Advertisement
  • Loading metrics

Shared water facilities and risk of COVID-19 in resource-poor settings: A transmission modelling study

  • Michael A. L. Hayashi ,

    Contributed equally to this work with: Michael A. L. Hayashi, Savannah Boerger

    Roles Conceptualization, Formal analysis, Supervision, Writing – review & editing

    Affiliation Department of Epidemiology, School of Public Health, University of Michigan, Ann Arbor, Michigan, United States of America

  • Savannah Boerger ,

    Contributed equally to this work with: Michael A. L. Hayashi, Savannah Boerger

    Roles Project administration, Writing – original draft, Writing – review & editing

    Affiliation Department of Epidemiology, School of Public Health, University of Michigan, Ann Arbor, Michigan, United States of America

  • Kaiyue Zou,

    Roles Formal analysis, Writing – review & editing

    Affiliation Department of Epidemiology, School of Public Health, University of Michigan, Ann Arbor, Michigan, United States of America

  • Sophia Simon,

    Roles Formal analysis, Writing – review & editing

    Affiliation Department of Epidemiology, School of Public Health, University of Michigan, Ann Arbor, Michigan, United States of America

  • Matthew C. Freeman ,

    Roles Conceptualization, Funding acquisition, Investigation, Supervision, Writing – review & editing

    ‡ These authors are joint senior authors on this work.

    Affiliation Gangarosa Department of Environmental Health, Rollins School of Public Health, Emory University, Atlanta, Georgia, United States of America

  • Joseph N. S. Eisenberg

    Roles Conceptualization, Funding acquisition, Investigation, Methodology, Supervision, Writing – review & editing

    jnse@umich.edu

    ‡ These authors are joint senior authors on this work.

    Affiliation Department of Epidemiology, School of Public Health, University of Michigan, Ann Arbor, Michigan, United States of America

Abstract

Shared water facilities are widespread in resource-poor settings within low- and middle-income countries. Since gathering water is essential, shared water sites may act as an important COVID-19 transmission pathway, despite stay-at-home recommendations. This analysis explores conditions under which shared water facility utilization may influence COVID-19 transmission. We developed two SEIR transmission models to explore COVID-19 dynamics. The first describes an urban setting, where multiple water sites are shared within a community, and the second describes a rural setting, where a single water site is shared among communities. We explored COVID-19 mitigation strategies including social distancing and adding additional water sites. Increased water site availability and social distancing independently attenuate attack rate and peak outbreak size through density reduction. In combination, these conditions result in interactive risk reductions. When water sharing intensity is high, risks are high regardless of the degree of social distancing. Even moderate reductions in water sharing can enhance the effectiveness of social distancing. In rural contexts, we observe similar but weaker effects. Enforced social distancing and density reduction at shared water sites can be an effective and relatively inexpensive mitigation effort to reduce the risk of COVID-19 transmission. Building additional water sites is more expensive but can increase the effectiveness of social distancing efforts at the water sites. As respiratory pathogen outbreaks—and potentially novel pandemics—will continue, infrastructure planning should consider the health benefits associated with respiratory transmission reduction when prioritizing investments.

Introduction

To control transmission of COVID-19, policies have centered around staying at home to reduce contact, supplemented by vigilant personal hygiene, mask wearing, and social distancing in shared spaces [1]. However, people living in resource-constrained contexts, specifically those living in households without access to private water, sanitation, and hygiene (WASH) facilities may be at disproportionate risk for COVID-19 transmission due to unavoidable close contact which may occur in these settings [2].

Nearly 30% of the global population relies on communal facilities of some type to fulfill their daily water needs [3]. This necessity, unyielding in the face of a pandemic, highlights the indirect impacts of resource constraints on infectious disease transmission, even when other prevention recommendations are followed. Individuals in these settings may be required to congregate at space-limited shared water sources to meet critical household needs [4,5]. Water gathering activities account for substantial amounts of time—greater than 2 hours in some rural communities [68]. As a result, users of communal water sites, who are often low-income, are disproportionately at risk for disease through other avenues of exposure including places of employment and through caretaking.

Due to the potential for COVID-19 transmission in public water collection sites, developing prevention and control measures is an urgent and evolving task. The Hierarchy of Controls, a framework originally developed for occupational health hazards [9], categorizes prevention strategies into those that are the most effective (but require more resources and appropriate infrastructure) and those that are less effective (but require fewer resources and therefore are more readily accessible). This hierarchy has been applied to the COVID-19 response [10,11], and funnels from highly effective elimination strategies, such as vaccination, down to the use of personal protective equipment (PPE) in the form of masks and administrative controls in the form of social distancing (Fig 1). Interventions at the bottom of the hierarchy are subject to considerable variability in behavior and adherence. Effective social distancing at water sites must take into account the functional size of the site itself and the density of users. Infrastructural changes, such as increased ventilation, lie in between vaccines and PPE. We consider increasing access to water to be higher up the Hierarchy of Controls as it is an infrastructural means of risk mitigation. While access to water is often considered in the context of reducing enteric disease risk, we argue that it may serve as a mitigation measure for COVID-19 as well by reducing the amount of social mixing and waiting that occurs at water sites.

thumbnail
Fig 1. Hierarchy of Controls as applied to COVID-19 control strategies.

