Effectiveness of the upscaled use of a silver–ceramic (silver ionization) technology to disinfect drinking water in tanks at schools in rural India

In many low- and middle-income countries, schools serve untreated drinking water that has been pumped into storage tanks. Water flows from the tanks (by gravity) to taps for consumption by children. Because the water is often of poor quality, pathogens may be present that cause gastrointestinal infections. This can lead to missed school days, stunted growth, cognitive impairment, and in extreme cases, death (Bain et al. 2014; McMichael 2019). While water can be dosed with chlorine, some communities lack the capacity to manage such systems as they require dosing appropriate masses of Ca(OCl)₂ on a daily basis coupled with managing purchasing and supply processes (Harris 2005). Furthermore, chlorination often faces resistance because it causes a noticeable change in taste and aesthetics (Crider et al. 2018; Spackman & Burlingame 2018), not to mention the possible generation of hazardous disinfection byproducts (Kali et al. 2021).

This study deployed an emerging silver–ceramic technology called MadiDrop to disinfect drinking water in storage tanks at schools in rural India. MadiDrop is a 52-g porous ceramic tablet embedded with micro-patches of metallic silver that can disinfect 10–20-L batches of water per day for up to 1 year. The tablet is placed in a water storage container where silver ions are passively released into the water. Users fill the containers at night, and in the morning the water is biologically safe to drink (Ehdaie et al. 2017; Hill et al. 2020). The advantages of MadiDrop are that it is low-cost (0.2 US cents/L), simple to use, requires no additional infrastructure (simply a water container), and produces water that is aesthetically acceptable since there is no change in taste, or odor, or other physical characteristics (Pacey 1977; Bhardwaj et al. 2021).

Silver ionization, typically through ceramic silver-impregnated pot filters, is effective at the household scale (e.g., 10–20 L batches) (Gupta et al. 2018; Rivera-Sánchez et al. 2020). However, very little laboratory or field research has been conducted on its effectiveness at larger scales (e.g., 100–500 L). To address the research gap, this pilot study implemented silver ionization treatment with MadiDrop tablets at the community scale by engaging schools that serve drinking water to marginalized, at-risk populations via storage tanks that are several hundred liters in volume.

The objective of this study is to assess the effectiveness of MadiDrop at removing total coliforms (TCs) and Escherichia coli (EC) from drinking water at the community scale (i.e., tanks at schools) while maintaining effluent silver concentration levels below the World Health Organization (WHO) drinking water quality guideline of 100 μg/L (WHO 2018). Additionally, this study seeks to identify various barriers and caveats that affect the successful use of MadiDrop and other silver ionization treatment methods at the community scale.

The three schools are located in the rural settlements of Damdama, Kiranki, and Saramthala with total school populations of approximately 165, 205, and 170 students and staff, respectively. The schools exhibit a variety of tank volumes, dimensions, materials, and drinking water sources, as shown in Table 1. This diversity in tanks and source waters – including one with bio-sand filtration – provided valuable insight into the performance of the MadiDrop technology. Furthermore, variations in water usage rates and the presence of weekends and holidays (when the MadiDrops had extended contact time) facilitated a realistic understanding of managing MadiDrop vis-à-vis real-world practicalities. All research and data collection activities were performed according to approved institutional ethical standards (Virginia Tech IRB #22-371) as well as an additional layer of standards deemed important by the schools and research team.

Next, to establish a baseline for each tank, triplicate influent and effluent water samples were collected and tested from each tank over the course of one week according to the methods described below. Baseline tests were followed by the insertion of the appropriate number of MadiDrops along with a Heron dipperLog vented pressure transducer. The pressure transducer logged tank inflows and outflows at continuous 15-min intervals to enable the monitoring of water usage and tank filling times as well as to inform water treatment, water quality testing, and refinements to the computer model. The tanks were then emptied and refilled every 1–3 days to mimic actual water usage along with weekends, holidays, and other anticipated irregularities.

