Systematic review of existing behaviour change models and frameworks in WASH

Rainey and Harding [28] examined the adoption and use of solar disinfection for water treatment in Nepal using a modified version of the Health Belief Model [29, 30], in which the relationship between individual perceptions and behavioural outcomes are linked by modifying factors, such as: perceived threat, individual socio-demographics, and cues to action. Two models focus on the decision making process related to the adoption and/or sustained use of specific WASH enabling technologies, one on water treatment technologies in Malawi [31] and the other on household sanitation in Benin [32]. Both models either implicitly or explicitly draw from Prochaska’s Transtheoretical (Stages of Change) Model [33].

Curtis et al. [34] present a conceptual model of handwashing behaviours based on an 11-country review of formative research findings in support of national-level handwashing promotion programs. The authors propose that three separate but interacting mental processes – planning, motivations, and habits - shape behaviour. Each of these three is, in turn, shaped by environmental determinants that include the biological, the social, and the physical. The specific determinants are further elaborated in a review paper [35].

The RANAS model by Mosler [36] divides behavioural determinants into five primary factor “blocks” – risk factors, attitudinal factors, normative factors, ability factors, and self-regulation factors. Of the models identified, the RANAS model is one of the few that is intended to be applicable across multiple WASH practices and interventions. The RANAS model associates specific intervention strategies with each of the identified factor blocks – information interventions with risk factors; persuasive interventions with attitudinal factors; infrastructural and ability interventions with ability factors.

Our review also included three primarily programmatic frameworks. Figueroa and Kincaid [37] developed a model of communication for water treatment and safe storage practices. According to their model, interventions influence behavioural outcomes via a set of multi-level intermediary outcomes. Individual-level outcomes included in the framework are further divided into cognitive elements, emotional factors, and social interactions. Household factors include time allocation, household-decision making practices, and household income. Community-level factors include community action and resources, community cohesion, and community leadership.

FOAM (for handwashing) and SaniFOAM (for sanitation behaviours) are behavioural frameworks developed by the World Bank’s Water and Sanitation Program [38, 39]. They organize behavioural determinants into three main domains – Opportunity to improve a particular behaviour, Ability to change behaviour, and Motivation to change behaviour. Product attributes - such as comfort, convenience, and smell - are considered opportunity determinants. Willingness-to-pay, particularly for sanitation improvements, is viewed as a critical component of motivational determinants.

The Hygiene Improvement Framework, developed by the Environmental Health Project in partnership with USAID, UNICEF, and others, proposes a framework for combating diarrhoeal disease that consists of three main components: improving access to water and sanitation hardware, promoting hygiene, and strengthening the enabling environment [1].

Discussion of systematic review findings

Each of the models identified in this review offers valuable theoretical and conceptual determinants that help us understand the nature of WASH behaviours and WASH behaviour change; however a number of other factors were not emphasized in the existing models. Characteristics of the type of water and sanitation technology (e.g. handwashing station, water treatment technology, latrine), including the cost and complexity of using it were under-represented in existing frameworks. Contextual factors – i.e.: gender, age, socio-economic status, household structure and availability of resources - were explicit in only a limited number of existing frameworks, such as the model proposed by Figueroa and Kinkaid; or selected contextual factors, such as roles in decision making, and affordability (FOAM/SaniFOAM), were grouped with other psychological or social determinants of behavioural outcomes.

Most existing models concentrated almost exclusively on individual-level factors that influence behavioural outcomes rather than utilized a broader ecological model approach that positions individual behaviours within a multi-level causal framework [40, 41], a trend that is consistent across much of public health research [42]. Those frameworks that do incorporate a multi-level perspective do so only for psychological factors related to behaviour change [34, 37]. While intervening at individual, community, and policy levels may be beyond the scope of many WASH interventions, understanding WASH behaviours within their spheres of influence is an important step in creating and sustaining behaviour change. In their application of the Health Belief Model to solar disinfection in Nepal, Rainey and Harding [28] discuss how structural factors, such as gender roles and involvement in agricultural production limit the time women have and are willing to commit to household water treatment. Characteristics of the physical environment such as chemical and microbiological composition of available sources of water, level of the water table, pattern of precipitation, population density, and available space were largely ignored.

Attributes or characteristics of WASH behaviours, such as the steps to be followed, and when and where the behaviour should be carried out in order to have a health impact [43] and factors affecting whether a behaviour becomes habitual [44] were rarely considered in existing frameworks. Habit may be particularly important in relation to WASH practices. Improved water, sanitation, and hygiene practices are not one-time behaviour changes – they require significant repetition across both time and space. The frameworks proposed by Jenkins and Scott [32] and Wood et al. [31] presented decision making models related to adoption of specific technologies, but only the model from Wood et al. explicitly addressed factors related to sustained and continued use and maintenance of technologies^b.

