Introduction

Diarrhoeal diseases remain a leading cause of mortality and morbidity worldwide, with aetiology varying by age and setting [1]. Cholera is an acute diarrhoeal disease caused by toxigenic Vibrio cholerae O1 or O139. Modelling suggests that 2·9 million cholera cases (95% CI: 1·3–4·0) and 95 000 deaths (95% CI: 21,000–143,000) occur annually, primarily in Africa and Asia [2]. In sub-Saharan Africa, over half of the cases are concentrated in high-burden “hotspots”, which represent less than 4% of the population in the region [3]. Frequent underreporting and a low-specificity standard case definition, combined with limited surveillance capacity, make it challenging to accurately estimate the respective burdens of cholera and other acute diarrhoeal diseases [2,4].

The Global Task Force on Cholera Control has developed a roadmap to achieve a 90% reduction in cholera mortality by 2030 [5]. The multisectoral cholera elimination strategy prioritises targeted interventions in cholera hotspots along three pillars: a) surveillance and health systems strengthening; b) oral cholera vaccination (OCV); and, c) long-term improvements in water, sanitation and hygiene (WASH) access. The third component aligns closely with the United Nations’ Sustainable Development Goal (SDG) indicators 6.1 and 6.2 to achieve universal access to safely managed water and sanitation services by 2030 [6].

Globally, in 2019, 69% of the diarrhoeal disease morbidity burden and 1·04 (95%CI: 0·93–1·16) million deaths due to diarrhoeal diseases were estimated to be attributable to inadequate WASH access [7]. In 2020, 771 million people lacked access to basic drinking water services, including almost half in Sub-Saharan African countries with a high burden of diarrhoeal diseases [8]. Improving access to safely managed water services can prevent waterborne transmission of diarrhoeal diseases through microbiologically safe drinking water and increased water availability for personal and domestic hygiene. The latter may be critical for cholera control given the importance of within-household transmission during outbreaks [9,10].

Although historical evidence suggests that water and sanitation infrastructure contributed to mortality reductions from infectious diseases in Europe and North America, there is limited scientific literature on the impacts of water supply infrastructure improvements intended to reduce the burden of diarrhoeal diseases and cholera [11,12]. The health impacts of WASH interventions in complex emergency settings also remain under-researched [13]. Besides accessibility challenges, these gaps may be due to the methodological, logistical and financial challenges related to the study of complex interventions, which are exacerbated in the resource-limited and often unstable contexts where the cholera burden is concentrated.

In 2014, a major infrastructure programme was initiated to improve the piped water supply in Uvira, a town located within a cholera hotspot in Eastern Democratic Republic of the Congo (DRC), in line with the national multisectoral cholera elimination plan 2013–2017 [3,14]. The programme was led by REGIDESO, the national water utility, with support from the Agence Française de Développement (AFD), Veolia Foundation (VF), and the European Union. We investigated the relationship between changes in the piped water supply service and the incidence of suspected and confirmed cholera cases admitted to the main cholera treatment facilities of the town.

Methods

Participants

The population of interest for the study was the residents of the town of Uvira, including all age groups, as the water supply improvement programme was intended to benefit all Uvira inhabitants. Population numbers per avenue (the smallest administrative unit) were available for November 2017 from Uvira authorities [22]; the population distribution between avenues was assumed to remain constant over time and combined with more recent official records in order to estimate the population size by cluster in subsequent years (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Estimatead population size, suspected and confirmed cholera cases by cluster.

https://doi.org/10.1371/journal.pntd.0012265.t001

In order to estimate suspected and confirmed cholera incidence in the town of Uvira, anonymised patient information as recorded in the CTC/CTU registry, including age, sex, address, dates and length of stay were shared with the study team by Uvira health authorities. All patients admitted to the CTC and CTU with acute watery diarrhoea (i.e. passing 3 or more loose or watery stools in 24 hours) were considered suspected cholera cases and eligible for participation in the study. From April 2016, participants were recruited continuously to collect a rectal swab for enrichment in alkaline peptone water (Himedia, India) for 6 hours followed by cholera confirmation using rapid diagnostic tests (RDT) (Crystal VC O1/O139, Arkray Inc., India). Rectal swabs were selected to minimise the risk of contamination and the enrichment procedure was estimated in a previous study to have a sensitivity of 92% and a specificity of 91% [27,28]. A recent systematic review found pooled estimates of 83% (67–92%) and 98% (94–99%) for sensitivity and specificity, respectively, for testing of APW-enriched samples using VC Crystal kits [29]. Enrolment was on a continuous basis, 7 days a week, with testing performed during daytime, typically within 24 hours or less following admission. No restriction was placed on patient age. Written consent was obtained from all participants prior to sample collection and written parental consent was sought for patients under 15 years old.

