Acute encephalitis syndrome (AES) is a disease condition characterized be presence of fever, and altered consciousness with or without presence of seizures or a neurological deficit. This definition is broad, and AES may be caused by a wide variety of neurotropic viruses, bacteria, protozoa, fungi or may even be non-infectious in etiology. Most AES is however considered to be due to a viral encephalitis, a condition which is common in many developing countries. The epidemiologic features of AES, especially in adults are still not well defined, and this study is first of its kind from India.

This dissertation consists of three original studies and two systematic literature reviews. We first performed a retrospective chart review of all patients with fever presenting to a single hospital in central India, to determine proportions of patients with AES, and to understand diagnostic and management practices in this condition. We found that AES accounts for about 20% of all acute fevers, has a high mortality, remains frequently undiagnosed, and is mostly empirically treated as malaria. In the subsequent year we designed a descriptive study to understand the temporal and spatial profile, and survival characterictics. We also conducted a case-control study to understand risk factors for AES. We found that most AES occurs in hot and humid months of the year, and proximity to a river or stream may be associated with development of the disease. About half of all patients with AES died within 30 days of disease, and low Glasgow coma score, and need for assisted ventilation significantly predicted hazard for mortality. In our extensive laboratory testing of cerebrospinal fluid and serum samples obtained from 152 AES cases suspected of having viral encephalitis, we found 31 (17%) patients who had a confirmed viral etiology. Enteroviruses were the commonest etiology (9.2% of all AES cases) followed by flaviviruses (4.3% of all AES cases). Based on serology 16 (8.8%) patients had probable leptospirosis. In a third of all cases etiology remained unknown. Low socioeconomic status was the only risk factor significantly associated with AES. These studies have provided novel insights into epidemiology of AES in India, and it is likely that most adult AES is due to water borne enteroviruses rather than vector borne flaviviruses as previously believed

Context: Fractures of the femoral shaft are common and have potentially serious consequences in patients with multiple traumatic injuries. The appropriate timing of fracture repair is controversial.

Objective: To assess the effect of timing of internal fixation on mortality in multi-system trauma patients

Design: Retrospective cohort.

Setting: Data from public and private Trauma Centers throughout the United States reported to the National Trauma Data Bank (version 5.0 for 2000-2005).

Patients: 3069 multi-system trauma patients (Injury Severity Score >15) who underwent internal fixation of femoral shaft fractures.

Intervention: Time to treatment was defined in categories as the time from admission to internal fixation: t0<12 hours; 12

Main Outcome Measure: The relative risk of in-hospital mortality comparing the four later periods to the earliest one was estimated using logistic regression, multivariate standardized risk ratios, and Inverse Probability of Treatment Weighting (IPTW). Subgroups with serious head, chest, abdominal and additional extremity injury were investigated.

Results: The mortality risk estimated by IPTW was significantly lower in several time categories: t1 (RR=0.45, 95%CI: 0.15 - 0.98, p=0.03), t2 (RR=0.83, 95%CI: 0.43 - 1.44, p=0.49), t3 (RR=0.58, 95%CI: 0.28 - 0.93, p=0.03) and t4 (RR=0.43, 95%CI: 0.097 - 0.94, p=0.03). Those with serious abdominal trauma (Abbreviated Injury Score, AIS>3) experienced the greatest benefit from delay of internal fixation beyond 12 hours (RR=0.82 for AIS<3, 95%CI: 0.54 - 1.35; versus RR=0.36 for AIS>3, 95%CI: 0.13 - 0.87; p-value for interaction =0.09). The results of the logistic regression and multivariate standardized risk ratios agreed with IPTW estimates. Qualitative and quantitative testing of modeling assumptions supported the validity of the marginal IPTW estimates.

Conclusion: Delayed repair of femoral shaft fractures beyond 12 hours in multi-system trauma patients significantly reduced mortality for three of the four treatment periods studied. Patients with serious abdominal injury benefit most from delayed treatment. This work applies causal inference methodology and assumption testing that is well suited to orthopaedic therapeutic studies when randomization is not possible or ethical.

