Rapid evidence assessment of the impacts of sewerage, drainage, and piped water chlorination in urban settings of low- and middle-income countries

Rapidly synthesize the available evidence regarding the expected impacts of piped water chlorination, drainage, and sewerage on individuals living in low- and middle-income urban settings. A systematic search was conducted in bibliographic databases and library catalogs. Impact evaluations that included a comparator and one of the three interventions of interest, attempted to control for confounding and selection bias, and took place in urban settings in a low- or middle-income country were considered. Outcomes related to health, well-being, economic growth, and the bacterial content of water were considered. 1,483 articles were identi ﬁ ed, with 18 included after ﬁ nal screening. Most studies were case-control and took place in Brazil, the Paci ﬁ c Islands, or south-western Asia. Fifteen studies considered sewerage, ﬁ ve considered chlorination, and two studies considered drainage with some studies considering multiple interventions. All primary outcomes were related to health. When implemented alone, the sewerage and chlorination interventions were largely successful. The evidence regarding the effects of chlorination, drainage, and sewerage interventions is limited and generally has a high risk of bias. When properly implemented, sewerage and chlorination are likely to have positive health impacts. However, when implemented poorly, all three interventions can have negative health impacts.


INTRODUCTION
Access to clean water and sanitation is a human right (United Nations a). However, three billion people lack access to basic sanitation (United Nations a). Only 36% of the world's population has access to sewerage (WHO/UNICEF JMP ). Although 90% of the world's population has access to basic drinking water sources (defined as sources that are protected from outside contamination such as fecal matter where water collection takes less than 30 min roundtrip), water quality remains low around the world (United Nations b; WHO/UNICEF JMP ). In addition, about two million people die from diarrheal diseases each year because of poor hygiene and sanitation (United Nations a). These numbers are slightly better in urban settings, where 96% of the population have access to an improved water supply and 79% have access to improved sanitation (United Nations ).
However, the health impacts of poor water and sanitation can be accentuated in urban settings where people are in close contact with one another (Mackinnon et al. ).
Despite this, the evidence regarding the health effects of water, sanitation, and hygiene (WaSH) interventions in urban settings remains poor (Mackinnon et al. ). Of the seven studies identified by Mackinnon et al. () that assessed the impact of urban WaSH interventions on health, only one study found a statistically significant reduction in diarrhea and dysentery; the remaining studies either did not find an effect or were focused on behavioral or microbial contamination outcomes. There is even less evidence regarding what works in decreasing the health impacts of poor-quality water and sanitation in urban settings (Mackinnon et al. ).
The challenges to providing clean and sustainable water and sanitation to a population are compounded within cities (United Nations ; Sosa-Rodriguez et al.

).
Because of this and other factors, improvements in access to WaSH resources have been slower in urban settings than rural settings (United Nations ).
Demographic and health trends continue to make urban settings more crowded, worsening the health impacts of poor water and sanitation in these settings (Mackinnon et al. ). In 54 of the 120 countries for which data are available, growth in access to sewerage has not kept pace with urban growth (United Nations ). Due to how cities are structured, cities need a separate drainage system and sewerage system that is resilient to floods Here, we present the results from a rapid evidence assessment (REA) in which we systematically reviewed and summarized the available, impact evaluations of three major urban WaSH infrastructure interventions (piped water chlorination, drainage, and sewerage) in low-and middle-income countries (LMICs). The primary objective of this review was to identify the expected health, social, and economic effects of these three interventions in urban LMICs. An REA uses a systematic process to search and screen studies but limits the scope and search strategy of the review to ensure that it is 'rapid' (Barends et al. ).
This review is an REA as we have limited the scope of the review to focus only on three interventions (piped water chlorination, drainage, and sewerage) in urban settings.
We have also narrowed the search strategy to five databases and exclude sector-specific databases. We also comment on the factors affecting the impact of the intervention and additional considerations that should be incorporated into the design of these interventions. We hope this work is used by policymakers and practitioners in informing the design and implementation of their initiatives to improve access to water and sanitation.
In the subsequent sections, we review the theory of change for the three interventions included in this study.
We then provide the methods used to produce this REA and summarize the results from the included studies. We leverage the literature search that was conducted for this REA in order to comment upon common challenges in the implementation, sustainability, and evaluation of WaSH infrastructure interventions. We conclude by discussing the synthesized results, limitations, and implications.

