Ensuring access to safe drinking water is a key strategy for reducing waterborne illness. The WHO/UNICEF Joint Monitoring Programme for Water Supply and Sanitation (JMP) differentiates between unimproved and improved sources to universally classify water access. This classification, however, is based on the type and location of the water source and does not take into account water quality; even sources classified as improved can have compromised water quality and pose a health risk from waterborne illness.

One type of improved water source in urban settings is piped drinking water on premises. However, the presence of a piped connection gives little information about the quality, quantity and frequency of water delivery. Even in settings with centralized water treatment, piped water distribution systems are vulnerable to performance deficiencies that can cause (re)contamination of treated water and plausibly lead to increased risk of gastrointestinal illness (GII) in consumers. It is well established that large system disruptions in piped water networks can cause GII outbreaks. We hypothesized that routine network problems can also contribute to background levels of waterborne illness and conducted a systematic review and meta-analysis to assess the impact of distribution system deficiencies on endemic GII. We reviewed published studies that compare direct tap water consumption to consumption of tap water re-treated at the point of use (POU) and studies of specific system deficiencies such as breach of physical or hydraulic pipe integrity and lack of disinfectant residual. In settings with network malfunction, consumers of tap versus POU-treated water had increased GII (IDR = 1.34, 1.00-1.79). The subset of non-blinded studies showed a marked association between GII and tap versus POU-treated water consumption (IDR = 1.52, 1.05-2.20); there was no association in studies that blinded participants to their POU water treatment status (IDR = 0.98, 0.90-1.08). Among studies focusing on specific network deficiencies, increased GII was associated with temporary water outages (RR = 3.26, 1.48-7.19) as well as chronic outages in intermittently operated distribution systems (OR = 1.61, 1.26-2.07). These findings suggest that tap water consumption is associated with GII in malfunctioning distribution networks. System deficiencies such as water outages increase the risk of GII, presenting a potential health risk for consumers served by piped water networks.

Additionally, intermittent delivery of piped water is a common form of water supply in low-income countries. Intermittent supply can lead to waterborne illness through contamination of water in pipelines or in household storage, use of unsafe water sources during intermittencies and limited water availability for hygiene. To assess the health impact of intermittent water delivery as a particular type of breach of hydraulic pipe integrity, we conducted a matched cohort study to assess the impact of switching from intermittent to continuous water supply in Hubli-Dharwad, India, on child diarrhea and mortality, and severe waterborne illness in tap water consumers. We used multivariate matching to match continuous supply areas to intermittent supply areas with comparable baseline characteristics in Hubli-Dharwad. We followed 3919 households with children under five over 15 months. In continuous supply areas, we observed 42% reduction in the percentage of households with at least one case of typhoid fever since the program was implemented (CIR = 0.58, 0.41-0.78) and potentially 49% reduction in the percentage of households with death of a child under the age of two (CIR = 0.51, 0.22-1.07). Consistently with these reductions, our findings also suggested reductions in seven-day prevalence of diarrhea (PR = 0.93, 0.83-1.04) blood or mucus in stool (PR = 0.78, 0.60-1.01) in children under five in continuous supply areas. These reductions were more pronounced in low-income households (diarrhea PR = 0.89, 0.76-1.04; blood/mucus PR = 0.63, 0.46-0.87). The effect of rainfall on the impact of continuous supply on child waterborne illness was inconclusive. Our findings indicate that switching from intermittent to continuous water supply reduced waterborne illness in Hubli-Dharwad.

Another type of improved water source, commonly used in rural settings, is tubewells that draw groundwater from shallow aquifers. Shallow tubewells are the primary drinking water source for the majority of rural Bangladeshis. While groundwater is often considered microbiologically safe, fecal contamination has been detected in tubewell water, typically at low concentrations at the source and at higher levels at the point of use. The magnitude of the waterborne disease burden associated with consumption of tubewell water is not well understood. We conducted a randomized controlled trial to assess whether improving the microbial quality of tubewell drinking water by household water treatment and safe storage would reduce diarrhea in children <2 y in rural Bangladesh. We randomly assigned 1800 households into one of three arms: chlorine plus safe storage, safe storage and control. We longitudinally followed households with monthly visits for one year to promote the products and collect health outcomes. Both interventions had high uptake in the study population. Safe storage, alone and in combination with chlorination, reduced heavy contamination of stored water. In the chlorine plus safe storage arm, 2% of stored water samples had E. coli concentrations exceeding 100 CFU/100 mL, compared to 7% in the safe storage arm and 21% in the control arm. Compared to controls, diarrhea prevalence in children <2 y was reduced by 36% in the chlorine plus safe storage arm (PR = 0.64, 0.55-0.73) and 31% in the safe storage arm (PR = 0.69, 0.60-0.80); there was no difference between the two intervention arms (PR = 0.92, 0.79-1.08). Our findings suggest that safe storage significantly improved drinking water quality and reduced child diarrhea in rural Bangladesh. There was no added benefit from combining safe storage with chlorination.

