Project Management IV Research Paper

Systems Science Approaches for Global Environmental Health Research: Enhancing Intervention Design and Implementation for Household Air Pollution (HAP) and Water, Sanitation, and Hygiene (WASH) Programs Joshua Rosenthal,1 Raphael E. Arku,2 Jill Baumgartner,3 Joe Brown,4 Thomas Clasen,5 Joseph N.S. Eisenberg,6 Peter Hovmand,7 Pamela Jagger,8 Douglas A. Luke,9 Ashlinn Quinn,1 and Gautam N. Yadama10 1Division of Epidemiology and Population Studies, Fogarty International Center, National Institutes of Health, Department of Health and Human Services, Washington, DC, USA 2Department of Environmental Health Sciences, Department of Environmental Health Sciences, University of Massachusetts, Amherst, Amherst, Massachusetts, USA 3Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Montreal, Quebec, Canada 4Department of Environmental Engineering, Georgia Institute of Technology, Atlanta, Georgia, USA 5Department of Environmental Health, Emory University, Atlanta, Georgia, USA 6Department of Epidemiology, University of Michigan, Ann Arbor, Michigan, USA 7Center for Community Health Integration, School of Medicine, Case Western Reserve University, Cleveland, Ohio, USA 8School for Environment and Sustainability, University of Michigan, Ann Arbor, Michigan, USA 9Center for Public Health Systems Science, Brown School, Washington University in St. Louis, St. Louis, Missouri, USA 10School of Social Work, Boston College, Boston, Massachusetts, USA

BACKGROUND: Two of the most important causes of global disease fall in the realm of environmental health: household air pollution (HAP) and poor water, sanitation, and hygiene (WASH) conditions. Interventions, such as clean cookstoves, household water treatment, and improved sanitation facilities, have great potential to yield reductions in disease burden. However, in recent trials and implementation efforts, interventions to improve HAP and WASH conditions have shown few of the desired health gains, raising fundamental questions about current approaches.

OBJECTIVES: We describe how the failure to consider the complex systems that characterize diverse real-world conditions may doom promising new approaches prematurely. We provide examples of the application of systems approaches, including system dynamics, network analysis, and agent- based modeling, to the global environmental health priorities of HAP and WASH research and programs. Finally, we offer suggestions on how to approach systems science.

METHODS: Systems science applied to environmental health can address major challenges by a) enhancing understanding of existing system structures and behaviors that accelerate or impede aims; b) developing understanding and agreement on a problem among stakeholders; and c) guiding interven- tion and policy formulation. When employed in participatory processes that engage study populations, policy makers, and implementers, systems sci- ence helps ensure that research is responsive to local priorities and reflect real-world conditions. Systems approaches also help interpret unexpected outcomes by revealing emergent properties of the system due to interactions among variables, yielding complex behaviors and sometimes counterin- tuitive results. DISCUSSION: Systems science offers powerful and underused tools to accelerate our ability to identify barriers and facilitators to success in environ- mental health interventions. This approach is especially useful in the context of implementation research because it explicitly accounts for the interac- tion of processes occurring at multiple scales, across social and environmental dimensions, with a particular emphasis on linkages and feedback among these processes. https://doi.org/10.1289/EHP7010

Introduction Two environmental risk factors underlie the substantial global dis- ease burden affecting young children and vulnerable populations around the world: household air pollution (HAP) exposure and deficiencies in water, sanitation, and hygiene (WASH). Both HAP and WASH face challenges to underlying assumptions regarding the viability of known interventions to achieve intended and sus- tained reductions in disease burden. Observational studies of HAP have consistently shown strong associations with adverse health conditions throughout the life course, including pneumonia, hyper- tension, and adverse birth outcomes (Thakur et al. 2018), as well as

all-cause mortality (Hystad et al. 2019). Poor WASH is also associ- ated with a range of adverse health outcomes, including enteric vi- ral, bacterial, and protozoan pathogen infections and diarrheal disease, schistosomiasis, and soil-transmitted helminth infections (Clasen et al. 2015; GBD 2016 Risk Factors Collaborators 2017). Technological solutions designed, investigated, and evaluated in laboratory settings, including clean cookstoves and water filters, offer promising solutions, but seldom deliver expected results when implemented in real-world settings (Clasen and Smith 2019; Sesan et al. 2018). Recently completed randomized controlled tri- als (RCTs) have returned disappointing results for interventions that are already in widespread use, raising questions regarding ba- sic assumptions of efficacy, effectiveness, potential for scale-up, and technology and policy options to address these priority envi- ronmental health concerns (Luby et al. 2018; Mortimer et al. 2017; Null et al. 2018).

In the present paper we consider HAP and WASH together, despite their different exposures and etiologies, because of four important commonalities. First, the greatest disease burden for both is concentrated in the poorest populations of low- and middle-income countries (LMICs), especially in rural and peri- urban areas. Second, although both HAP and WASH are subject to important policy-based social service and infrastructure chal- lenges, they are commonly addressed with interventions at the household and community levels (e.g., cleaner cookstoves, latrines, and water filtration devices). Third, these technology-

Address correspondence to Ashlinn Quinn, Division of Epidemiology and Population Studies, Fogarty International Center, National Institutes of Health, Building 16, Room 306, Washington, DC. Email: [email protected] The authors declare they have no actual or potential competing financial

interests. Received 9 March 2020; Revised 13 September 2020; Accepted 15

September 2020; Published 9 October 2020. Note to readers with disabilities: EHP strives to ensure that all journal

content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact [email protected]. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

Environmental Health Perspectives 105001-1 128(10) October 2020

A Section 508–conformant HTML version of this article is available at https://doi.org/10.1289/EHP7010.Commentary

focused interventions have both struggled to achieve the high lev- els of reach, adoption, and adherence necessary to reduce expo- sure and improve health. Finally, the emissions/effluents from one home have wider ecological impacts or spillover effects, sug- gesting that high levels of adoption are required in order to achieve favorable health outcomes at individual, household, and community scales (Sesan et al. 2018).

Here we argue that a systems science approach offers powerful, underused tools to develop guidance for intervention design and implementation in HAP and WASH. Systems science is the appli- cation of scientific methods to the understanding of complex sys- tems. It allows for consideration of problems and solutions across multiple dimensions, at various scales, and is dynamic in scope. A focus on systems challenges reductionist explanations that seek to explain behavior in terms of properties of constituent parts (Bunge 1997; Capra and Luisi 2014) and considers emergent properties that cannot be understood by reducing behavior to constituent terms. Challenges to implementing and sustaining evidence-based interventions, especially in complex socio-environmental arenas, are often systemic in nature; and thus, systems science methods may be especially appropriate (Luke et al. 2018). Systems science is an important complement to the randomized controlled trial approach that has dominated both HAP and WASH research in recent years. Unexpected and often disappointing trial results sug- gest that a different approach to studying implementation processes and outcomes is needed.

In public health, systems science methods have been applied to a wide range of issues including tobacco policy and regulation, cancer prevention, infectious disease, obesity and diabetes pre- vention, intentional injury, health systems, drug abuse/addiction, and others (Luke and Stamatakis 2012). To date, their application to environmental health challenges is limited (Currie et al. 2018).

Where HAP Programs Have Underdelivered Since the 1950s, many efforts at scales ranging from the local to the national have sought to improve household access to clean cook- stoves (Manibog 1984). The vast majority of efforts prior to 2000 focused on decreasing consumption of solid fuels to reduce per- ceived pressure on forest resources, and subsequently, global warming contributions from the same fuels accelerated interest in the field (Rosenthal 2015; K Smith et al. 2000). With recognition of the enormous health burden of HAP over the past decade, the household energy sector has turned its focus to promoting meas- ures with the greatest potential for mitigating this adverse health burden. Focusing on health goals has raised expectations for clean cooking technologies and their implementation considerably, but supposedly clean cookstove interventions have not always deliv- ered desired improvements in health. Despite air pollution reduc- tions of 50% or more in the most successful programs, post- intervention levels of fine particulate matter (PM ≤ 2:5 lm in aero- dynamic diameter) generally remained 2–40 times higher than the World Health Organization interim air quality guideline of 35 lg=m3 average annual concentration in homes with the new stoves (Quansah et al. 2017).

Randomized trials with cleaner stoves or fuels have also returned mixed results (Alexander et al. 2017; Mortimer et al. 2017; Smith et al. 2011; Thompson et al. 2011; Tielsch et al. 2016). Some of the shortfalls are related to inherent limitations of the technologies. A well-understood limitation of previous house- hold energy programs was that air pollution emissions from improved cookstoves in community settings greatly exceeded what was expected based on laboratory testing (Coffey et al. 2017; Eilenberg et al. 2018). Clean fuel interventions with lique- fied petroleum gas (LPG), electric induction, ethanol, biogas, and pellet-fueled gasifier stoves have consistently outperformed

traditional and improved solid-fuel stoves in both laboratory and field studies (Champion and Grieshop 2019; Jagger et al. 2017; Wathore et al. 2017) and are the most likely to achieve the desired air quality and health benefits.

Beyond these technology lessons, several implementation issues have emerged, leading HAP interventions to fall short of expectations. These include low initial adoption rates for clean fuels and stoves (Lewis and Pattanayak 2012; Mobarak et al. 2012; Troncoso and Soares da Silva 2017), concomitant use of polluting stoves (stove stacking) among adopters (Masera et al. 2000; Puzzolo et al. 2016), and issues with supply chains and cost for both stoves and fuels (Jagger and Das 2018; Puzzolo et al. 2019).

Finally, because most relevant LMIC homes are relatively open to outdoor air, ambient air quality sets the floor for house- hold exposure levels. Evidence is growing that household cook- ing can be a major contributor to poor local outdoor air quality (Butt et al. 2016; Chowdhury et al. 2019a; Snider et al. 2018) and its associated health effects (Conibear et al. 2018). Still poorly understood are the degree and conditions under which greater density and coverage of clean cooking interventions at scale can reliably reduce background ambient pollution (although see Chowdhury et al. 2019b for estimates for India).