Adapted from National Institute for Occupational Safety and Health (NIOSH) guidelines [9].

https://doi.org/10.1371/journal.pwat.0000011.g001

It is important to consider the context-dependent nature of water site access and community structure when assessing the risk posed by social mixing, such as along the urban-rural continuum. Urban sites tend to have more intra-community mixing beyond shared water facilities, whereas rural inter-community mixing may be more discrete and infrequent, occurring largely through shared water sites and markets. Within the rural scope, different regions of the world have different geographic clustering norms, potentially influencing the density of users at shared water facilities. This wide variation in social structure highlights the need for context-informed decision making so that recommendations could be assessed and adapted to the local context for most effective outcomes.

COVID-19 research thus far has focused on shared WASH facilities largely in the context of hygiene and viability of SARS-CoV-2, the causative agent, on fomites. The potential impact of distinct intervention types, ranging from simple social distancing (immediate change in conditions) to an actual increase in water sites in a community (long term, resource dependent change in conditions) is unknown. The degree of exposure events to SARS-CoV-2 in a communal water site may seed outbreaks in the general community. The Pareto Principle, suggesting that around 20% of people may cause up to 80% of outbreaks [12], has been demonstrated recently for COVID-19 [13] and previously in relation to other coronaviruses [14]. In particular, this idea could be related to high risk, essential venues like shared water facilities. Although it is clear shared water sites have the potential to incur risks of COVID-19 transmission, very little is known about the relative influence of this venue.

The dearth of empirical data surrounding water facilities and COVID-19 transmission, especially in low-income settings, underscores the utility of mathematical approaches to model the potential impact of shared sites in different types of communities worldwide. The purpose of this study is to investigate how COVID-19 specifically, and respiratory infectious disease transmission broadly, may be impacted by shared water conditions in rural and urban contexts. Mathematical modeling is an important strategy used to highlight the risk incurred and explore potential mitigation strategies. To this end, we examine a range of theoretical water site scenarios. We present a deterministic, compartmental transmission model which we use to estimate the proportional impact of water-gathering contexts in COVID-19 transmission and potential reduction through water-facility-related interventions.

Our primary research question is: What are the synergistic effects of social distancing and increasing the number of water sites for COVID-19 transmission in different shared water contexts worldwide? Our secondary question is: How can we conceptualize the risk associated with shared communal water sites (i.e., public water taps) in the spread of COVID? This study seeks to provide guidance on the most effective interventions for COVID-19 transmission in different shared water contexts worldwide and recommend best practices in water infrastructure development for broad disease reduction.

Methods

Overview

We conducted modeling simulations to estimate the contribution of water sharing intensity to COVID-19 transmission in low-resource communities. Communities are defined as distinct residential units which undergo mixing only at inter-community, shared venues. Our water site model represents outdoor sites such as taps, wells, or kiosks surrounded by a clearing of varying size, referred to as surface area for modeling purposes. Individuals interact and may transmit infection both in the community and at water sites. We assessed several factors which impact contact rate to determine their relative transmission contribution, including 1.) potential for social distancing (defined as surface area at water site), and 2.) degree of water sharing (defined as number of water sites within a community of 2500 persons for urban settings, or as the number of communities of 250 persons sharing one water site for rural settings). We also explore social distancing as an effect modifier of degree of water sharing. Various levels of population density, defined as the number of persons per m2, are considered to simulate a range of real-life contexts.

Water sharing patterns vary by population density and water availability. We consider two key features of shared water obtainment from the perspective of COVID-19 risk: the time spent obtaining water and the density of users at a water facility (Table 1). Public water facilities are associated with higher user densities and longer queueing times when compared with private water sources. This analysis focuses on public shared water sources, given their high user densities, high access frequencies, and potential to introduce novel mixing events. Transit to water sources and waiting times at water sources are modeled as decreasing functions of the number of water sites available, representing shorter distances to any given water site and distributed demand across all available water sites.

thumbnail
Table 1. Characteristics of water sites based on their relative contact rate and time spent at location.

https://doi.org/10.1371/journal.pwat.0000011.t001

In many water-scarce, rural contexts, household members must travel greater distances to fetch water [3], and these water sites may be shared among distinct communities, potentially introducing mixing events that would otherwise not occur. In urban informal environments and more densely populated settlements, residents may use multiple water sites within a given community. We created these generalized urban and rural water sharing contexts to highlight what we hypothesized drive population mixing and lead to COVID-19 transmission.

SEIR model framework

We developed deterministic, compartmental transmission models to evaluate the potential effects of shared water source scenarios on COVID-19 transmission. Our models are based on an SEIR framework where individuals in the population may be either susceptible (S), exposed (E), infectious (I), or recovered (R) [15,16]. To address transmission at shared water sources, we stratified our SEIR compartments to represent individuals at a water site, in transit between a community and water site, and in a community (Fig 2). To illustrate the models, we consider the case where one urban community uses two water sites (Fig 2A), and where two rural communities share one water site (Fig 2B). We assume that individuals in transit to water sites do not come into meaningful contact with each other that would result in exposure, and do not spend enough time in transit to progress through disease states. Individuals at water sites may expose each other, but do not progress further through disease states while at the water sites. These models utilize density-dependent transmission, where the per-capita contact rate and thus, transmission, is directly influenced by population density at a transmission venue (water site or community).

thumbnail
Fig 2. SEIR transmission model diagram for (A) the urban model (multiple water sites within one community) and (B) the rural model (multiple communities sharing one water site).

