Skip to main content
  • Loading metrics

Low-cost domestic rainwater harvesting in rural southeast Madagascar: A process and outcome evaluation

  • Jasmine Kelly ,

    Roles Data curation, Formal analysis, Investigation, Project administration, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Department of Water, Sanitation, and Hygiene, SEED Madagascar, Fort Dauphin, Anosy, Madagascar

  • Mamonjisoa Tsilahatsy ,

    Contributed equally to this work with: Mamonjisoa Tsilahatsy, Tolotra Carnot, Ramanantsiadiana Wilmin Fidelos, Giona Randriamanampy, Andrianarivelo Zafindrazana Charlier

    Roles Investigation, Project administration

    Affiliation Department of Water, Sanitation, and Hygiene, SEED Madagascar, Fort Dauphin, Anosy, Madagascar

  • Tolotra Carnot ,

    Contributed equally to this work with: Mamonjisoa Tsilahatsy, Tolotra Carnot, Ramanantsiadiana Wilmin Fidelos, Giona Randriamanampy, Andrianarivelo Zafindrazana Charlier

    Roles Investigation, Project administration

    Affiliation Department of Water, Sanitation, and Hygiene, SEED Madagascar, Fort Dauphin, Anosy, Madagascar

  • Ramanantsiadiana Wilmin Fidelos ,

    Contributed equally to this work with: Mamonjisoa Tsilahatsy, Tolotra Carnot, Ramanantsiadiana Wilmin Fidelos, Giona Randriamanampy, Andrianarivelo Zafindrazana Charlier

    Roles Investigation, Project administration

    Affiliation Department of Water, Sanitation, and Hygiene, SEED Madagascar, Fort Dauphin, Anosy, Madagascar

  • Giona Randriamanampy ,

    Contributed equally to this work with: Mamonjisoa Tsilahatsy, Tolotra Carnot, Ramanantsiadiana Wilmin Fidelos, Giona Randriamanampy, Andrianarivelo Zafindrazana Charlier

    Roles Investigation, Project administration

    Affiliation Department of Water, Sanitation, and Hygiene, SEED Madagascar, Fort Dauphin, Anosy, Madagascar

  • Andrianarivelo Zafindrazana Charlier ,

    Contributed equally to this work with: Mamonjisoa Tsilahatsy, Tolotra Carnot, Ramanantsiadiana Wilmin Fidelos, Giona Randriamanampy, Andrianarivelo Zafindrazana Charlier

    Roles Conceptualization, Investigation, Project administration, Resources, Supervision

    Affiliation Department of Water, Sanitation, and Hygiene, SEED Madagascar, Fort Dauphin, Anosy, Madagascar

  • Emilie Kowalczewski,

    Roles Data curation, Formal analysis, Investigation, Methodology, Project administration, Software, Supervision, Writing – review & editing

    Affiliation Department of Water, Sanitation, and Hygiene, SEED Madagascar, Fort Dauphin, Anosy, Madagascar

  • Lomba Hasoavana,

    Roles Resources, Supervision

    Affiliation Department of Water, Sanitation, and Hygiene, SEED Madagascar, Fort Dauphin, Anosy, Madagascar

  • Mamy Andriatsihosena,

    Roles Conceptualization, Methodology, Project administration, Resources, Supervision

    Affiliation Department of Water, Sanitation, and Hygiene, SEED Madagascar, Fort Dauphin, Anosy, Madagascar

  • Harry Chaplin,

    Roles Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing – review & editing

    Current address: Tatirano Social Enterprise, Fort Dauphin, Anosy, Madagascar

    Affiliation Department of Water, Sanitation, and Hygiene, SEED Madagascar, Fort Dauphin, Anosy, Madagascar

  • Hugo Legge

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Department of Water, Sanitation, and Hygiene, SEED Madagascar, Fort Dauphin, Anosy, Madagascar


In settings where communities rely on unimproved water sources, household rainwater harvesting (HRWH) may improve the quality and quantity of water available. This research presents results from a two-year controlled before-and-after study that evaluated the impact of low-cost HRWH on household water collection habits, hygiene practices and prevalence of childhood diarrhoea in rural Madagascar. The study assessed system functionality, water quality and the acceptability of requesting household financial investment (16–20 USD). Surveys were administered to enrolled intervention households (n = 138) and control households (n = 276) at baseline and endline. Water quality tests at endline compared microbial contamination in a sub-sample of HRWH systems (n = 22) and public water sources (n = 8). Difference-in-difference analyses were used to compare changes in outcomes between study arms at baseline and endline. At endline 111 (75%) of systems were functional with an average age of 1.25 years. Microbial contamination was 39.3 TTC/100ml in community water sources compared with 23.3 TTC/100ml in the HRWH systems (coef: -16.0, 95CI: -37.3 to 5.2, p = 0.133). 85 (57%) of households completed their repayment plans while remaining households owed on average 3.7 USD. There was weak evidence to suggest that intervention households collected more water per capita day than controls (adj coefficient: 3.45; 95CI: -2.51 to 9.41, p = 0.257). Intervention households had 11% higher absolute risk of owning a handwashing station compared against controls (95CI: 0.00 to 0.23; p = 0.06). There was no evidence of differences in ownership of soap or prevalence of childhood diarrhoea between study arms. Overall, operation and maintenance of the systems remained high, users demonstrated willingness to pay, and there was weak evidence that water provision at the household increased domestic consumption. However, the systems did not provide contaminant-free water. We conclude that HRWH using low-cost, locally available materials can increase household access to water in areas reliant on limited communal water sources.