The accumulating body of evidence from experiments, obser- vational studies, and models over the past two decades has led to a consensus in the health science community that achieving major reductions in levels of HAP will require at least three important shifts in the way HAP programs and policies are conceived. First, the technology, including fuels, must emit little to no particulate matter (Wathore et al. 2017; World Health Organization 2014). Second, individual households must rely on clean fuels and stoves for the overwhelming majority of their energy needs, with only very occasional combustion of polluting fuels for any purpose (Johnson and Chiang 2015). Third, local ambient conditions need also be relatively clean, probably requiring high, effective cover- age of clean cooking at the community and larger levels (see, e.g., Weaver et al. 2019).

Where WASH Programs Have Underdelivered Through the first half of the last century, higher-income settings implemented large-scale infrastructural solutions such as sewer- age, centralized water and wastewater treatment, and piped water to homes, achieving significant improvements in health (Cutler and Miller 2005). In contrast, in rural LMIC settings, such invest- ments were (and continue to be) insufficient, especially in sewer- age. As such, improvements in water and sanitation were traditionally focused on expanding access to improved, but low- cost, technologies (protected wells and springs, pit latrines) that were suitable for scaling up in rural settings. Other household- scale approaches, including behavior change programs to promote handwashing and promotion of water filtration devices to provide clean water at the point of use, were steadily advanced by develop- ment agencies and local nongovernmental organizations (Dangour et al. 2013; Darvesh et al. 2017). These approaches were believed to be cost-effective solutions because they did not require substan- tial expenditures on infrastructure. With limited resources, lower- income countries continued to implement these low-cost approaches to WASH, with some affirming evidence of efficacy in randomized and nonrandomized trials and systematic reviews through the early 2000s (Esrey et al. 1985, 1991).

In more recent years, a number of rigorous randomized trials of low-cost interventions in WASH yielded findings of little or no effect for most outcomes of interest, including diarrhea, linear growth, stunting, and helminth infections (Clasen et al. 2015). Examples include large sanitation trials in India (Clasen et al. 2014; Patil et al. 2014) and trials of multiple WASH and nutrition

Environmental Health Perspectives 105001-2 128(10) October 2020

interventions in Bangladesh (Luby et al. 2018), Kenya (Null et al. 2018), and Zimbabwe (Humphrey et al. 2019). Except for the Bangladesh trial, which reported protective effects from the sanita- tion and handwashing interventions, none of these trials showed that WASH interventions were protective against diarrhea or stunt- ing. Some of these studies were effectiveness trials of program- matically delivered interventions in which investigators explored the effects of improved water supplies, household water treatment, improved sanitation, and handwashing with soap (Clasen et al. 2014; Dangour et al. 2013; Patil et al. 2014; Sinharoy et al. 2017). In those cases, the investigators generally found poor intervention quality, coverage, uptake, or use—failures that could explain the lack of protective health effects because of incomplete interruption of exposure to human enteric pathogens.

Correct, consistent, and sustained adoption of healthy WASH practices, including handwashing with soap, was also shown to present important behavior change barriers (Freeman et al. 2014; Martin et al. 2018). Increasingly, poor compliance was shown to be fundamentally limiting for most WASH interventions (Brown and Clasen 2012). However, evenin efficacy trials in which the interven- tions achieved high levels of compliance (Humphrey et al. 2019; Luby et al. 2018; Null et al. 2018), results were disappointing.

These recent studies highlight both our limited understanding of which modes of diarrheal disease transmission are dominant and how these might vary by environmental context. Some pathways, such as food and animals, are not effectively addressed with basic WASH infrastructure. Furthermore, we are learning that many water supply systems periodically deliver unsafe water, often due to intermittent operation (Bivins et al. 2017). Studies of low-cost approaches to drinking water revealed that, even if safe at the point of collection, water was often contaminated with fecal pathogens during storage in the home (Levy et al. 2008; Wright et al. 2004). In some cases, exposures may also arise from zoonotic agents not addressed by conventional sanitation interventions (Berendes et al. 2018; Daniels et al. 2016; Penakalapati et al. 2017).

WASH is considerably more mature than the HAP field and offers experience with more types of interventions. The HAP community can learn a great deal from WASH experiences (Clasen and Smith 2019; Sesan et al. 2018). However, results from these recent WASH trials also suggest that major challenges underlie the successful implementation of environmental health interventions in low-income settings. Despite a much deeper evi- dence base to draw upon, WASH continues to be a major public health concern for most low-income countries.

Diagnosing the Problem When outcome measures fall short of expectations, commentators frequently focus on the inadequacies of the technology. As described in the preceding two sections, there are numerous exam- ples of projects that have underdelivered because of shortcomings of the technology. Many improved biomass stoves do not deliver the air pollution reductions required to greatly minimize adverse health impacts, although they do reduce fuel use and may provide some reduction in air pollution (Jagger et al. 2017). Improved water supplies do not necessarily provide water that is safe at the point of collection (Bain et al. 2014), in part because most house- hold water treatment options fail to cover the full array of water- borne pathogens or fail to keep water safe after it is treated (Shaheed et al. 2014). Even if water is effectively treated, the bene- fits of household-level water quality interventions are realized only if they are used exclusively (Enger et al. 2013).

Furthermore, focusing primarily on the household level may not always be the appropriate scale of implementation and evalua- tion. Household–to–ambient air pollution interactions (Huang et al. 2015), and herd immunity to diarrheal disease demonstrate

substantial indirect effects at the community level (Fuller and Eisenberg 2016). Associated nonlinear exposure–response rela- tionships (Burnett et al. 2014; Jung et al. 2017) mean that density and coverage of the intervention may be critical. In some cases, this becomes evident during efficacy-stage testing under relatively controlled conditions. However, it is more commonly apparent when the interventions are deployed in communities at scale.

Importantly, even effective, scaled interventions may fall short when attempted in a new environment. LPG cooking interventions may be very effective where air pollution exposure is dominated by indoor sources, but in highly polluted urban environments, they may not reduce fine particulate and other air pollutants sufficiently to reach the threshold levels thought necessary to achieve health gains (Liu et al. 2019), but see Chowdhury et al. (2019b) for an esti- mate of ambient effects of a large-scale clean fuel intervention in India. A WASH intervention that is effective in one setting may be wholly ineffective in another where there is a different dominant pathogen or transmission pathway (Eisenberg et al. 2007). Relevant social variables include individual, community, or insti- tutional behaviors. This may also reflect larger-scale systems prob- lems related to supply chains, availability of substitutes (e.g., freely available biomass fuels), price instability, and regulatory challenges that these technologies confront in real-world settings, especially in LMICs (Thomas 2016).

Finally, sustained effectiveness of most public health interven- tions is a major challenge. In low resource settings, program leaders often find themselves trapped in cycles of iterative attempts to main- tain or improve outcomes in the original communities. Decline in adherence to proven use protocols is common without continual reinforcement of behavior change messages over time (Brown and Clasen 2012) or adequate supply or servicing of new technologies required to achieve exposure reductions (Jagger and Das 2018). For example, wide-scale efforts to encourage community-led total sani- tation in rural low-income settings have suffered from backsliding (slippage) to open defecation (Venkataramanan et al. 2018). Importantly, even if an intervention is effective initially, exogenous conditions may change over time. For example, economic factors such as price increases or supply chain failures make clean fuels (Puzzolo et al. 2019) and water treatment devices (Schmidt and Cairncross 2009) and their maintenance less accessible. Migration to and from communities may fundamentally affect social or envi- ronmental patterns necessary to maintain the health benefits of the intervention (Eisenberg et al. 2006). Climate change brings unfore- seen flooding, potentially distributing contaminants across the household and the village (Mertens et al. 2019). For each of these likely developments, the intervention no longer works in the way it did under prior conditions.

RCTs have enormous power to provide internal validity (i.e., high causal inference/attribution) and insights into adoption and adherence in tightly controlled experimental settings. RCTs also have a critical role in defining the scalable unit of an intervention (e.g., a combined technology and behavior change package). However, external validity of trial results can be often limited by contextual factors that underlie success or shortfalls in programs, including environmental and social differences of the sort described in the preceding paragraphs. Given the recent history of failures of larger effectiveness and translational trials for these environmental health interventions, some may conclude that they just are not effec- tive, or alternatively, that they were inadequately deployed.

However, the importance of both endogenous and contex- tual factors as critical determinants of sustainability and scal- ability suggests that different analytical approaches are needed to support the development of effectiveness and translational studies whether these are conducted through RCTs or other approaches.

Environmental Health Perspectives 105001-3 128(10) October 2020

Cycles in Environmental Health Intervention Policies

At the policy level, we often see a pattern of initial success, excitement, and hype around a potentially effective interven- tion, followed by real-world failures or shortfalls and subse- quent loss of political and financial support to the next big thing. Even potentially important interventions are often rolled out before there is sufficient evidence to guide them, and a wave of premature enthusiasm and associated financing effec- tively set them up for failure (Little et al. 2012). Clean cooking and water filtration interventions illustrate this pattern ele- gantly. Gartner’s Hype Cycle (Fenn and Raskino 2008) is a widely employed conceptual tool from the technology business community to describe and predict the path of development to establishment of a new technology in industry. In Figure 1 we map the history of HAP and WASH interventions onto Gartner’s Hype Cycle to illustrate how a similar pattern has unfolded in the environmental health intervention science and policy communities.

The clean cooking movement received new impetus and rapid increases in investment beginning in 2010 when the HAP disease burden was recognized, and improved cookstoves were perceived by the policy community to offer win-win-win solutions for health, climate, and women’s empowerment (Bhattacharyya and Light 2010). As the scientific and development communities gradually uncovered significant shortcomings in cookstove programs enthu- siasm began to wane and international funding for this work declined precipitously (Figure 1A). Today, some believe that we may have lost important momentum because of these incautious efforts (Ezzati and Baumgartner 2017). Ironically, this has

occurred as the field is maturing scientifically and greater under- standing of what benefits are achievable as well as a greatly improved understanding of the socioeconomic and environmental conditions for successful and sustainable interventions are accru- ing rapidly.