Interventions and theoretical model
We considered three interventions which are likely to be impactful in urban settings: chlorination of the centralized water distribution system, expansion of the drainage system, and installation of the sewerage system. These interventions were chosen because the Millennium Challenge Corporation, an implementing organization, indicated a lack of sufficient, synthesized information in the available literature. The interventions are theorized to decrease population-level exposure to pathogens by improving drinking water, decreasing the amount of standing water, and decreasing exposure to fecal contamination ( Figure 1

METHODS
The protocol for the REA was finalized a priori to decrease risk of selection bias (Supplementary Appendix A). The following section outlines the search process and inclusion criteria for this review.

Screening
The selection of studies for data extraction as part of the review was managed using EPPI-Reviewer 4 software (EPPI) and completed by implementing the standard steps of de-duplication, title and abstract screening, and then full-text screening (Supplementary Appendix D). Screening was done independently by two reviewers. Disagreements were reconciled through conversation between the two reviewers.

Population (types of study participants)
The review includes any study with participants residing in urban settings in LMICs. Countries were defined to be LMIC using the WHO definition (WHO ). We collected data on differential effects and experiences for sub-populations as far as it is possible and useful to do so using (Where Progress stands for place of residence, race/ethnicity, occupation, gender, religion, education, socioeconomic status, and social capital, and 'Plus' refers to personal characteristics associated with discrimination, features of relationships, and time-dependent relationships.).

Interventions
The interventions considered here are limited to the chlorination of drinking water in the centralized water distribution system, expansion of storm drainage systems, and installation of a sewerage system. Other WaSH infrastructure, including on-site treatment and disposal, behavior change, or household-level interventions, were not considered for inclusion. Although we recognize their importance, they fall outside the scope of this REA, which was meant to respond to the expressed needs of an implementing organization. Inclusion was based on the intervention implemented and not whether or not it achieved safely managed water or sanitation, in accordance with JMP guidelines because this would require significantly more information regarding the relative success of implementation. We were not able to assess from the studies if these interventions actually achieved 'safely managed'.

Comparison group and study design
We include impact evaluations that employ an experimental or quasi-experimental design and/or analysis method, which seek to robustly measure the net change in outcomes that are attributed to an intervention or policy as compared with some appropriate counterfactual. We include randomized studies and non-randomized studies that attempted to address issues of confounding and selection bias.

Outcomes
The review considers outcome(s) that assess a change in some indicator of health, well-being, economic growth, and/or the bacterial content of water (Supplementary Appendix E).

Date, language, and form of publication
Only studies published in English are included. Studies are included if their publication date is between 1900 and December 2019, when the search was concluded. Only peer-reviewed publications are included.

Data extraction
All articles included after full-text screening underwent the same data extraction process. Two reviewers read the manuscripts and recorded the information in a template similar to Supplementary Table S1. The extracted information from the two reviewers was then combined by a single reviewer with verification by the second reviewer. The results tables were generated by the second reviewer from the combined, extracted data. Narrative summaries of results were written by a single reviewer and reviewed by the other. Due to the heterogeneity in the interventions, outcomes, and comparison groups, a meta-analysis was not conducted.

Reporting bias
Although there was no formal assessment of bias, we expect that there is considerable risk of reporting bias. Organizations which conducted interventions that did not have positive impacts may choose not to publish their results. Furthermore, journals may not wish to publish articles on projects that had null impacts. In addition, because there is a strong prior hypothesis that these interventions will improve health, results that do not support this hypothesis may be dismissed as erroneous even if they are accurate.
As such, results should be interpreted cautiously. Even interventions with several studies indicating positive effects could be subject to this bias as there could be many more unpublished interventions showing a null or even negative effect.

Inputs regarding implementation, sustainability, and evaluation
In an effort to support the adoption of evidence-informed policy, we have provided additional inputs regarding the implementation, sustainability, and evaluation of chlorination, drainage, and sewerage interventions. These comments were developed alongside the REA. During the screening process for this REA, studies were flagged if they provided relevant information regarding the implementation, maintenance, or evaluation of the relevant interventions but otherwise did not meet the inclusion criteria for this assessment. For the most part, these studies had inappropriate comparators and/or insufficient evaluation design to meet our inclusion criteria. Actionable information for the use of practitioners in implementing, maintaining, and evaluating these interventions was extracted from these studies and is presented after the main results.

Search results
Through the search for WaSH infrastructure interventions, we identified 1,920 papers, of which 1,483 remained after  Table S1). Three studies reviewed evidence from multiple countries. Most studies were either case-control (8) or open-cohort (4). Three systematic reviews were included, and the remaining studies were closed cohort (1), cross-sectional (1), or a combined cross-sectional design with a closed cohort (1). Eight studies focused on the installation of a sewerage system alone. Five studies con-  Table S2). However, typhoid (2), malaria (1), dengue (1), and visceral leishmaniasis (1) were also considered.