After introduction of the MadiDrop, inflow and treated outflow samples from each tank were collected three times per week for 5 weeks from December 2022 to January 2023. Samples were collected using sterile Nasco Whirl-Pak bags and stored on ice between 2 and 8 °C according to Standard Methods (Lipps et al. 2023). Samples were tested for TC and EC concentrations at the S M Sehgal Foundation water quality laboratory within 5 hours of collection. Duplicate samples from each tank (two inflow and two treated outflow samples) were diluted with deionized water to contain a total of 1, 5, or 10 mL of inflow and/or treated outflow water guided by baseline data collection. The membrane filtration method utilizing MilliporeSigma m-Coliblue24 broth culture media was used to determine colony forming units (cfu) per 100 mL. Samples were incubated for 24 h at 35 °C. All units were counted as TC and only dark units were counted as EC. A control test of deionized water was also conducted each time.

Silver concentrations of treated samples were measured at the TERI School of Advanced Studies using a Hach DR 5000 UV–vis spectrophotometer. A standard curve was generated using the Hach silver standard solution and tests were performed using the Hach silver reagent and sodium thiosulfate pillows according to the colorimetric method. Results were confirmed by remeasuring acidified samples using an Agilent 7900 inductively coupled plasma mass spectrometer (ICP-MS) at the University of Virginia. A Vernier turbidity sensor was used to determine nephelometric turbidity units (NTU) in both inflow and treated outflow samples. After the intervention, MadiDrops were removed from the tanks, the tanks were emptied, and a second pair of three inflow and three outflow baseline tests were conducted for each tank over the course of 1 week.

Results were analyzed using both descriptive and inferential statistics. Descriptive statistics such as standard deviation (SD) and percent reduction in TC and EC concentrations were computed in Microsoft Excel for tank inflow, treated outflow, and baseline samples. For inferential statistics, the normality of the data was found in R Studio using a Shapiro–Wilk test. To compare TC and EC concentrations between inflow samples and outflow samples treated with MadiDrop, a one-tailed Wilcoxon rank-sum test (Mann–Whitney U test) was performed in JMP Pro 16.0.0 for each tank and for the overall set of three tanks. The p-values were calculated using the chi-squared approximation method. The strength and direction of relationships between percent reduction in TC concentrations and tank residence time, silver concentration, and turbidity were assessed for each tank and the overall set of three tanks using Spearman's rank correlation coefficient (Spearman's rho) in JMP Pro. A coefficient closer to −1 suggests a more pronounced negative correlation, meaning a stronger tendency for the variables (e.g., TC and silver concentration) to move in opposite directions.

Statistical analyses determined that TC concentrations were significantly different (p ≤ 0.05) − in this case, reduced – when comparing outflow samples treated with MadiDrop to inflow samples in two of three intervention tanks and among the overall set of three tanks (Table 2). Additionally, TC concentrations were significantly different (i.e., reduced) when comparing treated outflow samples to baseline outflow samples across all three tanks and as an overall set. For EC, concentrations were reduced in all treated outflow samples (relative to inflow) overall and in all tanks with EC detected, although none of the comparisons were significant. However, when comparing treated outflow samples to baseline outflow samples, EC reduction occurred across the board and was significant overall and in two of three intervention tanks.

This study sought to evaluate the effectiveness of MadiDrop at removing TC and EC from drinking water tanks at three schools in rural India while maintaining effluent silver concentrations below the drinking water quality guideline of 100 μg/L. Additionally, this study aimed to identify barriers and caveats that affect the successful deployment of MadiDrop and similar water treatment methods at the community scale. This section dissects these topics while also providing study limitations and future research directions.

Results determined that TC and EC concentrations were reduced in every tank and when considering all samples as an overall set (Table 3). TC was significantly reduced in two of three tanks and overall, and reductions in TC and EC were significant compared to baseline data in seven of eight comparisons (Table 2). Although some EC comparisons were insignificant, we contend a governing factor may be that EC concentrations were not highly elevated to begin with and therefore there was not a large amount to remove. (Not to mention that silver concentrations remained at a relatively low level, which will be discussed later.) This is attested to by the tank at Saramthala, which had no EC to remove since the influent undergoes biosand filtration prior to intervention disinfection. Despite cases of insignificant EC reduction, Table 3 and the plots in Figures 5 and 7 demonstrate that EC reduction still eclipsed 70% overall and that reduction occurred across all tanks, especially in the tank at Kiranki that exhibited the highest EC concentrations. In fact, because EC concentrations were consistently quite low, outliers from untreated samples at Kiranki were manually removed from the box and whisker plots in Figure 5 solely to better visualize the data. If the outliers were included, the plots became so compressed that there was virtually nothing to discern. Note that including the outliers in Figure 5 − which was done in all statistical analyses and descriptive statistics – would have actually made EC reduction look greater in the plots (i.e., this is not a misrepresentation or exaggeration of data). However, doing so would have compromised visualization of the plots for Kiranki as well as Damdama to a lesser extent.

