2.1. Ethics

This study obtained ethical approval from the National Committee on Research in the Social Sciences and Humanities (NCRSH) in Malawi (P.10/18/326) and the Human Research Ethics Board at the University of Victoria (18–1129). Details of the questionnaire methodology and findings of the household water quality testing are presented in separate publications [25,26].

2.2. Overview of water quality test kits

The objectives of this work were addressed via water quality sampling survey in April 2019 in three sites in Southern Malawi (more details below in Section 2.3). Some laboratory work was conducted in advance in Victoria, Canada, mainly to orient the researchers to the LCK and as a preliminary evaluation of the 18-hour (reduced) incubation time for water quality test results (detailed below in Section 2.3).

The standard kit (SK) and the LCK for membrane filtration are pictured in Fig 1. The LCK costs approximately USD $60, a significant cost reduction compared to the SK (approximately USD $1200; see S1 Table). Most of this cost reduction was achieved by substituting the SK Microfil stainless steel base (Millipore, USA) with a 250 mL glass vacuum filter flask (Thomas Scientific, USA) and bespoke filter support (provided by UNICEF). The main functionality and all auxiliary components (e.g., forceps) and consumables (e.g., membrane filters) remained the same from the SK to the LCK, with two exceptions. First, in the SK, the 150 mL polypropylene syringe was used to directly collect the sample filtrate as the plunger was drawn, whereas in the LCK, the sample filtrate was drawn into the flask by the vacuum generated by the syringe. The standard MICS catalogue uses a 100 mL syringe; however, a 150 mL syringe was used for this work, to prevent the potential need to detach and reattach the syringe to reapply the vacuum. Second, filter funnels were dried, decontaminated with 70% isopropyl alcohol wipe (150 mm by 170 mm size, PDI Healthcare Inc., United Kingdom) and re-used between water quality tests as described and validated in Zimmer et al. [27]. The Computer Aided Design (CAD) files for a version of the filter support unit which can be 3D printed are available with the supporting material and suppliers for each component of the kit are listed in S1 Table.

For the LCK, large plastic cutting boards were successfully used as surfaces on which the tests were performed for the duration of the water sampling campaign in Malawi (described in Section 2.3), with the aim to provide stability on rough ground and a clean surface on which to preform tests. Stability was further aided by using a 0.75 m length of PVC tube (inner diameter 8 mm) to connect the glass flask to the 150 mL syringe so that any movement of the syringe by applying the vacuum did not destabilize the vacuum flask. Also, a cotton hand towel (laundered weekly) was used to wrap the glass flask for protection against breakage while transporting the LCK between test locations.

Both the SK and the LCK use the membrane filtration technique [11] to enumerate E. coli, in which a volume of sample water (100 mL in this work) is passed through a membrane filter of 0.45 µm pore size by application of suction (vacuum) pressure. The filter is then placed into a petri dish containing nutrients for E. coli growth and an enzyme substrate, X-glucuronide (Compact Dry EC plates, Nissui Pharmaceutical Co., Ltd., Japan), which are widely used by MICS programmes [11], have been shown to be comparable to reference methods [12] and are approved by the Association of Official Analytical Collaboration [28,29], although other nutrient formulations exist [19]. Petri dishes are incubated for 24 ± 2 hours at 37 °C. In the Canadian laboratory, a full-scale incubator was used (Forced Air Microbiological Incubator 6.3CF, VWR, Radnor, Pennsylvania, USA); when piloting in Malawi, belt incubators were worn by enumerators during the day while gathering samples (UNICEF catalogue number S0000593, Fig 2) and plates were gathered from enumerators and incubated overnight in a portable incubator connecting to the mains power supply (Discovery MX10 incubator, Lynd Products Ltd., UK). The Discovery MX10 incubators come with an alcohol thermometer that allows monitoring/adjustment of incubation temperature. Following incubation, it is assumed that E. coli have multiplied to form blue colonies because of the X-glucuronide and are differently coloured from other bacteria that may be present on the filter to allow differential identification. The colonies are visible to the naked eye and thus can be counted in terms of colony forming units (CFU), with microbiological water quality results stated in terms of CFU/100 mL.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. An incubation belt for plated samples, worn by an enumerator.