The Integrated Behavioural Model for Water, Sanitation, and Hygiene

Our framework has three intersecting dimensions that influence WASH-behaviours: the contextual dimension, psychosocial dimension, and the technological dimension. The Contextual Dimension includes determinants related to the individual, setting, and/or environment that can influence behaviour change and adoption of new technologies. The Psychosocial Dimension comprises the behavioural, social, or psychological determinants that influence behavioural outcomes and technology adoption. The specific attributes of a technology, product, or device that influence its adoption and sustained use constitute the Technological Dimension. These three interacting dimensions reflect the concept of reciprocal determinism in Social Cognitive Theory, which describes mutual interactions between the individual, the behaviour, and the environment in which the behaviour is practiced [45].

For the purposes of our framework we have identified five aggregate levels generally analogous to levels that exist in multi-level models. The Societal / Structural Level of our framework refers to the broad organisational, institutional, or cultural factors that influence behaviours in each of our three dimensions. This includes factors such as laws, policies, climate, geography, geology, and manufacturing and commercial distribution of products. The Community Level includes the physical and social environment in which individuals are nested, as well as the formal and informal institutions that shape individual experiences. It is analogous to both the Institutional and the Community factors as proposed by McLeroy et al. [40]. The Interpersonal / Household Level represents interactions between individuals and the people they intimately associate with, including household members, close friends and neighbours. At this level, factors include roles and responsibilities in the household, household wealth, injunctive and descriptive norms, aspirations, shame, sharing access to a product, and behavioural modelling. The Individual Level includes sociodemographic factors – such as age and gender, individual cognitive factors, and attitudes toward the product, hardware, or behaviour. The final level of our model is the Habitual Level. This level, nested within the individual, reflects the fact that the opportunity and necessity for WASH-related behaviours are repeated over the course of the day, and the multiple processes or events that can result in the specific behavioural outcomes. Given emerging evidence on the importance of habit and automaticity in promoting and maintaining behaviour change, our framework also focuses on factors related to habit formation.^b

Dimensions of IBM-WASH: contextual factors, psychosocial factors, and technology factors

Theory derived from practice

Our framework emerged not only through a review of existing theories and theoretical models, but was iteratively adapted based on its application during formative research and intervention piloting in support of two large-scale cluster randomized trials in Bangladesh. For example, earlier versions were focused on individual-level psychosocial determinants with relatively little focus on other levels or the complex interactions between psychosocial determinants and context and technology. The technological component was initially based solely on individual-level determinants as well, and the application to technologies shared among multiple households – such as chlorine dispensers – informed the multi-level approach to each of the three dimensions of our framework. Initial iterations consisted of a traditional “boxes and arrows” approach to presenting behaviour change theories. As our framework expanded, we modified our presentation to a matrix format, focusing on relationships between and among determinants rather than causal pathways. Further feedback informed the inclusion of the habitual level. It is beyond the scope of this manuscript to provide details on each stage of the evolution of IBM-WASH, however this process was used to identify redundancies in behavioural determinants in our initial summary model, identify factors that required more or less elaboration, and inform the explicit use of a multi-level framework. Detailed results from several of the concurrent pilot projects include technologies for child faeces management [71], handwashing station designs and selection [23], and point-of-use water treatment technologies (manuscript in preparation).

One key application of the IBM-WASH model is to guide researchers and practitioners in acknowledging the various levels of influence that may shape behavioural-level outcomes, through the use of individual cells as a “checklist” in order to ensure that the full set of determinants for a specific behavioural intervention have been considered prior to implementing an intervention. Such a checklist can help identify areas in which additional qualitative data collection will help to understand the technological, psychosocial, and contextual dimensions that shape specific behavioural outcomes. In a similar manner, our framework can be applied to the development of data collection tools and instruments intended to better understand WASH-behaviours. The structure of our matrix easily translates to the development of qualitative codebooks and analysis plans for qualitative data (see Hulland et al. 2013 for an example).

Transcending the individual-level

Structured as a multi-level model, the IBM-WASH facilitates the process of developing interventions that operate beyond the individual or household level. Interventions that operate at the structural level not only have the capacity to reach large sections of the population, but are also highly cost-effective [72].