Parallels in WASH include household water filtration technol- ogy (Figure 1B) and community-led total sanitation (CLTS), both potentially transformative interventions initially supported by substantial research and programmatic efforts, only to have rigor- ous trial results reveal limitations to the approaches (Brown et al. 2019). Enthusiasm for household-level chlorination has been tempered by challenges in supply chains, low adoption, and new evidence on the prevalence of chlorine-resistant diarrheagenic agents such as Cryptosporidium (Kotloff et al. 2019). The fact that interventions may not live up to the initial hype should not mask the real benefits such solutions can deliver: for example, household water treatment technologies have shortcomings, but systematic reviews of the evidence base reveal the approach’s potential as an interim solution that can improve water quality and reduce diarrheal disease in vulnerable populations (Clasen et al. 2015).

Whether intervention shortfalls are due to efficacy, effective- ness, implementation challenges, or changes to underlying ena- bling conditions, failures may result in enormous social and financial costs. Before the Clean Cooking Alliance focused their efforts toward clean fuels, approximately 40 million homes received improved stoves (ISO tiers 1–2) (Global Alliance for Clean Cookstoves 2017) for which there is little to no evidence of health benefits (although some reductions in greenhouse gas emissions, quantity of fuel used, and time spent collecting fuels

Technology trigger

Peak of inflated expecta�ons

Trough of disillusionment

Plateau of produc�vity

Slope of enlightenment

time

expectations

Gartner’s Hype Cycle HAP – Clean Cooking History

Stove R&D

Epidemiologists raise profile of HAP health burden

Companies and NGOs develop improved cookstoves (ICS)

Third-genera�on products and interven�ons

High-growth adop�on phase begins: 20 to 30 percent of the poten�al audience has adopted the innova�on

Cookstove programs proliferate

Methodologies and best prac�ces developing

Less than 5 percent of the poten�al audience has adopted fully

Cri�cal reviews and press proliferate

Evidence that ICS aren’t clean enoughMass media

hype begins

Cookstove Alliance forms

Exploring cleaner technologies and more sophis�cated approaches

Be�er understand role of ambient condi�ons

Discover that most people don’t use ICS exclusively

WHO issues Indoor Air Quality Guidelines

A

Figure 1. (A) Household air pollution (HAP)—clean cookstove history mapped onto Gartner’s Hype Cycle (adapted from Fenn and Raskino 2008). (B) Water, sanitation, and hygiene (WASH)—point-of-use water treatment history mapped onto Gartner’s Hype Cycle (adapted from Fenn and Raskino 2008). Note: HWTS, household water treatment and safe storage; NGO, nongovernmental organization; R&D, research and development; WHO, World Health Organization.

Environmental Health Perspectives 105001-4 128(10) October 2020

are possible). Worse are projects that have led to large negative unintended consequences, usually due to incomplete knowledge about the context. A tragic example is the arsenic poisoning epi- demic in Bangladesh that resulted from a decades-long program to reduce diarrheal disease by installing tube wells into naturally contaminated soil strata (AH Smith et al. 2000). A consequence is that today populations in Bangladesh are still grappling with diarrheal disease from exposure to contaminated surface water, simultaneous with poisoning from exposure to arsenic from tube wells (Yunus et al. 2016) and social conflict regarding access to safe water (Sultana 2011).

What Implementation Science Offers Implementation science (IS) is a relatively new field that provides a set of theoretical, analytical, and experimental methods to under- stand the processes that make interventions successful and sustain- able in service delivery programs at scale and over time (Colditz and Emmons 2019; Glasgow et al. 2004; Madon et al. 2007; Rosenthal et al. 2017; Yamey 2011). The field draws on conceptual frame- works that focus analysis on those types of organizational and com- munity processes that are required for broad and sustained uptake of new evidence-based programs and policies (Damschroder et al. 2009; Tabak et al. 2012). Increasingly, implementation science frameworks are linked to RCTs that aim to evaluate hypotheses regarding the importance of specific approaches to, for example, individual or household behavior change, or adoption of an inter- vention by institutions in a given setting (Curran et al. 2012).

The ability to understand the aspects of programs that enable successful transfer from one setting to another is a key compo- nent of successful implementation at scale. Because multisite RCTs are expensive, programs are sometimes launched based on efficacy studies or limited range effectiveness trials (Madon et al.

2007) without examining implementation variables such as context-specific adoption or maintenance needs that may be criti- cal to the success of the intervention. The implicit assumption in these cases is that future projects can invest in adapting effective interventions to new conditions elsewhere in the world. Implementation research uncovers factors that may be influential to success by offering experimental and analytical frameworks to systematically consider and evaluate individual and institutional behaviors and environmental variables that are often obscured or assumed to be constant (Glasgow et al. 1999; Rosenthal et al. 2017).

Systems Science Expands Implementation Science Approaches Systems science offers powerful, underused tools to develop les- sons for intervention design and implementation in both HAP and WASH. Broadly speaking, systems science is the application of scientific methods to the understanding of complex systems (Galea et al. 2010). Systems science has been embraced by some in the implementation science community because of its utility to describe, analyze, and simulate complex systems that defy tradi- tional methods (Burke et al. 2015; Hammond and Dubé 2012). Importantly, systems science does not make an a priori commit- ment that all phenomena are best understood in terms of systems. Rather, it provides tools for consideration of phenomena that may not be amenable to analyses focusing on decomposed sys- tem elements and their properties. Challenges to implementing and sustaining evidence-based interventions, are often systemic in nature and, thus, systems science methods may offer much promise (Luke et al. 2018).

System science methods include a variety of tools that gener- ally fall under the following categories: network analysis (NA),

Successful use of HWTS in controlling cholera outbreaks

Early trials and systema�c reviews sugges�ng drama�c reduc�ons of diarrheal disease

Rapid development of new technologies and approaches by public and private sectors

Reduced donor investments in HWTS strategy

Studies show large declines in use over �me

Refinement of HWTS roles in specific se�ngs, including in disaster response

Focus on user behaviors and the technology/user interface

Re-framing use from “silver bullet” to one strategy in WASH tool box

Con�nued itera�on on technology design and delivery

Research on sustainability and mul�- household approaches that reduce need for behavior change

Release of WHO tes�ng data

Evidence of poor adherence to water quality interven�ons

Cri�cism of unblinded trials with subjec�ve outcome measures

Evidence of water quality improvements in reducing diarrheal disease

Large-scale rollout of HWTS at scale in mul�ple countries

Technology trigger

Peak of inflated expecta�ons

Trough of disillusionment

Plateau of produc�vity

Slope of enlightenment

time

expectations WHO establishes performance guidelines for HWTS

Gartner’s Hype Cycle WASH – Water Filtration History B

Figure 1. (Continued.)

Environmental Health Perspectives 105001-5 128(10) October 2020

system dynamics (SD), agent-based modeling (ABM), and in some cases Geographic Information Systems (GIS). Some of these methods are more descriptive and oriented toward statistical analysis and visualization of data (e.g., network analysis, geo- graphic information systems), whereas others are more focused on theory development and computational modeling (e.g., system dynamics, agent-based modeling) (Table 1). The use of GIS is al- ready well developed in the environmental health community (see, e.g., Nuckols et al. 2004; Peng et al. 2018). Here, we focus primarily on opportunities for advancing HAP and WASH with three modeling approaches, network analysis, system dynamics, and agent-based modeling. In Table 1 and the following para- graphs we provide a brief description of each of these methods and references for more detailed information.

Network analysis focuses on the relationships among sets of actors. The actors can be any type of entity that can have a rela- tionship or tie with others: point sources, persons, animals, organ- izations, countries, websites, documents, and even genes. These networks provide information on social structure that can play an important role in either promoting or mitigating disease proc- esses. Almost all NA makes use of one or more of three different analytic modes: network visualization, network description, and statistical modeling of networks.

System dynamics (SD) is based on the idea that system behavior (e.g., the frequency of exposures over continuous time) results from the interplay of a set of feedback mechanisms or loops relating accu- mulations and their corresponding rates of change or flows. Models of the feedback system can be described informally using a series of causal diagrams or more formally as a system of nonlinear ordinary differential equations that can be simulated on a computer. System behavior is then explained in terms of an explicit set of balancing and reinforcing feedback loops, whether that is at an aggregate, indi- vidual, or multilevel system (Richardson 2020; Sterman 2018). The visual conventions of SD (e.g., casual loop diagrams, stock and flow diagrams) have evolved into a set of participatory methods for involving communities and other stakeholders in the process of con- ceptualizing, formulating, and analyzing the results of models called group model building (e.g., Hovmand 2014; Richardson and Andersen 1995; Vennix 1996).

Agent-based modeling (ABM) uses computer simulation to study complex systems from the ground up, by analyzing how

individual elements of a system (agents) behave as a function of individual properties, their environment, and their interactions with each other. Through these behaviors, emergent properties of the overall system are revealed. Compared with SD, this results in a form of decentralized modeling where there is no formalized defini- tion of global system behavior; that is, feedback mechanisms are implicit (vs. explicit in SD) and emerge through the interaction of agents within their environment.

Transparency and replicability remain cornerstones of science, and systems science tools are no exception (Barton et al. 2020). In principle, computational models used in systems science (e.g., sys- tem dynamics and agent-based modeling) offer an added level of transparency relative to more traditional statistical approaches by making assumptions fully explicit (e.g., as a set of differential equa- tions or programming code) that can be independently explored and modified to test the implications of assumptions, measurement errors, research designs, and so on through sensitivity analysis. This has the advantage over traditional statistical approaches in that one can explore the robustness of a policy to system states beyond what has been historically observed or collected as part of an experiment— something that is highly relevant when we consider structural changes in environmental health triggered by global trends such as climate change, pandemics, and forced displacement of populations due to conflict and environmental disasters. However, this potential to be more transparent is often lost when computational models become overly complicated, lack adequate documentation, or require computational resources with limited access (Meadows and Robinson 1985; Pilkey and Pilkey-Jarvis 2007). A number of stand- ards have, therefore, emerged for reporting guidelines (e.g., Caro et al. 2012; Rahmandad and Sterman 2012).

Although applications of systems approaches are growing in other health implementation arenas, to date, systems applications in environmental health have been limited. Eisenberg et al. (2012) reviewed the history of diarrheal disease research and identified how systems approaches would be helpful in understanding interde- pendencies across multiple enteric disease transmission pathways. Zelner et al. (2012) used network analysis to understand connected- ness between and within communities in relation to reported diarrheal events and found that high connectedness between com- munities enhances risk of disease, as expected. However, high con- nectedness within a community reduces risk of diarrheal disease.