Background: Every child deserves an equal chance at good health and cognitive development, but growth failure from undernutrition and disease rob many poor children of this opportunity in the first 2 years of their life. Child malnutrition kills roughly three million children every year, and many more suffer long-term health consequences from impaired physical growth and cognitive development, especially if malnutrition occurs during the first 1000 days of life. Wasting and stunting, two markers of child malnutrition, are strongly associated with increased mortality and decreased cognitive development, and are still highly prevalent in developing countries, even though their reduction are targets of the World Health Organization Millennium Development Goals. Trials on nutritional interventions have had null results or not clinically meaningful effect sizes. Most existing research on exposures associated with stunting and wasting rely on small cohorts or large cross-sectional surveys focused on stunting and wasting prevalence. Pooled analyses of data from many longitudinal studies offer unique opportunities to identify novel patterns in child undernutrition across combinations of time, space, and human characteristics, which is not possible in smaller datasets. The objective of this research was to answer the following questions by pooling information across multiple cohorts: when and where growth failure occurs, what characteristics best identify children at risk, and how early growth failure affects later growth and mortality. Answering these questions may help inform generalizable rules for child monitoring and develop preventative intervention to stop the initial onset of wasting and stunting.

Methods: In this dissertation, I conduct an individual-participant-data meta-analysis of 35 different longitudinal cohorts and trials from 15 different countries to identify patterns in wasting incidence and recovery across child ages, across regions, and across seasons. I use machine learning and targeted learning methods to identify key child, parental, and household characteristics associated with wasting and stunting, adjusting for potential confounders. In addition to estimating relative risks of wasting between levels of characteristics, I estimate more policy-relevant population attributable fractions and a novel variable importance metric. In the first chapter I introduce the motivation for the analyses and give an overview of individual participant data meta-analysis methods and advantages. In the second chapter I examine longitudinal patterns in wasting incidence, recovery, seasonality, and concurrence with stunting by using a subset of 18 monthly-measured cohorts, and I compare differences in the epidemiology of wasting across regions. In the third chapter I estimate the relative importance of the causes of early life growth failure using targeted learning methodologies. Using machine learning to flexibly adjust for measured confounders, I estimate and rank order associations between 36 child, parental, and household characteristics and measures of growth failure in 31 cohorts and trials. In the fourth chapter I estimate the associations between early growth failure and subsequent adverse outcomes, including mortality. In the fifth chapter I discuss key subject-area advancements in knowledge from these analyses and offer the public health implications.

Significance: This dissertation is currently the largest scale individual-participant data analysis of longitudinal patterns in and causes of child malnutrition in the first 1000 days of life. Using longitudinal, high quality anthropometry measurements, I examined patterns in wasting onset, recovery, seasonality, and concurrence with stunting not possible from national health surveys or accurate from small cohorts. The data were large in scope as well as scale, with data from 35 separate research sites (94,019 children and 645,869 total measurements), 18 with monthly measured cohorts (10,854 children and 187,215 anthropometry measurements). I was able to examine effect modification of wasting incidence by region and season. The results from chapter two found a high prevalence of wasting incidence at birth and the first three months of life, especially in South Asia. This is earlier than the paradigm in the nutrition field that most wasting episodes occur in children older than six months, after the cessation of exclusive breastfeeding and the start of crawling behavior. My findings suggest that interventions to prevent child acute malnutrition may have the largest impact if delivered earlier, including maternal interventions during pregnancy and early lactation and child interventions through the first months of life. Development of maternally focused, prenatal interventions represents an important opportunity to prevent child growth failure in lower-or-middle income countries (LMICs), especially as I found that early growth failure had a strong relationship with mortality. Future research into potential tradeoffs in prioritizing exclusive breast feeding versus the prevention or treatment of wasting in the first 6 months is needed. I also found high seasonality in wasting in synchrony with rainfall patterns, possibly due to seasonality in food availability and infectious diseases, indicating the timing of intervention delivery should target seasonal peaks in wasting. The strongest determinants of child growth failure were maternal anthropometry and child birth size. Future research should use the key determinants identified to find the optimal subgroups of children who would benefit most from postnatal interventions to start and reduce the substantial global health burden of child growth failure.

