Faecal Coliforms and Escherichia coli Contamination in Drinking Water Sources in Cholera Hotspot Areas of Lusaka District, Zambia: A Cross-Sectional Study

Abstract

The October 2023 to 2024 cholera outbreak demonstrates significant challenges related to water quality and sanitation, especially in peri-urban areas with limited access to clean water. This study assesses the presence of faecal coliforms and Escherichia coli (E. coli) in drinking water sources across five townships, identified as cholera transmission hotspots, two months post the cholera outbreak in the Lusaka District. A total of 169 water samples were collected from protected sources, treated piped water, and unprotected sources, including dams and shallow wells. Faecal coliforms and E. coli were detected across all source types. Among unprotected sources, 92.3% (12/13) of samples contained ≥100 CFU/100 mL of both faecal coliforms and E. coli. Protected sources showed variable contamination, with 18.3% exceeding ≥100 CFU/100 mL for faecal coliforms and 15.4% for E. coli. Treated water sources showed the lowest contamination, with 88.5% of samples having no detectable faecal coliforms and 90.4% having no detectable E. coli. Zero-inflated negative binomial regression showed that treated water sources were associated with substantially lower faecal coliform counts compared with protected sources (PR = 0.11, 95% CI: 0.03–0.35), while unprotected sources exhibited higher contamination intensity (PR = 1.77, 95% CI: 0.94–3.31). Treated sources were significantly more likely to be structurally free of contamination, whereas unprotected sources had an extremely low probability of yielding zero counts. These findings indicate that current water safety conditions in Lusaka’s cholera hotspot areas remain inadequate for preventing faecal-oral transmission.

1. Introduction

Globally, access to safe drinking water remains a critical public health challenge, with faecal contamination representing a major pathway for diarrheal diseases like cholera. The presence of faecal coliforms in drinking water is an indicator of such contamination and signals an increased risk of enteric infections [1,2,3]. Cholera is responsible for up to 4 million cases and 143,000 deaths worldwide each year due to unsafe water and sanitation [4]. Cholera outbreaks disproportionately affect low and middle-income countries with inadequate water, sanitation, and hygiene (WASH) conditions [4]. Zambia recently experienced one of its most severe cholera outbreaks with over 20,000 cases and 740 deaths [5]. This outbreak exposed critical challenges related to water quality and sanitation, especially in peri-urban areas characterised by limited access to safe drinking water [6].

Inadequate water and sanitation infrastructure in urban informal settlements has increased the threat of waterborne diseases such as cholera. Lusaka, Zambia, exemplifies this challenge, experiencing recurrent cholera outbreaks that disproportionately affect densely populated peri-urban communities lacking access to safe water [6,7]. Recent studies have detected faecal contamination in community drinking water sources in these unplanned settlements [6,8], highlighting the public health hazard posed by unsafe water supplies. Faecal indicator bacteria like Escherichia coli (E. coli) are widely used to assess microbial water quality because their presence indicates faecal contamination and the potential presence of pathogenic microorganisms [9,10]. The World Health Organization (WHO) Guidelines for Drinking-water Quality require that no E. coli or other faecal coliforms be detectable in any 100 mL sample of drinking water [3], and Zambia’s national drinking water standards echo these criteria with a zero tolerance for faecal coliforms in water intended for human consumption [11]. Systematic assessment of microbial contamination in Lusaka’s water is therefore essential to inform interventions to meet these water safety standards and strengthen cholera prevention in urban informal settlements.

This study characterises the microbiological water quality in cholera-affected peri-urban communities of Lusaka District by quantifying faecal coliforms and E. coli in different drinking water sources, comparing contamination patterns across source types and townships, and assessing graded contamination and associated infection risk classifications. These findings aim to inform public health interventions, as unsafe drinking water in low-income peri-urban settings has been consistently linked to recurrent cholera outbreaks.