Individuals in the community can progress through all stages of disease while individuals at the water site only undergo exposure events.

https://doi.org/10.1371/journal.pwat.0000011.g002

For the rural and urban models, the base SEIR model is parameterized with rate of leaving the exposed state (α) (where 1/α is latency), the rate of recovery from being infectious (δ) (where 1/δ is the duration of infectiousness), and the transmission rate. The transmission rate is assumed to be different at the water site (βw) and the community (βc) (Table 2). These two transmission rates are the product of an infectivity parameter (p) and a contact rate parameter (either cw or cc).

thumbnail
Table 2. Pathogen- and location-specific parameters used to calibrate urban and rural transmission models.

https://doi.org/10.1371/journal.pwat.0000011.t002

Because COVID-19 is spread through droplets and aerosols, we have chosen to model contact rate as density-dependent. Therefore, as density in the general community or shared water sites increases, the contact rate increases at a slope of k (kw or kc) [25]. Our transmission rate parameters are expressed as follows: where i represents either c for the community or w for water sites. For the community sites, we obtained information on population density, representing Nc/Ac. For the water sites, we assumed that Aw varied based on the scenario modeled, chosen based on estimates of the area surrounding a water collection point where people may gather for obtainment purposes (5-50m2), and Nw varied as determined by model dynamics.

The rural and urban models have 4 transit parameters (Table 2). For the urban model we define the number of water sites in the urban sector (Xw) and for the community model we define the number of communities (Xc) that share a water site. For both models there is the time spent in transit from or to the water site (tT; assume to be 30 minutes), time spent at the water site (tw) and time spent in the community (tc). Here we assume that the time spent at the water site is a function of the number of water sites for the urban model and the number of communities for the rural model. In the urban model, the more water sites the less waiting that occurs to obtain water. For the rural model the more communities sharing a water site results in more waiting at the site. The remainder of the day is assumed to be spent in the community. The Appendix contains model equations (S1 Fig).

Modeling approach

To evaluate the impact of social distancing and water site availability on COVID-19 transmission, we vary two conditions in both urban and rural contexts: 1) the degree of potential social distancing at a water site, defined by effective area of the site and number of users, and 2) the number of water sites within a defined community. We examine water sites with effective surface areas ranging from 5m2 to 50m2 (Table 2). This range encompasses kiosks and hand pumps on the lower end and wells surrounded by large clearings at the higher end. As noted above, we model transit time to- and from- water sites as well as the time spent gathering water as decreasing in the number of water sites. Gathering time ranges from 15 minutes to 60 minutes in urban scenarios, and 30 minutes to 120 minutes in rural scenarios (Table 3).

thumbnail
Table 3. Transit-related parameters for moving through compartments in the SEIR framework.

https://doi.org/10.1371/journal.pwat.0000011.t003

Since the relative transmission contributions of the shared water site vs. the general community are highly variable, we define 4 scenarios that allow us to explore this impact as well as potential interactive effects. The 4 scenarios comprise: 1) community transmission is dominant, but water site transmission is minimal, 2) community transmission is dominant, but water-site transmission is significant, 3) water-site transmission is dominant but community transmission is significant, and 4) water-site transmission is dominant but community transmission is minimal. Note that minimal and significant transmission are characterized by the ability for transmission at this site to sustain an outbreak on its own. These scenarios are defined by values of the scalar transmission parameters kc and kw (Table 4). We simulate our model until equilibrium to calculate risk measures.

thumbnail
Table 4. Parameter values used to define 4 scenarios, where the scalar quantities kc and kw are directly proportional to the transmission, i.e., when kc is high community transmission is high, and when kw is high water site transmission is high.

https://doi.org/10.1371/journal.pwat.0000011.t004

We focus on within-community dynamics for contexts where there are one or more shared water sites within a given community, which most often applies to areas considered to be urban or peri-urban. In these contexts, we scale the number of water sites available to a given simulated community from 1 to 10, reflecting a range from low water availability to higher water availability. We chose community population densities to reflect three urban sites: Mumbai, India (specifically Dharavi, referred to hereafter as Mumbai); Nairobi, Kenya (specifically Kibera, referred to hereafter as Nairobi); and Pretoria, South Africa (Table 2). For each of these contexts, we model populations of 2500 sharing one or more water sites (up to 10 sites). This means that the maximum daily demand on a given water site can range from 250 to 2500 people per day. Two sites (Mumbai, India and Nairobi, Kenya) reflect informal settlements with two degrees of high density, while one site (Pretoria, South Africa) is included to simulate a moderately dense, more general urban environment (Table 2).