Globally, diarrhea is one of the leading causes of morbidity and mortality. Populations living in developing countries disproportionally experience the burden of diarrhea; in these areas contaminated drinking water remains one of the primary environmental pathways for the transmission of diarrheagenic pathogens.

Centrally treated "piped" water is largely recognized as the best method to provide human populations with consistent access to safe drinking water. Piped water requires little additional input from end users, and therefore has few additional barriers to the adoption and sustained consumption of water that is free of microbial contamination. In areas where the construction of piped water systems is not feasible due to capital or maintenance costs, or dispersed rural populations, the promotion of household water treatment and safe storage (HWTS) strategies have been recommended as the best method to deliver to safe water to households that lack affordable alternatives.

Several HWTS strategies have been tested in field efficacy trials, including treatment by chlorine, filtration, boiling and solar disinfection. While various meta-analyses of these trials have suggested that HWTS strategies improve drinking water quality, and reduce diarrhea, there is some concern that: 1) the design and implementation of these trials may have biased results, and relatedly; 2) that these strategies themselves are not being used by target populations outside the context of these intensive trials. Specifically, each of the HWTS strategies listed above comes with small barriers to use: chlorine changes the taste and odor of the water and is not effective against spore forming pathogens (e.g., cryptosporidium), filters are slow and are have limited effectiveness against viruses, boiling changes the temperature of water and requires significant time and fuel inputs, and solar disinfection requires significant time inputs from end users to effectively produce safe water. In efficacy trials with intensive implementation activities and frequent follow up visits, there is concern that the evaluation itself is impacting participant usage of HWTS technologies thereby affecting the generalizability of results. In order to develop and evaluate strategies that will effectively provide safe drinking water in regional contexts it will be necessary to consider additional HWTS strategies and to use evaluation designs that yield results that are generalizable beyond the research context of a specific trial.

In this dissertation, I present three chapters that further the evidence regarding HWTS strategies, trial methodologies, and outcome measures that are available to the HWTS sector. In Chapter 1, I report on the evaluation of a safe water program that promotes a novel ultraviolet-based HWTS strategy to rural Mexican households. Ultraviolet disinfection, while more expensive than the other HWTS technologies mentioned above, and requiring a nominal source of electricity, has the potential to reduce the barriers to adoption and sustained use: it is fast, it inactivates all classes of pathogens (bacteria, viruses, parasites), and does not change the taste or odor of the water. The study presented in this chapter represent the first health evaluation of an ultraviolet-based HWTS strategy in a low- or middle-income country. The study design used for this evaluation, a randomized stepped wedge design, is also detailed; its advantages with respect to producing generalizable results are discussed. Specifically, the design allows an HWTS intervention to be rolled-out as it would have been under non-trial conditions, while maintaining the rigor of a randomized experimental design. The intention-to-treat (ITT) results from the trial show that while there were large improvements in drinking water quality (a 19 percentage point reduction [95% CI: −27%, −14%] in drinking water contamination, measured as the most probable number of Escherichia coli /100 ml [MPN EC]), there was not a significant reduction in the seven-day prevalence of self-reported diarrhea symptoms [Relative Risk (RR): 0.80 (95% CI: 0.51, 1.27)]. It is likely that a lower than expected prevalence in the control group reduced the power of the study to detect an effect of the intervention on diarrhea; however, the large impact of the intervention on drinking water contamination justifies expansion of the intervention to new regions. It is further concluded that future studies of potential health effects would be warranted, but in a larger study, or in an area with a higher burden of diarrheal illness.