2. Materials and Methods

2.1. Study Design

This was a cross-sectional observational study of microbiological drinking water quality, with samples collected from different types of water sources (protected, treated, and unprotected) across five cholera-affected peri-urban townships (Chawama, Chipata, Kanyama, Matero, and Mtendere) in Lusaka district (Figure 1). These densely populated settlements are characterised by informal housing, limited access to safely managed water and sanitation services, and reliance on shallow wells, communal boreholes, and shared pit latrines, which increase vulnerability to waterborne disease transmission. We classified the drinking water sources into three categories: protected sources (protected wells and boreholes), treated sources (piped tap water), and unprotected sources (dams and unprotected shallow wells).

Figure 1.

Map indicating the peri-urban townships from Lusaka District where samples were collected.

2.2. Data Collection Procedures

Grab samples (500 mL) were collected and transported at 2–8 °C to the Centre for Infectious Disease Research in Zambia (CIDRZ) campus laboratory for further analysis. Glassware and materials required for water sample collection from the five townships were prepared. The two-liter sampling bottles and filter units were rinsed with distilled water and air-dried. To prevent contamination, the bottles were secured with aluminium foil over the cap and neck, and components of the filter units were similarly covered with aluminium foil. Both the bottles and filter units were sterilized in an autoclave at 121 °C for 20 min, then cooled to room temperature.

Sampling from taps or pump outlets involved cleaning the taps with methylated spirit for 30 s using flaming cotton wool. For plastic taps, 99% ethanol was used. The taps were turned on to maximum flow for 1–5 min to ensure a representative sample. The sterilized bottles were opened while maintaining contact with the aluminium foil to prevent dust entry, and filled to the shoulder level (500 mL). Free residual chlorine (FRC) was measured at the point of collection for treated (chlorinated) water sources using a portable colorimetric comparator (Palintest Comparator kit, Palintest, Gateshead, UK). Chlorine levels (mg/L) were determined immediately after sample collection by visual comparison against the manufacturer’s colour scale. This allowed confirmation of chlorination status and assessment of residual disinfectant levels at the time of sampling. For samples collected from treated (chlorinated) water sources, sampling bottles were pre-dosed with sodium thiosulfate (Na₂S₂O₃) to neutralise free residual chlorine immediately upon collection. This procedure was implemented to prevent continued disinfectant activity during transport and storage, which could otherwise inactivate bacteria and lead to underestimation of faecal coliforms and E. coli. Samples were then placed in a transport cooler box maintained at 2–8 °C and returned to the laboratory within 24 h. When sampling from wells, a string approximately 20 m long was attached to a sampling bottle and the bottle was lowered into the well without touching its sides; once filled, excess water was discarded to create air space before retrieval. For dams, bottles were dipped approximately 30 cm below the surface with the mouth facing slightly upwards.

2.3. Membrane Filtration for Indicator Organism Detection

Media for indicator testing was prepared according to manufacturer instructions, specifically Membrane Faecal Coliform (M-FC) (Himedia Laboratories, Thane, India), and MacConkey agar (Himedia Laboratories, India). Membrane filtration was done as described previously [12]; the filtration unit was assembled, and sterile forceps were used to place a 0.45 µm filter paper on the filtration unit. For each sample, 100 mL was poured into the unit while opening the vacuum to facilitate filtration. Once filtration was complete, the vacuum was closed, and the top half of the filter unit was removed. The membrane filter was aseptically transferred onto selective agar media for bacterial growth: M-FC agar for faecal coliform and MacConkey agar for E. coli differentiation. The filter unit was sterilized with 99% alcohol and left to air dry or flame-dried using a Bunsen burner flame. This process was repeated for each new sample. Distilled water blank samples were included in each analytical run and processed alongside field samples.