Table 1. Three systems science tools.

Tool Focus Key strengths Source data Key references

Network analysis (NA) Relationships between actors Visualization; identification of structure in social systems; can be empirical or model based

Surveys, observations, administra- tive data (e.g., membership ros- ters, emails), social media data

Luke and Harris 2007; Valente 2010, 2012

System dynamics (SD) Dynamic behavior generated by an explicit set of feed- back mechanisms over time

Identifying endogenous sour- ces of dynamic behavior in a set of feedback loops; ability to identify key lever- age points for system interventions

Time series to establish empirical basis for reference modes; em- pirical research results that pro- vide estimates for model parameters and initial condi- tions; key informant interviews, direct observation, group model building, expert panels, and grounded theory approaches to qualitative data analysis for structural relationships

Rahmandad et al. 2015; Sterman 2018

Agent-based modeling (ABM)

Emergent patterns from inter- action of actors with struc- tured exposures

Ability to examine interaction of individual or groups of actors with each other and with their social and physi- cal environments; can deal with actor and environmen- tal heterogeneity

ABMs can take advantage of all of the data sources mentioned for NA and SD. For example, theo- ries can be used to design agent rules; empirical data can be used to characterize agents and their physical/social environments

Epstein 2007

Environmental Health Perspectives 105001-6 128(10) October 2020

More recently, scientists applied systems approaches to understand user behavior in relation to interventions. For example, Kumar et al. (2017) have examined behavioral dynamics at the community scale to understand the process of adopting of LPG cooking, and Chalise et al. (2018) analyzed abandonment of biogas digesters and their stoves.

Systems science methods applied to environmental health can be useful for:

• Enhancing understanding of existing system structures and behaviors—and the types of feedback mechanisms causing those behaviors—that can accelerate or impede environmen- tal health aims, including potential unintended consequences.

• Developing understanding and agreement on a problem among stakeholders, which is particularly important for WASH and HAP because of their dependence on numerous stakeholders at the household, community, and policy maker levels. Group model building, for example, can help bring the diversity of choices and associated tradeoffs among relevant stakeholders to the surface, and can facilitate decision-making.

• Guiding intervention and policy formulation, where simula- tions can be particularly useful to estimate the effects of interventions before attempting them at scale in populations, potentially sidestepping some of the significant economic and social costs of testing the interventions in situ. For example, Mellor et al. (2014) simulated a water filter inter- vention to predict sustainability of effectiveness in reducing diarrheal incidence. If systems approaches are developed in an iterative participa-

tory process that engages study populations, policy makers, and implementers, they help ensure that research is responsive to local priorities and reflects real-world conditions, making research a more co-creative and less an extractive process (Dilling and Lemos 2011; Eisenberg et al. 2012; Israel et al. 2005; Mauser et al. 2013).

When a model fails to generate the expected outcomes, this can be due to faulty data or model structure. Faulty data are due to measurement error, either systematic or random. Faulty model structure is due to building a model incorrectly (discovered through verification testing) or building a model of a theory cor- rectly, but based on a theory that is faulty (discovered through validation testing). Weak measurement models, verification test- ing, or validation testing can contribute to a false confidence in a model of a system. A strong program of validating measurements and verifying that a model has been built correctly can, however, falsify a theory. In doing so, a model can typically be advanced that can offer a stronger alternative explanation for emergent behavior of a system. In either case (whether discovering and cor- recting measurement errors and specification errors or advancing a stronger alternative explanation), the process is consistent with a progressive program of scientific research (Lakatos 1970).

Some Suggested Applications of Systems Science to HAP and WASH Problems An important goal for the application of systems science in the fields of HAP and WASH is to accelerate learning and flatten the Hype Cycle (Figure 2) to achieve steadier gains in health out- comes. This might be characterized as accelerating the pace of innovation, partly through failing fast or quickly identifying strat- egies that may not result in changes in exposure. Using systems approaches in the early design and planning stages of a WASH or HAP interventional program can help scientists and policy- makers think more clearly about the larger socioeconomic and environmental context in which disease occurs and about how innovations may work at the development, intervention, program, and sustainment stages over time. In this way, we would hope to head off many pitfalls, reduce the temporal and financial costs of

failed trials and interventions, and achieve public health impacts more reliably.

Below are examples of topics in which systems methods might be applied productively to complex questions in WASH and HAP.

• Behavioral adherence requirements: Behavioral data—such as from stove use and water filter use monitors—is increas- ingly integrated into intervention deployment, behavior monitoring, and exposure assessment efforts. These rich, time-series data could be combined with network analysis and agent-based models (Ginexi et al. 2014) to help predict the effects of a program roll out on exposure for different levels of adherence to a HAP or WASH intervention. In this way, investigators could examine their assumptions and identify potential challenges prior to large-scale and expen- sive exposure and health assessments. Several investigators have done work in this area using quantitative microbial risk assessment (Hayashi et al. 2019).

• Community engagement and design: Community engage- ment is widely understood to be important for acceptability and sustainability of an intervention (Israel et al. 2005). It is especially important when complex individual behaviors have either social or exposure spillover effects on other households. Incorporating community-based system dynam- ics (Hovmand 2014) or another form of group model build- ing in designing a community-based WASH intervention can be extremely useful to identify critical behaviors and de- velop a consensus understanding of how the sanitation and hygiene practices of one group affects others. An interven- tion that emerges from this process is much more likely to be sustained and enforced by the community members indi- vidually not only because of acceptability borne of commu- nity participation but also because it is more likely to have correctly identified local facilitators and barriers to imple- mentation (Powell et al. 2017).

• Predicting effectiveness: Sanitation interventions aim to inter- rupt transmission of specific enteric infections, but threshold effects reflecting herd immunity can be important mediators of benefits (Fuller and Eisenberg 2016). An analog in HAP interventions would be the question of thresholds for coverage in community-scale programs that reduce ambient pollution sufficiently to yield health benefits for the population. In the

Technology trigger

Peak of inflated expecta�ons

Trough of disillusionment

Plateau of produc�vity

Slope of enlightenment

time

expectations

Aspirational Cycle - R&D to scaled effective interventions

Figure 2. Aspirational (flattened hype) cycle that accelerates research, devel- opment, and scale-up successes of effective environmental health interven- tions employing systems science methods in the context of implementation science (hypothetical curve). Note: R&D, research and development.

Environmental Health Perspectives 105001-7 128(10) October 2020

sanitation–infection domain, these thresholds have been explored with transmission models in a variety of contexts including drinking water. Population-scale models can be developed using system dynamics tools that force investiga- tors to specify their mechanistic assumptions and to investigate how changing these assumptions around exposures, interven- tion coverage, and adherence might impact disease incidence. These can be also be linked to networks to evaluate how differ- ent features of community structure affect outcomes.

• Dissemination/diffusion of interventions: Behaviors, includ- ing the adoption of new HAP and WASH technologies, often diffuse through communities through social connections and word of mouth (Rogers 2003; Valente et al. 2015). Social net- work analysis helps elucidate the underlying structure of social ties and could be useful in determining the patterns of information transfer in a community that will influence the adoption and use of an environmental health technology. Understanding the underlying social network structure, more- over, could shed light on potential interventions that could act through social networks to influence adoption and use. For example, social ties could be used to leverage group incentives for sustained use of LPG cooking (where each member of a group receives the incentive only if all members demonstrate the desired behavior) or to identify well-connected influencers who might be good targets for marketing and outreach efforts.

• Emergent patterns from individual variation: Fuel and stove stacking is ubiquitous in clean cooking programs, and yet we have a limited understanding of the choices people will make when multiple clean and traditional cooking options are available. Moreover, population-scale exposure results from hundreds of thousands of individual choices. Individual (agent-based) models based on survey data of the diverse choices people have made and their socioeconomic and envi- ronmental covariates may offer opportunities to explore what will happen at scale in relation to diverse influences, as well as the relative strength of these influences on outcomes.

• Scale-up, bottlenecks, and delays: Advanced gasifier stoves fueled by compressed wood pellets show promise as a clean cooking technology for both health- and climate-related objec- tives (Champion and Grieshop 2019). These stove/fuel combi- nations are being explored in development programs and commercial projects, but the technology faces significant chal- lenges in scale-up and maintenance over time because of feed- stock supply, pellet production, distribution, and costs to end users (Jagger and Das 2018). If enough demand is generated, stress to multiple points in the supply chain is likely to emerge and this could in turn dampen demand or create a new problem with unlawful harvesting of trees for feedstock. System dynam- ics models are particularly well suited to this type of problem because they allow us to model complex conditions and time delays across a wide variety of factors, including policy, eco- nomic, social, and physical variables and then to see how these play out over time in a given setting through simulation.

• Sustaining interventions: System science approaches are essential to understand and solve implementation problems that result from feedback between technologies, social and cul- tural norms, and livelihood options that are nonlinear and dynamic. Chalise et al. (2018) identified such feedback mecha- nisms in sustained use of improved biogas stoves in two com- munities; one community that sustained high levels of exclusive biogas use and an adjacent community that largely abandoned the technology. Using qualitative group model building and subsequent system dynamics simulation they traced the community interactions that led to solutions for tech- nical and maintenance problems with fuel digesters and thus

increased use of the cleaner technology in the one community, compared with frustration and abandonment by the other. The group model building and simulation highlight multiple house- hold, technical, and social factors that are interlinked in a feed- back structure and cannot be fully understood in isolation.

• Adaptability to environmental change: Behaviors and associ- ated technologies that may function well in avoiding or mitigat- ing environmental health challenges today may change quickly or lag in their ability to adapt to future changes. Climate change raises a host of challenges for WASH programs, in particular, andmanyofthese areinfluencedbysocial context.Cherng etal. (2019) illustrated the use of social networks in Ecuadorian communities to measure social cohesion across safe water sourcing practices, and agent-based models to understand how these structured communities will beable to adapt under chang- ing flooding and drought conditions.