In chapter 2, the effect of the intervention among households that "complied with the messages in the community intervention is explored, under the hypothesis that these households would have experienced greater reductions in drinking water contamination and diarrhea compared to the total population. The causal average complier effect (CACE) has been described for parallel arm trials, and is considered an estimate of treatment efficacy in effectiveness trials (i.e., a compliment to the ITT results reported in chapter 1). However, methods to estimate CACE in parallel arm trials require untestable assumptions. In this chapter a novel method to estimate the CACE parameter in a randomized stepped wedge trial is described. The primary insight from this chapter is that this method requires fewer assumptions than are necessary to estimate CACE in a parallel arm trial. To illustrate these methods, a definition of compliance is adopted, and effects of the intervention on household water quality and diarrhea are estimated among households that meet this definition. These stepped wedge CACE are compared to ITT results from the original trial and to CACE estimates from an instrumental variable (IV) estimator (a common method to estimate CACE in parallel arm trials). The stepped wedge CACE parameter estimated modest increases in reductions in drinking water contamination and diarrhea among complier households, compared to the ITT results from Chapter 1. In contrast, results from the IV estimator suggest a doubling in effect compared to the ITT estimates; however, it shown that violations of the assumptions necessary for IV estimation likely biased these results.

In the third and final chapter, the significance of measuring reductions in MPN EC (as a measure of water quality) in HWTS trials is considered. While most species of EC are not themselves pathogenic, they are excreted in large concentrations with human fecal material. Escherichia coli has therefore been a target indicator organism to test for the presence of fecal material and possible pathogen contamination in drinking water for over a century. For decades, in lieu of available methods to easily and cheaply test for EC in drinking water, there has been a reliance on surrogate indicators; the primary surrogate has been thermotolerant ("fecal") coliforms (FC). This chapter reports the results of a systematic review and meta-analysis that evaluates available evidence for a link between diarrhea and the presence of EC and FC in drinking water. The findings suggest that there is evidence to link diarrhea to EC (pooled RR: 1.54 [95% CI: 1.37, 1.74]) but not FC (pooled RR: 1.07 [95% CI: 0.79, 1.45]). It is concluded that this evidence supports the continued use of EC as an indicator organism to measure water quality in HWTS trials, and other field and research applications where an association with diarrhea is important; there is no evidence, however, to support the continued use of FC in these applications. The results from this review also serve to strengthen the findings and conclusion reported in chapter 1.