Photo credit: C. Dorea.

https://doi.org/10.1371/journal.pwat.0000367.g002

2.3. Piloting of the LCK in Malawi

For the present study, the LCK was piloted during the water quality testing component of a household survey study, conducted in April 2019 in three sites in Southern Malawi: Ndirande (peri-urban), Mbayani (peri-urban) and Chikwawa (rural). The sites were selected through local partner guidance and depended on local community and authority approval (i.e., local community chiefs) and are detailed in other publications [25,26]. In total, 375 randomly selected households were surveyed in peri-urban settlements and rural areas over a three-week period. On average, each enumerator surveyed five households per day, with a maximum of 20 households surveyed by each enumerator each week. For each survey, two water samples were collected and processed in situ: one from the PoU and one from the PoC, following the MICS protocol [11], except that the LCK was used (Fig 1b) to process all samples. Each LCK was used to conduct a maximum of 120 water quality tests over three weeks of water sampling. There were six enumerators who used one LCK each.

The six enumerators conducted the questionnaire and water quality tests after completing five days of training, which included approximately two days dedicated solely to water quality testing. The training followed the standard protocols used in MICS [11].

2.4. Evaluation of 18-Hour E. coli (reduced) incubation

The MICS-based water quality testing method recommends a minimum incubation period of 24 ± 2 hours [30]. It is standard practice in MICS programmes to place CompactDry EC plates containing processed membrane filters are in an incubation belt worn by the enumerator (Fig 2). This method is widely used by MICS programmes; it is intended to keep plates close to body temperature, which is considered to be sufficiently close to the prescribed incubation temperature of 37°C for CompactDry EC plates [11]. These customized belts can comfortably hold 18 plates; multiple belts cannot be worn to increase capacity as belts rely on proximity to body heat to maintain incubation temperature and wearing multiple belts would likely compromise user comfort, which was a key consideration in their design [31]. The limited capacity may be exceeded during prolonged surveys, which would then require the use of an electric incubator which is more expensive than a belt and requires continuous access to electricity. Thus, the feasibility of increasing the capacity in the incubation belt could be of interest during water sampling activities in areas with limited or unreliable electricity supply. To this end, a reduced incubation period of 18 hours was evaluated during the water quality testing in Southern Malawi, and additionally tested prior in a laboratory setting in Canada.

In both locations, E. coli was sourced from a commercially-available probiotic Mutaflor (Pharma-Zentrale GmbH, Germany) and incubated to stationary phase of growth before spiking water samples for processing. The details of water sample preparation can be found elsewhere [27]. Select water quality samples were enumerated after an abbreviated 18-hour incubation period, followed by a further 6 hours of incubation time and subsequent additional enumeration for comparison, to reach the prescribed 24-hour incubation period.

In the Canadian laboratory setting, the SK was used to prepare the E. coli plates and plates were incubated in a standard laboratory incubator [27]. During fieldwork in Malawi, the LCK was used to prepare the E. coli plates and samples were incubated in the UNICEF belts during the day and in a portable electric incubator overnight.

2.5. Validation of trained enumerator colony counts

During MICS-type water quality testing, enumerators with no prior training or experience in water quality are selected and trained in sampling and colony counting. To validate the ability of recently trained enumerators in Southern Malawi (i.e., trained novices) to correctly enumerate E. coli colonies, their results were compared against colony counts of an experienced counter. Specifically, once per week, the previous day’s plates were counted by the respective enumerator as well as an experienced counter (CZ), who had three years of experience interpreting water quality results. The comparison of the two counts (novice vs experienced counter) was the basis of the validation.

2.6. Assessment of blank test water

MICS guidelines recommend using a locally purchased, pre-tested bottled water brand for quality control blank testing (11). While MICS trainers typically test these bottles before enumerator deployment, this practice remains unvalidated. Given evidence that bottled water in low- and middle-income countries (LMICs) is more likely to contain fecal indicator bacteria than in high-income countries (25), there is a risk that positive blank samples may stem from contaminated bottled water rather than equipment malfunction or enumerator error. Therefore, to validate the MICS practice of using of bottled water for blank testing, we tested bottled water used for past MICS in 6 countries – Democratic People’s Republic of Korea, DPRK [32], Guinea Bissau [33], São Tomé and Príncipe [34], Fiji [35], Samoa [36], and Tonga [37] – as well as that used during this study’s water testing in Southern Malawi.

As our objective was to assess bottled water quality in conditions similar to what would be expected during a MICS-type household survey, our bottled water sampling and analysis did not follow all study criteria listed by Williams et al. [15], (i.e., representativeness) as it was not intended to represent the overall situation of bottled water in each country. At least three samples of the same brand were acquired from different vendors during the MICS surveys in each country between 2017 and 2021. Although the bottled water was not refrigerated was kept at ambient air temperature before testing (as blank test water is not typically refrigerated during MICS), care was taken to ensure samples were kept away from sunlight and analyzed before the stated expiry date. During MICS, as a measure to validate locally purchased water prior to enumerator training, each water bottle was opened immediately prior to analysis and a 100 mL aliquot was directly poured into the membrane filtration setup (SK or LCK). The water bottles were used only once, as would be the case during MICS fieldwork. This bottled water testing was conducted by an experienced trainer, typically in preparation for a MICS campaign to identify a brand of bottled water that could be used for blank testing during the campaign. Triplicate analysis of E. coli, as per methods described above, was performed on each sampled bottle in all countries except DPRK where thermotolerant coliforms were the faecal indicator bacteria of choice, as described in Dorea et al. [38].