Access to a safely managed water supply is a key part of Sustainable Development Goal (SDG) 6 and is associated with increased household consumption of water, improved hygiene practices and lower incidence of diarrhoeal disease [13]. While progress towards SDG6 continues to improve globally, this progress is not equitable and of the 86 countries currently with less than 95% access to a basic water supply, 71 are not on target to reach universal coverage by 2030 [4].

An important factor limiting access to safe water is the poor sustained functionality of water supply interventions [5, 6]. Interventions that require financial contributions from users have been shown to increase water source lifespan [7]. However, in some rural settings, these approaches are not always enough to make a community water supply sustainable [8]. In these instances, interventions at the household level that offer a supported self-supply, such as household rainwater harvesting (HRWH), offer a promising but under-researched alternative to community water sources, such as a public tap-stand or well [9, 10]. Household water supplies have the additional benefit of reducing the amount of time a household spends collecting water as well as increasing the amount of water consumed by the household, which in turn is associated with improved hygiene practices by household members [11, 12].

Rainwater harvesting (RWH) is a broad term used to describe the collection, storage and use of rainwater. It has been utilised in a variety of climates and socioeconomic settings as a solution to water scarcity, intermittence of water supply, and inadequate drinking water quality [1315]. The joint monitoring programme currently lists RWH as an improved water source and previous studies of HRWH interventions have indicated that the technology can yield potable water with low levels of microbial contamination [8]. However, water quality and yield are highly dependent on design specification and previous studies have shown that rainwater should not be consumed without treatment [8]. While widespread as a practice in some countries, HRWH has achieved only limited coverage in many settings within Sub-Saharan Africa and Asia where there are suitable climates as well as the requisite need for reducing reliance on surface water as a primary drinking water source. This under-utilisation of HRWH is in part due to the reliance on technology, materials and expertise that are not accessible in many rural, low-income settings [1, 2, 6].

This paper presents results from a process and outcome evaluation for a partially subsidized, domestic rainwater harvesting intervention delivered in rural southeast Madagascar over a 28-month period. The outcome evaluation assesses the impact of the intervention on water collection habits, including per capita household consumption and time spent collecting water, access to hygiene facilities, and prevalence of diarrhoea in children under five years of age. The process evaluation examined intervention acceptability, adherence to household payment schedules, HRWH system functionality, and the water quality parameters of water stored in the HRWH systems. The findings can help to inform national water, sanitation, and hygiene (WaSH) programmes designing alternative water supply financing and delivery mechanisms that seek to address sustainability of WaSH in Madagascar and other countries with similar profiles in Sub-Saharan Africa and beyond.


Study location

Madagascar is the fourth lowest ranked country in the world on water, sanitation and hygiene indicators [16] and this coincides with stagnated health indicators, such as high child and maternal mortality [17]. The Anosy region is located on the south-eastern most point of the island and is divided into three districts, and sub-divided into 64 communes. The majority of the region’s households (81%) are in rural communities. Despite a relative abundance of rainfall in the region (1,700 to 3,500 mm annually), access to improved water sources is limited [17, 18]. This study took place in eight target villages (fokontany) within the Mahatalaky Rural Commune (MRC), which were selected based on their accessibility and the known reliance of households on limited or unimproved communal water sources such as unprotected dug wells, unprotected springs, and rivers.

Intervention background

The intervention was implemented from September 2017 to December 2019 by the British non-governmental organisation (NGO), SEED Madagascar, to address limited access to drinking water sources and hygiene in the Anosy region. The full intervention design has been previously described in detail [19]. In brief the study was conducted with beneficiary and control households in eight target villages (Fig 1). Households that expressed interest, demonstrated a demand for the intervention, and were willing to pay an initial instalment of 2.5 USD for the HRWH system were enrolled in the study as intervention households. Control households were enrolled into the study on a 2:1 ratio with intervention households and were matched by date of enrolment and randomly by geographic location (i.e., the closest, eligible households). Specifically, control households were sampled by spinning a pen at the location of intervention households and selecting the two closest households in the direction the pen was pointing. This approach was based on the EPI sampling method [20]. Only households that had not previously purchased or were in current ownership of a HRWH system were invited to participate as control households. Following enrolment, intervention households had a 250L freestanding HRWH system fitted onto either metal or palm leaf (ravinala) roofs (Figs 2 and 3). In addition, three hygiene promotion and operations and maintenance seminars were delivered to each household over the course of the intervention. Households continued to pay monthly instalments (average 2.5 USD) over a 6-month period until the subsidised cost of the system had been repaid. For palm leaf systems the total amount paid was 16 USD and for metal systems the total amount was 20 USD, representing 26% of the material cost for each system type.