Taken together, this body of evidence confirms previous findings that, even for water sources categorized as improved by the JMP, there are water quality problems that can pose a public health threat. Our findings highlight deteriorations in water quality associated with deficiencies in the distribution and handling of drinking water from the point of source to the point of consumption, rather than contamination at the water source. These findings suggest that efforts to improve drinking water quality should place emphasis on preventing contamination at each step of the chain leading from the water source to the point of consumption, including the distribution system and household storage containers, to maximize the protection against waterborne illness.

Evaluating the Role of Sanitation in Improving Child Health and Nutrition: Does it Matter and Can We Make it Count?

by

Sumeet Rajshekhar Patil

Doctor of Philosophy in Epidemiology

University of California, Berkeley

Professor John M. Colford, Jr., Chair

Poor water, sanitation and hygiene (WASH) infrastructure and behaviors are believed to be the major contributors to the worldwide burden of diarrheal diseases and parasite infections; the second most leading cause of deaths among children under 5 years of age in developing and under-developed countries. India alone accounts for a third of those without improved sanitation (814 million), nearly 60% of those who practice open defecation (626 million), 25% of the world’s deaths from diarrheal diseases among children under 5 years of age and approximately one third of the stunted children globally. Compared to this staggering burden of poor WASH in India, research on which WASH interventions are efficacious is limited particularly on sanitation.

This dissertation seeks to bridge some of the gaps in sanitation public health research by answering following three questions.

Does India’s Total Sanitation Campaign (TSC) – one of the largest rural sanitation program in the world – deliver the hypothesized health benefits of improved sanitation to pre-school age children in India? I use a cluster randomized controlled trial to evaluate the effectiveness of the TSC in terms of prevalence of diarrhea, highly credible gastrointestinal infections, parasitic infections, anemia and child growth in terms of age adjusted height, weight and arm circumference. The sample consists of 5239 children under the age of 5 years from 80 villages in Dhar and Kargone districts of Madhya Pradesh state. I find that while the TSC almost doubled the coverage of private toilets (41% in the intervention group vs. 19% in the control group), the relative reduction in the open defecation rate was small and remained high in absolute magnitude (73% in the intervention group vs. 83% in the control group). Possibly due to inadequate reduction in open defecation levels, the TSC did not improve health of children under the age of 5 years in terms of above health outcome indicators.

Can the private toilet coverage increase substantially by reducing the price of the toilets through subsidies? Subsidies are a cornerstone of India’s TSC to increase private toilet coverage however little is known whether and to what extend these subsidies can increase the toilet coverage. I estimate the arc price elasticity of demand for private toilets using the data from the TSC trial in Madhya Pradesh. Taking advantage of variation in the level of subsidies offered by the Government of Madhya Pradesh to build private toilets, I find that the price elasticity of toilet demand is 0.91 so that if the price of a private toilet is reduced from ₹18000 to ₹6000 as per the new sanitation program norms in India, the private toilet coverage in rural India can increase from 30% (as per Census 2011 data) to 80%. However, using data from another experimental efficacy trial in Odisha of a pilot sanitation program that consisted of intensive behavior change, I find that the price elasticity is 0.26 and statistically not different from 0. The findings provide an evidence that the demand for private toilets is inelastic and reducing the price of toilets through subsidies may not be enough to increase the toilet coverage. Whether the built toilets be used regularly resulting in drastic reductions in open defecation levels, and whether this reduction in open defecation will result in improved health outcomes for children still remain unanswered.