Overall, TC concentrations were negatively correlated with tank residence time (i.e., contact time with MadiDrops) and effluent silver concentration (Table 4). These correlations are intuitive as it was hypothesized that longer durations of contact time with the silver–ceramic technology and higher silver concentrations would positively affect disinfection. While it was also hypothesized that turbidity may negatively impact disinfection (i.e., greater turbidity would lead to lower TC and EC removal), the statistical tests were insignificant. This result contradicts Zhao & Stevens (1998) and Harris et al. (2023) but is in agreement with those of Ehdaie et al.’s (2017) fieldwork in South Africa. We contend that this result is (partly) due to the source waters (rainwater and groundwater) not being very turbid. Considering all tanks, the mean and median were 4.6 and 3.6 NTUs, respectively. While these NTUs are not ideal, we contend that they are not exceedingly elevated and therefore turbidity might not have been great enough to significantly impede disinfection. That being said, turbidity is an impediment to disinfection, and turbidity has been shown to reduce the effectiveness of silver ionization. This is especially a factor when treating surface water, which is generally more turbid compared to rainwater and groundwater and also tends to contain more biological contaminants. Thus, the source water should be tested before deploying MadiDrop (or any disinfection technology), with high NTUs signifying that more tablets or greater contact time may be required.

An extremely important result was sustaining silver concentrations below the WHO drinking water quality guideline in every treated outflow sample. The ability to do so was aided by the model that was developed. However, although the model called for 29.6, 31.5, and 50.2 MadiDrops in the tanks at Damdama, Kiranki, and Saramthala, respectively, we elected to insert only 26, 26, and 31 tablets based on public health concerns and water consumption dynamics observed during the baseline. Since this was a pilot study with schools that serve vulnerable populations, we consciously chose to deploy fewer MadiDrops out of an abundance of caution. In particular, we wanted the water to be safe if a school happened to leave water in the tank over the weekend or for an extended period of time. Second, fewer tablets were deployed based on actual water usage observed during baseline data collection (which was less than was estimated in the model). As a result, overall silver concentrations averaged only 10.8 μg/L with a median of 12.1 μg/L (Table 5). This leads to three contentions: (1) the model appears to be practical; (2) model results are roughly in line with estimated calculations of one MadiDrop tablet per 10–20 L (i.e., the intended volume to be treated by a single tablet); and (3) TC and EC were still reduced in every case (often significantly) with considerable room to increase silver concentrations to enhance disinfection. In fact, maintaining 70–90 μg/L would be ideal, meaning that silver concentrations could be increased many times over. Additionally, the cost per liter treated is only US $0.0021 while it is estimated (based on current market prices) that chlorine tablets like Aquatabs cost $0.02 per liter and P&G Purifier sachets cost $0.075 per liter, rendering MadiDrop relatively cost-effective compared to other common PoU disinfection methods.

While this pilot intervention could be considered successful, several barriers emerged. First, users must be able to deploy and manage the disinfection technology appropriately. This requires basic data (e.g., tank volume, dimensions, and estimates of inflow–outflow per day) in order to simulate how many MadiDrops are needed. In turn, this requires guidance from an expert or access to the model and the ability to use it. Alternatively, a simple ‘back-of-the-envelope’ calculation that estimates how many MadiDrops are needed based on tank volume and general/assumed usage would be valuable. Next, the ability to measure silver concentrations was identified as a barrier. Silver standard solution, reagents, and other consumables needed to measure silver concentrations were unavailable in India, and therefore these materials had to be brought into the country. This presents a significant barrier as monitoring silver concentrations requires access to a laboratory, equipment (e.g., spectrophotometer), and associated consumables that may be unavailable or cost-prohibitive in low-resource contexts.