How to Approach Systems Modeling Productively in Environmental Health Environmental health scientists are increasingly turning from assessing risks to designing and testing interventions. Research outcomes can catalyze programs and policies that operate at large spatial scales, and as in HAP and WASH, with great dependence on individual and community behaviors, it behooves us to approach interventions with an appreciation of the complexity of these systems. Environmental health research provides rigorous methodologies and a wide variety of large data sets from both inter- ventional and observational studies on population-based expo- sures, behaviors, and correlated health outcomes. Combining rich data sources with systems modeling may be especially productive to explore the sensitivity of hoped-for outcomes to assumptions regarding implementation variables such as household, commu- nity, or institutional adoption and adherence. This may also help us consider potential effects of adapting effective interventions to locally important contextual variables such as other exposure sour- ces, cost constraints, policy, and infrastructure influences.

As in any interdisciplinary team, depth in both analytical meth- ods and the environmental health challenge is necessary for suc- cess. System scientists come from diverse backgrounds, including sociology, ecology, physics, operations research, management, en- gineering, public health, and computer science, to name a few. Spending time at the outset of a collaboration between environ- mental health, systems scientists, and where relevant, policy experts, is critical to familiarizing one another with basic concepts and terminology in their respective fields and to agreeing on the aims, value, and limitations of various approaches.

For the systems science community, it is important to recog- nize that most environmental health scientists are trained to undertake population-based risk assessments using inferential sta- tistics. Systems approaches often include expert opinion, key in- formant inputs, or other soft evidence in addition to experimental results to parameterize a working model. Furthermore, systems modeling is often conducted with significant data gaps [indeed, this can be a primary reason for undertaking such modeling, see Wallace et al. (2015)]. Significant discussions at the outset may be necessary to ensure that environmental health scientists under- stand and are comfortable with the approach.

For environmental health scientists, it may be useful to approach systems modeling as a means of exploring and testing their own assumptions. We recommend, where possible, an itera- tive process including participatory or co-creative activities to frame questions, conceptualizing an initial model, collecting data to parameterize it, evaluating model output in relation to observa- tions, improving the model, and then testing interventions in sil- ico before beginning larger population-based studies. Such

Environmental Health Perspectives 105001-8 128(10) October 2020

approaches can allow for evaluation of the plausibility of health effects via changes in exposure at levels that may be health-rele- vant—before undertaking large trials using distal outcomes. Engagement with social, behavioral, and political scientists in the conceptualization, parameterization, and analysis of systems models can be important to avoid the omission of critical socio- demographic, economic, institutional, and macro-level processes that shape implementation. Participatory methods, such as group model building, may be particularly useful, especially with community-based research activities.

There are a variety of ways to build an interdisciplinary team that include systems scientists within the constraints of project budgets. At the most basic level, integrating systems modeling into environmental health research can be done by bringing a single experienced systems modeler into a project at the design stage and continuing through analyses and interpretation of results. At this level, the additional expertise may not be a large cost burden for a project, and it could conceivably fit within, for example, a National Institutes of Health R21 exploratory grant designed to assess basic feasibility of an intervention. Of course, modelers may raise ques- tions that require new data, including, for example, relationships between community members, expectations of policy-makers, or economic influences. Modelers may also suggest participatory processes, surveys, and focus group exercises that the environmen- tal health researchers did not anticipate. These in turn may require engagement of more disciplinary expertise or significant computa- tional coding time and, thus, further expansion of the team to the R01 or larger level. However, one of the strengths of modeling gen- erally is the ability to work within the constraints posed by the available data by drawing on a variety of existing or easily acquired sources to parameterize factors that may be especially challenging to estimate from what can be collected within a given project. The key here is understanding which factors require greater certainty for a given question and end-use plan, and which factors may allow for greater uncertainty.

Summary and Conclusion Systems science approaches offer important and underexploited tools for environmental health, especially in complex environ- ments that change over time. We work in a time of rapid innovation and with a public health policy community that increasingly seeks evidence to support decision-making. Both observational and ex- perimental studies in environmental health generate large data sets that can be used in both a priori and post hoc systems modeling. These data have often been gathered using extended, multiyear efforts, often with high collection costs. Increased use of systems modeling in implementation research for environmental health may offer cost-effective approaches with both heuristic value and practical output to improve design and deployment of interventions to improve public health. HAP and WASH science are particularly suitable to systems modeling given the complexity of the socioeco- nomic and environmental contexts that often regulate their effec- tiveness, implementability, and scale-up potential. Many other environmental health interventions may also benefit from their use. Increased collaboration among systems and environmental health scientists has the potential to accelerate lessons and flatten the Hype Cycle that slows and sometimes dooms promising interven- tions for public health before we understand their potential.

Acknowledgments Financial support from the NIH Common Fund’s Global

Health program for the Clean Cooking Implementation Science Network and from the Boston College School of Social Work enabled the 2018 workshop where these ideas were initially

developed. The findings and conclusions in this report are those of the authors and do not necessarily represent the official views of the U.S. National Institutes of Health or Department of Health and Human Services.

References Alexander D, Northcross A, Wilson N, Dutta A, Pandya R, Ibigbami T, et al. 2017.

Randomized controlled ethanol cookstove intervention and blood pressure in pregnant Nigerian women. Am J Respir Crit Care Med 195(12):1629–1639, PMID: 28081369, https://doi.org/10.1164/rccm.201606-1177OC.

Bain R, Cronk R, Hossain R, Bonjour S, Onda K, Wright J, et al. 2014. Global assessment of exposure to faecal contamination through drinking water based on a systematic review. Trop Med Int Health 19(8):917–927, PMID: 24811893, https://doi.org/10.1111/tmi.12334.

Barton CM, Alberti M, Ames D, Atkinson J-A, Bales J, Burke E, et al. 2020. Call for transparency of COVID-19 models. Science 368(6490):482–483, PMID: 32355024, https://doi.org/10.1126/science.abb8637.

Berendes DM, Yang PJ, Lai A, Hu D, Brown J. 2018. Estimation of global recoverable human and animal faecal biomass. Nat Sustain 1(11):679–685, https://doi.org/10. 1038/s41893-018-0167-0.

Bhattacharyya A, Light A. 2010. United States joins alliance to promote clean cooking in developing countries. Center for American Progress. 22 September 2010. https:// www.americanprogress.org/issues/green/news/2010/09/22/8353/united-states-joins- alliance-to-promote-clean-cooking-in-developing-countries/ [accessed 15 July 2020].

Bivins A, Sumner T, Kumpel E, Howard G, Cumming O, Ross I, et al. 2017. Estimating infection risks and the global burden of diarrheal disease attributable to intermit- tent water supply using QMRA. Environ Sci Technol 51(13):7542–7551, PMID: 28582618, https://doi.org/10.1021/acs.est.7b01014.

Brown J, Albert J, Whittington D. 2019. Community-led total sanitation moves the needle on ending open defecation in Zambia. Am J Trop Med Hyg 100(4):767– 769, PMID: 30860017, https://doi.org/10.4269/ajtmh.19-0151.

Brown J, Clasen T. 2012. High adherence is necessary to realize health gains from water quality interventions. PLoS One 7(5):e36735, PMID: 22586491, https://doi.org/ 10.1371/journal.pone.0036735.

Bunge M. 1997. Mechanism and explanation. Philos Soc Sci 27(4):410–465, https://doi.org/10.1177/004839319702700402.

Burke JG, Lich KH, Neal JW, Meissner HI, Yonas M, Mabry PL. 2015. Enhancing dissemination and implementation research using systems science methods. Int J Behav Med 22(3):283–291, PMID: 24852184, https://doi.org/10.1007/s12529- 014-9417-3.

Burnett RT, Pope CA III, Ezzati M, Olives C, Lim SS, Mehta S, et al. 2014. An inte- grated risk function for estimating the global burden of disease attributable to ambient fine particulate matter exposure. Environ Health Perspect 122(4):397– 403, PMID: 24518036, https://doi.org/10.1289/ehp.1307049.

Butt EW, Rap A, Schmidt A, Scott CE, Pringle KJ, Reddington CL, et al. 2016. The impact of residential combustion emissions on atmospheric aerosol, human health, and climate. Atmos Chem Phys 16(2):873–905, https://doi.org/10.5194/ acp-16-873-2016.

Capra F, Luisi PL. 2014. The Systems View of Life: A Unifying Vision. Cambridge, UK: Cambridge University Press.

Caro JJ, Briggs AH, Siebert U, Kuntz KM, ISPOR-SMDM Modeling Good Research Practices Task Force. 2012. Modeling good research practices—overview: a report of the ISPOR-SMDM Modeling Good Research Practices Task Force-1. Value Health 15(6):796–803, PMID: 22999128, https://doi.org/10.1016/j.jval.2012.06.012.

Chalise N, Kumar P, Priyadarshini P, Yadama GN. 2018. Dynamics of sustained use and abandonment of clean cooking systems: lessons from rural India. Environ Res Lett 13(3):03510, https://doi.org/10.1088/1748-9326/aab0af.

Champion WM, Grieshop AP. 2019. Pellet-fed gasifier stoves approach gas-stove like performance during in-home use in Rwanda. Environ Sci Technol 53(11):6570–6579, PMID: 31037940, https://doi.org/10.1021/acs.est.9b00009.

Cherng ST, Cangemi I, Trostle JA, Remais JV, Eisenberg JNS. 2019. Social cohe- sion and passive adaptation in relation to climate change and disease. Glob Environ Change 58:101960, PMID: 32863604, https://doi.org/10.1016/j.gloenvcha. 2019.101960.

Chowdhury S, Chafe ZA, Pillarisetti A, Lelieveld J, Guttikunda S, Dey S. 2019a. The con- tribution of household fuels to ambient air pollution in India—a comparison of recent estimates. New Delhi, India: Collaborative Clean Air Policy Centre. https:// ccapc.org.in/policy-briefs/2019/5/30/the-contribution-of-household-fuels-to-ambient- air-pollution-in-india-a-comparison-of-recent-estimates [accessed 15 July 2020].

Chowdhury S, Dey S, Guttikunda S, Pillarisetti A, Smith KR, Di Girolamo L. 2019b. Indian annual ambient air quality standard is achievable by completely mitigat- ing emissions from household sources. Proc Natl Acad Sci USA 116(22):10711– 10716, PMID: 30988190, https://doi.org/10.1073/pnas.1900888116.

