellent paper, they just need to add additional references.

Reviewer #2: (No Response)

Reviewer #3: Overall well done.

Reviewer #4: Conclusions are sounds. The discussion of weaknesses in the data at every step is commendable but detracts from the significance of their findings.

The authors clearly discuss what could be taken away from this study for efforts to disrupt the parasite lifecycle

--------------------

Editorial and Data Presentation Modifications?

Use this section for editorial suggestions as well as relatively minor modifications of existing data that would enhance clarity. If the only modifications needed are minor and/or editorial, you may wish to recommend “Minor Revision” or “Accept”.

Reviewer #1: (No Response)

Reviewer #2: (No Response)

Reviewer #3: Accept

Reviewer #4: See comments below

--------------------

Summary and General Comments

Use this section to provide overall comments, discuss strengths/weaknesses of the study, novelty, significance, general execution and scholarship. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. If requesting major revision, please articulate the new experiments that are needed.

Reviewer #1: The title indicates it is a study of re-infection, which is misleading. In order to measure re--infection they would have to identify infected children, treat them, re-test them, and then have them exposed to water and determine re-infection

Reviewer #2: In this paper, Lund and colleagues address the interesting question of the interaction between hazard, exposure and vulnerability associated with infection and infection intensity of S. haematobium in Senegal. They found that infection presence was mostly driven hazard and vulnerability, while infection intensity is determined by exposure and hazard. I think this is an interesting paper that deserves to be published in PLoS NTDs when my comments would have been addressed.

I see two main concerns with the statistical analysis. First, they select one variable for each component (exposure, vulnerability and hazard), and then perform model selection to identify which one is the most parcimonious. Doing things like that can introduce biases. Indeed, if two variables included in their "hazard" category could explain the whole pathogen presence (or intensity), this cannot be identified because only one variable by component is selected and confronted to the others. I understand the rationale behind that, but I think it would be more robust to perform the analysis without such variable grouping.

Second, in the case where they select only one variable by group, I think including the interactions between variables could be interesting, since exposure and hazard could definitely interact between them.

On a minor note, from table 5, I don't understand the difference between model 16 and 19.

Reviewer #3: (No Response)

Reviewer #4: This research article highlights the importance of environmental hazards, exposure and vulnerability on the risks of urogenital schistosomiasis affecting school-aged children in a region of northern Senegal. Their findings extend work that has primarily focused on the demographic features of a population that affect the risk of infection. Consideration for each component of risk seems highly logical but rarely reported in an integrated way as they have reported here. This is a key, novel feature of this article (in my opinion). Discussion highlights several weaknesses in the approach used herein. Although this has limited the power of the analyses (as mentioned in the discussion), it also highlights that there is a lot more work to do in planning future epidemiological work. Overall, the methods look quite comprehensive. I was unable to review the DRYAD information using the url provided for dryad.s7h44j15p. Is it publicly available? You mention three studies that were undertaken to obtain the data used herein. Some abbreviation of previously published methods may be warranted here. Please make sure they are cited, similar to the PNAS manuscript (reference 18). This will make it clear what methods are unique to this manuscript.

I have some other suggestions that I think are worth considering.

The introduction suggests a broad range of hazards should be considered, including floods and extreme weather events. Despite this being a focus in the introduction, it wasn’t a focus on the analysis and no data reported on extreme/catastrophic events in the region studied during the time of sampling. For this reason, I suggest simplifying the introduction to focus on the risks identified in a more usual seasonal cycle in the region (as presented in the results and discussed).

Narrative in the introduction can be confusing. I would consider starting with schistosomiasis or significantly reduce the background at the start of the introduction. I don't think an extended justification for looking at risk (exposure, hazard, and vulnerability) is required (start of the introduction). Most readers would be aware of this topic. Brief explanation would be better with an introduction that is focussed on the subject (schistosomiasis).

Lines 144 to 147. Please make it clear in the introduction why demographic features were included as control here. It is in the methods, results and discussion, but the approach could be better presented in the last paragraph of the discussion.

Line 222 to 223. Something missing in this sentence

Lines 529 to 532.Please include references to support these statements.

Line 618 and Line 648. Manuscript is well written but please check for grammatical errors throughout

Discussion. Would it be better to incorporate the limitations of this study into one section and discuss how this can be resolved in the future. It feels repetitive at times and detracts from the significant results obtained in this study.

--------------------

PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Jay Richard Stauffer, Jr.

Reviewer #2: No

Reviewer #3: No

Reviewer #4: No

Figure Files:

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org.

