Access to safe drinking water remains a significant challenge in sub-Saharan Africa (SSA), with a substantial portion of the population exposed to contaminated water sources. While improved water sources like boreholes with handpumps are often employed, they are not immune to contamination. The World Health Organization (WHO) recommends both sanitary inspection (a visual assessment of potential contamination sources) and microbial water quality analysis to assess water safety risks. However, the relationship between these two approaches is not well understood. This study aims to clarify this relationship, investigating whether sanitary inspection scores effectively predict the presence of E. coli, a key indicator of fecal contamination. The research questions address the nature of the relationship between sanitary risk scores and E. coli occurrence, the best method for quantifying and modeling this relationship, and how these findings can improve drinking water service delivery and monitoring in rural SSA. The study uses data from a large sample of handpumps in diverse locations across SSA, providing a robust basis for analysis and generalization.

Existing literature exploring the relationship between sanitary inspection and water quality has yielded mixed results, with some studies finding no significant association. This apparent contradiction has led to questions about the effectiveness of sanitary inspections. Previous research has often relied on untested assumptions about the positive and linear association between sanitary risk scores and E. coli concentrations. Statistical methods used have varied, further complicating comparisons. This study aims to address these limitations by using a more nuanced approach to modeling the relationship.

The study analyzed data collected from 1028 boreholes with handpumps in 12 sub-Saharan African countries as part of a larger World Vision water, sanitation, and hygiene (WASH) program evaluation. Data collection involved sanitary inspections using an adapted WHO form, generating a sanitary risk score for each handpump (ranging from 0 to 10). Water quality analysis determined the presence or absence of E. coli in 100-mL samples. The researchers employed both established statistical models (logistic regression, ordinal logistic regression, Pearson’s chi-square) and a novel alternative model. The established models directly compared sanitary risk scores or classes with E. coli occurrence. The alternative model categorized sanitary risk factors into contamination sources, carriers, and barrier breakdowns, assessing the relationship between combinations of these factors and E. coli occurrence. Logistic regression was used to analyze the alternative model, exploring all possible combinations of source, carrier, and barrier breakdown presence/absence. Rainfall in the past two days was also incorporated as a potential carrier in a separate analysis. Model fit was compared using Akaike's Information Criterion (AIC).

Descriptive statistics showed that 78% of handpumps had no detectable E. coli, with sanitary risk scores ranging from 0 to 9 and averaging 3.4. The most common risk factor was a missing fence (78%), while the least common was a latrine within 10 meters (1%). Established statistical models revealed no significant association between sanitary risk score and E. coli occurrence. Logistic regression, showing the best model fit among the established models, considered individual sanitary risk factors and E. coli presence/absence; it identified significant associations between E. coli presence and specific factors such as a latrine on higher ground (OR = 0.60, p = 0.046), a broken drainage channel (OR = 1.8, p = 0.005), and cracks in the apron (OR = 0.56, p = 0.002). The alternative model demonstrated a slightly better fit than most established models; the model fit further improved when rainfall in the past two days was added as a carrier. However, even with this improvement, no single risk scenario was consistently and significantly associated with E. coli occurrence at a 95% confidence level. The alternative model indicated that only the scenario with a barrier breakdown present but the source and carrier absent was significantly associated with E. coli occurrence (OR = 3.6, p = 0.019).

The lack of a strong correlation between sanitary risk scores and E. coli occurrence validates the use of the alternative model which better reflects the causal pathways of contamination. The weak associations found may be due to several factors: the inherent variability of E. coli levels over time and space; the incomplete capture of all relevant contamination sources, carriers, and barrier breakdowns by current sanitary inspection tools; and the limitations of using a model that does not fully account for the complex interplay of risk factors in anomalous microbial events. The finding that barrier breakdowns were more strongly linked to E. coli than sources or carriers suggests either that crucial sources and carriers are missing from the inspection forms, that these are harder to identify visually, or that barrier breakdowns are the most significant factor in contamination events. The study's findings highlight that sanitary inspection and microbial water quality analysis provide complementary but distinct types of information.

This study reinforces the value of both sanitary inspection and water quality analysis as complementary tools for assessing water safety. Sanitary inspection, being simpler, less resource intensive, and providing actionable information, should be prioritized in settings where frequent water quality testing is infeasible. The alternative model offers a more nuanced framework for understanding contamination pathways. Future research should focus on improving sanitary inspection forms by including additional relevant risk factors, particularly carriers, and exploring the impact of weighting risk factors. Longitudinal studies incorporating meteorological, water quality, and health outcome data are also recommended to strengthen our understanding of the relationship between these factors and water safety.

The study's limitations include its reliance on a dataset aggregated from 12 geographically diverse countries, precluding detailed country-level analysis. The models used were relatively simple, omitting other potentially relevant water system variables (e.g., well depth, soil characteristics) or community-level factors (e.g., sanitation practices). The classification of some risk factors as either sources, carriers, or barrier breakdowns was not always straightforward, which could affect the interpretation of the results. Future studies should address these limitations by employing more complex models and incorporating additional variables to create a more complete picture of the factors influencing water contamination.