We focus on inter-community dynamics for contexts in which multiple communities share a single water source but otherwise do not (or infrequently) mix with each other, which most often applies to areas considered to be rural. In these contexts, we scale the number of communities sharing a single water site from 1 to 10. Similar to urban and peri-urban contexts, this range is intended to capture different degrees of water availability in rural areas. Three rural villages: Mahasingi (Kandhamal, Orissa), India; Kisima (Samburu County), Kenya; and Ingele, South Africa were chosen to contextualize the models, again using location-specific model parameters (Table 2). For these simulations, we define a community to be a grouping of 250 residents. As in the urban context scenarios, this effective population size means that total demand for the water site varies from 250 people per day to 2500 people per day. To better reflect the effective shared water scenarios in more rural environments where inter-community resource sharing may be more common, this model explores variations in the number of communities sharing one water site. Analogous to the urban models, these simulations assess degree of social distancing, in addition to exploring the relative impact of water site vs. community-driven transmission and potential interactive effects between social distancing and degree of water sharing.

Model assumptions

We make the following assumptions in our modeling approach. (1) We only include respiratory transmission of COVID-19 where the primary transmission route for COVID-19 is through droplet or aerosol spread [26]. Although transmission through fomites and other mechanisms may be feasible, these pathways are likely lower contributors to overall transmission for water fetching activities [27,28]. This lack of fomite or environmental routes may be particularly true in under these conditions, as opposed to more enclosed areas such as sanitation facilities. (2) We assume that there is no reinfection with SARS-CoV-2. Although little is known about the duration of immunity, given the relatively short duration of the outbreaks simulated here, those who have recovered are unlikely to be re-infected during the simulation period. (3) We assume people are homogeneous within each compartment with respect to risk factor behavior, such as movement rates to water sites and contact rates among others in the population. (4) For the main results, we assume no other interventions, including community lockdown orders, which are explored separately. (5) In urban settings, we assume that the number of water sites is inversely proportional to both time spent at the water site and water site transit time due to the direct effect of water availability on queuing time and catchment area, respectively. In the rural settings, we assume that the number of communities sharing a water site is proportional to time spent at the water site. (6) As in other density dependent transmission models, we assume that the risk of infection at water sites scale linearly with population density. For airborne pathogens such as SARS-CoV-2, we argue that physical mobility within a given space is less relevant than the amount of shedding individuals in that space. Risk, therefore, is less likely to fall off at high densities due to physical constraints. In addition, the average dose inhaled for airborne pathogens such as COVID-19 is relatively low, suggesting that a linear dose response function is a reasonable approximation. (7) For simplicity we assume that transit and water obtainment rates remain constant throughout the day.

Ethical considerations

This study was deemed exempt from human subjects IRB approval by the University of Michigan Institutional Review Board as no empirical data are included. This theoretical model was built using literature-informed parameters regarding the pathogen and specific cities and no empirical or personally identifiable data was involved.

Results

Impact of increasing social distancing at shared water sites

We consider the impact of social distancing at water sites on the attack rate based on population density and potential for exposure at the water site, holding degree of water sharing constant at a moderate level (3 within-community water sites for urban populations; 3 between-community water sites for rural populations) and assuming community-driven transmission where water-driven transmission is still significant (kc = 11, kw = 20; Table 4). We observe a non-linear decrease in attack rate as the potential for social distancing increases (modeled by increasing the area that people congregate from 5 m2 to 50 m2). This relationship is strongest at low community density (Pretoria, 0.0036 people/m2; Mahasingi, 0.00048 people/m2; and Ingele, 0.00019 people/m2), where increasing water site surface area from 5 m2 to 50 m2 results in a 100% reduction in attack rate (Fig 3A and 3B). At highest community density (Mumbai, 0.1 people/m2), the same change in water site surface area reduces the attack rate by only 0.25%. (Fig 3A). Similar effects are observed for peak outbreak size in both urban and rural contexts (S2A and S2B Fig).

thumbnail
Fig 3. COVID-19 attack rate (proportion infected) using a community driven transmission scenario (kc = 20, kw = 1; see Table 4) for A) each urban site (number of water sites, Xw = 3) and B) each rural site (number of communities, Xc = 3) as a function of the size of water gathering surface area (ranging from 5 to 50m2).

https://doi.org/10.1371/journal.pwat.0000011.g003

Impact of reducing degree of water sharing

We consider the impact of degree of water sharing on the attack rate based on population density and potential for exposure at the water site, holding water site surface area constant at a moderate value of 25m2 and assuming community-driven transmission (kc = 11, kw = 20). As water sharing decreases, either through increasing intra-community water sites in the urban contexts (Fig 4A) or through decreasing number of communities sharing an inter-community water site in rural contexts (Fig 4B), we observe decreases in attack rate that interact with community density. In low- to moderate-density urban settings like Pretoria and Nairobi, increasing from 1 to 10 water sites per 2500 persons community produces a strong nonlinear effect and reduces the attack rate by 99.95% and 95.6%, respectively. In high-density urban settings like Mumbai, we observe an attenuated, linear interaction which reduces the attack rate by only 0.9% across the same water site sharing conditions. All sites in urban and rural contexts, including Mumbai, experience a decrease in outbreak peak prevalence with decrease in shared water site user load (S3A and S3B Fig).

thumbnail
Fig 4. COVID-19 attack rate (proportion infected) using a community driven transmission scenario (kc = 20, kw = 11; see Table 4) for A) each urban site as a function of the number of water sites within a community (ranging from 1 to 10) and B) each rural site as a function of the number of communities sharing water site (ranging from 1 to 10), holding the surface area of the water sites constant at 25m2.

