Schistosomiasis, caused by parasitic flatworms, affects over 200 million people globally, primarily in tropical low- and middle-income countries (LMICs). The disease burden is concentrated in rural poor communities, where the interplay of pathogen exposure, infection acquisition, and socio-demographic and environmental factors complicates identifying determinants of exposure. Schistosome transmission is complex, driven by human behavior, water and sanitation access, occupation, and ecological conditions for freshwater snails (intermediate hosts). Exposure occurs during water contact, with cercariae penetrating the skin. The main control strategy is mass drug administration (MDA) with praziquantel, but treatment doesn't prevent reinfection, and concerns exist about drug resistance. The 2030 WHO roadmap for neglected tropical diseases necessitates complementing MDA with additional interventions like improved water, sanitation, and hygiene (WASH), environmental control, or behavior change. However, knowledge on identifying high-exposure groups and key exposure determinants is limited. Water contact is a well-established proxy for exposure, but previous studies have been largely cross-sectional, using self-reported data or crude binary indicators to predict current infection. Many studies lacked snail abundance data to account for environmental factors and lacked systematic variable selection and out-of-sample validation. Regular MDA complicates studying exposure's role due to acquired immunity. No standardized exposure measurement tools exist. This study aimed to comprehensively characterize water contact using statistical models, investigating the determinants of water contact and comparing them with infection determinants.

A systematic review and meta-analysis by Reitzug et al. (2023) found that only 21.8% of studies collected snail abundance data to account for environmental factors influencing parasite acquisition risk. Attempts to integrate water contact with environmental variables haven't consistently improved infection prediction, potentially due to limitations in data (e.g., rough immersion estimates), assumptions about cercarial density, or methodological limitations (lack of systematic variable selection and out-of-sample validation). Studies have shown varying infection prevalence based on household distance to water bodies, but the explanation through exposure trends remains unclear. Statistical models have predominantly focused on water contact as a predictor of infection without characterizing water contact itself as an outcome. Current mathematical models infer age-specific water contact levels from infection trends, assuming infection levels are a nonlinear function of cumulative water contact and adult schistosome survival within the host. There's a need to characterize water contact determinants and assess whether infection and water contact follow similar age trends.

This population-based study used baseline data from the SchistoTrack cohort, collected in January-February 2022 in 38 diverse villages across three districts in rural Uganda (Pakwach, Buliisa, and Mayuge). 2867 individuals (aged 5-90) from 1444 households were surveyed. Data collected included socio-demographics, biomedical information, WASH information, and environmental data. Exposure was measured by asking individuals about their engagement in 11 water contact activities, their weekly frequency, and duration. *Schistosoma mansoni* infection status was ascertained using Kato-Katz stool microscopy and point-of-care circulating antigen tests. Malacological data on snails and waypoints of water sites, households, schools, and village centers were used to capture environmental and spatial variables. Bayesian variable selection (BVS) was employed on 27 socio-demographic, biomedical, WASH, and environmental candidate variables to predict water contact. Variables with inclusion probabilities ≥0.5 were used in multivariable logistic regression models. The same methods were used to select predictors of infection status, with the addition of six water contact variables. Generalized additive models (GAMs) were used to examine age-specific patterns of water contact and infection. Logistic and negative binomial regressions were used to model water contact and infection, respectively. Ten-fold cross-validation assessed model predictive performance. Direct water contact observations were used to validate self-reported data.

46.7% of participants reported weekly water contact. The median frequency was six times per week, and the median duration was eight hours per week. Getting drinking water, washing clothes, and fishing were the most common activities. Water contact was most common between 6-9 am, varying by activity type. Water contact declined rapidly with distance from water sites; 80% of individuals with water contact lived within 0.43 km. Females had higher water contact frequency than males, but males had longer durations. Age-dependent water contact peaked at age 30, while infection prevalence peaked at age 15. BVS identified age, gender, occupation, distance to public latrines, water site type, number of water sites, and village infection prevalence as predictors of water contact. Age, occupation, number of water sites, and village infection prevalence were shared predictors of both water contact and infection. Only age was significantly associated with both outcomes in logistic regressions. Females had higher odds of domestic water contact, while males had higher odds of occupational water contact. Village-level infection prevalence was uncorrelated with the village-level proportion of individuals with water contact. Self-reported water contact data closely resembled observed patterns from direct observation. The water contact model had better predictive performance (auROC 0.777) than the infection status model (auROC 0.693).

This study's findings highlight the complex relationship between water contact and *Schistosoma mansoni* infection. The lack of strong association between current water contact and current infection underscores the role of acquired immunity, particularly the impact of past infections and MDA. The 15-year gap between the peaks of water contact and infection suggests that immunity acquired through repeated exposure or successful treatment reduces susceptibility to reinfection. The differing age trends highlight the limitations of assuming a direct correspondence between age-specific water contact and infection prevalence in modeling studies. Gender differences in water contact emphasize the need for interventions targeting both domestic and occupational water contact. The study shows that environmental factors influence water contact patterns and may also influence environmental hazards (local cercariae abundance). Household distance to freshwater bodies showed a stronger gradient with water contact than with infection prevalence. The study validates the use of self-reported water contact data for population-level measures. The study shows that human-environmental interactions are group-specific, and suggests target groups for WASH interventions.

This study provides a comprehensive characterization of water contact patterns and their determinants in rural Uganda, revealing a complex interplay between exposure, immunity, and infection. The findings emphasize the need for multifaceted control strategies for schistosomiasis, combining MDA with interventions targeting specific at-risk groups and behaviors. Future research should focus on longitudinal studies integrating exposure histories with immunological data and more granular exposure measurements using wearable technology to refine our understanding of schistosomiasis transmission dynamics.

The cross-sectional study design limits causal inferences about the relationship between water contact and infection. The reliance on self-reported data, despite validation efforts, may introduce some measurement error. The study focuses on a specific geographical area in Uganda, limiting the generalizability of findings to other settings with different environmental conditions or cultural practices. The study period was limited to the dry season, potentially missing variations in water contact patterns during other seasons.