Reliance on traditional research methods in situations where innovative methodologies are superior can hinder inference. In global health research, where wasted resources can be equated with lives lost, non-adaptation of promising methods has dire consequences. Given the high cost of randomized controlled trials (RCTs), one major obstacle in these studies is the Type II error, in which an investigator fails to reject the null hypothesis when the null hypothesis is false. Unexpected disease transmission rates may lead studies to be underpowered and yield Type II errors. Furthermore, even if an RCT correctly rejects or fails to reject the null, data may not be fully utilized if investigators only analyze and report the intervention effect in the total study sample. Only focusing on quantifying the overall impact of the intervention of interest may neglect investigation into important observational relationships between covariates of interest or heterogeneity of the intervention effect in population subgroups, although these observational analyses and subgroups of interest should be pre-specified to maintain transparency and replicability. As participants were not randomized to receive exposures represented by covariate values, additional analytic rigor and inferential caution must be exercised in order to enable meaningful interpretation. My dissertation aims to explore three examples of potential methodological solutions to these challenges to inference in RCTs: a systematic review of the ring trial design, a targeted learning analysis of treatment heterogeneity, and an observational analysis of multiple biomarkers using trial data. Chapter 1. Spatiotemporal clustering: ring trials. For cluster RCTs in emergent and elimination disease settings, unpredictable spatiotemporal clustering often leads to imbalanced clusters, or even clusters with zero incident cases, which can reduce study power and limit inference. Ring trials, a trial design in which the units of randomization are responsively-defined clusters of individuals in social or physical proximity to an index case, may improve investigators’ ability to make inferences in these settings. Despite this potential utility, this RCT design remains under-examined and under-utilized. Investigators conducted a systematic review of the ring trial design to examine the existing applications of this study design as well as its benefits and drawbacks. We identified 26 ring trials, 15 cluster-randomized trials that used ring interventions, five trials that used ring recruitment and randomized within rings, and one individually-randomized trial that used a ring intervention. Ring trial designs require strong disease surveillance and contact tracing mechanisms, rapid intervention delivery systems, and a treatment with a strong post-exposure prophylactic effect. In these settings, ring trials can retain power despite unpredictable spatiotemporal clustering of the outcome of interest. Chapter 2. Heterogeneity in treatment effect: targeted learning analysis of treatment heterogeneity. Even if there is an effect of the intervention on the outcome among certain individuals, a study may fail to detect this relationship if the effect is heterogeneous in the study sample. Investigators can gain additional insight through analysis of the conditional average treatment effect, which is the treatment effect based on individual covariate status. Using data from the WASH Benefits study, which enrolled pregnant mothers and young children in rural Bangladesh, analysis of treatment heterogeneity can improve our understanding of child growth in low and middle-income countries. Despite the widespread use of water, sanitation, hygiene (WSH), nutrition (N), and combined (N+WSH) interventions, investigators have found mixed evidence regarding these interventions’ impact on child growth and development. Insufficient reduction of pathogens may explain WSH’s lack of impact, and environmental enteric dysfunction (EED), a condition of impaired intestinal permeability and inflammation, may modify the impact of WSH and N interventions on child growth. This study applied targeted machine learning methods to assess treatment heterogeneity of N+WSH, WSH, and N interventions on child growth by pathogen and EED biomarker status. We found that children with greater levels of myeloperoxidase, a gut inflammation biomarker associated with EED, and Campylobacter, a genus of bacteria that is associated with EED onset, had a greater effect of all treatments on growth. These results contribute to the body of literature characterizing individual predictors of N+WSH, WSH, and N intervention effectiveness as well as our understanding of EED. Chapter 3. Maximizing data utilization: observational analysis of high-dimensional data nested within a randomized controlled trial. Trials devote enormous resources to evaluating the effect of the randomized intervention in the study sample, but limiting analyses to only include this intervention may neglect the wealth of data that these RCTs can provide. In addition to analysis of the effect of the intervention, investigators can conduct observational analyses nested within an RCT, although these analyses will require additional methodological rigor in order to limit confounding and bias to enable meaningful inference. Data from the WASH Benefits study provide an opportunity to assess the relationship between stress neurobiology, an exposure that could not be ethically randomized, and child development. Stress has been implicated as a key pathway by which adverse circumstances can lead to developmental impairment, and prior studies have indicated a possible link between stress and subsequent development. This study evaluated the relationship between stress and development through an observational analysis nested within an RCT. We assessed physiologic measures of stress using measures of the hypothalamic-pituitary-adrenal (HPA) axis, the sympathetic-adrenal-medullary (SAM) system, and oxidative status. We constructed generalized additive models to compare development outcomes of children at the 75th and 25th percentiles of stress biomarker distributions while adjusting for potential confounders. We found that measures of HPA axis activity were associated with poor development outcomes. These observations support the use of HPA axis biomarkers, particularly cortisol and glucocorticoid receptor methylation, to indicate children who are at risk of poor developmental outcomes. This study explores and provides applied examples of RCT design and analysis methods that may improve efficiency in global epidemiologic research. This research serves to 1) improve our understanding of a neglected trial design and explore innovative methods of analyzing trial data; 2) identify which characteristics define amenability to N+WSH, WSH, and N interventions, providing insights that EED may be associated with treatment effectiveness; 3) evaluate the relationship between stress biomarkers and child development through observational analyses nested within an RCT, which supports the use of HPA axis biomarkers to indicate children at risk for poor developmental outcomes.