Fig 1. Map of the Anosy region of Madagascar and eight districts in the Mahatalaky Rural Commune.

Map showing the study location in the Anosy region and eight target communities in the Mahatalaky Rural Commune. Base map and data from OpenStreetMap and OpenStreetMap Foundation (—). Contains information which is made available under the Open Database License: Any rights in individual contents of the database are licensed under the Database Contents License: [21].

Fig 2. HRWH system fitted to a metal roof.

[Image of a domestic rainwater harvesting system supplied by the “Tatirano” project, fitted onto the metal roof of a beneficiary household [22]. Photo taken by Harry Chaplin.

Fig 3. HRWH system fitted to a palm leaf roof.

[Image of a domestic rainwater harvesting system supplied by the “Tatirano” project, fitted onto the palm leaf roof of a beneficiary household [22]. Photo taken by Harry Chaplin.

Study design

The study was comprised of two components; 1) an impact evaluation assessing the degree to which the intervention impacted water collection behaviours, hygiene related outcomes, and childhood diarrhoea, and 2) a process evaluation examining the intervention’s durability, implementation fidelity and community acceptability (Fig 4).

Fig 4. Timeline of study in line with relevant project activities.

Visual showing the study timeline relative to the “Tatirano” project activities.

Impact evaluation design.

The evaluation took the form of a controlled before and after study where data were collected pre-intervention and post-intervention from participating households (intervention group) and non-participating households (control group). The intervention group consisted of 138 households that purchased either one HRWH system (n = 128 households) or two HRWH systems (n = 10 households) and participated in two hygiene promotion and system maintenance workshops. Prospective control households were enrolled on a 2:1 ratio with intervention households (n = 276) and matched by geographic proximity and date of enrolment. To establish the final control group, propensity score matching (PSM) was used to match prospective control households with intervention households on a 1:1 basis using nearest neighbour matching (Fig 5). Scores were calculated based on a composite wealth index score generated at baseline, which was hypothesised as being predictive of enrolment in the intervention arm as wealthier households had greater financial means to invest in a HRWH system.

Fig 5. Consort flow diagram of the outcome evaluation including propensity score matching.

Sample size calculations were based on the outcome of per capita household consumption of water. In the absence of data from the study site or Madagascar as a whole, we used results from a systematic review of studies measuring unmetered domestic water consumption to assume an increase of 10 litres per capita day (lpcd) as a result of the intervention [23, 24]. Assuming a 10% loss to follow up and an intra-class coefficient of 0.05, we calculated that 133 households per study arm would have a power of 80% and a significance of 0.05 to detect a true difference of this magnitude between study arms.

Impact evaluation outcomes.

The impact of the intervention was evaluated by comparing results for two primary and three secondary outcomes between study groups at post-intervention. These included primary impact outcomes:

  1. Mean household daily per-capita water collection time
  2. Mean household daily per-capita water consumption for domestic purposes

And secondary impact outcomes:

  1. Prevalence of diarrhoea in children under 5 two weeks prior to survey
  2. Proportion of households with a handwashing station within 10 meters of the house
  3. Proportion of households with soap on the premises

Impact evaluation sampling and data collection.

Data collection methods to measure these outcomes involved a household survey administered to all participating households at baseline and then again at endline, administered using the open data kit (ODK) mobile data collection platform, as described in S1 Text [36].

Impact evaluation data analysis.

Household surveys were cleaned and analysed for statistical difference with Microsoft Excel and STATA16. Difference-in-difference analyses using ordinary-least squares regression with robust standard errors were used to estimate the effect of the intervention on outcomes of interest. All models included wealth index as a covariate to control for residual differences in socioeconomic status between study groups following PSM. Wealth index scores were calculated using principal component analysis (PCA) using 12 survey items including occupation, literacy, materials of the household structure, and ownership of common to expensive household items.

Process evaluation design.