Results from impact evaluations
Chlorination has mixed effects on communicable disease outcomes. One of the three studies that looked at the impact of chlorinating the municipal water on health found that chlorination reduces the likelihood of contracting a communicable disease (Supplementary Table S3).  Sewerage reduces disease. Generally, sewerage reduced disease in the identified studies. In Brazil, Barreto et al. () found that the prevalence of diarrhea decreased by 19% after the installation of a sewerage system (Prevalence Ratio: 0.81; 95% CI: 0.78-0.86). In Iran, after the sewerage system was installed, 9% of the reduction in diarrhea in the intervention group was due to the intervention (Kolahi Combining sewerage with chlorination or drainage interventions does not increase effectiveness. Combined sewerage and chlorination interventions did not consistently reduce disease burden. In Yemen, Klasen et al. () found that having chlorinated piped water increased disease burden in both the mountain and coastal regions (mountain β ¼ 0.0399, t-value ¼ 1.98, p < 0.05; coastal β ¼ 0.0455, t-value ¼ 2.76, p < 0.01). However, the improved sanitation system increased disease burden in the mountain regions but there was a non-statistically significant decrease in coastal Gasem et al. () found that households that did receive the chlorination and sewerage interventions had increased odds of typhoid fever compared with the control households (chlorination OR: 7.19, 95% CI: 1.33-38.82; sewerage OR: 29.18, 95% CI: 2.12-400.8).
Similarly, sewerage and drainage interventions did not have consistent effects on disease burden. Butala et al.
() found that health insurance claims for waterborne illnesses decreased in slums that received the slum upgrading intervention relative to those that did not (β ¼ -0.62, SE ¼ This study found that the risk of diarrhea among intervention households was 0.30 relative to controls. Turley et al. () considered a suite of interventions referred to as 'slum upgrading'. These included interventions related to sewerage and drainage. They conclude that water-related interventions are successful in reducing diarrheal disease incidence.

Heterogeneity in effects
Reports of effects by subgroup were generally limited.

Hypothesized sources of contamination
Since these studies considered the spread of disease, many identified various sources of contamination. The failure to address these other sources of contamination could be the reason why many of these studies found null results.
Hypothesized sources of contamination fell into four broad categories: general lack of improved water and sanitation, sources in the nearby environment, household hygiene practices, and issues with maintaining infrastructure systems.

Improved water and sanitation
Five studies highlighted lack of improved water and sanitation as sources of contamination. Having an inadequate sewage system may attract flies and therefore increase the This could be due to proximity to standing water or that low-income households tend to live closer to these exposures. Household hygiene practices such as unsafe storage of water, eating unwashed produce, or lack of handwashing behaviors also increased risk for bacterial con-

Study quality and risk of bias
Although a formal risk of bias assessment was not con-  Besides pipe maintenance, the entire infrastructure system needs to be properly constructed and maintained.
In Hyderabad, Pakistan, the sewage sludge treatment plants were not properly maintained and sewage was allowed to discharge directly into the river, contaminating household water sources (Yousafzai et al. ). Electrical pump failures in water distribution can lead to water contamination as water pressure drops (Moraes et al. ; Werber et al. ; Klasen et al. ). In addition, chlorination is only effective if the chlorine has enough time in the water system for it to disinfect. If the water distribution system does not have a holding tank, the water will not be adequately chlorinated for it to actually reduce pathogen exposure (Ries et al. ; Werber et al. ). For chlorination to be effective, the residual chlorine level needs to be greater than 0.3 mg/L. Lu & Zhang () found that when residual chlorine in water samples increased from <0.1 to 0.3 mg/L, bacteria cell counts decreased from 10,000 to 1,000 CFU/mL. Lu et al. () found that residual chlorine levels between 0.5 and 1.5 mg/L were able to suppress bacterial growth even when there was sufficient organic substrate.
To ensure the success and sustainability of projects that implement these interventions, these challenges should be addressed during the design phase of the project.