https://doi.org/10.1371/journal.pwat.0000011.g004

Impact of community- vs. Shared water site-dominant transmission

We explore the impact of community transmission on attack rate, specifically relative to transmission at shared water sites. When the general community is the dominant transmission pathway and the water site transmission is low (kc = 20, kw = 1), decreasing water sharing from high to low levels (2500 to 500 people per water site) results in only a 4% reduction in attack rate. As the relative contribution of shared water sites to transmission increases either slightly (kc = 15, kw = 5) or to the point where water site is the dominant mode of transmission (kc = 1, kw = 20), the same decrease in water sharing results in a 40% reduction in attack rate or elimination of COVID-19, respectively (Fig 5).

thumbnail
Fig 5. Attack rate (proportion infected) as a function of degree of water sharing (measured by the number of residential clusters sharing one water site) and the relative transmission rate of the community compared to the shared water sites (see Table 4).

The degree of water sharing varies from 10 residential clusters sharing one water site (High) to one residential cluster with a single water site (Low). Community transmission and water site transmission are scaled by adjusting the parameters kc and kw from 1 (Low) to 20 (High).

https://doi.org/10.1371/journal.pwat.0000011.g005

In our main analyses, we assume that transmission from community venues (markets, religious gatherings, household transmission, etc.) is substantial, but that water sites are also meaningful venues for transmission (kc = 11, kw = 20). While less realistic, scenarios assuming dominant transmission in shared water sites (kc = 1, kw = 20) isolate the effects of water infrastructure on transmission from background community-level dynamics and are included in the Appendix (S4 and S5 Figs).

Interactive impacts of social distancing and degree of water sharing

We quantify the potential for interaction when considering varying our two shared water site conditions concurrently. When considering both mitigation efforts together, we estimate that the degree of water sharing modifies the impact of social distancing on the attack rate. For example, in an urban setting (e.g., Nairobi, Kenya), increasing the potential for social distancing from low to high (5m2 to 50m2 effective water site area) is somewhat effective when there is a high degree of water sharing (1 site per 2500 persons) [9.9% attack rate reduction] (Fig 6 and S6 Fig). However, when there is a moderate degree of water sharing (5 sites per 2500 persons), social distancing has a much stronger impact [80.9% attack rate reduction]; and when degree of water sharing is minimal, social distancing can effectively eliminate transmission. Outbreak peak prevalence is more sensitive to changes in both conditions (S8 Fig).

thumbnail
Fig 6. Attack rate (proportion infected) as a function of degree of water sharing (measured by the number of residential clusters sharing one water site) and social distancing (measured by the water site surface area), for the scenario using Nairobi, Kenya population density (9.86x10-2/m2).

The degree of water sharing varies from 10 residential clusters sharing one water site (High) to one residential cluster with a single water site (Low). The degree of social distancing at the water site is characterized by the effective surface area of the water site ranging from 5 m2 (Low) to 50 m2 (High).

https://doi.org/10.1371/journal.pwat.0000011.g006

In rural Mahasingi, India, we observe that degree of water sharing and potential for social distancing have similar effects on the attack rate as in the urban model, albeit to a reduced degree. Increasing potential for social distancing from low to high (2m2–50m2 effective water site surface area) is negligibly effective when there is a high degree of water sharing (10 communities of 250 sharing 1 site) [0.28% attack rate reduction] (Fig 7 and S7 Fig). When there is a moderate degree of water sharing (6 communities sharing 1 site), increasing social distancing from low to high has a moderate effect [24.2% attack rate reduction]. At minimal sharing, social distancing reduces an already low attack rate [40% attack rate reduction]. Similar to the urban sites, we see that outbreak peak prevalence is more sensitive to changes in both conditions than attack rate, following more of a smooth trend across the matrix (S9 Fig).

thumbnail
Fig 7. Attack rate (proportion infected) as a function of degree of water sharing (measured by the number of residential clusters sharing one water site) and social distancing (measured by the water site surface area), for the scenario using Mahasingi, India population density (4.87x10-4/m2).

The degree of water sharing varies from 10 residential clusters sharing one water site (High) to one residential cluster with a single water site (Low). The degree of social distancing at the water site is characterized by the effective surface area of the water site ranging from 5 m2 (Low) to 50 m2 (High).

https://doi.org/10.1371/journal.pwat.0000011.g007

Discussion

Main findings

In many communities in low- and middle-income countries (LMICs), the lack of household water connections necessitates congregation at public water points. This daily need to fetch water leads to intra- and inter-community mixing, and at times long queuing times that may result in added risks during the COVID-19 pandemic or other future pandemics. Prior empirical work has found that waiting times for drinking water across multiple East African countries can vary from under 15 minutes to over 2 hours with prevalent queueing behavior [68]. While evidence suggests that SARS-CoV-2 transmission outdoors is relatively limited when contact times are short, these extended contact periods may lead to increased transmission. Given that most households will need to visit these water sites at a daily basis—a unique characteristic of shared water points as a driver of infectious disease transmission—there is potential for significant transmission, which in turn can result in transmission within households and between communities.