Faecal coliforms were detected by incubating M-FC agar plates at 44.5 °C for 18–24 h. This elevated temperature selectively promotes the growth of faecal coliforms while inhibiting non-faecal bacteria. Following incubation, colonies exhibiting characteristic blue to dark blue coloration were identified as faecal coliforms and enumerated using a hand-held magnifier. Results were expressed as colony-forming units per 100 mL (CFU/100 mL). MacConkey agar plates were incubated aerobically at 35–37 °C for 18–24 h. Presumptive E. coli colonies were identified based on their characteristic lactose-fermenting morphology, appearing as pink to red colonies due to acid production and neutral red indicator uptake [11,12]. Non-lactose fermenting colonies appeared colorless and were excluded from E. coli counts. To confirm E. coli, representative lactose-positive colonies were sub-cultured onto nutrient agar and subjected to indole testing using Kovac’s reagent (HiMedia Laboratories, India). Indole-positive isolates were considered confirmed E. coli, and the counts were reported as CFU/100 mL. No growth was observed in blank controls. For samples exhibiting colonies too numerous to count (TNTC) on initial filtration, serial dilutions were performed to obtain countable colonies within the recommended enumeration range. For samples that remained too numerous to count after serial dilution, results were reported as ≥100 CFU/100 mL.

2.4. Variables

2.4.1. Outcome

The bacterial levels were calculated as colony-forming units per 100 mL (CFU/100 mL) using the formula: CFU/100 mL = [(number of colonies counted)/(volume of sample filtered (mL)] × 100.

2.4.2. Covariates

We defined the types of drinking water sources as “protected” if the source was protected wells and boreholes; “treated” if the source was piped water from taps; and “unprotected” if the source was dams or unprotected shallow wells.

2.5. Statistical Analysis

We summarised the characteristics of the water samples using descriptive statistics. Categorical variables were presented as frequencies and proportions. For faecal coliform and E. coli counts (CFU/100 mL), which were highly skewed with many zeros, we reported medians and interquartile ranges overall and within categories of water source and township. To aid visual interpretation of these distributions, we produced bar charts of sample type and township, as well as overlaid scatter and box plots of faecal coliforms and E. coli by water source and township.

Because the raw CFU counts exhibited both over-dispersion and an excess of zeros, we modelled faecal coliform and E. coli counts using zero-inflated negative binomial (ZINB) regression. For each faecal coliform and E. coli count, we fitted a ZINB model with water source type and township as main predictors in both the negative binomial count component and the zero-inflation (logit) component. The count part estimated Prevalence ratios (PRs) for expected CFU counts among samples at risk of contamination, while the zero-inflation part estimated the log-odds of belonging to a “structural zero” group (samples assumed to be always uncontaminated). Model choice was guided by comparing alternative count models (including Poisson, negative binomial, zero-inflated Poisson, and two-part hurdle model composed of a logistic regression for the probability of any contamination (zero versus > 0 CFU) and a truncated negative binomial model for the positive counts) using Akaike’s Information Criterion (AIC) and the Bayesian Information Criterion (BIC), and the ZINB specification was retained where it provided the lowest AIC/BIC values and best captured the data structure. We reported PRs and 95% confidence intervals from the count component and log-odds coefficients with 95% confidence intervals from the zero-inflation component, together with p-values. Robust standard errors were used for all models, and statistical significance was assessed at the 5% level.

Although WHO drinking water compliance is defined as the absence of E. coli in 100mL, these ordered categories (0, 1–9, 10–<100, and ≥100 CFU/100 mL) were used for descriptive risk stratification consistent with environmental health studies [13,14]. We then cross-tabulated these categories by water source type and township, and compared distributions using Pearson’s chi-square tests or Fisher’s exact tests, depending on the distribution of observations across categories. These contingency tables provided the basis for describing the proportion of samples falling into each contamination band for both indicators across the five townships and three source types. All regression analyses were conducted in Stata version 19 (StataCorp, College Station, TX, USA). To estimate the potential risk of disease associated with the microbiological results, we applied established interpretation frameworks that correlate indicator organism levels with the probability of infection or disease outcomes. Microbiological contamination levels were therefore translated into risk categories (no, low, high) using thresholds previously applied in environmental health and water-quality risk assessments [13,14]. Using R (v4.5.2) with the sf, ggplot2, and ggspatial packages [15], we generated a map showing the spatial distribution of graded contamination and corresponding risk zones.