“Accessible” to practitioners

While specific determinants and their relative contribution to behaviour change and habit formation may differ between behaviours, technologies, or populations, the IBM-WASH model is intended to be simple, adaptable, and accessible to both practitioners and researchers. Materials used for presenting and training researchers on the IBM-WASH framework are included as an on-line supplemental file (Additional file 1). The framework can help guide monitoring and evaluation strategies for both large- or small-scale WASH programs by identifying intermediate outcomes linking enabling technologies and hygiene promotion to specific behavioural outcomes. Tracking these intermediary outcomes provides a possible shorter-term outcome that can be used in on-going monitoring to evaluate the effectiveness of longer-term intervention activities or refine messaging and promotion strategies to influence key determinants more directly.

The intended accessibility of our framework to practitioners, however, underscores a number of specific research gaps in the WASH sector. The framework identifies a number of testable hypotheses applicable to WASH-related behavioural determinants, and rigorous approaches to the measurement of such determinants across all three domains of the framework are needed. Rigorous measurement of determinants and the application of measurement theory has been largely lacking within the WASH community. A handful of studies have focused on the validity and reliability of self-reported behavioural outcomes in both domestic and institutional contexts, and there is general agreement that alternative approaches to assessing behavioural outcomes are required [73–75]. There has been little to no attention given to the measurement of behaviour determinants within a domestic environment. The RANAS [36] model includes a suggestion for standardized questions and analyses that can be used to measure the behavioural factors included in the model; however, the measurement system proposed is based almost entirely on individual-level behavioural factors, nor does it address the complex inter-relationships among psychosocial, technological, and contextual determinants. Applications of the RANAS model have operationalized many of its determinants through single questions in a survey [76] and more information is required on the reliability and validity of the proposed measurement system. Studies have addressed the reliability and validity (or psychometric properties) of measures for specific behavioural determinants, including self-efficacy for hand washing in an institutional setting [63], disgust sensitivity [77] and habit strength [44]. Understanding the validity and reliability of scaled measures in low literacy populations – where there is a general lack of familiarity with both eliciting and providing scaled or Likert-type responses - would allow for a better understanding of their relative contribution to behaviour change and habit formation, and allow for comparisons among multiple behavioural outcomes.

We developed our model based on formative research for a select number of water, sanitation, and hygiene technologies, a potential limitation. These included various improved models of handwashing station design, technologies for cleaning child faeces, multiple point-of-use water treatment systems, and communal or household-level chlorination systems. Household water filters are one type of technology not included in the development of our model. Future research is needed in which the IBM-WASH framework is applied to and adapted for filter technologies. Regarding the Contextual Dimension of our framework, the durable nature of household filters associated with higher capital costs compared to chemical household water treatment systems may limit their appeal within lower income – and thus – higher need populations; while both the costs and visual appeal of filter technologies may appeal to the aspirational factors included in our Psychosocial Dimension. It is the Technology Dimension in which filters will likely differ the most from chemical-based treatment systems. Filters typically require fewer steps and less labour – thus easier to use - than chemical treatment options, which may facilitate adoption and habit formation. However, this perceived ease will be mediated by other strengths and weaknesses of the design including flow rate, wait time, and filtration effectiveness.

The determinants outlined in the IBM-WASH framework should be viewed as dynamic and inter-related. They may differ across place, setting, and time. There is general consensus, for example, on the fact that disgust may be a major determinant of WASH practices [34, 35, 52, 55], yet there has been little attention to the ways in which the disgust response is shaped by factors related to the physical and natural environment. Exposure to open sewerage pits, industrial run-off, and crowded living conditions in an urban slum may result in populations with less sensitivity to disgust stimuli. Behaviours in schools, the street, and places of employment may be influenced by a different constellation of factors than in the home. At a minimum, behaviours outside of the home will be subject to a different set of contextual factors than in the domestic environment. In all cases, it is important to adapt the framework by defining the specific determinants operating at each level and dimension through formative and pilot research. The multiple levels of the IBM-WASH model ensure that any individual behavioural outcome must be understood within the larger societal and communal contexts in which it occurs. The potential complexities of these dynamic relationships among determinants are simplified by specifying a behaviour and/or a population of interest, a concept that is embedded in many programmatic behaviour change approaches and frameworks [38, 39, 78]. The focus of our framework is on behaviours and habit formation, and not explicitly on exposure reduction. It is assumed that improving WASH practices will reduce pathogen exposure, yet it is important to understand the determinants of these behaviours independent of their direct link with transmission pathways. Incorporating a dynamic view of the role of context in habit formation – specifically identifying opportunities and moments throughout the day within the community in which behaviours can or cannot be repeated – may guide practitioners in identifying other settings in which new behaviours and/or technologies are required. These other contexts within a community may also provide novel entry points for introducing new behaviours and technologies to a population.