What are the importance of risk factors including owning a private toilet in explaining linear growth faltering among children aged 6-24 months? I propose and apply a variable importance analysis method using SuperLearner –– a machine learning based ensemble algorithm –– to objectively and non-parametrically model the relationship between HAZ and 51 risk factors related to child nutrition, pre- and post-natal care, mothers’ health and nutrition, household socioeconomics, and water and sanitation. I also apply a new estimator called Targeted Maximum Likelihood Estimator to estimate the magnitude and standard error of variable importance measures.

I apply the proposed method the nationally representative Demographic and Health Survey data from India as a case study application. Subject to the available data and model limitations, I find that the following are main risk factors for stunting: mother of short stature (< 145 centimeter height); child not fed as per the WHO recommended guidelines; boiling drinking water; and children second or later in the birth order. I find that access to private sanitation explained -0.09 Z loss in HAZ which is a much smaller importance than above variables. However, the importance of sanitation may be underestimated because access to private toilet is an underestimate or poor indicator of reduction in open defecation or the reduction in exposure to enteric pathogens in the community.

I conclude my dissertation by underlying the need for more evidence based advocacy, design and implementation of sanitation programs than what was done over the past 20 years, and flag some of the important consideration in design of future studies based on the insights gained in developing this dissertation.

Many infectious bacterial and viral agents exist in the world and are located in areas where humans may come into contact with them. Food, water, and environmental locations may be contaminated with infectious material, and detecting the presence of these harmful biologic agents is of import to public health agencies. One method that has been used to determine if infectious agents may be present in food or water is measurement of "indicator" bacteria or viruses. Indicator organisms are easily measured bacteria or viruses whose presence in water or food is thought to parallel the potential presence of infectious agents in the same food or water samples. Because there are so many potential bacteria (or viruses) that may infect a sample, it is impractical to test for all of them; rather, measurement of a single indicator organism may be more feasible.

Indicator bacteria have been used to determine if marine waters at beaches across the United States are safe for swimming. Guidelines issued by the U.S. Environmental Protection Agency (U.S. EPA) have focused on determining when recreational waters may pose a risk of excess gastrointestinal illness among swimmers when compared to non-swimmers. However, marine environments are very complex, and tidal patterns, solar inactivation, water temperature, and many other factors all can influence the presence or absence of indicator and infectious microorganisms in the water. Research has indicated that indicator organisms may be useful in predicting gastrointestinal illness in marine environments, but other health outcomes have been less studied. In order to verify that indicator organisms do track well with infectious organisms, a systematic review and meta analysis was conducted to determine if indicator organisms can predict a different health outcome, skin infection. Once the link between indicator organisms and health outcomes was established, the next goal was to explore different methods to strengthen the relationship between indicator organisms and health. Currently, the U.S. EPA advises that a single bacterial indicator, Enterococcus, be measured in marine waters. A binary cutoff of above or below 104 colony forming units per 100 mL is used to advise whether a beach is unsafe or safe for swimming. In order to improve prediction of illness at beaches using indicator organisms, several methods were considered. Flexible statistical modeling techniques, such as SuperLearning, were used, as well as consideration of multiple biological and physical indicators at the same time. The final aim was to examine the potential sources of the infectious agents, as well as the indicator bacteria, at Avalon beach in Southern California.