3. Results

3.1. Distribution and Characteristics of Water Samples

A total of 169 drinking water samples were collected from five peri-urban townships in Lusaka District identified as cholera transmission hotspots. Water sources were categorised as protected sources (N = 104), treated piped water sources (N = 52), and unprotected sources (N = 13), including shallow wells and surface water. Samples were distributed across the five townships as follows: Chawama (N = 30), Chipata (N = 30), Kanyama (N = 41), Matero (N = 27), and Mtendere (N = 41). Within each township, samples were collected from available source types to reflect local water access patterns. Unprotected sources were sampled only where present and accessible, resulting in smaller numbers of samples for this category. The total number of samples per source type and per township is explicitly reported in Table 1, providing a clear cross-tabulation of sampling coverage.

Table 1.

Distribution of water samples by source type and township.

Characteristic	Number of Water Samples (% Total); N = 169
Sample source
Protected water	104 (61.5)
Treated water	52 (30.8)
Unprotected water	13 (7.7)
Township
Chawama	30 (17.8)
Chipata	30 (17.8)
Kanyama	41 (24.3)
Matero	27 (16.0)
Mtendere	41 (24.3)

Percentages are calculated as a proportion of the total sample (N=169)

Of the 169 water samples analysed, the majority were drawn from protected water sources (61.5%), with treated sources accounting for nearly one-third of samples (30.8%) and only a small proportion originating from unprotected sources (7.7%). The samples were collected across five high-density townships, with Kanyama and Mtendere each contributing 24.3% of the samples, while Chawama and Chipata Compound each contributed 17.7%, and Matero 16.0%. This distribution indicates that the dataset is dominated by samples from relatively improved water sources and is reasonably balanced across the selected urban townships, with slightly higher representation from Kanyama and Mtendere (Table 1).

3.2. Zero-Inflated Negative Binomial Models of Microbial Contamination by Water Source

Overall, the distributions of both faecal coliforms and E. coli were highly skewed, with substantial zero counts in most groups and marked elevation in unprotected sources and specific townships. For faecal coliforms, the median level was 0 CFU/100 mL overall among protected and treated water sources, although the upper quartile reached 35 CFU/100 mL for protected sources, indicating contamination in a subset of samples, while all quartiles for unprotected sources clustered around 99 CFU/100 mL, indicating consistently high contamination levels. A similar pattern was observed for E. coli: medians were 0 CFU/100 mL for protected and treated sources, but 99 CFU/100 mL across the 25th, 50th, and 75th percentiles for unprotected sources. By township, faecal coliform and E. coli levels were highest and most widely distributed in Chawama, where the median values were 37.5 and 43 CFU/100 mL respectively and the upper quartiles reached 99 CFU/100 mL, while the other townships generally showed medians at or near zero with lower upper quartiles, indicating more modest but still variable contamination (Figure 2).

Figure 2.

Distribution of E. coli and faecal coliforms in water samples by source type and township. (a) Faecal coliforms by sample source type; (b) Faecal coliforms by township; (c) E. coli by sample source type; (d) E. coli by township.

In the zero-inflated negative binomial models, treated water sources had markedly lower faecal coliform counts than protected sources in the count component (PR = 0.11, 95% CI 0.03–0.35; p < 0.001), while unprotected sources showed a tendency toward higher faecal coliform counts, although this did not reach conventional statistical significance (PR = 1.77, 95% CI 0.94–3.31; p = 0.076). There was no clear evidence of differences in faecal coliform or E. coli counts across townships, with all township incidence rate ratios having wide confidence intervals spanning unity (Table 2). For E. coli, neither treated nor unprotected sources differed significantly from protected sources in the count part, reflecting substantial uncertainty around these effects. In contrast, the zero-inflation component suggested that treated water was substantially more likely than protected sources to belong to the “always-zero” group (i.e., structurally uncontaminated) for both faecal coliforms and E. coli (positive, statistically significant coefficients), whereas unprotected sources were strongly and inversely associated with zero inflation, indicating a very low probability of being structurally free of detectable contamination (Table 2).

Table 2.

Zero-inflated negative binomial regression models of faecal coliform and Escherichia coli contamination by water source type and township.