A process evaluation was conducted during the project to assess intervention fidelity and acceptability among participants. The process evaluation addressed four primary outcomes:

  1. Proportion of beneficiary households with functional HRWH systems at project close
  2. Reported satisfaction of beneficiary households with the HRWH system at project close
  3. Proportion of beneficiaries fully paying for their HRWH system as agreed during enrolment at project close
  4. System water quality measured as number of thermotolerant coliforms per 100ml,

And two secondary outcomes:

  1. Proportion of HRWH systems in use at project close compared to project start
  2. Degree of HRWH system functionality at project close compared to project start

Process evaluation data collection instruments.

Assessment of these outcomes relied on three primary data collection instruments. Endline spot checks of HRWH systems were used to ascertain the level of operations and maintenance of systems (outcome 1). A question module included in the endline household survey was used to assess user satisfaction (outcome 2). Payment records were used to evaluate household adherence to repayment plans (outcome 3). Water quality tests measuring thermotolerant coliforms were used to evaluate the level of microbial contamination present in the HRWH systems at endline (outcome 4), as described in S2 Text [37].

System functionality was defined as the ability of the system to capture, store, and dispense water. To assess this, system spot checks of HRWH systems (n = 138) were conducted in October, during the dry season. Spot checks assessed six core binary criteria, all of which were required to pass in order to record the system as being functional. These included the presence/absence of holes in the catchment area, the position and location of the gutter, the position and location of the first flush system, and the functionality of the tap. A further 15 binary items were assessed to determine factors that were likely to impact water quality, for example presence of debris in the gutter, listed in S1 Table. Enumerators also made a qualitative assessment of the amount of water present in the tank which was marked as being full, half-full, or empty. All assessments were completed by direct observation.

Participant satisfaction was assessed at endline through a satisfaction question module included in the household survey described above. Items used a 5-point Likert scale from “very satisfied” to “very unsatisfied.” Other items included multiple choice options on specific areas to improve and open-ended short answer questions.

Beneficiary payment records were collected in Google Sheets. Each enrolled beneficiary started with a principal balance and payments were recorded monthly as they were collected, including early and late payment information.

Water for thermotolerant coliform tests was sampled from 22 randomly selected HRWH systems used by intervention households and 8 randomly selected primary drinking water sources used by control households. HRWH systems were eligible for sampling if they were accessible by the team at the time of testing and had water present. To select samples, all HRWH were randomly numbered. If the first system on the list did not meet the eligibility criteria it was discarded, and the next system was selected until 22 samples were taken. Similarly, a list of alternate water sources used by control households was generated and eight were chosen through simple random sampling. Alternate water sources were eligible for sampling if they were not HRWH systems provided by the pilot. Alternate water sources sampled included rivers, natural springs, boreholes, pumps, and wells.

Process evaluation data analysis.

For water quality analysis, final TTC counts were compiled in Microsoft Excel and analysed for mean differences by group in STATA16.

Observational system check data was cleaned and analysed in STATA16. Core functionality of HRWH systems was coded as a binary variable (functional VS not functional). Two discrete continuous variables counting the number of functionality faults, and the number of cleanliness faults that HRWH systems scored at endline were also created.

Household surveys were cleaned in Microsoft Excel and analysed in STATA16. Satisfaction was determined for each item in this module of the survey. Participants who answered “satisfied” or “very satisfied” were counted as being satisfied with that aspect of the system.

Ethical considerations

Each member of the research team was trained in research ethics concerning human subjects and obtaining informed consent. All participant-driven data was collected following either written or verbal informed consent. Participants with limited or no literacy had the contents of the information and consent form was read to them in the presence of a literate individual not connected to the study. The participant was then audio recorded providing their consent to participate. Ethical approval for this study was obtained in writing from the Madagascar Ministry of Health through the Medical Inspector of the Fort Dauphin District Health Service.


HRWH system functionality

Between September 2017 and January 2019 73 metal-roofed systems and 75 palm leaf systems were installed across eight target villages (total n = 148) (Table 1). At endline, 75% of the installed HRWH systems were functional, 17.6% were not functional and 7.5% had been repurposed for other activities or were missing. Functionality was higher among metal HRWH systems (84.9%) when compared with palm leaf systems (65.33%). Water was present in the tanks of 26.9% (30/111) of functional systems at endline. The average age of systems at endline was 457.6 days (1.25 years).

Table 1. Status of HRWH systems throughout duration of project and at the endline assessment.

Water quality assessments indicated a mean average of 23.3 thermotolerant coliforms per 100ml (TTC/100ml) (SD: 18.5, median: 24, IQR: 8–34.3) among HRWH systems at endline (Table 2). TTC counts were on average lower among metal roofed systems (mean TTC/100ml: 16.6, SD: 16.6, median: 12.9, IQR: 2.5–22.5) compared with palm leaf systems (mean TTC/100ml: 28.5, SD: 18.7, median: 29.5, IQR: 19–34.5) (Fig 6). Compared with existing community water sources both metal (coef: -16.0, 95CI: -37.3 to 5.2, p = 0.345) and palm leaf systems (coef: -22.8m, 95CI: -47.2to 1.6, p = 0.067) had lower levels of microbial contamination. However, these results provide only weak evidence that the observed differences are not due to chance.