Oceans and human health is an emerging interdisciplinary field. The oceans modulate our climate and provide food resources, recreational opportunities, and esthetic enjoyment. Marine ecosystems are under threat due to population growth, anthropogenic pollutants, tourist activities, aquaculture production, and global shipping transportation. Additionally, harmful algal blooms (HABs), although naturally occurring, have increased dramatically worldwide. Microbial and chemical contamination of seawater and seafood has had a negative effect on human health. Preservation and restoration of the oceans is critical to the health of marine organisms, marine mammals, and humans.

This dissertation comprises three research chapters. The research question of interest is "what acute health risks, if any, are associated with exposure to seawater?" Chapter one presents a meta-analysis that reviewed acute human health effects associated with harmful algal blooms. Karenia

brevis, a microalgae, produces a suite of natural toxins known as brevetoxins. These brevetoxins were found to be associated with acute health symptoms of nasal congestion, eye irritation, and cough. Specifically, this meta-analysis revealed there was a 3.58 increase in relative risk of eye irritation [95% confidence intervals (CI), 2.00 - 6.42], 2.45 increase in relative risk of nasal congestion (95% CI 1.51 - 3.98), and 2.24 increase in relative risk of cough (95% CI 1.49 - 3.38) associated with exposure of aerosolized brevetoxins.

Chapter two and three examined the acute health effects among lifeguards and surfers. They were ideal participants for these studies due to their high levels of seawater exposure. In both studies upper respiratory symptoms were the most commonly reported acute health symptoms at 21% and 29% respectively. Findings from the surfer health study, in chapter two, demonstrated that surfers

who previously experienced acute health symptoms while surfing during a HAB were 1.63 times more likely to have upper respiratory symptoms than surfers who have not experienced any acute health symptoms during a HAB event (95% CI = 1.09, 2.44). Surfers with housemates with upper respiratory symptoms were 1.72 times more likely to have upper respiratory symptoms as compared to surfers with healthy housemates (95% CI = 1.22, 2.43). Surfers with a physician diagnosis of allergies were 1.41 times more likely to have upper respiratory symptoms in comparison to surfers without allergies (95% CI = 1.01, 1.97).

Chapter three describes the national cross-sectional study of ocean lifeguard. Lifeguards with a history of surfers ear were 2.95 times more likely to have upper respiratory symptoms than lifeguards who did not have a history of surfers ear (95% CI = 1.33, 6.54). Lifeguards who reported they have had acute health symptoms after exposure to a HAB were 1.5 times more likely to upper respiratory symptoms than lifeguards with no such history (95% CI = .72, 3.47).

Gastrointestinal symptoms were 2.49 times more likely among lifeguards with a history of acute health symptoms after HAB exposure than lifeguards who reported no history of acute health symptoms after exposure to HABs (95% CI = 0.76, 8.16). Lifeguards with a history of physician diagnosed anxiety were 1.69 times more likely to have acute stress symptoms after a rescue than lifeguards without a history of anxiety (95% CI = 1.31, 2.17). Lifeguards with a history of physician diagnosed asthma were 1.66 times more likely to experience acute stress symptoms after a rescue as compared to lifeguards without a history of asthma (95% CI = 1.30, 2.11). Additionally, lifeguards with acute health symptoms after exposure to a HAB were 1.42 times more likely to experience acute stress symptoms after a rescue than lifeguards without a history of acute health symptoms after a HAB (95% CI =1.08, 1.86).

In conclusion, evidence from these studies suggests that there is a positive association between HAB exposure and acute health symptoms in humans. Future studies are necessary to assess the human health risks of the different algal toxins produced during a HAB. Additionally, co-occurring bacteria and viruses, with exponential growth, during these HAB events need to be further investigated.

Diarrhea remains a significant cause of morbidity and mortality worldwide, especially for children and persons with compromised immune status. This dissertation considers two key subpopulations that are higher risk for gastrointestinal illness due to immunocompromise, persons living with Human Immunodeficiency Virus (HIV), and older adults over 55 years of age. Three investigations with different methodologies (population survey, meta-analysis, and case-control study) on different aspects of diarrhea burden, risk patterns and etiology in these two populations are presented.