Several practical barriers also arose. First, some of the tank inflows were difficult to access and it can be difficult to gain access to the inside of tanks to add and position the MadiDrop tablets. Next, tank management is of foremost concern. Water should reside in the tank in contact with the tablets overnight, and the tank should be regularly emptied (if not fully or mostly used) and refilled to safely manage silver concentrations. This demands a dependable, trustworthy staff who can manage the tank in cases of water sitting over weekends, holidays, school closures, or other circumstances. For this reason, and as mentioned earlier, we elected to deploy fewer MadiDrops than the model stipulated to err on the side of caution. Vertically positioning tablets inside the tank can also be difficult. To accomplish this, we hung fishing lines into the tanks with tablets tied on at the simulated heights. Finally, users of this technology must still remember to implement other water, sanitation, and hygiene (WASH) best practices. While MadiDrop has been shown to improve water quality, maintaining a clean tank, sludge removal, and other safe storage and serving activities should still be practiced to promote water quality.

In terms of caveats, an interesting yet expected finding was that the tanks themselves presented TC and EC hazards that were difficult to account for. We sought to understand the tank ecosystems, as explored by Salehi (2022) in the case of storage tanks and Juran & MacDonald (2014) in the case of post-treatment drinking water storage, with the hypothesis that TC and EC already exist in the tanks and thus water entering the tanks – without any treatment – would be better quality (at best the same) than after it flows from the tanks. As expected, the tank ecosystems – level of cleanliness, presence of biofilm, sludge, and other residues and constituents, and instances of low dissolved oxygen in times of prolonged storage – negatively affected the initial quality of inflow water. For example, baseline tests of inflow and untreated outflow samples established that outflow samples from the tank at Kiranki averaged an increase of 993 TC and 1,182 EC compared to inflow samples. At Damdama, baseline samples averaged an increase of 22 TC and 3 EC after flowing through the tank, while outflow samples from Saramthala averaged an increase of 40 TC (no EC detected in inflow or outflow). These data demonstrate and further stress the criticality of maintaining a clean tank, regularly removing biofilm and sludge, and conducting other household water treatment and safe storage (HWTS) activities to promote and maintain water quality. Regarding the use of MadiDrop in the tanks, we therefore contend that the tablets were removing TC and EC that were present in the inflow as well as TC and EC that were already present in the tanks. This means that MadiDrop not only produced effluent with fewer TC and EC than were present in the source water, but it did so while also eliminating additional, unaccounted for TC and EC that were part of the tank ecologies and premise plumbing.

Turbidity represents a potential caveat. While turbidity was not found to be a significant factor in this study – perhaps due to relatively low (though non-ideal) NTUs – it could impede bacterial reduction in other settings or at different treatment scales/volumes. This is particularly true in cases when surface water, which is likely to be more turbid, is used. Moreover, the variable of turbidity emphasizes the importance of testing the source water in advance of deploying MadiDrop (or any treatment technology) and looking for noticeable changes in inflow turbidity throughout the year/seasons.

A final caveat is that this study only tested for coliform bacteria and Escherichia coli. While neither of these tests specifically identifies a pathogen, their presence suggests that the water has come into contact with mammalian feces and is therefore at higher risk of having pathogenic microorganisms. In fact, the WHO (2017) recommends <1 EC per 100 mL. Greater than 73% of treated effluent samples met this guideline and our mean EC removal rate was >70%, which is comparable to results reported in other published PoU treatment interventions in low-resource settings (e.g., Rosa et al. 2010; Juran & MacDonald 2014; Francis et al. 2016; Aumeier et al. 2017). Furthermore, any decrease in microorganisms improves the (relative) safety of drinking water.

This study demonstrates some limitations and there are several avenues for future research. The water quality intervention was a pilot study and deployment at the community scale was novel; these factors enabled the research team to garner much knowledge and lessons learned. First, the length of study and number of schools/tanks involved could be greater. To that point, future research should engage more schools/tanks for a longer duration of time. Temporally, conducting the intervention at different times of the year (e.g., during the monsoon and dry season and periods when source water quality fluctuates) would also be valuable. Pragmatically, in order to vertically position the MadiDrop tablets in the tanks, we suggest utilizing 3D printing to develop long ‘sleeves’ that the tablets can be inserted into at incremental distances.

