Fecal Contamination of Drinking-Water in Low- and Middle-Income CountriesA Systematic Review and Meta-Analysis

If a household uses an improved water source, such as a protected well, a piped connection, or a borehole, then according to the global monitoring standard, they have safe drinking water. If they have safe drinking water, they are counted as protected. If they are counted as protected, then the world is making progress. Now consider this number: in more than a quarter of water samples taken from those improved sources, researchers found fecal contamination. The chain breaks at the very first link. That finding comes from a systematic review and meta-analysis by Bain and colleagues, published in PLOS Medicine. This study pooled data from three hundred nineteen studies covering nearly ninety-seven thousand water samples across low- and middle-income countries. It represents the most comprehensive attempt yet to test the assumption embedded in global water monitoring — that the type of source a person uses tells you something meaningful about the safety of what they drink. The monitoring system in question is run by the World Health Organization and UNICEF through their Joint Monitoring Programme, or JMP. The JMP tracks progress on safe drinking water using a single proxy: whether people use an improved source. The definition is infrastructure-based. Improved sources include piped water into a dwelling, public standpipes, boreholes, protected dug wells, protected springs, and rainwater harvesting. Unimproved sources, such as unprotected wells, unprotected springs, and surface water, are presumed to offer less protection from contamination. The JMP itself acknowledged that this indicator does not take water quality measurements into account. However, it was adopted because it can be derived from household surveys and national censuses, making it scalable across dozens of countries. Actual microbial testing cannot. Measuring water safety properly means testing for fecal indicator bacteria. The World Health Organization recommends Escherichia coli, or E. coli, as the gold standard, with thermotolerant coliforms as an acceptable alternative. The guideline value is clear: zero E. coli detected in any one hundred milliliter sample. Bain and colleagues used these indicators as their benchmark across all three hundred nineteen included studies. To assemble that evidence base, the team searched PubMed, Web of Science, the Global Health Library, and grey literature sources, pulling studies in five languages — Chinese, English, French, Portuguese, and Spanish — published between 1990 and 2013. From over seven thousand eight hundred initial records, they arrived at three hundred nineteen studies reporting on ninety-six thousand seven hundred thirty-seven water samples. To synthesize across studies, they used meta-analysis, pooling odds ratios to estimate the overall effect, and meta-regression, a technique that allows you to ask which factors predict variation between studies. Heterogeneity was high, with an I-squared value of eighty point three percent. Bain and colleagues treat that variation not as a problem but as the most important signal in the data. So what did the data actually show? The headline result is real and worth stating clearly: improved sources were substantially less likely to harbor fecal contamination than unimproved ones. The pooled odds ratio was zero point fifteen, with a confidence interval running from zero point ten to zero point twenty-one. That means the odds of detecting fecal indicator bacteria in an improved source were roughly eighty-five percent lower than in an unimproved source, within the same study. That represents a meaningful protective effect. Improved sources are genuinely better. However, the second finding is where the story gets uncomfortable. In thirty-eight percent of the one hundred ninety-one studies that reported data on improved sources, at least one quarter of samples from those improved sources were contaminated. Protected dug wells, Bain and colleagues note, were rarely free of fecal contamination. Even boreholes and piped supplies occasionally showed high contamination levels. The picture within the improved category is uneven. Piped supplies were less likely to be contaminated than other improved types, with an odds ratio of zero point fifty-three. Treatment made a dramatic difference: treated piped supplies versus untreated piped supplies showed an odds ratio of zero point zero seven, meaning treated systems were ninety-three percent less likely to test positive. Groundwater protection mattered too, with an odds ratio of zero point twenty-six for protected versus unprotected groundwater sources. The label improved collapses a wide range of actual safety into a single category. Two factors predict contamination risk more powerfully than almost anything else: where you are and how wealthy your country is. Water sources in low-income countries had two point thirty-seven times the odds of contamination compared to other countries. Rural sources had two point thirty-seven times the odds compared to urban ones. In both cases, the confidence intervals ran from roughly one point five to three point eight, indicating the effect is real and consistent. The same borehole or protected well carries a meaningfully higher risk of fecal contamination depending on whether it sits in a rural village in a low-income country or in an urban area in a middle-income one. Then there's the problem of where samples were actually collected. Most studies sampled water at the source — the tap, the well, or the borehole itself. However, people don't drink water at the source. They carry it home in containers, store it in vessels, and drink it hours or days later. Where Bain and colleagues could compare stored household water against source water, the result was stark: stored water was more than twice as likely to be contaminated, with an odds ratio of two point zero nine. This is a compounding problem. Not only does source water quality vary, but quality deteriorates further between collection and consumption, and almost none of the three hundred nineteen studies tracked that degradation. Sanitary risk inspections were also rarely reported. The review is thus likely underestimating the true burden of contamination reaching people's mouths. That underestimation has direct consequences for global health accounting. The Global Burden of Disease two thousand ten assumed zero risk from improved sources. Bain and colleagues argue that assumption likely substantially undercounts the diarrheal disease burden. When the JMP reports on global progress toward safe water access, it counts every household using an improved source as protected, regardless of whether that source tested clean last week, last year, or ever. Given that seven hundred eighty-three million people still lacked an improved source at the time of this study, and that piped supplies, which serve roughly sixty-three percent of the world's population, are themselves not uniformly safe, the scale of overestimation is significant. The paper is specific about what better monitoring would require. Microbial testing should be paired with sanitary inspection rather than being substituted by infrastructure classification alone. Treated and untreated piped supplies should be reported separately, as their contamination rates differ by an order of magnitude. Stored water should be sampled alongside source water. Testing should happen more frequently to catch seasonal and episodic failures that single-visit surveys miss. None of this is technically impossible. What's been missing is the institutional will to move beyond the convenience of a binary infrastructure indicator. The authors conclude with the stakes framed plainly: access to safe drinking water is a human right. This is not rhetorical decoration — it is the standard against which the data should be measured. Improved sources are genuinely safer than unimproved ones, and that matters for health and equity across hundreds of millions of households. However, safer than unimproved is not the same as safe. A monitoring system that treats those two things as equivalent is not measuring progress toward a human right. It is measuring something easier to count and calling it the same thing. Bain and colleagues have now quantified the gap between those two claims across nearly one hundred thousand water samples in dozens of countries over more than two decades. The gap is not narrow. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.