The results of this investigation suggest that indicator bacteria can be quite useful in predicting human illness, but perform better under certain conditions. The systematic review and meta-analysis showed that there was a strong relationship between certain indicator organisms and skin infections in marine water settings. Higher concentrations of total coliform, fecal coliform, E. coli, Enterococcus, and fecal Streptococci were associated with increased risk of skin related illness in marine waters. These findings support the biological plausibility of using indicator organisms to predict illness, even in a complicated, dynamic environment such as a marine beach. The second investigation found that application of the U.S. EPA guidelines at Avalon Beach did not accurately predict when waters were unsafe for swimming. However, use of flexible statistical methods (SuperLearner) greatly improved prediction of gastrointestinal illness over traditional modeling methods, such as logistic regression. Further improvements were seen when, instead of using a single indicator organism, combinations of biological and physical indicators were used. By combining physical and biological indicators, it was possible to identify circumstances when elevated concentrations of Enterococcus predicted excess gastrointestinal illness in swimmers. When solar radiation levels were low, indicator bacteria concentrations were more strongly associated with adverse health outcomes, whereas higher solar radiation levels were protective. This finding is biologically plausible because it is thought that solar radiation can directly damage indicator bacteria as well as pathogens and render them non-viable. Thus, under high solar radiation conditions, indicator organisms as well as infectious organisms would be inactivated. The final analysis examined groundwater flow as a potential risk to swimmers at Avalon beach. Because of a leaking sewage infrastructure at Avalon, it is thought that groundwater flux might be transporting raw sewage contents into the ocean water. Sewage is known to carry potentially high levels of pathogenic organisms, and thus groundwater flow levels might pose a direct threat to swimmers. When groundwater flow was higher, the incidence of gastrointestinal illness was elevated among swimmers who swallowed water, relative to swimmers who swallowed water on days when groundwater flow was lower. Additionally, the relationship between groundwater flow and solar radiation was similar to that seen with Enterococcus and solar radiation. When solar radiation levels were high, groundwater flow was less predictive of excess gastrointestinal illness, as would be expected. When traditional analysis methods were used to relate traditional and rapid indicators to illness, relationships were much stronger when groundwater flow was high versus when groundwater flow levels were lower. In conclusion, the results of these analyses suggest that indicator organisms can be used to predict health outcomes in recreational water settings, but that their performance may be greatly improved by using flexible modeling techniques as well as other indicators, such as solar radiation.

Background: Scientists and development stakeholders argue that health interventions proven effective in randomized efficacy trials should be translated into large-scale programs to benefit public health. Substantive evidence supports the scale-up of numerous health interventions, such as water, sanitation, and deworming interventions, and since the establishment of the Millennium Development Goals (MDGs) the funding and motivation for such scale-up has grown. In the field of water and sanitation, numerous interventions have been demonstrated to be efficacious in the reduction of diarrhea and soil-transmitted helminth infection. However, scaling up these interventions to regional or national levels frequently presents implementation challenges, and systematically studying the reasons for scale-up success or failure is essential to refine and sustain public health programs. Another important feature of scaling up interventions is determining how best to integrate interventions at scale and whether intervention delivery should be focused at the individual, household, or community level. Population attributable fraction (PAF) parameters and a new class of parameters which build upon the PAF can be used to estimate the effect of large-scale programs on population health. Evaluation of interventions at scale poses unique questions, and epidemiologic designs and analyses need to be tailored to answer these particular questions. Modern approaches to PAF estimation allow for parameter definition to be tailored one's particular research question and are well suited to the evaluation of population-level effects of large-scale health interventions.

Methods: In this dissertation, I illustrate and apply methods to evaluate population-level effects of water, sanitation, and hygiene interventions. I specifically focus on methods for and applications with observational, cross-sectional data, and I discuss generalizations to other study designs. In the first chapter, I quantify the association between deworming, improved sanitation, and hygiene interventions and soil-transmitted helminths in a population in rural Bangladesh. I assess the potential for interactions between these interventions and explore associations at both the individual and village level. In the second chapter I assess the quality of implementation of a large-scale water, sanitation, and hygiene intervention implemented by UNICEF and the Government of Bangladesh in rural Bangladesh. It was found that this intervention did not meet most of its health and behavior targets in an interim evaluation. To help understand why, I envision a scenario in which implementation had been better in all areas, and I estimate how much outcomes may have changed under this scenario compared to the outcomes that were observed. In the third chapter, I discuss parameters appropriate for estimating population-level effects of health interventions. Specifically, I describe the estimation of the PAF and two modern parameters which build upon the PAF: the population intervention model and stochastic intervention model parameters. I provide a didactic description of the estimation of these parameters.

Significance: This dissertation illustrates the use of rigorous methods to systematically evaluate the effect of individual and combined interventions at scale. Rigorous assessment of water, sanitation, and hygiene interventions is difficult, even for small-scale interventions, and very few large-scale WASH interventions have been evaluated rigorously. The parameters I illustrated and estimated in this dissertation have broad applicability to similar assessments of other large-scale public health programs. My findings contribute to the growing empirical evidence base describing best practices for and barriers to delivering interventions at scale. This evidence may contribute to improvements in design, delivery, and prioritization of interventions which in turn could increase the health impact of such interventions when delivered at scale.