Results

A total of 4556 suspected cholera cases were admitted to cholera treatment facilities in Uvira and enrolled in the study between January 2017 and December 2021 (Table 1). The majority (92·3%) sought care at the General Hospital CTC and 50.0% were recorded as female. Age was available for 4463 (97·9%) patients; the median was 20.0 years (IQR 9.0–40.0) and patients aged five years or younger represented 16·4% of admitted cases (compared to an estimated 12.8% of the Uvira population based on 2019 official records). Twenty-three deaths were recorded among patients admitted to cholera treatment facilities between 2017 and 2021, including seven in patients who tested positive for cholera.

Rectal swabs were collected from 3309 patients (72·6%). The proportion of patients tested varied between 63% in 2018 and 89% in 2021 (Table B in S1 Table). The main reason for missing tests was the absence of the responsible laboratory technician. Among samples subjected to rapid diagnostic tests, 45·4% were positive (Fig 1). The average monthly incidence rate of suspected cholera across the entire town of Uvira was 4·0 per 10 000 inhabitants. When conservatively categorising untested suspected cases as rapid test negative, the mean monthly incidence rate of confirmed cholera was estimated to be 1·3 per 10 000. Variations in incidence rates were observed between clusters and over time (Table C in S1 Table and Fig A in S1 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Monthly incidence of suspected and confirmed cholera in Uvira, DRC by rapid diagnostic test outcome (colour).

https://doi.org/10.1371/journal.pntd.0012265.g001

The water supply service quality as described by an index from 0 to 100%–the main exposure of interest in our analysis–ranged from 6·7% (“very low”) to 38·1% (“medium-low”) by cluster in January 2017, prior to the start of construction works in Uvira. The observed change between January 2017 and December 2021 ranged from +1·8% to -18.3%. The change was negative or null in 14 out of 16 clusters, with an average of -5·4%, a median of -3·0%. In December 2021, the index ranged from 6·7% (very low) to 25·0% (low). This evolution was primarily driven by the limited quantity of water available and increasing frequency of water supply interruptions. The estimated volume of water delivered remained under 20 L per capita per day during >95% of the observation period and showed a decreasing trend, as the number of connections to the network substantially increased (with the “accessibility” component of the index reaching 100% in four clusters) while production capacity remained limited (Fig B in S1 Fig). Pumps operated 58% of the time on average and electricity supply deteriorated over time, leading to reduced hours of operation and more frequent service interruptions [16].

Overall service quality was associated with confirmed cholera incidence (RR 0·84, 95% CI 0·73–0·97, p = 0·024), suggesting that a 5-percentage point increase in the water service quality index could reduce confirmed cholera incidence by 16%. Concretely, in Uvira, such an increase in the index value could in most cases be achieved by increasing the quantity of water available to users. Water quantity (RR 0·76, 95% CI 0·61–0·94, p = 0·020) and service continuity (RR 0·75, 95% CI 0·69–0·81, p<0.001) were also individually associated with confirmed cholera (Fig 2), meaning that increasing the quantity of water available by 5 L per user per day or increasing monthly pumping hours by 5% (i.e. approximately 1.2 hour per day) were associated with a 24% and 25% reduction in the incidence of confirmed cholera, respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Rate ratios for 5-unit increases in accessibility, continuity, affordability, and overall service quality, or a 5 L/p/day increase in water quantity for both all suspected cholera cases and those who were positive with a rapid diagnostic test (“confirmed”).

https://doi.org/10.1371/journal.pntd.0012265.g002

Rate ratio estimates were similar for suspected cholera incidence and the strongest association was with service continuity, a 5% increase in pumping hours associated with a 19% rate reduction (RR 0·81, 95% CI 0·77–0·86, p<0.001) (Fig 2).

Neither confirmed nor suspected cholera incidence was associated with accessibility or affordability of the water service (Fig 2).

A sensitivity analysis to assess the impact of modelling secular trend using a categorical form, compared to the quadratic form used in the main model, suggested that the only association that would substantially change and become weaker is that between continuity of supply and suspected cholera cases (Fig C in S1 Fig).

Discussion

We studied a large-scale intervention to improve piped water supply in Uvira (DRC). Our results support the case that water supply service improvements can reduce acute diarrhoeal diseases and, in particular, cholera. They provide evidence based on five years of water supply monitoring and clinical surveillance in a town with endemic cholera and a high burden of diarrhoeal diseases. Challenges in Uvira are typical of those found across cholera hotspots in the broader region, including vulnerability to climate-related natural hazards, protracted conflict, constant population movements, and multiple health emergencies [37].

The comparison of different water supply service components underlines the importance of water quantity and service continuity. The association with continuity, although sensitive to assumptions about seasonality, is consistent with previous analyses of data from the same setting [38]. It is a reminder that the quality of water supply services, beyond infrastructure improvements, is critical to reduce the disease burden, and suggests that an incremental approach to improving water supply services is unlikely to yield significant progress in reducing cholera and other acute diarrhoeal diseases in settings such as Uvira. Our results thus provide support for the higher level of service envisaged under SDG 6. Concretely, in order to achieve public health and other benefits, programmes aiming to prevent diarrhoeal disease and cholera in urban settings through an improved piped water supply should: (i) design and invest in infrastructure solutions that can deliver a “safely managed” water service in the long term, resilient to a changing climate; and (ii) integrate sufficient technical and organisational support to water utilities to ensure an adequate system operation. In the Uvira context, where water service intermittence is closely linked to electricity supply issues, our results also call for enhanced cross-sectoral planning and coordination [16].