Environmental Health Perspectives 105001-9 128(10) October 2020

Clasen T, Boisson S, Routray P, Torondel B, Bell M, Cumming O, et al. 2014. Effectiveness of a rural sanitation programme on diarrhoea, soil-transmitted hel- minth infection, and child malnutrition in Odisha, India: a cluster-randomised trial. Lancet Glob Health 2(11):e645–e653, PMID: 25442689, https://doi.org/10. 1016/S2214-109X(14)70307-9.

Clasen T, Smith KR. 2019. Let the “A” In WASH stand for air: integrating research and interventions to improve household air pollution (HAP) and water, sanita- tion and hygiene (WaSH) in low-income settings. Environ Health Perspect 127(2):25001, PMID: 30801220, https://doi.org/10.1289/EHP4752.

Clasen TF, Alexander KT, Sinclair D, Boisson S, Peletz R, Chang HH, et al. 2015. Interventions to improve water quality for preventing diarrhoea. Cochrane Database Syst Rev 2015(10):CD004794, PMID: 26488938, https://doi.org/10.1002/ 14651858.CD004794.pub3.

Coffey ER, Muvandimwe D, Hagar Y, Wiedinmyer C, Kanyomse E, Piedrahita R, et al. 2017. New emission factors and efficiencies from in-field measurements of tradi- tional and improved cookstoves and their potential implications. Environ Sci Technol 51(21):12508–12517, PMID: 29058409, https://doi.org/10.1021/acs.est.7b02436.

Colditz GA, Emmons KM. 2019. The promise and challenges of dissemination and implementation research. In: Dissemination and Implementation Research in Health. Brownson RC, Colditz GA, Proctor EK, eds. 2nd ed. Oxford, UK: Oxford University Press, 1–18.

Conibear L, Butt EW, Knote C, Arnold SR, Spracklen DV. 2018. Residential energy use emissions dominate health impacts from exposure to ambient particulate matter in India. Nat Commun 9(1):617, PMID: 29434294, https://doi.org/10.1038/ s41467-018-02986-7.

Curran GM, Bauer M, Mittman B, Pyne JM, Stetler C. 2012. Effectiveness-implemen- tation hybrid designs: combining elements of clinical effectiveness and imple- mentation research to enhance public health impact. Med Care 50(3):217–226, PMID: 22310560, https://doi.org/10.1097/MLR.0b013e3182408812.

Currie DJ, Smith C, Jagals P. 2018. The application of system dynamics modelling to environmental health decision-making and policy—a scoping review. BMC Public Health 18(1):402, PMID: 29587701, https://doi.org/10.1186/s12889-018-5318-8.

Cutler D, Miller G. 2005. The role of public health improvements in health advances: the twentieth-century United States. Demography 42(1):1–22, PMID: 15782893, https://doi.org/10.1353/dem.2005.0002.

Damschroder LJ, Aron DC, Keith RE, Kirsh SR, Alexander JA, Lowery JC. 2009. Fostering implementation of health services research findings into practice: a consolidated framework for advancing implementation science. Implement Sci 4:50, PMID: 19664226, https://doi.org/10.1186/1748-5908-4-50.

Dangour A, Watson L, Cumming O, Boisson S, Che Y, Velleman Y, et al. 2013. Interventions to improve water quality and supply, sanitation and hygiene practices, and their effects on the nutritional status of children. Cochrane Database Syst Rev 2013(8):CD009382, PMID: 23904195, https://doi.org/10.1002/ 14651858.CD009382.pub2.

Daniels ME, Smith WA, Schmidt W-P, Clasen T, Jenkins MW. 2016. Modeling Cryptosporidium and Giardia in ground and surface water sources in rural India: associations with latrines, livestock, damaged wells, and rainfall pat- terns. Environ Sci Technol 50(14):7498–7507, PMID: 27310009, https://doi.org/10. 1021/acs.est.5b05797.

Darvesh N, Das JK, Vaivada T, Gaffey MF, Rasanathan K, Bhutta ZA, et al. 2017. Water, sanitation and hygiene interventions for acute childhood diarrhea: a systematic review to provide estimates for the Lives Saved Tool. BMC Public Health 17(suppl 4):776, PMID: 29143638, https://doi.org/10.1186/s12889-017-4746-1.

Dilling L, Lemos MC. 2011. Creating usable science: opportunities and constraints for climate knowledge use and their implications for science policy. Glob Environ Change 21(2):680–689, https://doi.org/10.1016/j.gloenvcha.2010.11.006.

Eilenberg SR, Bilsback KR, Johnson M, Kodros JK, Lipsky EM, Naluwagga A, et al. 2018. Field measurements of solid-fuel cookstove emissions from uncontrolled cooking in China, Honduras, Uganda, and India. Atmos Environ 190:116–125, https://doi.org/10.1016/j.atmosenv.2018.06.041.

Eisenberg JNS, Cevallos W, Ponce K, Levy K, Bates SJ, Scott JC, et al. 2006. Environmental change and infectious disease: how new roads affect the trans- mission of diarrheal pathogens in rural Ecuador. Proc Natl Acad Sci USA 103(51):19460–19465, PMID: 17158S216, https://doi.org/10.1073/pnas.0609431104.

Eisenberg JNS, Scott JC, Porco T. 2007. Integrating disease control strategies: bal- ancing water sanitation and hygiene interventions to reduce diarrheal disease burden. Am J Public Health 97(5):846–852, PMID: 17267712, https://doi.org/10. 2105/AJPH.2006.086207.

Eisenberg JNS, Trostle J, Sorensen RJD, Shields KF. 2012. Toward a systems approach to enteric pathogen transmission: from individual independence to community interdependence. Annu Rev Public Health 33:239–257, PMID: 22224881, https://doi.org/10.1146/annurev-publhealth-031811-124530.

Enger KS, Nelson KL, Rose JB, Eisenberg JNS. 2013. The joint effects of efficacy and compliance: a study of household water treatment effectiveness against child- hood diarrhea. Water Res 47(3):1181–1190, PMID: 23290123, https://doi.org/10. 1016/j.watres.2012.11.034.

Epstein JM. 2007. Generative Social Sciences: Studies in Agent-Based Computation Modeling. Princeton, NJ: Princeton University Press.

Esrey SA, Feachem RG, Hughes JM. 1985. Interventions for the control of diar- rhoeal diseases among young children: improving water supplies and excreta disposal facilities. Bull World Health Organ 63(4):757–772, PMID: 3878742.

Esrey SA, Potash JB, Roberts L, Shiff C. 1991. Effects of improved water supply and sanitation on ascariasis, diarrhoea, dracunculiasis, hookworm infection, schisto- somiasis, and trachoma. Bull World Health Organ 69(5):609–621, PMID: 1835675.

Ezzati M, Baumgartner JC. 2017. Household energy and health: where next for research and practice? Lancet 389(10065):130–132, PMID: 27939060, https://doi.org/10.1016/ S0140-6736(16)32506-5.

Fenn J, Raskino M. 2008. Mastering the Hype Cycle: How to Choose the Right Innovation at the Right Time. Boston, MA: Harvard Business Press.

Freeman MC, Clasen T, Dreibelbis R, Saboori S, Greene LE, Brumback B, Muga R, Reingans R. 2014. The impact of a school-based water supply and treatment, hygiene, and sanitation programme on pupil diarrhoea: a cluster-randomized trial. Epidemiol Infect 142(2):340–351, PMID: 23702047, https://doi.org/10.1017/ S0950268813001118.

Fuller JA, Eisenberg JNS. 2016. Herd protection from drinking water, sanitation, and hygiene interventions. Am J Trop Med Hyg 95(5):1201–1210, PMID: 27601516, https://doi.org/10.4269/ajtmh.15-0677.

Galea S, Riddle M, Kaplan GA. 2010. Causal thinking and complex system approaches in epidemiology. Int J Epidemiol 39(1):97–106, PMID: 19820105, https://doi.org/10. 1093/ije/dyp296.

GBD 2016 Risk Factors Collaborators. 2017. Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet 390:1345–1422, PMID: 28919119, https://doi.org/10.1016/ S0140-6736(17)32366-8.

Ginexi EM, Riley W, Atienza AA, Mabry PL. 2014. The promise of intensive longitu- dinal data capture for behavioral health research. Nicotine Tob Res 16(suppl 2):S73–S75, PMID: 24711629, https://doi.org/10.1093/ntr/ntt273.

Glasgow RE, Klesges LM, Dzewaltowski DA, Bull SS, Estabrooks P. 2004. The future of health behavior change research: what is needed to improve translation of research into health promotion practice? Ann Behav Med 27(1):3–12, PMID: 14979858, https://doi.org/10.1207/s15324796abm2701_2.

Glasgow RE, Vogt TM, Boles SM. 1999. Evaluating the public health impact of health promotion interventions: the RE-AIM framework. Am J Public Health 89(9):1322–1327, PMID: 10474547, https://doi.org/10.2105/ajph.89.9.1322.

Global Alliance for Clean Cookstoves. 2017. Clean Cooking: Key to Achieving Global Climate and Development Goals, 2016 Progress Report. Washington, DC: Global Alliance for Clean Cookstoves. https://www.cleancookingalliance. org/binary-data/RESOURCE/file/000/000/495-1.pdf [accessed 15 July 2020].

Hammond RA, Dubé L. 2012. A systems science perspective and transdisciplinary models for food and nutrition security. Proc Natl Acad Sci USA 109(31):12356– 12363, PMID: 22826247, https://doi.org/10.1073/pnas.0913003109.

Hayashi MAL, Eisenberg MC, Eisenberg JNS. 2019. Linking decision theory and quantitative microbial risk assessment: tradeoffs between compliance and effi- cacy for waterborne disease interventions. Risk Anal 39(10):2214–2226, PMID: 31529800, https://doi.org/10.1111/risa.13381.

Hovmand PS. 2014. Community Based System Dynamics. New York, NY: Springer. Huang W, Baumgartner J, Zhang Y, Wang Y, Schauer JJ. 2015. Source appor-

tionment of air pollution exposures of rural Chinese women cooking with bio- mass fuels. Atmos Environ 104:79–87, https://doi.org/10.1016/j.atmosenv.2014. 12.066.

Humphrey JH, Mbuya MNN, Ntozini R, Moulton LH, Stoltzfus RJ, Tavengwa NV, et al. 2019. Independent and combined effects of improved water, sanita- tion, and hygiene, and improved complementary feeding, on child stunting and anaemia in rural Zimbabwe: a cluster-randomised trial. Lancet Glob Health 7(1):e132–e147, PMID: 30554749, https://doi.org/10.1016/S2214-109X (18)30374-7.