Characteristics	Number of Samples (% of Total)	Faecal Coliforms (CFU/100 mL)	E. coli (CFU/100 mL)
		PR [95% CI]; p	PR [95% CI]; p
Sample source type	N = 169	Negative binomial part CFU/100 mL
Protected water	104 (61.5)	1.00	1.00
Treated water	52 (30.8)	0.11 [0.03, 0.35] 0.000	1.44 [0.40, 5.22] 0.581
Unprotected water	13 (7.7)	1.77 [0.94, 3.31] 0.076	1.75 [0.87, 3.54] 0.117
Township
Chawama	30 (17.8)	1.00	1.00
Chipata	30 (17.8)	0.76 [0.35, 1.66] 0.493	0.75 [0.32, 1.72] 0.492
Kanyama	41 (24.2)	1.03 [0.50, 2.13] 0.937	0.81 [0.34, 1.93] 0.638
Matero	27 (16.0)	1.01 [0.39, 2.57] 0.989	0.68 [0.25, 1.81] 0.439
Mtendere	41 (24.2)	0.56 [0.26, 1.20] 0.136	0.50 [0.20, 1.28] 0.149
Sample source type		Zero-inflation part (log-odds scale)
Protected water	104 (61.5)	0.00	0.00
Treated water	52 (30.8)	2.30 [1.29, 3.30] 0.000	2.65 [1.62, 3.68] 0.000
Unprotected water	13 (7.7)	−20.10 [−21.58, −18.63] 0.000	-19.80 [-21.36, -18.24] 0.000
Township
Chawama	30 (17.8)	0.00	0.00
Chipata	30 (17.8)	−0.45 [−1.99, 1.09] 0.567	−0.41 [−2.01, 1.19] 0.617
Kanyama	41 (24.2)	−0.16 [−1.67, 1.34] 0.833	−0.15 [−1.73, 1.43] 0.852
Matero	27 (16.0)	−1.01 [−2.78, 0.77] 0.266	0.01 [−1.70, 1.71] 0.993
Mtendere East	41 (24.2)	−0.32 [−1.81, 1.17] 0.672	−0.84 [−2.49, 0.81] 0.318

PR = Prevalence Ratio;Estimates from the zero-inflation component are reported on the log-odds scale; CI, confidence interval.

3.3. Faecal Coliform and E. coli Contamination Levels

The faecal contamination levels in water samples collected from the five different townships were determined. For the water samples analysed, marked variation in faecal contamination level intensity was observed per township. The highest level of faecal coliform contamination was recorded in the unprotected water sources (particularly in the samples collected near dams), with 92.3% of samples (12/13) containing more than 100 CFU/100 mL (Table 3). In comparison, protected water sources showed high contamination, with 18.3% exceeding 100 CFU/100 mL, while treated water sources had substantially lower contamination, with 88.5% of samples showing no detectable faecal coliforms in all the townships. Other sites, such as Matero and Mtendere, had 14.8% and 7.3% water sources showing high contamination, respectively. Chipata and Kanyama showed comparatively lower contamination, with most samples containing fewer than 10 CFU/100 mL (Table 3).

Table 3.

Faecal coliforms and E. coli levels recorded in the water samples collected from the various Sampling sites.