Data Requirements:

Please note that, as a condition of publication, PLOS' data policy requires that you make available all data used to draw the conclusions outlined in your manuscript. Data must be deposited in an appropriate repository, included within the body of the manuscript, or uploaded as supporting information. This includes all numerical values that were used to generate graphs, histograms etc.. For an example see here: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5.

Reproducibility:

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Submitted filename: ms.pdf

Click here for additional data file.

17 Jun 2021

Submitted filename: Lund_EHV_responseLetter.pdf

Click here for additional data file.

25 Aug 2021

Dear Ms. Lund,

Thank you very much for submitting your manuscript "Exposure, hazard, and vulnerability all contribute to Schistosoma haematobium re-infection in northern Senegal" for consideration at PLOS Neglected Tropical Diseases. As with all papers reviewed by the journal, your manuscript was reviewed by members of the editorial board and by several independent reviewers. The reviewers appreciated the attention to an important topic. Based on the reviews, we are likely to accept this manuscript for publication, providing that you modify the manuscript according to the review recommendations.

There are very few changes that need to be made to the manuscript. The reviewer 1 has attached a ms with some edits. In addition please include some additional minor changes from me and then I am sure the manuscript will be accepted straight away. It is a lovely paper and I really enjoyed reading it again.

Please prepare and submit your revised manuscript within 30 days. If you anticipate any delay, please let us know the expected resubmission date by replying to this email.

When you are ready to resubmit, please upload the following:

[1] A letter containing a detailed list of your responses to all review comments, and a description of the changes you have made in the manuscript.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out

[2] Two versions of the revised manuscript: one with either highlights or tracked changes denoting where the text has been changed; the other a clean version (uploaded as the manuscript file).

Important additional instructions are given below your reviewer comments.

Thank you again for your submission to our journal. We hope that our editorial process has been constructive so far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.

Sincerely,

Poppy H L Lamberton

Deputy Editor

PLOS Neglected Tropical Diseases

Anna Last

Associate Editor

PLOS Neglected Tropical Diseases

***********************

Reviewer's Responses to Questions

Key Review Criteria Required for Acceptance?

As you describe the new analyses required for acceptance, please consider the following:

Methods

-Are the objectives of the study clearly articulated with a clear testable hypothesis stated?

-Is the study design appropriate to address the stated objectives?

-Is the population clearly described and appropriate for the hypothesis being tested?

-Is the sample size sufficient to ensure adequate power to address the hypothesis being tested?

-Were correct statistical analysis used to support conclusions?

-Are there concerns about ethical or regulatory requirements being met?

Reviewer #1: yes

Reviewer #2: (No Response)

Reviewer #3: As suggested in reviews for the 1st submission, the methods are overall conceived and designed. Minor issues were appropriately addressed in the revision.

--------------------

Results

-Does the analysis presented match the analysis plan?

-Are the results clearly and completely presented?

-Are the figures (Tables, Images) of sufficient quality for clarity?

Reviewer #1: yes

Reviewer #2: (No Response)

Reviewer #3: Well done!

--------------------

Conclusions

-Are the conclusions supported by the data presented?

-Are the limitations of analysis clearly described?

-Do the authors discuss how these data can be helpful to advance our understanding of the topic under study?

-Is public health relevance addressed?

Reviewer #1: yes

Reviewer #2: (No Response)

Reviewer #3: Yes.

--------------------

Editorial and Data Presentation Modifications?

Use this section for editorial suggestions as well as relatively minor modifications of existing data that would enhance clarity. If the only modifications needed are minor and/or editorial, you may wish to recommend “Minor Revision” or “Accept”.

Reviewer #1: There are many places in the ms that nouns are used as adjectives. I marked many of these, but the authors should do a global search. One cannot use scientific names as adjectives! The introduction should be expanded slightly to indicate that there are many types of schistosomes that infect other mammals and birds. Should clarify that the paper deals with urinary schistosomes of humans, but there are also schistosomes that infect the bowels of humans.

Reviewer #2: (No Response)

Reviewer #3: (No Response)

--------------------

Summary and General Comments

Use this section to provide overall comments, discuss strengths/weaknesses of the study, novelty, significance, general execution and scholarship. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. If requesting major revision, please articulate the new experiments that are needed.

Reviewer #1: This is an excellent study and should be published.

Reviewer #2: I am happy with author's replies and therefore recommend acceptance

Reviewer #3: The authors have done a good job in addressing the reviewer's comments.

--------------------

PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Jay R. Stauffer, Jr.

Reviewer #2: Yes: Benjamin Roche

Reviewer #3: No

Editor Poppy Lamberton comments:

In line 61 in the authors summary please change from ‘is acquired, where increasing numbers of worms lead to increasing severity of disease.’ To ‘is acquired. Increasing numbers of worms is known to be positively associated with increasing severity of disease.’ As you don’t actually measure the severity of disease in your study.