Our results were similar for confirmed and for suspected cholera (i.e. acute diarrhoeal diseases), with slightly smaller effects for the latter that could reflect the variable susceptibility of different pathogens to water supply service improvements. Case confirmation via rapid diagnostic tests suggests that a substantial proportion of the patients admitted to cholera treatment facilities in Uvira are not infected by Vibrio cholerae O1 [27]. A wide range of enteric pathogens circulate in Uvira, with enterotoxigenic Escherichia coli, Cryptosporidium spp. (resistant to chlorination), and Campylobacter (predominantly foodborne) detected most frequently among patients admitted to the CTC after V. cholerae [32]. This diversity is typical of settings with a high burden of diarrhoeal disease [1].

There are three main limitations to our study related to (i) implementation challenges, (ii) documentation of the exposure of interest, and (iii) care-seeking behaviour.

We had initially planned to conduct a SW-CRT whereby new taps would be installed in a randomised order across the town of Uvira [15]. However, due to multiple disruptions (extreme flooding, lake level rise, Covid-19), adherence to the intervention protocol was not possible. The selection of a SW-CRT design was based on the considerations that it would strengthen our ability to draw causal inferences and a priori be compatible with a sequential roll-out of the construction works. As a consequence of the implementation challenges and of using an observational approach for the analysis, we could neither clearly distinguish situations before and after the intervention nor evaluate the impact of the intervention. Instead, the focus was on observed variations in the water supply service (exposure) and estimated cholera and diarrhoeal disease incidence (outcomes), amidst complex influences. Population movement is an important factor which could not be documented systematically and integrated into our modelling. However, our experience suggests that realistic expectations should be set, and pragmatic approaches adopted, for the rigorous evaluation of non-clinical, complex interventions in challenging emergency settings. In particular, pre-specifying risks and analysis strategies for different scenarios as we did in the published protocol for this study [15] reduces potential bias.

A second limitation lies in the documentation of the exposure of interest, namely changes in the water supply service. The index used in our analyses is primarily based on utility data, which were collected retrospectively and may not always reflect users’ experiences. Continuity was approximated for the entire town by a centralised, indirect measure (hours of operation of the pumps). The continuous monitoring of water quality was not possible–water samples were only collected and tested during household surveys–and water quality is a critical dimension of the water supply service quality that is not included in our analyses.

A third limitation is the reliance of our study on clinical surveillance based at the official cholera treatment facilities, which overlooks the multiplicity of private and informal options available for diarrhoeal disease and cholera care in Uvira and may have been impacted by changes in care-seeking behaviour over time linked to the Covid-19 pandemic. Among respondents who reported seeking care for diarrhoea in a survey conducted in September 2021 in Uvira (n = 160), 70% said that they went to a pharmacy and less than 5% mentioned a public healthcare facility [16,26]. While it seems reasonable to assume that the most severe cases are referred to the main cholera treatment centre located at the Uvira General Hospital, three community deaths likely due to acute diarrhoea were reported in Uvira between September 2021 and January 2023 [39], and many cases certainly remained undetected in our study. While decreases in attendance of health services have been documented during the Covid-19 pandemic, including in Kinshasa [40], we were not able to assess such changes in our study. If possible, clinical surveillance should be complemented by community-based cholera detection in future research projects in order to provide a more reliable picture of the cholera and diarrhoeal disease burden.

We assessed the relationship between changes in the piped water supply service and diarrhoeal diseases in an urban setting with endemic cholera and a high diarrhoeal disease burden in the context of a largescale programme to improve water supply, in Eastern Democratic Republic of the Congo. Our results provide insights into the potential challenges and effects of water supply interventions in similar settings, where large investments in water supply and sanitation services will be necessary to accelerate progress towards SDG6 and the global target of cholera elimination by 2030. We found strong associations between the continuity of water supply and both confirmed and suspected cholera, as well as between the quantity of water delivered and confirmed cholera. This implies that water infrastructure programmes must achieve high standards of service–providing a sufficient, continuous water supply–in order to materialise expected public health benefits in terms of diarrhoeal disease prevention.

Supporting information

Acknowledgments

We would like to thank the Uvira Health Zone and OXFAM International in DRC for their support to this project. We are grateful to the French Agency for Development and the Veolia Foundation, AGC, Arab Contractors, REGIDESO, and the teams of construction workers, all of whom worked and are working tirelessly to improve the Uvira water supply network. We would like to thank all study participants.

Lastly, and most importantly, we would like to acknowledge the contributions of the late Dr Jeroen Ensink. Dr Ensink was instrumental in the conception of this study and it would not have been possible without him. He remains very much in our minds as we conclude this study and is still sorely missed as a colleague and as a friend.