Our modeling study suggests that indeed community density, and thus potential congregation density at the water point, plays an important role in determining the impact of reliance on community water points on transmission. Both social distancing and increased water site availability can reduce COVID-19 transmission through both independent and synergistic effects. Above certain densities, increasing water point access does little to mitigate transmission. Yet there are threshold levels of both population density where better access to water—and thus less contact time at water points—could have an impact.

These effects are evident in our simulations of realistic population density scenarios. In communities with higher population densities (Mumbai, Nairobi, and Pretoria), increases in social distancing and access to shared water sources result in a relatively smooth reduction in the COVID-19 attack rate. However, in communities with lower population densities (Mahasingi, Ingele, and Kisima), social distancing and shared water access produce a threshold effect—above a certain degree of distancing and water access, the COVID-19 attack rate falls to near 0. Transmission at water sites is more pronounced in urban areas with high density and community mixing, but notable in all sites modelled in this analysis. Substantially reducing attack rates requires an increased number of water sites in urban contexts or a decreased number of communities sharing one site in rural communities, as social distancing alone only produces minimal changes under low levels of background community transmission.

Successful management of COVID-19 transmission in communities that depend on shared drinking water sources requires consideration of both existing water infrastructure and community characteristics. For example, highly dense communities with fewer small water sources may require more aggressive interventions to control transmission from essential water gathering activities than less dense communities, or those with more space for social distancing available at the water sites themselves. Our model does not explicitly calculate the attributable fraction of transmission at water sites relative to community spread. However, as lockdowns or other social distancing measures are imposed, we would expect the community contact rate may decline and the contact rate at critical water sources to remain the same, thus increasing the attributable fraction of transmission related to shared water sources.

Implications of population density

In urban settings, population density has a large impact on COVID-19 attack rate and peak outbreak size, due to the model’s assumptions that in these high density areas there are ample contact points outside of water fetching that might lead to transmission. In extremely dense contexts, adjusting shared water sites to allow for maximal social distancing at sites does not impact the overall attack rate but does lower the outbreak peak. While this suggests that water sharing is less significant as a driver of overall disease burden in dense urban environments, the timing and magnitude of outbreak peaks has a substantial impact on healthcare capacity. Flattening and widening an outbreak peak would reduce the burden on the health system, likely improving treatment outcomes and reducing excess deaths due to unavailability of care.

In less dense urban contexts, water sharing appears to play a much greater role in overall COVID-19 transmission. Reducing water sharing through capacity increases (adding water sites) or improving the capacity for social distancing at water sites in these contexts can yield dramatic reductions in both COVID-19 attack rates and the magnitude of outbreak peaks, due to model assumptions that in these contexts, contact at community water points represents a larger portion of inter-household contact and congregation that could lead to transmission. These findings suggest that in moderate density communities, increasing water access could support reductions in the risk and burden of severe respiratory disease transmission.

In rural contexts, variation in within-community density is less influential when considering shared water site impact on COVID-19 transmission. In these models, rural communities are considered to only mix at shared water venues. Even so, the COVID-19 attack rates decrease as degree of water sharing decreases, with the greatest attenuation occurring when moving from 2 to 1 community. This demonstrates that in rural areas, reducing inter-community mixing which may otherwise not occur by improving water availability is an impactful mitigation strategy for COVID-19.

Limitations

This theoretical modeling analysis provides a helpful framework to assess various shared water site conditions and disease transmission, but several factors may influence these findings. First, although COVID-19 may also transmit through fomites and other pathways, this model is limited to exploring droplet and aerosol spread, to allow simplification of the model to focus on qualitative differences in a deterministic framework. Second, to focus on the impact of transmission from shared water venues, all other interactions are condensed into one contact rate. This rate is meant to account for other potential high-risk contacts within the community but could be higher or lower in different contexts and is also highly variable by person. Third, this model does not explicitly consider preference for a specific water site or utilization of multiple water sites by a single individual, which could increase mixing and attenuate the reductions in COVID-19 burden demonstrated here. Finally, our framework does not consider the impact of other reduction strategies including mask-wearing or self-quarantine after potential exposure or isolation when symptomatic, which could exaggerate the relative impact of the conditions considered here.

Future work

Our model highlights community water conditions that may amplify or mitigate COVID-19 transmission in relatively urban or rural contexts, which could be used to develop policies and adherence guidelines. Leadership at multiple administrative levels may leverage this information to assess their unique local context and how implementing short- and long-term conditions may best fit their Hierarchy of Controls. Emerging infectious diseases and pandemics of this kind are expected to increase in frequency [29,30], warranting preventative measures and prospective infrastructure development. This modeling framework for shared water site assessment may be parameterized with other respiratory pathogens. Shared sanitation facilities are also an essential service which may introduce close contact as well as fomite transmission of COVID-19 and should be explored further using a modeling approach to assess potential risk and mitigation factors.