Characteristic	Total N (%)	Faecal Coliforms (CFU/100 mL)				p Value	E. coli (CFU/100 mL)				p Value
	169 (100%)	0	1–9	10–<100	≥100		0	1–9	10–<100	≥100
Sample Source
Protected	104 (61.5%)	41 (39.4%)	16 (15.4%)	28 (26.9%)	19 (18.3%)	<0.001	41 (39.4%)	21 (20.2%)	26 (25.0%)	16 (15.4%)	<0.001
water
Treated water	52 (30.8%)	46 (88.5%)	4 (7.7%)	2 (3.8%)	0 (0.0%)		47 (90.4%)	3 (5.8%)	0 (0.0%)	2 (3.8%)
Unprotected water	13 (7.7%)	0 (0.0%)	0 (0.0%)	1 (7.7%)	12 (92.3%)		0 (0.0%)	0 (0.0%)	1 (7.7%)	12 (92.3%)
Township
Chawama	30 (17.8%)	13 (43.3%)	1 (3.3%)	3 (10.0%)	13 (43.3%)	0.004	13 (43.3%)	1 (3.3%)	3 (10.0%)	13 (43.3%)	0.014
Chipata	30 (17.8%)	14 (46.7%)	5 (16.7%)	6 (20.0%)	5 (16.7%)		14 (46.7%)	5 (16.7%)	6 (20.0%)	5 (16.7%)
Kanyama	41 (24.3%)	27 (65.9%)	4 (9.8%)	4 (9.8%)	6 (14.6%)		27 (65.9%)	5 (12.2%)	4 (9.8%)	5 (12.2%)
Matero	27 (16.0%)	12 (44.4%)	5 (18.5%)	6 (22.2%)	4 (14.8%)		17 (63.0%)	5 (18.5%)	2 (7.4%)	3 (11.1%)
Mtendere	41 (24.3%)	21 (51.2%)	5 (12.2%)	12 (29.3%)	3 (7.3%)		17 (41.5%)	8 (19.5%)	12 (29.3%)	4 (9.8%)

p-values are derived from Pearson’s chi-square or Fisher’s exact tests, as appropriate, comparing distributions across categories; WHO drinking-water compliance is defined as the absence of E. coli in 100 mL; p-values are derived from Pearson’s chi-square or Fisher’s exact tests, as appropriate.

The E. coli contamination levels by water source and township were also profiled. Unprotected sources had high contamination levels, with 92.3% of samples containing more than 100 CFU/100 mL of E. coli. Protected sources exhibited moderate contamination as 15.4% of samples exceeded 100 CFU/100 mL, while treated sources showed minimal contamination, with 90.4% registering no detectable E. coli and only 3.8% exceeding 100 CFU/100 mL. By township, Chawama again exhibited high E. coli counts, with 43.3% of samples surpassing 100 CFU/100 mL. Mtendere East and Matero recorded high E. coli levels in 9.8% and 11.1% of samples, respectively. Although Chipata and Kanyama recorded lower counts, both areas had detectable E. coli in several “protected” or “treated” water sources (Table 3).

3.4. Risk Assessment for Waterborne Infections

Water samples drawn from unprotected sources, which were collected only from Chawama, exhibited a high risk of waterborne infection, with bacterial indicators exceeding acceptable limits (Figure 3). In contrast, protected water sources exhibited a mixed risk profile, with samples distributed across no, low, and high-risk categories, indicating intermittent contamination rather than consistent safety. Treated water sources were predominantly classified as no risk, with most samples showing no detectable faecal coliforms or E. coli. However, a small number of treated samples fell into the low-risk category. Importantly, no treated water samples were classified as high risk. Overall, Figure 3 highlights that infection risk increases sharply with decreasing source protection and visually confirms that unprotected sources pose the greatest and most consistent public health risk, while treated sources provide the highest level of protection.

Figure 3.

Risk of Waterborne infection in the different types of water sources.

4. Discussion

This study presents critical findings on water contamination levels from five peri-urban townships in Lusaka District identified as cholera hotspots during the 2023/24 outbreak. Our results demonstrate that faecal coliform and E. coli contamination was unevenly dispersed across water sources and locations, with a large proportion of samples recording zero counts alongside a smaller subset showing very high contamination. This pattern was particularly evident in unprotected water sources in Chawama, where contamination levels frequently exceeded WHO guideline thresholds, while treated water sources were predominantly free of detectable bacteria. We confirmed that treated sources were substantially more likely to be structurally uncontaminated, whereas unprotected sources had an extremely low probability of being contamination-free, highlighting stark differences in microbial safety across source types. WHO guidelines recommend that there should be no detectable faecal coliforms in any 100 mL sample of drinking water [3].