2.7. Statistical methods

Relevant geometric mean colony counts and upper/lower 95% Confidence Intervals (CIs) were computed for all colony count data, using a value of 0.5 CFU/100 mL, half of the lower detection limit (1 CFU/100 mL), for non-detects. Colony counts greater than 1 ∙ 10² CFU/100 mL were recorded as 1.01 ∙ 10² CFU/100 mL in accordance with MICS protocols [11].

The assumption of data normality was tested via the Shapiro-Wilk test. Colony counts after 18 hours and 24 hours of incubation time at 37°C were compared using a two-sided, paired Wilcoxon rank sum test, as the data violated the assumption for normality. The data were also converted to risk categories which were again compared using a two-sided, paired Wilcoxon rank sum test. Colony counts for enumerators were compared to those done by an experienced counter using a two-sided, paired t-test, as those data were normally distributed.

Results from all statistical tests used in the analysis were considered significant at the α ≤ 0.05 significance level. In the case of statistically significant (p ≤ 0.05) results, we further examined if the result would have an impact on the categorization of results into such risk categories. All statistical tests were performed with R statistical software, in RStudio, version 3.6.3. Where appropriate, water quality data compared to the a priori waterborne risk categories according to the WHO Guidelines for Drinking Water Quality [39]: “low risk” (< 1 CFU/100 mL); “intermediate risk” (1–10 CFU/100 mL); “high risk” (11–1 ∙ 10² CFU/100 mL); “very high risk” (1.01 ∙ 10² to 1 ∙ 10³ CFU/100 mL).

3.1. Piloting of the LCK in Malawi

User feedback on the LCK during the water sampling activities in Malawi was gathered by asking the enumerators about their general experiences during weekly debriefings. Although enumerators did not have any previous experience with either the SK or the LCK with which to compare, they reported the LCK to be stable, with no reported incidences of the filtration base tipping over; and robust, with no glass flasks broken during the 4-week campaign (training and sampling – including during transit from Canada to Malawi). The implementation of plastic cutting boards successfully provided stability on rough ground, and an easy-to-clean surface on which to perform tests. If properly cleaned (i.e., wiping with 70% alcohol), the board can provide a clean surface on dusty ground.

Enumerators reported that while using the LCK they did not have trouble with loss of suction while producing a vacuum using the 150 mL syringe, and the rare occasions where they did have to detach and reattach the syringe to reapply the vacuum, it did not cause significant delays to their work. No water leakage was reported by the enumerators between the LCK filtration base and funnel seal; each funnel was used a maximum of 24 times during the water sampling campaign. There were no issues reported by the enumerators regarding vacuum flask stability.

3.2. Assessment of blank test water

A summary of the colony counts found in the selected brands of bottled water tested is shown in Table 1. Of all the brands tested, there were no samples that produced positive counts (i.e., each brand tested resulted in a triplicate of non-detect results). Full test results are displayed in S2 Table.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Counts found in selected brands of bottled water.

https://doi.org/10.1371/journal.pwat.0000367.t001

3.3. Evaluation of 18-Hour E. coli (reduced) incubation

The comparison of colony counts after 18 and 24 hours of incubation during Canadian laboratory work and the fieldwork in Malawi is presented in Table 2. In the Canadian laboratory setting, the difference in counts was not statistically significant using water spiked with E. coli (p = 0.06). None of the samples which were free from E. coli after 18 hours of incubation (N = 49) developed countable colonies after 6 further hours of incubation.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Summary of paired Wilcoxon rank sum (non-parametric) analysis of counts.

https://doi.org/10.1371/journal.pwat.0000367.t002

In Malawi, there was a statistically significant difference between counts after 18 hours of incubation compared to 24 hours of incubation (p < 0.01); however there was no significant difference in the risk categories resulting from the counts (p = 0.15; Table 2). Two plates (6%, N = 34) that showed no detection after 18 hours of incubation developed countable colonies (1 and 4 CFU, respectively) after a further 6 hours of incubation. Also, in one case (3% of N = 34) a further 6 hours of incubation time changed a 10 CFU count into a 12 CFU count, thereby changing the risk category [39] of the resulting count. Therefore, in total, three samples (9% of the 34) changed risk category with the additional incubation time, though this result was not statistically significant (p = 0.15). One of these occurrences took place during each of the 3 weeks of water testing, suggesting that they were not attributable to site-specific water quality parameters. There were no such occurrences in the laboratory.