The pathogenesis of diarrhea is complex and associated with multiple etiologies including bacterial, parasitic and viral pathogens, adverse events from medications, tumors and underlying immune status. The relative contributions of these risk factors to the overall burden of diarrheal illness is unknown and may vary by geographic location, immune status and access to anti-retroviral medications. Gastrointestinal adverse events are the most frequently cited reason for discontinuation for highly active antiretroviral therapy (HAART) and significantly influence patient adherence. As treatment decisions are increasingly made with regard to tolerability, probability of adherence, and patient quality of life, the burden of diarrhea associated with various HAART regimens may be critical in determining an individual's optimal treatment course. A systematic review to estimate the rates of diarrhea among HIV-infected individuals associated with specific HAART regimens was conducted showing protease inhibitor containing regimens to confer the highest rates of diarrhea. Diarrhea presents a persistent challenge to those living with HIV under anti-retroviral treatment and varies by regimen composition. Additionally, a prospective case control study of acute, chronic and asymptomatic diarrhea was conducted among an urban HIV+ clinic cohort to evaluate the relative risk of diarrhea due to medication and infectious pathogens.

Lastly, a random digit dial serial cross-sectional survey among 2163 adults over 55 years of age was conducted in Sonoma County, California from September 2001 to September 2005 to estimate the prevalence of gastrointestinal illness. The average monthly prevalence of gastrointestinal illness (vomiting or diarrhea) was 7.34% (6.23, 8.63), corresponding to an incidence rate of 0.99 (95% CI: 0.84, 1.48) episodes per person per year. Of those reporting gastrointestinal illness, 30.0% experienced vomiting, 23.4% sought medical care, and 10.8% took antibiotics.

This dissertation presents a quasi-experimental study design to evaluate non-random\-ized, pre-existing community interventions not originally designed to gather rigorous data about sustainability and impact. The core components of the design include selection of a control group using propensity score matching with pre-intervention (baseline) secondary data and post-intervention follow-up in the field. The main advantages of the design include measurement of interventions implemented under actual field conditions (independent of scientific research), as well as the design's ability to gather information about the long term impacts and sustainability of interventions without years of costly prospective follow-up. Studies of non-randomized, pre-exisiting interventions must address threats to validity, principal among them: unmeasured confounding and informative censoring.

I outline the main strengths and weaknesses of the study design using simulation and empirical examples. I also apply the design in two sustainability field studies: a 3-year household water treatment and hygiene promotion intervention in rural Guatemala and a 5-year community led total sanitation, water supply and hygiene education intervention in rural India. In both studies, the design leads to samples of intervention and control groups with highly comparable baseline characteristics.

A principal finding of both field studies is poor initial impact and sustainability of the behavioral components of the interventions. In Guatemala, I find a small, five percentage point increase (8.7% vs. 3.3%) in the proportion of households that treat their water six months after the promotion intervention, but no differences in hygiene knowledge or practice, and no detectable differences in child health based on acute illness or growth. In India, I find a large, 33 percentage point increase (48% vs. 15%) in private toilet construction as a result of the intervention, but open defecation persists in 40% of households with functional private toilets. In India, diarrhea is rare in both intervention and control communities (1.8% over 14,259 child weeks), but most children show growth faltering by international standards (mean height-for-age Z-score: -1.98). Despite no signifiant differences in health between children living in intervention and control villages, I observe important non-health benefits: private toilets increase privacy and safety during defecation for women and girls by 28 percentage points (81% vs. 53%), and private water taps reduce water collection time by a median of 25 minutes per day relative to public taps (50 vs. 75 minutes). I also find that hardware improvements are highly sustainable up to five years after implementation with more than 94% of private toilets and 96% of private water taps in use during repeated visits over one year.

Studies of non-randomized, pre-existing interventions are a rapid, low-cost alternative to prospective intervention studies for evaluating intervention sustainability. The study design and methodology developed in this dissertation are applicable to evaluating a broad range of pre-existing, community interventions beyond the water, sanitation and hygiene sector.