The treated water sources had markedly lower faecal coliform levels compared to protected sources, reflecting the effectiveness of treatment in reducing microbial loads. In contrast, unprotected sources showed a tendency toward higher contamination intensity, consistent with their exposure to surface runoff, shallow groundwater vulnerability, and proximity to sanitation infrastructure. These findings are consistent with reports from other low-resource settings, including Pakistan, Nepal, Nigeria and Malawi, where unprotected sources are commonly associated with high faecal contamination, while treated sources show lower but non-zero risks due to post-treatment contamination or infrastructure failures [13,14]. Furthermore, a study in Malawi found widespread faecal contamination in water sources, emphasizing the regional consistency of these challenges [16]. In the Zambian context, contamination observed in unprotected sources in Chawama, particularly near Ngwenya and Blue Water Dam, likely reflects local environmental and infrastructural conditions that facilitate contamination events and elevated bacterial loads when contamination occurs. However, the absence of strong township-level effects in the count models suggests that, once contamination risk is present, source type plays a more dominant role than geographic location, while spatial factors may primarily influence the likelihood of contamination rather than its magnitude. This interpretation aligns with geospatial analyses of cholera transmission in Lusaka, which highlight the role of environmental exposure and sanitation rather than administrative boundaries in shaping outbreak risk [17,18,19]. The devastating 2023–2024 cholera outbreak in Zambia, with over 23,000 reported cases and 740 fatalities, clearly reveals the urgent need for improved water quality management and sanitation practices to mitigate these immediate risks.

The observed contamination patterns translate directly into an elevated risk of waterborne diseases within these peri-urban communities. The frequent detection of faecal coliforms and E. coli, particularly in unprotected and some protected water sources, indicates evidence of faecal exposure and a high likelihood of pathogen transmission [20]. The zero-inflated modelling further emphasises this risk by demonstrating that unprotected sources are highly unlikely to be structurally free from contamination, while treated sources are more often truly uncontaminated. Accordingly, the zero-inflated negative binomial model was used for explanatory purposes to characterise relative associations and contamination patterns in the observed data, rather than to generate predictive or population-level estimates. This distinction is critical, as it suggests that while some water sources pose a high infection risk, others may present intermittent contamination risk. Such patterns are epidemiologically important in cholera-endemic settings, where repeated low-level exposure or sporadic high-dose exposure can sustain transmission. These findings reinforce the link between contaminated drinking water and increased susceptibility to cholera and other diarrhoeal diseases, particularly in densely populated settlements with limited sanitation infrastructure. Beyond cholera, inadequate access to safe drinking water contributes significantly to the burden of other diarrheal diseases, which the WHO estimates account for approximately 2.2 million deaths annually [21]. In Lusaka’s peri-urban areas, where sanitation infrastructure is often lacking or poorly maintained [6,22], the risk of disease transmission is increased, particularly affecting children under five years old who are highly vulnerable to severe dehydration from diarrhoea. Poverty, rapid urbanization, and the growth of informal settlements in Lusaka increase the challenges of providing access to clean water and sanitation. Limited access to basic services, combined with overcrowded living conditions, creates an environment conducive to the spread of infectious diseases.

This study provides a scientific basis for prioritizing investments in safe water infrastructure and hygiene promotion in cholera-prone peri-urban areas, contributing to sustainable disease control and improved health outcomes. However, several limitations should be acknowledged. The study was cross-sectional and conducted approximately two months post-outbreak, limiting the ability to assess temporal variability or seasonal effects on contamination patterns. As contamination levels in surface and shallow groundwater sources can vary substantially with rainfall, water table dynamics, and usage patterns, the results reflect conditions at the time of sampling rather than long-term trends. In addition, each water source was sampled once, restricting assessment of short-term variability or intermittent contamination events. Microbiological identification of E. coli was based on standard indicator-organism protocols using selective culture and biochemical confirmation, which, while widely applied in environmental surveillance, may not fully resolve all members of the Enterobacteriaceae. In addition, demographic and household-level data were not collected, restricting the ability to link contamination directly to behavioural or socioeconomic factors. Finally, the relatively small number of unprotected water sources limits the ability to draw strong township-level inferences for this category. These limitations should be considered when interpreting the findings, and the results primarily highlight contamination patterns rather than providing definitive spatial conclusions.