Conclusion

This current pandemic will likely not be the last, and there is a need to plan for both ongoing mitigation of COVID-19 and the control of future outbreak. This approach and the specific findings could be used to support planning by local and national governments to prioritize the control of outbreaks where water sharing poses a risk for mixing and exposure. Both social distancing and increasing the number of water sites have potential to attenuate respiratory outbreaks like COVID-19. Social distancing is an inexpensive intervention, suffers from non-compliance. Adherence often declines over time, and is often highly variable even under ideal conditions [31,32]. By contrast, infrastructural improvements such as adding water sites or transitioning to piped water systems, though costly, create long term infrastructure that will continue to provide health benefits beyond the outbreak period. Importantly, these two interventions exhibit synergistic relationships. For example, a relatively small decrease in water sharing can have a dramatic effect in increasing the efficacy of social distancing at the water site. Shared water conditions are not only important for the control of enteric pathogens, but also for respiratory pathogens. Future pandemic preparedness and response should incorporate the risks connected to respiratory infection when assessing the need for water infrastructure investment.

Supporting information

S1 Fig. Compartmental model equations for urban and rural analyses.

https://doi.org/10.1371/journal.pwat.0000011.s001

(TIFF)

S2 Fig. COVID-19 outbreak peak size considering community driven transmission (kW = 20, kC = 11) for all urban (A) and rural (B) sites when increasing social distancing potential.

https://doi.org/10.1371/journal.pwat.0000011.s002

(TIFF)

S3 Fig. COVID-19 peak outbreak size considering community driven transmission for all urban (A) sites when increasing number of water sites within community and rural (B) sites when increasing number of communities sharing water site.

https://doi.org/10.1371/journal.pwat.0000011.s003

(TIFF)

S4 Fig. Sensitivity analysis assessing (A) attack rate and (B) outbreak peak for urban sites parameterized for shared water venues (rather than general community) as dominant transmission site of COVID-19 (kC = 1, kW = 20).

https://doi.org/10.1371/journal.pwat.0000011.s004

(TIFF)

S5 Fig. Sensitivity analysis assessing (A) attack rate and (B) outbreak peak for rural sites parameterized for shared water venues (rather than general community) as dominant transmission site of COVID-19 (kC = 1, kW = 20).

https://doi.org/10.1371/journal.pwat.0000011.s005

(TIFF)

S6 Fig. 2-D heatmap of attack rate values when varying social distancing and water sites within a residential cluster in Nairobi, Kenya under general community-dominant transmission (kC = 20, kW = 11).

https://doi.org/10.1371/journal.pwat.0000011.s006

(TIFF)

S7 Fig. 2-D heatmap of attack rate values when varying social distancing and number of residential clusters sharing water sites in Mahasingi village, India under general community-dominant transmission (kC = 20, kW = 11).

https://doi.org/10.1371/journal.pwat.0000011.s007

(TIFF)

S8 Fig. Outbreak peak size with variations in social distancing and number of water sites, parameterized for urban Nairobi, Kenya population density (0.1 persons/m2).

https://doi.org/10.1371/journal.pwat.0000011.s008

(TIFF)

S9 Fig. Outbreak peak size with variations in social distancing and number of water sites, parameterized for rural Mahasingi, India population density (0.00048 persons/m2).

https://doi.org/10.1371/journal.pwat.0000011.s009

(TIFF)

Acknowledgments

Mondal Hasan Zahid and Hannah Van Wyk at the University of Michigan provided project support. Radu Ban from the Bill & Melinda Gates Foundation provided project feedback.