Fig 6. Thermotolerant coliform counts.

Thermotolerant coliform counts among alternative community water sources, palm leaf systems and metal roofed systems at endline.

Table 2. Levels of thermotolerant coliforms in HRWH systems and community water sources at endline.

At endline, the average number of faults recorded per system spot check that could impact the quantity of water collected were 0.6 for metal systems and 2.2 for palm leaf systems. The average number of faults relating to the cleanliness of the system that could impact water quality were higher among metal-roofed systems (1.3) compared with palm leaf (0.6) systems.

Intervention acceptability

Payment history for all 132 beneficiary households was compiled from September 2017 through December 2019. At project close, 63 (42.6%) of households had an outstanding balance and 85 (57.3%) of households were completely paid (Table 1). Households that failed to complete repayment by endline had paid on average 79% of their payment plan and had on average 3.7 USD left to pay.

User satisfaction with the cost of the system was high with 85% of households satisfied or very satisfied with the amount paid. Satisfaction with the quality of water yielded by the systems was similarly high, with 92% of households satisfied or very satisfied with the quality of water supplied. The amount of water yielded, and the durability of the systems had lower approval with 54% and 69% satisfied or very satisfied respectively.

Behavioural and health outcomes

Between pre-intervention baseline surveys and post-intervention endline surveys, 84 (27.3%) of households were lost to follow up. All households lost to follow up were in the control group, were distributed evenly across study villages and did not significantly differ in socioeconomic and WaSH characteristics when compared with households retained in the control group, as shown in S2 Table. Endline surveys were recorded for 135 intervention households and 172 prospective control households.

Following PSM analysis, 135 control households were matched on a 1:1 ratio with intervention households while the remainder (n = 37) were dropped from the control group. Socioeconomic score was more balanced between study groups following matching (Table 3) while the distribution of control and intervention households across villages remained equal.

Table 3. Baseline sociodemographic and outcome variables in control and intervention groups before and after propensity score matching.

Among outcome indicators, per capita daily water collection, per capita time spent collecting water, presence of a handwashing station and primary source of drinking water were broadly similar prior to matching and equalised further following matching. There were significant differences between study groups in prevalence of childhood diarrhoea and presence of soap in the household prior to matching and these differences were only fractionally equalised following matching (Table 3). Intervention households at endline were disproportionately over-represented in higher SES quintiles despite the matching process and as a result SES was controlled for in all models (Table 4).

Table 4. Sociodemographic variables in matched control and intervention groups at endline.

Intervention households collected on average 28.4 litres of water per capita day (lpcd) at endline compared with 23.1 lpcd among control households (Table 5). A difference-in-difference (DiD) analysis that also controlled for SES demonstrated that intervention households collected more water than control households with an intervention effect of 3.45 lpcd. However, there was insufficient evidence to rule out the role of chance in this observed difference (95CI: -2.51 to 9.41, p = 0.257). Per capita time spent collecting water was slightly higher among intervention households (13 minutes per capita day) compared with control households (11.2 minutes per capita day), however this difference was not significant following DiD analysis (adj. coef: -3.02; 95CI: -2.73 to 5.99, p = 0.519). 25.9% of intervention households identified a handwashing station within 10 meters of the household, compared with 11.1% of control households. After DiD analysis there was an observed increase of 11% in absolute risk among intervention households for having access to a handwashing station and some evidence to reject the role of chance in observing this difference (95CI: 0.00 to 0.23, p = 0.061). Access to soap was higher among intervention households but not significantly, following the adjusted DiD analysis (adj. risk difference: 0.07; 95CI: -0.07 to 0.2; p = 0.324). Diarrhoeal prevalence in children <5 years of age was substantially lower in intervention households at endline. However, after taking into account differences at baseline and adjusting for SES the DiD demonstrated a null treatment effect (adj. risk difference: 0.02; 95CI: -0.16 to 0.2; p = 0.808).

Table 5. Summary statistics and adjusted difference-in-difference model outputs for behavioural and health outcomes in control and intervention groups at endline.


Results from this study demonstrate that a partially subsidised, low-cost HRWH system, built using locally available materials can offer an acceptable household-level water source in a low-income, water abundant environment. In this study, the majority of households continued to operate their systems until endline and among those that continued to use the system, operation and maintenance of the systems remained high at endline. Users also demonstrated a willingness to pay for the systems despite the high relative cost. However, despite the high levels of adherence to operation and maintenance, the systems were not capable of providing water free from microbial contamination. Additionally, limited storage capacity inhibited the ability of households to rely exclusively on the systems for their domestic water requirements, in turn resulting in only minor gains in household water consumption and no observed reduction in the time spent collecting water by household members.