3.4. Validation of trained enumerator colony counts

We found no significant difference in colony counts between novice and experienced counters (p = 0.14; N = 36). This result suggests that the typical practice of having newly trained enumerators conduct counts during household surveys or water sampling campaigns is acceptable and comparable to counts by experienced counters.

4.1. Piloting of the LCK

In Malawi, we received positive informal feedback from enumerators during the piloting of the LCK. We also observed good stability and minimal loss of suction, as well as no breakage of the vacuum flasks, suggesting it is suitable for widespread use in low-resource settings. The positive experiences in this work serve to increase the confidence in the practice of using membrane filtration techniques outside of a conventional laboratory setting [13] and reinforce findings from previous work to validate the water quality module in Belize [3]. Further, findings from our pilot in Malawi are consistent feedback with the experiences of using the LCK in Afghanistan, with no major design changes suggested by enumerators during that survey [21–23] and further adoption of the LCK in Pakistan [24]. Our findings are consistent with the feedback for improvement from those surveys, which focused not on the kit itself but on accessories such as the carrying bag, forceps, and incubator cap. Based on this evidence, the LCK should be considered suitable for conducting quantitative microbial water quality analysis in low-resource settings, with the possibility for up-scaling in national household surveys and other drinking water monitoring activities.

To estimate the potential reduction in the cost of water testing in household surveys, we estimated the costs of the SK and LCK for MICS surveys published in 2019 (before any disruption by the COVID-19 pandemic). MICS surveys published in 2019 employed anywhere from 8 to 39 enumerators, with each enumerator using their own SK. On this basis, the SKs for all eight MICS surveys published in 2019 are valued at $195,800 (see S3 Table). If UNICEF and the National Statistical Offices implementing MICS surveys had adopted the LCK to conduct all MICS surveys published in 2019, they would have saved an estimated total of USD $185,100, or 94.5% of the budget for water testing equipment (See S3 Table). These substantial savings could reduce financial barriers and thus enable more widespread drinking water quality monitoring in household surveys and support other drinking water monitoring activities by service providers, regulators or ministries responsible for water quality surveillance. In addition to the cost savings, the LCK is approximately five times lighter than the SK, making it easier for enumerators to carry – an important consideration in surveys like MICS where the water tester is often also responsible for anthropometry, which requires carrying a set of scales and height measurement board.

4.2. Assessment of blank test water

E. coli was not detected in any of the sampled bottled water, including those used in Malawi and other included MICS trainings. This work supports the current practice in MICS of using pre-tested locally purchased bottled water to undertake negative quality control testing. This indication is positive, as laboratory-based blank tests typically employ autoclaved water samples, which is unfeasible in the context of water quality testing in many low-resource contexts. Further work is required to determine the number of bottles that need to be pre-tested to ensure local brands are suitable for use for blank testing, as well as protocols to select suitable bottled water. Interpretation of blank water testing results also needs standardization to determine suitable thresholds for incidences of positive tests within MICS-type surveys.

4.3. Evaluation of 18-Hour E. coli (reduced) incubation

Using a reduced 18-hour incubation period, samples plated in the afternoon could be counted the following morning, thus clearing incubation space for new samples, which could double the number of tests that could be incubated in a given 24-hour period, provided the sampling is spread across a 6-hour period in a typical workday. In Malawi, a large majority (91%) of samples examined showed the same risk category between 18 and 24 hours of incubation time. However, if the purpose is to classify water samples according to risk categories, it would be advisable to incubate samples for an additional 6 hours if after 18 hours if they produce a count of only 1 or 2 CFU/100 mL lower than the cut-off for the next highest risk category (e.g., 9 CFU/100 mL or 1 ∙ 10² CFU/100 mL), or for water in the low risk category (< 1 CFU/100mL). It may be possible to report presence/absence results following an abbreviated 18-hour incubation period, but further work is needed to verify this. In Malawi, the incubation step was undertaken using an incubation belt (Fig 2) worn by the enumerator, which kept samples at approximately 37°C. In other contexts, incubators are often the most expensive equipment required for microbial analysis [40] and require the availability of a reliable and continuous supply of electricity for the duration, presenting logistical and cost barriers. In addition to the use of an incubation belt as an alternate to a conventional incubator, other studies have examined simple, low-cost incubators heated without electricity, using a warm water bath [40,41] or powered using a solar-voltaic array [42,43]. Despite the energy savings and reduction of logistical burdens achieved by such innovative setups, such off-grid methods (including the incubation belt) typically have limited capacity.