Hystad P, Duong M, Brauer M, Larkin A, Arku R, Kurmi OP, et al. 2019. Health effects of household solid fuel use: findings from 11 countries within the pro- spective urban and rural epidemiology study. Environ Health Perspect 127(5):57003, PMID: 31067132, https://doi.org/10.1289/EHP3915.

Israel BA, Parker EA, Rowe Z, Salvatore A, Minkler M, López J, et al. 2005. Community- based participatory research: lessons learned from the Centers for Children’s Environmental Health and Disease Prevention Research. Environ Health Perspect 113(10):1463–1471, PMID: 16203263, https://doi.org/10.1289/ehp.7675.

Jagger P, Das I. 2018. Implementation and scale-up of a biomass pellet and improved cookstove enterprise in Rwanda. Energy Sustain Dev 46:32–41, PMID: 30449968, https://doi.org/10.1016/j.esd.2018.06.005.

Jagger P, Pedit J, Bittner A, Hamrick L, Phwandapwhanda T, Jumbe C. 2017. Fuel ef- ficiency and air pollutant concentrations of wood-burning improved cookstoves in Malawi: implications for scaling-up cookstove programs. Energy Sustain Dev 41:112–120, PMID: 29731584, https://doi.org/10.1016/j.esd.2017.08.007.

Environmental Health Perspectives 105001-10 128(10) October 2020

Johnson MA, Chiang RA. 2015. Quantitative stove use and ventilation guidance for behavior change strategies. J Health Commun 20(suppl 1):6–9, PMID: 25839198, https://doi.org/10.1080/10810730.2014.994246.

Jung YT, Lou W, Cheng Y-L. 2017. Exposure–response relationship of neighbour- hood sanitation and children’s diarrhoea. Trop Med Int Health 22(7):857–865, PMID: 28449238, https://doi.org/10.1111/tmi.12886.

Kotloff KL, Nasrin D, Blackwelder WC, Wu Y, Farag T, Panchalingham S, et al. 2019. The incidence, aetiology, and adverse clinical consequences of less severe diarrhoeal episodes among infants and children residing in low-income and middle-income countries: a 12-month case-control study as a follow-on to the Global Enteric Multicenter Study (GEMS). Lancet Glob Health 7(5):e568– e584, PMID: 31000128, https://doi.org/10.1016/S2214-109X(19)30076-2.

Kumar P, Dhand A, Tabak RG, Brownson RC, Yadama GN. 2017. Adoption and sus- tained use of cleaner cooking fuels in rural India: a case control study protocol to understand household, network, and organizational drivers. Arch Public Health 75:70, PMID: 29255604, https://doi.org/10.1186/s13690-017-0244-2.

Lakatos I. 1970. Falsification and the methodology of scientific research pro- grammes. In: Criticism and the Growth of Knowledge. Lakatos I, Musgrave A, eds. New York, NY: Cambridge University Press, 91–196.

Levy K, Nelson K, Hubbard A, Eisenberg JNS. 2008. Following the water: a controlled study of drinking water storage in northern coastal Ecuador. Environ Health Perspect 116(11):1533–1540, PMID: 19057707, https://doi.org/10.1289/ehp.11296.

Lewis JJ, Pattanayak SK. 2012. Who adopts improved fuels and cookstoves? A sys- tematic review. Environ Health Perspect 120(5):637–645, PMID: 22296719, https://doi.org/10.1289/ehp.1104194.

Little M, Kaoukji D, Truesdale B. 2012. Achieving lasting impact at scale. Part two: assessing system readiness for delivery of family health innovations at scale. A convening hosted by the Bill & Melinda Gates Foundation. Synthesis and summary by the Social Research Unit at Dartington, UK. https://www. impactstrategist.com/wp-content/uploads/2019/08/Achieving-Lasting-Impact-at- Scale-Part-2.pdf [accessed 15 July 2020].

Liu J, Kiesewetter G, Klimont Z, Cofala J, Heyes C, Schopp W, et al. 2019. Mitigation pathways of air pollution from residential emissions in the Beijing-Tianjin-Hebei region in China. Environ Int 125:236–244, PMID: 30731373, https://doi.org/10.1016/ j.envint.2018.09.059.

Luby SP, Rahman M, Arnold BF, Unicomb L, Ashraf S, Winch PJ, et al. 2018. Effects of water quality, sanitation, handwashing, and nutritional interventions on diar- rhoea and child growth in rural Bangladesh: a cluster randomised controlled trial. Lancet Glob Health 6(3):e302–e315, PMID: 29396217, https://doi.org/10. 1016/S2214-109X(17)30490-4.

Luke DA, Harris JK. 2007. Network analysis in public health: history, methods, and applications. Annu Rev Public Health 28(1):69–93, PMID: 17222078, https://doi.org/ 10.1146/annurev.publhealth.28.021406.144132.

Luke DA, Morshed AB, McKay VR, Combs TB. 2018. Systems science methods in dis- semination and implementation research. In: Dissemination and Implementation Research in Health: Translating Science to Practice. Brownson RC, Colditz GA, Proctor EK, eds. Oxford, UK: Oxford University Press, 157–174.

Luke DA, Stamatakis KA. 2012. Systems science methods in public health: dynam- ics, networks, and agents. Annu Rev Public Health 33:357–376, PMID: 22224885, https://doi.org/10.1146/annurev-publhealth-031210-101222.

Madon T, Hofman KJ, Kupfer L, Glass RI. 2007. Public health. Implementation sci- ence. Science 318(5857):1728–1729, PMID: 18079386, https://doi.org/10.1126/ science.1150009.

Manibog FR. 1984. Improved cooking stoves in developing countries: problems and opportunities. Annu Rev Energy 9(1):199–227, https://doi.org/10.1146/annurev. eg.09.110184.001215.

Martin NA, Hulland KRS, Dreibelbis R, Sultana F, Winch PJ. 2018. Sustained adop- tion of water, sanitation and hygiene interventions: systematic review. Trop Med Int Health 23(2):122–135, PMID: 29160921, https://doi.org/10.1111/tmi.13011.

Masera OR, Saatkamp BD, Kammen DM. 2000. From linear fuel switching to multi- ple cooking strategies: a critique and alternative to the energy ladder model. World Dev 28(12):2083–2103, https://doi.org/10.1016/S0305-750X(00)00076-0.

Mauser W, Klepper G, Rice M, Schmalzbauer BS, Hackmann H, Leemans R, et al. 2013. Transdisciplinary global change research: the co-creation of knowledge for sustainability. Curr Opin Environ Sustain 5(3–4):420–431, https://doi.org/10. 1016/j.cosust.2013.07.001.

Meadows D, Robinson JM. 1985. The Electronic Oracle: Computer Models and Social Decisions. New York, NY: John Wiley & Sons.

Mellor J, Abebe L, Ehdaie B, Dillingham R, Smith J. 2014. Modeling the sustainabil- ity of a ceramic water filter intervention. Water Res 49:286–299, PMID: 24355289, https://doi.org/10.1016/j.watres.2013.11.035.

Mertens A, Balakrishnan K, Ramaswamy P, Rajkumar P, Ramaprabha P, Durairaj N, et al. 2019. Associations between high temperature, heavy rainfall, and diar- rhea among young children in rural Tamil Nadu, India: a prospective cohort study. Environ Health Perspect 127(4):47004, PMID: 30986088, https://doi.org/10. 1289/EHP3711.

Mobarak AM, Dwivedi P, Bailis R, Hildemann L, Miller G. 2012. Low demand for nontraditional cookstove technologies. Proc Natl Acad Sci USA 109(27):10815– 10820, PMID: 22689941, https://doi.org/10.1073/pnas.1115571109.

Mortimer K, Ndamala CB, Naunje AW, Malava J, Katundu C, Weston W, et al. 2017. A cleaner burning biomass-fuelled cookstove intervention to prevent pneumonia in children under 5 years old in rural Malawi (the Cooking and Pneumonia Study): a cluster randomised controlled trial. Lancet 389(10065):167–175, PMID: 27939058, https://doi.org/10.1016/S0140-6736(16)32507-7.

Nuckols JR, Ward MH, Jarup L. 2004. Using Geographic Information Systems for exposure assessment in environmental epidemiology studies. Environ Health Perspect 112(9):1007–1015, PMID: 15198921, https://doi.org/10.1289/ehp.6738.

Null C, Stewart CP, Pickering AJ, Dentz HN, Arnold BF, Arnold CD, et al. 2018. Effects of water quality, sanitation, handwashing, and nutritional interventions on diarrhoea and child growth in rural Kenya: a cluster-randomised controlled trial. Lancet Glob Health 6(3):e316–e329, PMID: 29396219, https://doi.org/10. 1016/S2214-109X(18)30005-6.

Patil SR, Arnold BF, Salvatore AL, Briceno B, Ganguly S, Colford JM Jr, et al. 2014. The effect of India’s total sanitation campaign on defecation behaviors and child health in rural Madhya Pradesh: a cluster randomized controlled trial. PLoS Med 11(8):e1001709, PMID: 25157929, https://doi.org/10.1371/journal.pmed.1001709.

Penakalapati G, Swarthout J, Delahoy MJ, McAliley L, Wodnik B, Levy K, et al. 2017. Exposure to animal feces and human health: a systematic review and proposed research priorities. Environ Sci Technol 51(20):11537–11552, PMID: 28926696, https://doi.org/10.1021/acs.est.7b02811.

Peng C, den Dekker M, Cardenas A, Rifas-Shiman SL, Gibson H, Agha G, et al. 2018. Residential proximity to major roadways at birth, DNA methylation at birth and midchildhood, and childhood cognitive test scores: project Viva (Massachusetts, USA). Environ Health Perspect 126(9):97006, PMID: 30226399, https://doi.org/10.1289/EHP2034.

Pilkey OH Jr, Pilkey-Jarvis L. 2007. Useless Arithmetic: Why Environment Scientists Can’t Predict the Future. New York, NY: Columbia University Press.