Testing for reductions in other health hazards, particularly viruses (e.g., rotavirus), protozoa (e.g., Cryptosporidium), and mosquito larvae, is another avenue for more in-depth research. Additionally, the approach could be expanded to similarly situated contexts where drinking water is served from tanks, such as hospitals, restaurants, government buildings, dormitories, office towers, daycares, apartment buildings, and independent houses that have private tanks (often located on the roof). Finally, augmenting the disinfection range and capacity of the MadiDrop technology itself represents a pathway forward in environmental engineering. MadiDrop currently releases only silver ions. However, ongoing research is investigating the delivery of silver in combination with copper ions (Harris et al. 2024) as well as hypochlorous acid. The synergistic capabilities of these disinfectants would allow for the elimination of a greater range of pathogens while simultaneously allowing for each disinfectant to be managed at a low-to-moderate level (since all would be working together) in efforts to more easily maintain safe drinking water quality.

In many low- and middle-income countries, schools serve untreated drinking water that has been pumped into storage tanks. Because the water is often of poor quality, pathogens may be present that cause gastrointestinal infections, potentially leading to missed school days, stunted growth, cognitive impairment, and in extreme cases, death. Given this backdrop, we evaluated the effectiveness of MadiDrop, a novel silver–ceramic technology, at disinfecting drinking water served from tanks at schools in rural, water-scarce India. While silver ionization has been proven effective at the household scale, relatively scarce research has been conducted on its effectiveness at the community level.

Statistical analyses determined that TC concentrations were significantly reduced when comparing inflow samples to outflow samples treated with MadiDrop in two of three intervention tanks and among the overall set of tanks. Additionally, TC was significantly reduced when comparing treated outflow samples to baseline outflow samples across all tanks and as an overall set. For EC, while concentrations were reduced in treated outflow samples compared to inflow samples both overall and in two of two tanks, none of the comparisons were significant. However, EC reduction was significant in two of three intervention tanks and overall when comparing treated outflow samples to baseline outflow samples. In terms of bacterial reductions, TC was reduced across all intervention tanks and by >63% overall, while EC concentrations were reduced across two of two tanks and by >70% overall. TC concentrations were negatively correlated with tank residence time (i.e., contact time with MadiDrops) and silver concentration levels. No relationship was identified between turbidity and MadiDrop's ability to remove TC and, importantly, no effluent samples exceeded the WHO drinking water quality guideline of 100 μg/L of silver ions.

Several practical and technological barriers were identified, including modeling or estimating the number (and vertical placement) of MadiDrops required and the capability to measure effluent silver concentrations. The latter requires access to a laboratory, silver standard solution, and various equipment and consumables, all of which may be unavailable/inaccessible or cost-prohibitive in low- and middle-income countries. Tank management is critical, especially if water sits in a tank over the weekend, a holiday, or similar school closure. Implementing silver ionization at the community scale also surfaced the issue of ‘tank ecology’ and that pathogens and health hazards may already be present in the tank and premise plumbing in the form of biofilm and sludge. Therefore, the best practice of a joint-treatment approach (e.g., combining filtration and disinfection) signifies a promising HWTS method for localized, community-level water treatment, and such an approach also imbues water with residual bactericidal capacity. Several suggestions for future field-based research were posited, including a longer study period that includes more schools and/or community contexts such as hospitals, restaurants, dormitories, and private rooftop tanks. A temporal scale that incorporates different seasons (i.e., when there are changes in source water quality) would also be insightful. Ultimately, in the Anthropocene that we inhabit, hydrological perturbations and extreme weather resulting in disasters, electrical outages, and intermittent water supplies are all too real. In such cases, community-scale deployment of silver ionization and similar technologies could represent a palatable and contextually appropriate coping mechanism.

Many thanks to the three participating schools, S M Sehgal field staff, TERI School of Advanced Studies, Sushant School of Art and Architecture, Dillon Mungle, and Joshua Starner.

Institute for Society, Culture, and Environment (ISCE) Scholars Program; Stephen H. Schoenholtz Water Undergraduate Research Fund (SWURF).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.