Considerations for the design of an impact evaluation
The design of a rigorous evaluation of the causal impacts of large-scale infrastructure projects is challenging due to logistical, political, and ethical concerns. It is rarely possible to randomize such projects. Therefore, the analytical approach taken in order to accurately estimate the impact of largescale infrastructure projects on the health and well-being of their beneficiaries should be established during the study design phase. It is recommended that a pre-analysis plan is written before the beginning of data collection.
Such a plan can be registered at RIDIE or one of the journals supported by the Center for Open Science. The benefit of such an approach is that it ensures that the necessary data are collected to allow for a successful evaluation after the study is completed.  Muti et al. () found that the number of typhoid fever cases dropped after the municipal water supply was chlorinated. Cholera has also been found to be associated with poor drainage systems and sewage overflow (Rebaudet et al. a). Although broader impacts including increased economic output and improved measures of well-being are expected, it may take several years for these to develop. As such, these are likely to be inappropriate primary outcomes for any study under 5 years.
There is likely to be an additive effect in which access to sewage and chlorinated water may improve health outcomes more than access to each one individually (Baltazar et al. ). If there is a desire to quantify the impacts of each intervention separately and additively, a multi-arm trial design is necessary. Failure to implement such a design results in an inability to separate the effects of many different interventions which were implemented simultaneously, and may result in future projects replicating ineffective portions of the larger project (Butala et al. ).
Given the seasonality of rainfall in most regions of the world, there is often also a seasonal variation in the incidence of cholera and likely other waterborne diseases (Rebaudet et al. b). As such, it is recommended that studies include matched data collection periods during different seasons over at least a 2-year span. Other climate variations such as unusually heavy flooding may also affect results, but will be harder to account for (Rebaudet et al. b).
The microbial content of the same piped water system varies by location: residential, industrial, sewage hub, and more (Bohra et al. ). Therefore, if water quality is of interest, it may be appropriate to collect water samples at several different types of locations.
There is also a need to collect information on the reliability of water and household coping strategies for intermittent water as these may affect the interventions' impact (Lee et al. ). If the piped water source is not reliable, households are more likely to rely on stored water and rely on multiple water sources. Household hygienic practices should also be collected. Household hygiene practices such as unsafe storage of water, eating unwashed produce, or lack of handwashing behaviors also increased risk for bacterial contamination (Ries et  Potential confounders identified in the literature that may be on the pathway between WaSH interventions and health/well-being outcomes can be found in Table 1.

DISCUSSION
Although there was considerable heterogeneity in results, sewerage and chlorination interventions tended to be successful in

Implications
Generally, sewerage and chlorination interventions were found to have positive impacts by reducing disease burden when adequately implemented. The impacts of drainage interventions and combined interventions were mixed. Many of the interventions included in this REA were not implemented successfully, which may have contributed to the mixed results.
Household hygiene practices, the surrounding neighborhood/environment, and lack of improved water and sanitation modified the impact of these interventions.

Strengths, limitations, and future directions
The primary strength of this work is that it is a result of a rigorous and systematic search of the available peer-reviewed literature. However, due to the rapid nature of this assessment, the search was limited to only six databases. We did not perform any searches in the grey literature or in specialist websites. There is a possibility that there may be additional evidence related to the three WaSH infrastructure interventions that was not captured in our search. However, we were still able to identify 18 unique studies through our systematic search. We found that the studies included in our REA considered the impacts on intestinal diseases. We did not find any studies looking at other health outcomes, social well-being, or economic growth nor any studies looking at the impact of drainage interventions alone on health or societal benefits. None of these studies took place in the African continent which is a limitation of the generalizability of this work to that region.
Additionally, only a few of the included studies looked at heterogeneity and reported effects by subgroups. The subgroup analyses did not take into account wealth disparities or other equity measures that could impact WaSH programming. This suggests that these are areas where further research is needed. The majority of the studies included were case-control studies where the primary objective was not to evaluate an intervention but to identify potential causes of a disease outbreak. Rigorous studies that are designed a priori with the intent to evaluate impact are needed to understand if WaSH infrastructure interventions can improve health and social well-being.
We make no discussion of safely managed water or sanitation, as defined by the Joint Monitoring Programme, in this study. This is because we are unable to speak to whether or not the included studies achieved safely managed standards.
Although many of these interventions, if implemented properly, would have qualified, we have evidence that they did not, in fact, meet these standards. As such, we limit the discussion to the interventions and not these standards. Future impact evaluations on these interventions should report on intervention fidelity as well as these intermediate outcomes so that effectiveness results can be interpreted using these indicators.

CONCLUSION
Generally, chlorination and sewerage were successful in reducing disease burden when they were implemented alone.
Many of the authors of these papers indicated that implementation issues, such as intermittent water supply, leaky pipes, and ineffective chlorination, affected the impact of the interventions. These challenges likely had the dual effect of decreasing the impact of interventions and making it impossible to determine what the impact of interventions would have been had they been implemented properly. For these interventions to be successful, the infrastructure must be maintained and should be sustainable. We hope that this work will highlight the importance of planning for sustainability in the implementation of infrastructure works. But sustainable infrastructure is not sufficient. Behavior change interventions are needed to ensure that the household engages in proper hygiene practices in order to gain the maximum benefit from infrastructure works. We hope that practitioners use this work when designing their own WaSH infrastructure interventions.