Powell BJ, Beidas RS, Lewis CC, Aarons GA, McMillen JC, Proctor EK, et al. 2017. Methods to improve the selection and tailoring of implementation strategies. J Behav Health Serv Res 44(2):177–194, PMID: 26289563, https://doi.org/10.1007/ s11414-015-9475-6.

Puzzolo E, Pope D, Stanistreet D, Rehfuess EA, Bruce NG. 2016. Clean fuels for resource-poor settings: a systematic review of barriers and enablers to adoption and sustained use. Environ Res 146:218–234, PMID: 26775003, https://doi.org/10. 1016/j.envres.2016.01.002.

Puzzolo E, Zerriffi H, Carter E, Clemens H, Stokes H, Jagger P, et al. 2019. Supply considerations for scaling up clean cooking fuels for household energy in low- and middle-income countries. Geohealth 3(12):370–390, PMID: 32159025, https://doi.org/10.1029/2019GH000208.

Quansah R, Semple S, Ochieng CA, Juvekar S, Armah FA, Luginaah I, et al. 2017. Effectiveness of interventions to reduce household air pollution and/or improve health in homes using solid fuel in low-and-middle income countries: a systematic review and meta-analysis. Environ Int 103:73–90, PMID: 28341576, https://doi.org/ 10.1016/j.envint.2017.03.010.

Rahmandad H, Oliva R, Osgood N, eds. 2015. Analytical Methods for Dynamic Modelers. Cambridge, MA: MIT Press.

Rahmandad H, Sterman JD. 2012. Reporting guidelines for simulation-based research in social sciences. Syst Dyn Rev 28(4):396–411, https://doi.org/10.1002/sdr.1481.

Richardson GP. 2020. Reflections on the foundations of system dynamics. Syst Dyn Rev 27(3):219–243, https://doi.org/10.1002/sdr.462.

Richardson GP, Andersen DF. 1995. Teamwork in group model building. Syst Dyn Rev 11(2):113–137, https://doi.org/10.1002/sdr.4260110203.

Rogers EM. 2003. Diffusion of Innovations. 5th ed. New York, NY: Simon & Schuster.

Rosenthal J. 2015. The real challenge for cookstoves and health: more evidence. Ecohealth 12(1):8–11, PMID: 25691140, https://doi.org/10.1007/s10393-014-0997-9.

Rosenthal J, Balakrishnan K, Bruce N, Chambers D, Graham J, Jack D, et al. 2017. Implementation science to accelerate clean cooking for public health. Environ Health Perspect 125(1):A3–A7, PMID: 28055947, https://doi.org/10.1289/EHP1018.

Schmidt W-P, Cairncross S. 2009. Household water treatment in poor populations: is there enough evidence for scaling up now? Environ Sci Technol 43(4):986– 992, PMID: 19320147, https://doi.org/10.1021/es802232w.

Sesan T, Jewitt S, Clifford M, Ray C. 2018. Toilet training: what can the cookstove sector learn from improved sanitation promotion? Int J Environ Health Res 28(6):667–682, PMID: 30068235, https://doi.org/10.1080/09603123.2018.1503235.

Shaheed A, Orgill J, Montgomery MA, Jeuland MA, Brown J. 2014. Why “improved” water sources are not always safe. Bull World Health Organ 92(4):283–289, PMID: 24700996, https://doi.org/10.2471/BLT.13.119594.

Sinharoy SS, Schmidt W-P, Wendt R, Mfura L, Crossett E, Grépin KA, et al. 2017. Effect of community health clubs on child diarrhoea in western Rwanda: cluster-randomised controlled trial. Lancet Glob Health 5(7):e699–e709, PMID: 28619228, https://doi.org/10.1016/S2214-109X(17)30217-6.

Environmental Health Perspectives 105001-11 128(10) October 2020

Smith AH, Lingas EO, Rahman M. 2000. Contamination of drinking-water by arsenic in Bangladesh: a public health emergency. Bull World Health Organ 78(9):1093– 1103, PMID: 11019458.

Smith K, Uma R, Kishore V, Lata K, Joshi V, Zhang J, et al. 2000. Greenhouse gases from small-scale combustion devices in developing countries, Phase IIA: house- hold stoves in India. EPA/600/R-00/052. Washington, DC: U.S. Environmental Protection Agency. https://cfpub.epa.gov/si/si_public_record_Report.cfm?Lab= NRMRL&dirEntryId=55087 [accessed 15 July 2020].

Smith KR, McCracken JP, Weber MW, Hubbard A, Jenny A, Thompson LM, et al. 2011. Effect of reduction in household air pollution on childhood pneumonia in Guatemala (RESPIRE): a randomised controlled trial. Lancet 378(9804):1717– 1726, PMID: 22078686, https://doi.org/10.1016/S0140-6736(11)60921-5.

Snider G, Carter E, Clark S, Tseng JTW, Yang X, Ezzati M, et al. 2018. Impacts of stove use patterns and outdoor air quality on household air pollution and cardi- ovascular mortality in southwestern China. Environ Int 117:116–124, PMID: 29734062, https://doi.org/10.1016/j.envint.2018.04.048.

Sterman J. 2018. System dynamics at sixty: the path forward. Syst Dyn Rev 34(1– 2):5–47, https://doi.org/10.1002/sdr.1601.

Sultana F. 2011. Suffering for water, suffering from water: emotional geographies of resource access, control and conflict. Geoforum 42(2):163–172, https://doi.org/ 10.1016/j.geoforum.2010.12.002.

Tabak RG, Khoong EC, Chambers DA, Brownson RC. 2012. Bridging research and practice: models for dissemination and implementation research. Am J Prev Med 43(3):337–350, PMID: 22898128, https://doi.org/10.1016/j.amepre.2012.05.024.

Thakur M, Nuyts PAW, Boudewijns EA, Kim JF, Faber T, Babu GR, et al. 2018. Impact of improved cookstoves on women’s and child health in low and middle income countries: a systematic review and meta-analysis. Thorax 73(11):1026– 1040, PMID: 29925674, https://doi.org/10.1136/thoraxjnl-2017-210952.

Thomas EA, ed. 2016. Broken Pumps and Promises: Incentivizing Impact in Environmental Health. Cham, Switzerland: Springer.

Thompson LM, Bruce N, Eskenazi B, Diaz A, Pope D, Smith KR. 2011. Impact of reduced maternal exposures to wood smoke from an introduced chimney stove on newborn birth weight in rural Guatemala. Environ Health Perspect 119(10):1489–1494, PMID: 21652290, https://doi.org/10.1289/ehp.1002928.

Tielsch JM, Katz J, Khatry SK, Shrestha L, Breysse P, Zeger S, et al. 2016. Effect of an improved biomass stove on acute lower respiratory infections in young chil- dren in rural Nepal: a cluster-randomised, step-wedge trial. Lancet Glob Health 4(suppl 1):S19, https://doi.org/10.1016/S2214-109X(16)30024-9.

Troncoso K, Soares da Silva A. 2017. LPG fuel subsidies in Latin America and the use of solid fuels to cook. Energy Policy 107:188–196, https://doi.org/10.1016/j. enpol.2017.04.046.

Valente TW. 2010. Social Networks and Health: Models, Methods, and Applications. Oxford, UK: Oxford University Press.

Valente TW. 2012. Network interventions. Science 337(6090):49–53, PMID: 22767921, https://doi.org/10.1126/science.1217330.

Valente TW, Palinkas LA, Czaja S, Chu K-H, Brown CH. 2015. Social network analy- sis for program implementation. PLoS One 10(6):e0131712, PMID: 26110842, https://doi.org/10.1371/journal.pone.0131712.

Venkataramanan V, Crocker J, Karon A, Bartram J. 2018. Community-led total sanitation: a mixed-methods systematic review of evidence and its quality. Environ Health Perspect 126(2):026001, PMID: 29398655, https://doi.org/10. 1289/EHP1965.

Vennix JAM. 1996. Group Model Building: Facilitating Team Learning Using System Dynamics. New York, NY: John Wiley & Sons.

Wallace RB, Geller A, Ogawa VA. 2015. Assessing the Use of Agent-Based Models for Tobacco Regulation. Washington, DC: National Academies Press.

Wathore R, Mortimer K, Grieshop AP. 2017. In-use emissions and estimated impacts of traditional, natural- and forced-draft cookstoves in rural Malawi. Environ Sci Technol 51(3):1929–1938, PMID: 28060518, https://doi.org/10.1021/acs.est.6b05557.

Weaver AM, Gurley ES, Crabtree-Ide C, Salje H, Yoo EH, Mu L, et al. 2019. Air pollu- tion dispersion from biomass stoves to neighboring homes in Mirpur, Dhaka, Bangladesh. BMC Public Health 19(1):425, PMID: 31014315, https://doi.org/10. 1186/s12889-019-6751-z.

World Health Organization. 2014. WHO indoor air quality guidelines: household fuel combustion. http://www.who.int/iris/bitstream/10665/141496/1/9789241548885_ eng.pdf?ua=1 [accessed 17 September 2020].

Wright J, Gundry S, Conroy R. 2004. Household drinking water in developing coun- tries: a systematic review of microbiological contamination between source and point-of-use. Trop Med Int Health 9(1):106–117, PMID: 14728614, https://doi.org/ 10.1046/j.1365-3156.2003.01160.x.

Yamey G. 2011. Scaling up global health interventions: a proposed framework for success. PLoS Med 8(6):e1001049, PMID: 21738450, https://doi.org/10.1371/ journal.pmed.1001049.

Yunus FM, Khan S, Chowdhury P, Milton AH, Hussain S, Rahman M. 2016. A review of groundwater arsenic contamination in Bangladesh: the Millennium Development Goal era and beyond. Int J Environ Res Public Health 13(2):215, PMID: 26891310, https://doi.org/10.3390/ijerph13020215.

Zelner JL, Trostle J, Goldstick JE, Cevallos W, House JS, Eisenberg JNS. 2012. Social connectedness and disease transmission: social organization, cohesion, village context, and infection risk in rural Ecuador. Am J Public Health 102(12):2233–2239, PMID: 23078481, https://doi.org/10.2105/AJPH. 2012.300795.

Environmental Health Perspectives 105001-12 128(10) October 2020

Copyright of Environmental Health Perspectives is the property of Superintendent of Documents and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.