Household adherence to and satisfaction with payment schedules was high, with most households completing their payment plans and the remainder repaying on average 79% of their payment plans. This finding suggests that households were willing to allocate significant financial resources towards a household water supply, despite having access to existing community water sources in close proximity. Previous studies evaluating cost-recovery models for community water supplies in Kenya, Uganda, and Rwanda demonstrated similar findings [25], with users’ content to allocate financial resources to water supply if the convenience and quality of the water supplied is perceived to be high. However, despite target communities being located in a region where the World Bank estimates 96% of households live below 2 USD/day [18], households enrolled in the intervention arm of this study tended to be drawn from the higher socioeconomic strata within study communities. This indicates that less wealthy households may have been prevented from enrolling due to financial barriers, a finding that has been replicated by studies in Ghana and India [26, 27]. However, as indicated in Table 5, some control households did adopt the RWH systems as their primary source of drinking water by endline (2 control households, 1.5%). While it is outside the scope of this research, the possibility of exploring communal or shared RWH systems in this context could address the inherent inequality of affordability.

Durability and functionality of the HRWH systems over the two-year study period was high, despite low observed water levels at endline. Metal roofed systems functioned better over time than palm leaf systems, with 85% of metal systems functional at endline versus 65% of palm leaf systems, despite metal systems having a higher mean age at endline. This difference is likely due to the varied durability of the materials used in these roof types and indicates that, in this setting, the palm leaf systems demonstrated only a limited long-term functionality. In line with functionality, palm leaf systems also performed worse than metal systems in terms of water quality and had similar Levels of microbial contamination compared with existing community water sources. Metal systems showed markedly lower levels of contamination, but were not free from faecal contaminants, indicating that in this setting—even with high levels of adherence to operation and maintenance—low-cost HRWH systems do not provide water that is safe to drink at point of access. This is in line with findings from previous studies suggesting that domestic rainwater harvesting structures often require additional treatment after collection to meet health and safety standards [8, 2830]. A recent pooled cohort study from Bangladesh also indicates that rainwater has lower levels of minerals compared with groundwater, and as such exclusive use can have negative cardiometabolic health outcomes [31].

Impact outcomes

Results from this study did not provide strong evidence to support the widely reported phenomenon that provision of a water source at the household increases domestic water consumption [2, 32]. However, in this setting, only a quarter of functional systems were observed as having water in the tank at endline and few households in receipt of the intervention reported relying on the HRWH as their primary drinking water source. In addition to this, user satisfaction was at its lowest when considering the quantity of water delivered by the system, with almost half of intervention households reporting dissatisfaction. This suggests that the yield of the 250L HRWH system was insufficient to meet households’ domestic water requirements and may have contributed towards the lack of observed impact on prevalence of self-reported childhood diarrhoea among recipient households. For future HRWH interventions, steps should be taken to ensure that systems can meet the water demand of households. We attribute the lack of water in most systems to the season in which the endline data was collected. Average monthly climatology of Madagascar indicates that October is the last of a six-month dry season [33]. This suggests that water reserves would have been lowest at this time of year.

The borderline significant increase in observed number of handwashing stations within 10 meters of the household supports findings from previous studies that provision of a household water source can result in improved hygiene behaviours [11, 34, 35].

Study limitations

As with most research conducted in resource limited settings, this study had several limitations. First, due to ethical considerations, participant households were not randomised prior to intervention delivery, and as such findings comparing results between control and intervention households are vulnerable to confounding. Propensity score matching, controlling for socio-economic difference in multivariate models, and use of a difference-in-difference analysis approach was used to account for potential selection bias, but residual confounding may have persisted. Intervention households were, by definition, those that decided the cost of the HRWH system was affordable, as such our measure of satisfaction with the cost of the system could be biased due to self-selection. Additionally, limited resources prohibited more regular data collection on water consumption and water quality and as a result the high levels of seasonal variation in rainfall were not captured in the data. Endline surveys were collected during the dry season which may have caused an underestimation of water consumption and contributed to the low number of systems with observed water. Due to funding constraints, only a limited number of HRWH systems could be sampled to assess microbial water quality parameters, which negatively impacted the precision of these estimates and our ability to detect differences between HRWH systems and community water sources. Additionally, we acknowledge the limitations of TTC as an indicator of faecal contamination.


Findings from this study contribute towards the growing evidence base that supported self-supply intervention models offer a promising alternative to supply-driven interventions, even in extremely low-resource settings. Specifically, in this study we have demonstrated that demand exists for water supply infrastructure despite estimates that most of the population live on less than 2 USD/day and have abundant access to existing unimproved community water sources. Further to this we have shown that low-cost water supply infrastructure, manufactured using locally available materials can maintain functionality over an extended time-period. However, as programmes continue to weigh the relative benefits of decentralised, lower-cost water supply models, it is important to ensure that point of use water quality is protected and where this is not possible, provision for affordable and sustainable water treatment options are made available. It is also important to note that capacity of systems and seasonal rainfall should be considered in order to provide a consistent, year-round supply using HRWH systems in this setting.