From an environmental perspective, the findings highlight how inadequate source protection, shallow groundwater systems, and poor sanitation infrastructure interact to sustain faecal contamination in peri-urban settings. The observed contamination in unprotected and intermittently contaminated protected sources suggests ongoing environmental exposure pathways driven by surface runoff, informal waste disposal, and compromised drainage, rather than isolated point failures. While the study focused on selected cholera hotspot townships and was not designed to provide district-wide representativeness, it nevertheless captures conditions typical of many rapidly urbanising settlements in Lusaka and similar settings. The findings have important implications for cholera prevention and water safety interventions in peri-urban Lusaka. The strong association between unprotected water sources and both a low probability of being contamination-free and high bacterial loads shows the urgent need to reduce reliance on these sources through improved infrastructure, source protection, and alternative water provision. The presence of contamination in a subset of protected and treated sources reveals the need for continuous monitoring, effective chlorination, and community education on safe water handling and storage.

5. Conclusions

Our study shows that drinking water in peri-urban cholera hotspot townships of Lusaka is frequently contaminated with faecal coliforms and E. coli, with levels in many samples exceeding the WHO guidelines of zero detectable indicator organisms per 100 mL. Unprotected sources were almost universally contaminated, and while treated water sources had substantially lower faecal coliform counts and a much higher probability of yielding no detectable organisms, contamination was still observed in a non-trivial proportion of samples. Our findings indicate that current water safety conditions in Lusaka’s cholera hotspot areas remain inadequate for preventing faecal-oral transmission. Integrated interventions can improve protection and treatment of water at source, strengthen point-of-use treatment and safe storage at the household level, and embed routine microbial monitoring into cholera prevention and control strategies.

Abbreviations

The following abbreviations are used in this manuscript:

AIC	Akaike’s Information Criterion
BIC	Bayesian Information Criterion
CFU	Colony-Forming Units
CI	Confidence Interval(s)
CIDRZ	Centre for Infectious Disease Research in Zambia
°C	Degrees Celsius
E. coli	Escherichia coli
FRC	Free Residue Chlorine
M-FC	Membrane Faecal Coliform
mL	Milliliters
NHRA	National Health Research Authority
Na2S2O3	Sodium thiosulfate
PR	Prevalence Ratio
UNZABREC	University of Zambia Biomedical Research Ethics Committee
WASH	Water, Sanitation, and Hygiene
WHO	World Health Organization
ZINB	Zero-Inflated Negative Binomial

Author Contributions

Conceptualisation, H.N. and C.C.C.; methodology, S.S., R.K. and C.S.; formal analysis, C.C.L., B.P., D.N. and S.B.; investigation, F.L. and H.N.; resources, R.C. and N.R.T.; data curation, S.S., R.K., C.S. and E.W.; writing—original draft preparation, F.L., H.N. and B.P.; writing—review and editing, S.B., C.C.L., D.N., N.M., W.K., E.W., D.S., L.L. and D.S.; supervision, W.K., C.C.C., D.S., L.L., N.R.T. and N.M.; funding acquisition, S.B., R.C. and C.C.C. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data are available to qualified researchers upon request, subject to approval by the CIDRZ Ethics and Compliance Committee. Requests should be emailed to Ms. Hope Chinganya, Secretary to the Committee and Head of Research Operations (Hope.Chinganya@cidrz.org), and must specify the requester’s contact details, project title, planned analyses, and desired data format. Data use is restricted to purposes aligned with the original study; the Committee typically reviews submissions within 48–72 h (business days), notifying applicants of approval status or any required additional details.

Conflicts of Interest

The authors declare no conflicts of interest.

Funding Statement

This research was funded in part by the Science for Africa Foundation to the Developing Excellence in Leadership, Training and Science in Africa (DELTAS Africa) programme, grant number DEL-22-012. This study was also partly funded by the ELMA Cholera Relief Foundation.