Intro line 71: respond to, or recover from,…. Please add these two commas.

Line 126 change to hotspots and line 727, as for most of the references for it.

Methods: line 190, please clarify that you mean urine on two days at each time point, or a single urine sample at two time points? I assume the former, but good to clarify.

Line 235: Please change schistosome infections to Schistosoma haematobium as that was what you tested for.

Lines 237: please clarify at what level you mean the medians, a cross the whole cohorts at each time point? Or across multiple samples from each person at each timepoint? And you can justify the median also as I assume the data were not normally distributed? Likely negative binomial distribution?

Results

Line 432: best fit: I think this is an example of where you use a noun rather than the verb. Maybe rephrase to was the best fit for the s.h presence data.

Discussion:

Line 607: found ‘the’ area of snail habitat to be a…: please add either a ‘the’ or inverted commas: found ‘area of snail habitat’ to be a …

Line 616: change fit to fitted.

Line 617: found proximity to water access points: please write either found the proximity to water access points…. Or found ‘proximity to water access points’ as above

Line 623: fit to fitted

Line 662: represented access to, and use of, water and sanitation

Line 666: in ‘the’ highest and lowest SES quintiles

Line 716: household level, most contribute…. Please change to household level, contribute most…. And for line 718

Line 718: change parasites to S. haematobium parasites

Figure Files:

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org.

Data Requirements:

Please note that, as a condition of publication, PLOS' data policy requires that you make available all data used to draw the conclusions outlined in your manuscript. Data must be deposited in an appropriate repository, included within the body of the manuscript, or uploaded as supporting information. This includes all numerical values that were used to generate graphs, histograms etc.. For an example see here: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5.

Reproducibility:

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

References

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article's retracted status in the References list and also include a citation and full reference for the retraction notice.

Submitted filename: revision senegal.pdf

Click here for additional data file.

9 Sep 2021

Submitted filename: Lund_EHV_responseLetter_2.docx

Click here for additional data file.

10 Sep 2021

Dear Dr. Lund,

We are pleased to inform you that your manuscript 'Exposure, hazard, and vulnerability all contribute to Schistosoma haematobium re-infection in northern Senegal' has been provisionally accepted for publication in PLOS Neglected Tropical Diseases.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. A member of our team will be in touch with a set of requests.

Please note that your manuscript will not be scheduled for publication until you have made the required changes, so a swift response is appreciated.

IMPORTANT: The editorial review process is now complete. PLOS will only permit corrections to spelling, formatting or significant scientific errors from this point onwards. Requests for major changes, or any which affect the scientific understanding of your work, will cause delays to the publication date of your manuscript.

Should you, your institution's press office or the journal office choose to press release your paper, you will automatically be opted out of early publication. We ask that you notify us now if you or your institution is planning to press release the article. All press must be co-ordinated with PLOS.

Thank you again for supporting Open Access publishing; we are looking forward to publishing your work in PLOS Neglected Tropical Diseases.

Best regards,

Poppy H L Lamberton

Deputy Editor

PLOS Neglected Tropical Diseases

Anna Last

Associate Editor

PLOS Neglected Tropical Diseases

***********************************************************

30 Sep 2021

Dear Dr. Lund,

We are delighted to inform you that your manuscript, "Exposure, hazard, and vulnerability all contribute to Schistosoma haematobium re-infection in northern Senegal," has been formally accepted for publication in PLOS Neglected Tropical Diseases.

We have now passed your article onto the PLOS Production Department who will complete the rest of the publication process. All authors will receive a confirmation email upon publication.

The corresponding author will soon be receiving a typeset proof for review, to ensure errors have not been introduced during production. Please review the PDF proof of your manuscript carefully, as this is the last chance to correct any scientific or type-setting errors. Please note that major changes, or those which affect the scientific understanding of the work, will likely cause delays to the publication date of your manuscript. Note: Proofs for Front Matter articles (Editorial, Viewpoint, Symposium, Review, etc...) are generated on a different schedule and may not be made available as quickly.

Soon after your final files are uploaded, the early version of your manuscript will be published online unless you opted out of this process. The date of the early version will be your article's publication date. The final article will be published to the same URL, and all versions of the paper will be accessible to readers.

Thank you again for supporting open-access publishing; we are looking forward to publishing your work in PLOS Neglected Tropical Diseases.

Best regards,

Shaden Kamhawi

co-Editor-in-Chief

PLOS Neglected Tropical Diseases

Paul Brindley

co-Editor-in-Chief

PLOS Neglected Tropical Diseases