Supporting information

S1 Text. Survey methodology and demographics.

A description of how surveys were administered and demographic information collected [36].


S2 Text. Water quality test methodology.

A description of how water quality tests, measuring thermotolerant coliforms, were conducted [37].


S1 Table. System functionality tool.

Question items used by the data collection team during observational system checks. System functionality questions determined if the system was able to collect and store water. Water quality questions determined the condition of the water collected by the system.


S2 Table. Demographic and socioeconomic characteristics.

Characteristics of comparison households present at baseline but lost to follow up and comparison households present at both baseline and endline.


S1 Dataset. Study data.

Raw study data collected and analysed by the researchers and accompanying codebook.



We would like to acknowledge the support and contributions of SEED Madagascar, both at the headquarters and Fort Dauphin offices, Tatirano Social Enterprise, the Madagascar Ministry of Health and Medical Inspector of the Fort Dauphin District Health Service, and the communities and participants of Mahatalaky Rural Commune.


  1. 1. Fewtrell L, Kaufmann RB, Kay D, Enanoria W, Haller L, Colford Jr JM. Water, sanitation, and hygiene interventions to reduce diarrhoea in less developed countries: a systematic review and meta-analysis. The Lancet infectious diseases. 2005;5(1):42–52. pmid:15620560
  2. 2. Cassivi A, Guilherme S, Bain R, Tilley E, Waygood EOD, Dorea C. Drinking water accessibility and quantity in low and middle-income countries: A systematic review. Int J Hyg Environ Health. 2019;222(7):1011–20. pmid:31320308
  3. 3. Cairncross S, Cuff JL. Water use and health in Mueda, Mozambique. Transactions of the Royal Society of Tropical Medicine and Hygiene. 1987;81(1):51–4. pmid:3445322
  4. 4. WHO, UNICEF, The World Bank. State of the world’s drinking water: An urgent call to action to accelerate progress on ensuring safe drinking water for all [Internet]. Geneva: World Health Organization; 2022 [cited 2023 Sep 8]. Available from:
  5. 5. Sustainability Assessment of Rural Water Service Delivery Models: Findings of a Multi-Country Review. World Bank; 2017 p. 12. (Water Global Practice).
  6. 6. Taylor B. Addressing the Sustainability Crisis: Lessons from Research on Managing Rural Water Projects—Google Search [Internet]. WaterAid, Tanzania; 2009 [cited 2022 Mar 24]. Available from:
  7. 7. Andres L, Deb S, Gambrill M, Giannone E, Joseph G, Kannath P, et al. Sustainability of Demand Responsive Approaches to Rural Water Supply: The Case of Kerala. World Bank; 2017 Apr p. 30. Report No.: 8025.
  8. 8. DeBusk K, Hunt W. Rainwater harvesting: A comprehensive review of literature. Water Resources Research Institute of the University of North Carolina; 2014.
  9. 9. Campisano A, Butler D, Ward S, Burns MJ, Friedler E, DeBusk K, et al. Urban rainwater harvesting systems: Research, implementation and future perspectives. Water Res. 2017 May 15;115:195–209. pmid:28279940
  10. 10. Hunter PR, MacDonald AM, Carter RC. Water Supply and Health. PLOS Medicine. 2010 Nov 9;7(11):e1000361. pmid:21085692
  11. 11. Oswald WE, Hunter GC, Kramer MR, Leontsini E, Cabrera L, Lescano AG, et al. Provision of private, piped water and sewerage connections and directly observed handwashing of mothers in a peri-urban community of Lima, Peru. Tropical medicine & international health. 2014;19(4):388–97.
  12. 12. Cairncross S, Valdmanis V. Water Supply, Sanitation, and Hygiene Promotion. In: Jamison DT, Breman JG, Measham AR, Alleyne G, Claeson M, Evans DB, et al., editors. Disease Control Priorities in Developing Countries [Internet]. 2nd ed. Washington (DC): World Bank; 2006 [cited 2022 Mar 24]. Available from:
  13. 13. Islam MM, Afrin S, Tarek MH, Rahman MM. Reliability and financial feasibility assessment of a community rainwater harvesting system considering precipitation variability due to climate change. J Environ Manage. 2021 Jul 1;289:112507. pmid:33839610