References

  1. 1. Bruinen de Bruin Y, Lequarre A-S, McCourt J, Clevestig P, Pigazzani F, Zare Jeddi M, et al. Initial impacts of global risk mitigation measures taken during the combatting of the COVID-19 pandemic. Saf Sci. 2020 Aug;128. pmid:32296266
  2. 2. Freeman MC, Caruso BA. Comment on “Global Access to Handwashing: Implications for COVID-19 Control in Low-Income Countries”. Environ Health Perspect. 2020 Sep;128(9):098001.
  3. 3. United Nations Children’s Fund (UNICEF), World Health Organization (WHO). Progress on household drinking water, sanitation and hygiene 2000–2017: a special focus on inequities. [Internet] New York and Geneva: UNICEF and WHO. 2019. https://washdata.org/sites/default/files/documents/reports/2019-07/jmp-2019-wash-households.pdf.
  4. 4. Wasdani KP, Prasad A. The impossibility of social distancing among the urban poor: the case of an Indian slum in the times of COVID-19. Local Environ. 2020 May 3;25(5):414–8.
  5. 5. Staddon C, Everard M, Mytton J, Octavianti T, Powell W, Quinn N, et al. Water insecurity compounds the global coronavirus crisis. Water Int. 2020 Jul 3;45(5):416–22.
  6. 6. Gulyani S, Talukdar D, Kariuki RM. Universal (Non)service? Water Markets, Household Demand and the Poor in Urban Kenya. Urban Stud. 2005 Jul;42(8):1247–74.
  7. 7. Thompson J, Porras IT, Wood E, Tumwine JK, Mujwahuzi MR, Katui-Katua M, et al. Waiting at the tap: changes in urban water use in East Africa over three decades. Environ Urban. 2000 Jun;12(2):37–52.
  8. 8. Sarkar A. Everyday practices of poor urban women to access water: Lived realities from a Nairobi slum. Afr Stud. 2020;79(2):212–31.
  9. 9. Occupational Safety and Health Administration (OSHA). Recommended Practices for Safety and Health Programs. [Internet] 2016 Oct. https://www.osha.gov/sites/default/files/publications/OSHA3885.pdf.
  10. 10. COVID-19 Hierarchy of Controls | Environment, Health and Safety [Internet]. [cited 2020 Nov 19]. https://ehs.cornell.edu/campus-health-safety/occupational-health/covid-19/covid-19-hierarchy-controls.
  11. 11. Wong J, Goh QY, Tan Z, Lie SA, Tay YC, Ng SY, et al. Preparing for a COVID-19 pandemic: a review of operating room outbreak response measures in a large tertiary hospital in Singapore. Can J Anesth Can d’anesthésie. 2020 Jun 11;67(6):732–45.
  12. 12. Lloyd-Smith JO, Schreiber SJ, Kopp PE, Getz WM. Superspreading and the effect of individual variation on disease emergence. Nature. 2005;438:355–9. pmid:16292310
  13. 13. Gürsakal N, Batmaz B, Aktuna G. Drawing transmission graphs for COVID-19 in the perspective of network science. Epidemiol Infect. 2020 Nov 4;148. pmid:33143782
  14. 14. Kim KH, Tandi TE, Choi JW, Moon JM, Kim MS. Middle East respiratory syndrome coronavirus (MERS-CoV) outbreak in South Korea, 2015: epidemiology, characteristics and public health implications. J Hosp Infect. 2017 Feb 1;95(2):207–13. pmid:28153558
  15. 15. Vynnycky E, White R. An introduction to infectious disease modeling. Oxford, New York: Oxford University Press; 2010.
  16. 16. Anderson RM, May R. Infectious Disease of Humans: Dynamics and Control. Oxford; New York: Oxford University Press; 1991.
  17. 17. Byrne AW, McEvoy D, Collins AB, Hunt K, Casey M, Barber A, et al. Inferred duration of infectious period of SARS-CoV-2: rapid scoping review and analysis of available evidence for asymptomatic and symptomatic COVID-19 cases. BMJ Open. 2020 Aug;10(8). pmid:32759252
  18. 18. Peirlinck M, Linka K, Sahli Costabal F, Kuhl E. Outbreak dynamics of COVID-19 in China and the United States. Biomech Model Mechanobiol. 2020 Dec 27;19(6). pmid:32342242
  19. 19. Phucharoen C, Sangkaew N, Stosic K. The characteristics of COVID-19 transmission from case to high-risk contact, a statistical analysis from contact tracing data. EClinicalMedicine. 2020 Oct;27(100543).
  20. 20. Kaushal J, Mahajan P. Asia’s largest urban slum-Dharavi: A global model for management of COVID-19. Cities. 2021 Apr;111(103097). pmid:33519012
  21. 21. 2019 Kenya Housing and Population Census. Nairobi; 2019 Dec.
  22. 22. Pretoria Population [Internet]. 2021 [cited 2021 Apr 23]. https://worldpopulationreview.com/world-cities/pretoria-population.
  23. 23. Primary Census Abstract Data Tables—Odisha [Internet]. Office of the Registrar General & Census Commissioner, India. 2011 [cited 2021 Dec 4]. https://censusindia.gov.in/pca/pcadata/Houselisting-housing-OR.html.
  24. 24. Firth A. Ingele SP [Internet]. [cited 2021 Mar 20]. https://census2011.adrianfrith.com/place/504029.
  25. 25. Begon M, Bennett M, Bowers RG, French NP, Hazel SM, Turner J. A clarification of transmission terms in host-microparasite models: numbers, densities and areas. Epidemiol Infect. 2002 Aug 2;129(1). pmid:12211582
  26. 26. Zhang R, Li Y, Zhang AL, Wang Y, Molina MJ. Identifying airborne transmission as the dominant route for the spread of COVID-19. Proc Natl Acad Sci. 2020 Jun 30;117(26).
  27. 27. Wilson AM, Weir MH, Bloomfield SF, Scott EA, Reynolds KA. Modeling COVID-19 infection risks for a single hand-to-fomite scenario and potential risk reductions offered by surface disinfection. Am J Infect Control. 2021 Nov;49(6):846–8. pmid:33207258
  28. 28. Pitol AK, Julian TR. Community Transmission of SARS-CoV-2 by Surfaces: Risks and Risk Reduction Strategies. Environ Sci Technol Lett. 2021 Jan 6;8(3):262–9.
  29. 29. Zinsstag J, Utzinger J, Probst-Hensch N, Shan L, Zhou X-N. Towards integrated surveillance-response systems for the prevention of future pandemics. Infect Dis Poverty. 2020 Dec 7;9(1). pmid:33028426
  30. 30. Aguirre AA, Catherina R, Frye H, Shelley L. Illicit Wildlife Trade, Wet Markets, and COVID-19: Preventing Future Pandemics. World Med Heal Policy. 2020 Sep 5;12(3).
  31. 31. Coroiu A, Moran C, Campbell T, Geller AC. Barriers and facilitators of adherence to social distancing recommendations during COVID-19 among a large international sample of adults. PLoS One. 2020 Oct 7;15(10). pmid:33027281
  32. 32. Wolff W, Martarelli CS, Schüler J, Bieleke M. High Boredom Proneness and Low Trait Self-Control Impair Adherence to Social Distancing Guidelines during the COVID-19 Pandemic. Int J Environ Res Public Health. 2020 Jul 28;17(15). pmid:32731369