  14. 14. Zang J, Kumar M, Werner D. Real-world sustainability analysis of an innovative decentralized water system with rainwater harvesting and wastewater reclamation. Journal of Environmental Management. 2021 Feb 15;280:111639. pmid:33203578
  15. 15. Bernard B, Joyfred A. Contribution of Rainfall on Rooftop Rainwater Harvesting and Saving on the Slopes of Mt. Elgon, East Africa. The Scientific World Journal. 2020;2020. pmid:32733168
  16. 16. Madagascar WASH Investment case [Internet]. 2017 [cited 2020 Jun 20]. Available from:
  17. 17. Madagascar Multiple Indicator Cluster Survey (MICS) 2018 [Internet]. Antananarivo, Madagascar: INSTAT UNICEF; 2018 [cited 2020 Jun 20]. Available from:
  18. 18. Healy TM. The Deep South [Internet]. The World Bank; 2018 Jul [cited 2020 Jun 20] p. 1–48. Report No.: 127982. Available from:
  19. 19. Chaplin H, Legge H. Financing and design innovation in rural domestic rainwater harvesting in Madagascar. Waterlines. 2019 Apr 1;38:113–22.
  20. 20. Bostoen K, Chalabi Z. Optimization of household survey sampling without sample frames. International Journal of Epidemiology. 2006 Jun 1;35(3):751–5. pmid:16481364
  21. 21. OpenStreetMap [Internet]. [cited 2022 Feb 28]. OpenStreetMap. Available from:
  22. 22. SEED Madagascar [Internet]. [cited 2020 Jun 21]. Project Tatirano. Available from:
  23. 23. Tamason CC, Bessias S, Villada A, Tulsiani SM, Ensink JHJ, Gurley ES, et al. Measuring domestic water use: a systematic review of methodologies that measure unmetered water use in low-income settings. Trop Med Int Health. 2016 Nov;21(11):1389–402. pmid:27573762
  24. 24. Thomas MLH, Channon AA, Bain RES, Nyamai M, Wright JA. Household-Reported Availability of Drinking Water in Africa: A Systematic Review. Water. 2020 Sep;12(9):2603.
  25. 25. Angwec CA. Cost recovery for piped rural water supply systems in developing countries: case studies from Kenya, Rwanda and Uganda [PhD Thesis]. Loughborough University; 2015.
  26. 26. Opare S. Rainwater harvesting: an option for sustainable rural water supply in Ghana. GeoJournal. 2012 Oct 1;77(5):695–705.
  27. 27. Kettle S. User Fees for Rural Water Projects. University of Bristol; 2013.
  28. 28. Richards S, Rao L, Connelly S, Raj A, Raveendran L, Shirin S, et al. Sustainable water resources through harvesting rainwater and the effectiveness of a low-cost water treatment. Journal of Environmental Management. 2021;286:112223. pmid:33684801
  29. 29. Kirs M, Moravcik P, Gyawali P, Hamilton K, Kisand V, Gurr I, et al. Rainwater harvesting in American Samoa: current practices and indicative health risks. Environmental Science and Pollution Research. 2017;24(13):12384–92. pmid:28357803
  30. 30. Tran SH, Dang HTT, Dao DA, Nguyen VA, Nguyen LT, Nguyen VA, et al. On-site rainwater harvesting and treatment for drinking water supply: assessment of cost and technical issues. Environ Sci Pollut Res Int. 2020 Feb 19; pmid:32077016
  31. 31. Naser AM, Rahman M, Unicomb L, Parvez SM, Islam S, Doza S, et al. Associations of drinking rainwater with macro-mineral intake and cardiometabolic health: a pooled cohort analysis in Bangladesh, 2016–2019. npj Clean Water. 2020 Apr 24;3(1):1–11.
  32. 32. Fry LM, Cowden JR, Watkins DW, Clasen T, Mihelcic JR. Quantifying Health Improvements from Water Quantity Enhancement: An Engineering Perspective Applied to Rainwater Harvesting in West Africa. Environ Sci Technol. 2010 Dec 15;44(24):9535–41. pmid:21080624
  33. 33. World Bank Climate Change Knowledge Portal: Madagascar [Internet]. [cited 2023 Jun 18]. Available from:
  34. 34. Cairncross S, Hunt C, Boisson S, Bostoen K, Curtis V, Fung IC, et al. Water, sanitation and hygiene for the prevention of diarrhoea. International journal of epidemiology. 2010;39(suppl_1):i193–205. pmid:20348121
  35. 35. Howard G, Bartram J. Domestic Water Quantity, Service Level and Health [Internet]. World Health Organization; 2003 [cited 2020 Sep 18]. (Water, Sanitation, and Hygiene). Available from:
  36. 36. Open Data Kit [Internet]. 2018 [cited 2019 Jul 11]. Open Data Kit. Available from:
  37. 37. DelAgua—Product—10098—DELAGUA—DelAgua Single Incubator [Internet]. [cited 2020 Jun 20]. Available from: