Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A realist implementation evaluation of British Columbia's school food and beverage sales policy Levay, Adrienne Vanessa 2018

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2018_september_levay_adrienne.pdf [ 3.63MB ]
Metadata
JSON: 24-1.0366053.json
JSON-LD: 24-1.0366053-ld.json
RDF/XML (Pretty): 24-1.0366053-rdf.xml
RDF/JSON: 24-1.0366053-rdf.json
Turtle: 24-1.0366053-turtle.txt
N-Triples: 24-1.0366053-rdf-ntriples.txt
Original Record: 24-1.0366053-source.json
Full Text
24-1.0366053-fulltext.txt
Citation
24-1.0366053.ris

Full Text

  A REALIST IMPLEMENTATION EVALUATION OF BRITISH COLUMBIA’S SCHOOL FOOD AND BEVERAGE SALES POLICY  by  Adrienne Vanessa Levay  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENT FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Integrated Studies in Land and Food Systems)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  April 2018  © Adrienne Vanessa Levay, 2018 ii  Abstract  This dissertation explores social processes related to implementation of British Columbia’s (BC) school food and beverage sales policy as a food environment intervention. Using a realist approach to evaluation, the first phase of the research focused on development of a retrospective program theory. This was used to create the framework for an implementation evaluation conducted in the second phase. I used a multiple case study approach with three urban and two rural BC school districts to explore what about this intervention is working, in what contexts, and for whom. Data collection included semi-structured interviews and questionnaires with relevant heath, education, and private industry stakeholders, observations, document analysis and website scans. Data analysis focused on identifying (i) mechanisms influencing if and how stakeholders engage in implementation activities and (ii) specific dimensions of context influencing these mechanisms. I identified four mechanisms. The mandatory mechanism refers to the ways that the mandatory nature of the policy is effective for triggering implementation efforts, influenced by a normative acceptance of the education system hierarchy. The scofflaw mechanism refers to an opposite response to the mandate whereby expected implementers may ignore and/or ‘skirt’ around the policy, influenced by beliefs about the role of government, school food, and food in general. The money mechanism refers to the way in which vendors respond to school and district demand for compliant options, influenced by beliefs about food preferences of children, health and food, and the existence of competition. The resource constraint mechanism refers to how a lack of capacity triggers otherwise motivated stakeholders to not implement. These findings helped refine the initial program theory to include an articulation of specific dimensions of context influencing implementation, an emphasis on the mandatory nature of school food environment policy, and the role of private industry. Interventions to support implementation could include: monitoring systems and incentive schemes, targeted resources to motivated school communities, initiatives to increase availability of compliant options, and to improve the private food vending environment in the vicinity of schools. There is a need to reconsider the implications of using nutrient-based standards that enable reformulation and/or finding loopholes.   iii  Lay Summary  This research explores how and why implementation of a school food and beverage sales policy happens the way it does in British Columbia. The findings identify that the provincial government’s declaration of the policy as mandatory both helps and hinders implementation efforts, especially when it does not come with resources to support implementation. Findings also identify a process of financial incentivisation for private industry that leads private vendors to comply with the policy in different ways. There is a need for initiatives to support better implementation. These may include: (1) developing monitoring and incentive schemes, (2) targeting resources to support motivated school communities who want to offer policy-compliant options but feel they have a lack of capacity, (3) devising initiatives to stimulate an increase in options that comply with the policy, and (4) working with vendors in the vicinity of schools to address their food and beverage environments. iv  Preface Adrienne Levay was responsible for devising the research idea and designing the research program, undertaking recruitment and data collection, analysis, and preparation of dissertation.  Chapter 4. A version of this material has been accepted with minor revisions as: Levay, A.V., Chapman, G.C., Seed, B. and Wittman, H. W. ‘It’s just the right thing to do’: A theory of change for a school food and beverage sales environment intervention. Evaluation and Program Planning. Manuscript # EPP 2017 189. AL and GC derived the research idea. All authors contributed to the development of the research process. AL recruited participants, collected data, conducted the analysis and prepared the manuscript. GC, BS, and HW contributed in-depth intellectual input for the development of the manuscript.    This research was conducted with approval from the University of British Columbia’s Behavioural Research Ethics Board, under the project title “Implementation of School Food and Beverage Sales Guidelines”, certificate number: H15-01051.v  Table of Contents  Abstract ........................................................................................................................................... ii Lay Summary ................................................................................................................................. iii Preface............................................................................................................................................ iv Table of Contents………………………………………………………………………………….v List of Tables ............................................................................................................................... viii List of Figures ................................................................................................................................ ix List of Abbreviations ...................................................................................................................... x Acknowledgements ........................................................................................................................ xi Chapter 1 ~ Introduction ................................................................................................................. 1 1.1 School food and beverage sales policy in British Columbia, Canada ................................... 4 1.2 Organisation of the dissertation............................................................................................. 6 Chapter 2~ Public health policy to improve school food environments and a realist approach to implementation research ................................................................................................................. 8 2.1 Public health policy and food environments ......................................................................... 8 2.2 Individual behaviour change and ecological health interventions addressing nutrition ..... 11 2.3 School food and beverage environment interventions ........................................................ 15 2.3.1 School food and beverage sales standards .................................................................... 16 2.4 Implementation evaluations of school food and beverage sales policies in BC and beyond ................................................................................................................................................... 18 2.5 Potential for inequitable implementation of school food and beverage sales policies ........ 22 2.6 Implementation research and the realist approach .............................................................. 23 2.7 Purpose of the research ....................................................................................................... 25 Chapter 3~ Realist evaluation methodology ................................................................................. 28 3.1 Realist philosophy of science .............................................................................................. 28 3.1.1 Realist ontology ............................................................................................................ 30 3.1.2 Generative versus successive causation ....................................................................... 34 3.1.3 Realist conception of validity ....................................................................................... 35 3.2 Realist methodology and the realist evaluation cycle ......................................................... 36 3.2.1 Realist research cycle ................................................................................................... 37 vi  3.2.2 The realist research cycle for the implementation evaluation in this dissertation ........ 45 3.3 Locating the researcher ....................................................................................................... 49 Chapter 4~ Development of a retrospective logic model and program theory for British Columbia’s school food and beverage sales standards ................................................................. 51 4.1 Introduction and background .............................................................................................. 51 4.2 Methods ............................................................................................................................... 54 4.3 Findings ............................................................................................................................... 59 4.3.1 A logic model for British Columbia’s school food and beverage sales policy ............ 60 4.3.2 Program theory for British Columbia’s school food and beverage sales policy .......... 65 4.3.3 A broad program theory: combining BC-specific logic model and program theory .... 71 4.4 Discussion ........................................................................................................................... 73 Chapter 5~ Social processes associated with district-level implementation of British Columbia’s school food and beverage sales policy: what works and why? ..................................................... 79 5.1 Introduction ......................................................................................................................... 79 5.2 Research objective............................................................................................................... 81 5.3 Methods ............................................................................................................................... 82 5.4 Results ................................................................................................................................. 85 5.4.1 Mandatory mechanism: driver of demand creation ...................................................... 86 5.4.2 Money mechanism: driver of supply creation .............................................................. 88 5.4.3 Monitoring mechanism: a dynamic informal process of enforcement ......................... 91 5.5 Discussion ........................................................................................................................... 93 Chapter 6~ School-level implementation of the BC Guidelines ................................................ 101 6.1 Introduction ....................................................................................................................... 101 6.2 Methods ............................................................................................................................. 106 6.3 Findings ............................................................................................................................. 112 6.3.1 Mandatory mechanism and scofflaw mechanism: Opposite responses to the same intervention resource ........................................................................................................... 112 6.3.2 Money mechanism: Demand created leads to supply creation................................... 118 6.3.3 Resource constraint mechanism ................................................................................. 120 6.4 Discussion ......................................................................................................................... 123 Chapter 7 ~ Discussion and conclusions .................................................................................... 131 7.1 Considerations for program theory refinement ................................................................. 133 7.1.1 Mandatory mechanism versus hypothesised responsibilisation efforts ...................... 133 7.1.2 Motivated by money and potential for public procurement ....................................... 137 7.1.3 ‘Moral distress’ among responsibilised implementers ............................................... 139 vii  7.1.4 Refined program theory for school food and beverage environment interventions ... 140 7.1.4 The potential for exacerbating health inequities ......................................................... 144 7.2 Known-unknowns and future research .............................................................................. 146 7.3 Study limitations ............................................................................................................... 149 7.3.1 Influence of researcher positionality on data collection ............................................. 149 7.3.2 Inconsistency of data sources across cases ................................................................. 149 7.3.3 Data saturation ............................................................................................................ 151 7.3.4 The hearsay problem (?) and making inferences ........................................................ 151 7.4 Implications of this study .................................................................................................. 153 7.4.1 Implications for research ............................................................................................ 153 7.4. Implications for practitioners........................................................................................ 153 7.4.3 Implications for policymakers .................................................................................... 154 7.4 Conclusion ......................................................................................................................... 155 References ................................................................................................................................... 157 Appendix A: Descriptions of the five participating districts ...................................................... 181 Appendix B: Ethical considerations and challenges ................................................................... 186 Appendix C: Semi-structured interview guidelines (phase 1) .................................................... 188 Appendix D: Candidate CMOcs derived from phase 1 .............................................................. 193 Appendix E: Semi-structured interview guidelines (phase 2) .................................................... 198 Appendix F: Questionnaire for school-level implementers ........................................................ 200 Appendix G: Consent forms and information letters (phase 1) .................................................. 207 Appendix H: Consent forms and information letters (phase 2) .................................................. 213  viii  List of Tables Table 4.1 Documents reviewed..................................................................................................... 55 Table 4.2 Examples of inductive codes ........................................................................................ 58 Table 5.1 Data collection strategies in each school district and regional health authority (RHA) 83 Table 5.2 Questionnaires completed by district and school-type ................................................. 84 Table 5.3 Describing the mandatory mechanism at the district level ........................................... 87 Table 5.4 Describing the money mechanism at the district level ................................................. 91 Table 5.5 Describing the monitoring mechanism at the district level .......................................... 93 Table 6.1 Data collection strategies used by district and Regional Health Authority (RHA) .... 109 Table 6.2 Questionnaires completed by district and school-type ............................................... 110 Table 6.3 Context + mechanism  outcome configurations for the mandatory mechanism at the school level .......................................................................................................................... 115 Table 6.4 Context + mechanism  outcome configurations for the opposition to implementation ............................................................................................................................................. 117 Table 6.5 Context + mechanism  outcome configurations for demand and supply creation at the school level .................................................................................................................... 120 Table 6.6 Context + mechanism outcome configurations for the resource constraint mechanism ........................................................................................................................... 123 Table 7.1 Known-unknowns emerging from this research: suggestions for future research ...... 148 Table A.1 Basic information about participating school districts ............................................... 181 Table D.1 Complete set of candidate CMOcs derived from phase 1.......................................... 193       ix  List of Figures Figure 3.1 Realist evaluation cycle (Vareilles et al., 2015) .......................................................... 38 Figure 3.2 The research cycle for this research (adapted from Vareilles et al., 2015) ................. 45 Figure 4.1 Retrospective logic model specifically for the 2013 Guidelines for Food and Beverage Sales in BC Schools ............................................................................................................... 61 Figure 4.2 Hypothesised underlying program theory specifically for BC’s Guidelines ............... 66 Figure 4.3 A hypothesised program theory for school food and beverage sales environment interventions .......................................................................................................................... 72 Figure 7.1 Refined program theory for school food and beverage sales environment interventions ............................................................................................................................................. 143 x  List of Abbreviations  BC- British Columbia BCMoED- British Columbia Ministry of Education BCMoH- British Columbia Ministry of Health BNFL- Brand Name Food List C- Context CIT- Contextual Interaction Theory CMO- Context + Mechanism Outcome CMOc - Context + Mechanism  Outcome configuration CSH- Comprehensive School Health CSOs- Community Service Organisations GDP- Gross Domestic Product HBM- Health Belief Model HLBC- Health Link British Columbia JCSH- Joint Consortium for School Health LNED- Low Nutrient Energy Dense M- Mechanism MRT- Middle Range Theory O- Outcome PAC- Parent Advisory Council RCT- Randomised Control Trial RFP- Request For Proposal RHA- Regional Health Authority SD- School district SES- Socio-Economic Status SSBs- Sugar-Sweetened Beverages UBC- University of British Columbia US- United Statesxi  Acknowledgements  I would first like to start by expressing my immense gratitude for the support of the Faculty of Land and Food Systems administrative staff and faculty. In particular, Dr. Gwen Chapman, my tireless and supportive supervisor who provided an atmosphere of supervision that allowed me just enough space and support to try and work things out on my own but not fall flat on my face. Thank you to Dr. Hannah Wittman, who provided me excellent learning and work opportunities and willingly stepped into the shoes of co-supervisor when asked. I could not have asked for more encouraging and supportive supervisors to have challenged me intellectually. I would also like to acknowledge the hard work of my supervisory committee whose various perspectives contributed to strengthening this work.  I would also like to thank the Regional Health Authority staff involved in school food that eagerly responded to a call for interview participants and provided excellent insight into the atmosphere surrounding school food. In addition, stakeholders from the British Columbia Ministry of Health were always available to respond to random queries that emerged from my research, helping to clarify and verify.  I am sincerely grateful for the school districts who allowed me access to gather data in their district as well as the time given by school administrators. I recognise how challenging it is for school staff to find time to participate in research given the daily burdens of running schools. In short, I was grateful for all stakeholders who agreed to spend time with me in person or over the phone and who willingly shared their experiences and perspectives.  This study depended on two major sources of funding: the Canadian Institutes of Health Research Frederick Banting and Charles Best Canada Graduate Scholarships Doctoral Award (2015-2018), supplemented by the University of British Columbia’s 4-year fellowship (2014-2018). Not only did this funding support the actual work but it also enabled me to have wonderful experiences attending and presenting at a variety of international conferences and the space to have valuable life experiences while writing this dissertation.  Finally, special thanks are owed to my friends and family. Their unceasing support inspired me and their belief in my abilities remained constant and pulled me through when my belief in myself wavered. And to WH for giving me the opportunity to experience the full extent of my emotional resiliency and strength.1  Chapter 1 ~ Introduction   The current global food system is a complex and interconnected set of biological, economic, political, and social processes (Gamboa et al., 2016). A recent report from the Food and Agriculture Organisation defined food systems, generally, as:  ...all the elements (environment, people, inputs, processes, infrastructures, institutions, etc.) and activities that relate to the production, processing, distribution, preparation and consumption of food, and the outputs of these activities, including socio-economic and environmental outcomes. (High Level Panel of Experts [HLPE], 2017, p.11)  While heterogeneity exists as to what constitutes a food system in a given locale, this dissertation is concerned with a modern food system. This is a system in which there is a wide array of food options throughout the year, and processing and packaging is used to extend the shelf life of food. Markets can be easily accessible in high-income areas while lower-income areas may have less access to markets that provide nutritious options. Staple foods are less expensive than animal-source foods, fruits and vegetables, and other specialty foods like organic or locally sourced items. Customers’ access to information about products is promoted through detailed labeling. Measures to ensure foods and beverages are safe to consume are monitored and enforced. In addition, there is a vast and reliable storage and transport infrastructure that ensures the cold chain is unbroken when necessary (HLPE, 2017).  Moreover, production of staples, produce, and livestock within a modern food system tends towards intensive, large scale monocultures, subsidised for over-production (Patel, 2008; Pollan, 2007). This system typically involves globalised trade of food as a commodity which not only increases transportation needs, and therefore greenhouse gas emissions, but also contributes to perpetuating social and economic inequities between populations (Clapp & Cohen, 2009; Holt Giméénez & Shattuck, 2011). These characterisations are now widely acknowledged as not only environmentally and socially unsustainable but also a detriment to human (and other global occupants’) health (Fanzo, McLaren, Davis, & Choufani, 2017; Kickbusch, 2011; Lang, Barling, & Caraher, 2012; Rojas et al., 2011).   2  One way in which the modern food system has contributed to poor human health outcomes is the emergence of a food supply dominated by ubiquitous, low nutrient energy dense (LNED), highly subsidised, and inexpensive food products along with concentrated corporate power over how food is produced, and what information is provided to consumers (Food and Agriculture Organization of the United Nations [FAO], 2013; McRae, Szabo, Anderson, Louden, & Trillo, 2012; Moss, 2013; Patel, 2008; Stuckler & Nestle, 2012; The World Bank, 2017a). A recent review of data from 195 countries found the increasing incidence of diet-related morbidity, attributed to this food system, is associated with multiple leading causes of preventable death in the world (Afshin et al., 2017). Therefore, ‘fixing the food system’ has become one of the most important foci of public health efforts globally (FAO, 2013; HLPE, 2017; Kickbusch, 2011; World Health Organization [WHO], 2017a).  The public health problems associated with the modern, dominant food system are “wicked problems” (Rittel & Webber, 1973, p. 160). Wicked problems are:  ...a class of social system problems which are ill-formulated, where the [available] information is confusing, where there are many clients and decision makers with conflicting values, and where the ramifications in the whole system are thoroughly confusing...[such that] proposed ‘solutions’ often turn out to be worse than the symptoms. (Churchman, 1967, p. B141)   Addressing these types of complex problems requires complex interventions (Wong et al., 2016). Public health policy initiatives related to addressing the challenges created by the modern food system have been introduced at global, regional, national, and local levels. These range from high-level strategy documents, legislation, subsidies, disincentives, mandates, standards, guidelines, and recommendations (Lang et al., 2012). These interventions often have multiple, linked components delivered to individuals or communities where success of the intervention is dependent on how individuals respond to the intervention and the broader contexts in which they are launched.  Interventions addressing school food and beverage environments are a ‘family of public health intervention’ that have become increasingly prevalent across the globe.  Some examples from western nations include: farm-to-school programming to increase access to fruits and vegetables 3  (Bontrager Yoder et al., 2014), garden and cooking programs (Jaenke et al., 2012), more comprehensive school food systems change aiming to address all dimensions of the food cycle from production to food literacy to dealing with waste (Rojas et al., 2011), or re-arranging the options in a cafeteria to make the healthier choices the default choice (i.e., nudges) (Thaler & Sunstein, 2009). Nutritional criteria to improve the quality of foods and beverages available for sale in public schools are also a common example of a complex public health policy intervention used in a number of jurisdictions (e.g., Centers for Disease Control and Prevention [CDC], 2012; Dick et al., 2012; Government of Ontario, 2017). Evaluations measuring the influence of enacting such policies on the consumption practices and body mass indexes (BMIs) of school-aged children have found a variety of results, including changes in both expected and unexpected directions, as well as no significant changes at all (Chriqui, Pickel, & Story, 2014; Fung, McIsaac, Kuhle, Kirk, & Veugelers, 2013; Sanchez-Vaznaugh, Sanchez, Crawford, & Egerter, 2015).   While the mixed evidence regarding child-related outcomes of food and beverage sales policies in schools may be a product of the challenges associated with measuring these associations, it may also be a product of polices not being fully implemented: evidence shows that an important proximal implementation outcome of school food and beverage sales policies, to increase availability of healthy options and reduce or eliminate unhealthy ones, is rarely fully achieved. While the existence of a policy can contribute somewhat to achieving these implementation outcomes (Dick et al., 2012; Han-Markey et al., 2012; Hills, Nathan, Robinson, Fox, & Wolfenden, 2015; Kubik et al., 2010; Larson, Davey, Hoffman, Kubik, & Nanney, 2016; Mozaffarian et al., 2016; Ohri-Vachaspati, Turner, & Chaloupka, 2012; Phillips et al., 2010; Watts et al., 2014; Whatley Blum et al., 2007), the literature shows a wide variation in levels of compliance to government policies in western contexts (e.g., Dick et al., 2012; Gorski et al., 2016; Nanney, Davey, & Kubik, 2013; Taber et al., 2015). This body of literature will be elaborated on in Chapter 2. Therefore, because of the variation in compliance observed in different contexts, it can be inferred that expected implementers, in some contexts, are not taking steps to implement while in other contexts expected implementers are taking steps. So why is this case?   A number of qualitative studies have aimed to explore barriers and facilitators to implementation of school food and beverage sales policies (e.g., Government of Manitoba, 2015; MacLellan, Holland, Taylor, McKenna, & Hernandez, 2010; Nollen et al., 2007), which will also be elaborated on in 4  Chapter 2. These studies undoubtedly provide important insight about specific factors that may be influencing implementation efforts. However, they typically present facilitators and barriers as discrete components. This neglects the complexity of social processes shaping implementation: the relationships between identified facilitators and barriers, implementation outcomes, and the role of context in shaping these relationships. In other words, these types of studies typically create “catalogues” of implementation barriers and facilitators as opposed to “configurations” capable of explaining why implementation happens the way it does (Pawson, 2013, p. 21).  Aim of this study Complex interventions to address the “wicked” public health problems associated with the modern food system have been launched in many different jurisdictions around the world. One example of these are school food and beverage environment interventions, such as policies addressing food for sale in schools. It has been argued that to have more effective and consistent implementation of these types of important interventions, a deep understanding of implementation processes is necessary Wong et al., 2016). Input-output evaluations of school food and beverage sales environment interventions in many contexts have found that they are never implemented 100% of the time, in all places, and for all people. Qualitative studies exploring barriers and facilitators to implementation of these types of policies provide a foundation of insightful information to build upon. However, these studies often neglect to provide in-depth explanations of how barriers and facilitators lead to expected and unexpected implementation outcomes. This study, then, aims to begin to fill the gap in this body of important literature by articulating the social processes (i.e., configurations) influencing implementation of school food environment interventions. Given the complexity of public health policy interventions to address school food and beverage environments, and the need to deeply understand the nature of these initiatives (The PLOS Medicine Editors, 2013; Wong et al., 2016), evaluations at multiple-levels are required--high-level examinations of the effects of multiple initiatives, but also in-depth evaluations of specific interventions being implemented in specific contexts. This dissertation takes the latter approach. 1.1 School food and beverage sales policy in British Columbia, Canada This study will focus specifically on evaluating the implementation of a food and beverage sales policy in British Columbia (BC), Canada. This food environment intervention has the hallmarks of complexity: multiple stakeholders working at multiple levels and long and interconnected chains of implementation activities. As one of the first provinces in Canada to launch this type of intervention 5  over 12 years ago, the BC case is useful to explore given that the intervention theoretically would have had time to become institutionally embedded and fully implemented (Sabatier, 1988). In Canada, the province of BC has been one of the leaders in school food policy.  Previous to 2005, there was growing sentiment among dietitians, school teachers, and parents that action was required to improve school food environments, with a specific focus on eliminating sugar-sweetened beverages (SSBs) from vending machines (Plant, 2004). Dietitians from one of BC’s 5 Regional Health Authorities (RHAs) embarked upon media advocacy campaigns (e.g., McDonald, 2003; Plant, 2003; Richards, 2003) as well as a multi-year school food policy pilot project with schools within their catchment area (Interior Health Authority & The Vancouver Foundation, 2004). The intentions and efforts of this pilot intervention were greatly bolstered by the provincial government when Vancouver won their bid to host the 2010 winter Olympics (Government of British Columbia, 2008). The BC Government publically declared their goal to be the healthiest jurisdiction to ever host them. This public declaration translated into the provision of resources to develop interventions that would help support this goal. This directly resulted in the development of one of the first provincial school food and beverage sales policies in Canada in 2005, which the government committed to review and update every five years.1 BC’s  Guidelines for Food & Beverage Sales in BC Schools (“the Guidelines”)(BCMoH & BCMoEd, 2005) are a nutrient-based criteria aiming to reduce the amount of salt, sugar, and fat in items sold to students in all venues in schools, including vending machines, cafeterias, fundraisers, sports days, and school stores (Government of British Columbia, 2013). The Guidelines were made mandatory in 2008 but no plans were made for official enforcement. While this study will have specific relevance for school-food policymakers and those involved in implementation in BC, it will also have broader relevance. This dissertation will contribute to a broader understanding of social processes associated with implementation of complex public health policy interventions related to food environments more generally. It will provide a more nuanced understanding of the links between implementation barriers and facilitators, why implementers choose to take action or not, and how context influences this relationship. Furthermore, school food and beverage sales interventions have subsequently been launched in numerous jurisdictions around the world (e.g., Antilla, Rytkonen, Kankaanpaa, Tolvanen, & Lahti, 2015; CDC, 2012; Dick et al., 2012; S. W. van den Berg, Mikolajczak, & Bemelmans, 2013), making it part of  the family of school food and beverage environment intervention family (Pawson, 2016). While school food and                                                           1 In conversation with government employees involved in early development of the Guidelines. 6  beverage sales interventions may manifest differently in different contexts, they are founded upon the same program theory. Program theories, generally, are plausible explanations about how a program or intervention is expected to work (Bickman, 2000). While numerous implementation studies have been conducted at specific local levels, little has been done by way of understanding the program theory underlying school food and beverage sales environment interventions and how pre-existing dimensions of context, more broadly, interact with the intervention to influence implementation. Approaching implementation evaluation with the aim of elucidating broader conceptualisations of the inner workings of an intervention can provide guidance around how the same intervention may be successfully implemented in different contexts. 1.2 Organisation of the dissertation This dissertation follows a manuscript-style format. Chapter 2 provides a contextual framing for the research, situating it within public health policy theory, the extant literature examining implementation of school food sales policies as part of the school food environment ‘intervention family’, and the realist approach to implementation evaluation. Following this, in Chapter 3, I present the philosophical and methodological underpinnings of this work, realist philosophy of science and realist methodology, as well as a conceptualisation of the overall research process. As this dissertation is publication-based, to limit redundancies, the specific details of data collection and analysis are incorporated into each of the three findings chapters, Chapters 4, 5, and 6. In Chapter 4, I first present the development and articulation of a retrospective logic model and program theory for BC’s school food and beverage sales guidelines. The logic model outlines inputs, activities, outputs and outcomes specific to the BC case.  The program theory articulates the underlying theory of what was expected to drive expected implementers in BC to take action and create healthier school food and beverage sales environments. From this, a broad program theory for school food and beverage environment interventions is created. This broad program theory guided the lines of inquiry for the investigations presented in Chapters 5 and 6. These chapters examine the relationships between context and how expected-implementers respond to the intervention and how and why it drives them to engage in implementation activities or not. In other words, I describe the social processes related with implementation of this policy. It became clear during the analysis that different social processes are occurring at the district level compared with the school level. Therefore the findings of this phase of the research are presented in two chapters: Chapter 5, which focuses on the district level, and Chapter 6, which focuses on the school-level. Both of these chapters provide insight into how important intervention resources, depending on the contexts, can lead to different responses from 7  expected implementers. This can lead to some implementers choosing to engage in implementation activities, in different ways, while others choose not to engage. I propose that it is these variations in decisions to engage in implementation activities, as influenced by context that can lead to inequitable implementation of this public health policy. Finally, in Chapter 7, I step back from the microscope under which I put the case of BC’s policy and reflect on social processes of implementation more broadly, how they differ from the hypothesised broad program theory developed in Chapter 4, and the implications of these differences for refining the program theory and for public health research and practice. Additionally, Chapter 7 offers suggestions for future research and a detailed limitations section which discusses how the use of a realist perspective is a radical one that can challenge ways of thinking about conventionally-presented limitations of qualitative research.8  Chapter 2~ Public health policy to improve school food environments and a realist approach to implementation research  The aim of this chapter is to provide a contextual framing for this research by situating it within the existing literature related to implementation of school food sales policies and a realist approach to implementation evaluation. It also provides conceptual tools by first reviewing relevant theory related to public health policy and food environments. I then discuss food environment interventions, focusing on schools as a public institution often targeted by interventions and the challenge of equitable implementation. Following this, I present background on the specific focus of this dissertation, school food and beverage sales policies, and existing implementation literature from British Columbia and other international jurisdictions. A justification for why implementation research, using a realist evaluation approach, can help provide insight into implementation processes is provided. Finally, I outline the specific objectives of this research and the empirical, theoretical, and methodological contributions.  2.1 Public health policy and food environments Food environments, generally, have been proposed to be one of three core constituents of a food system, along with food supply chains and consumer behaviour (HLPE, 2017). The High Level Panel of Experts for the FAO (2017) define food environments as “the physical, economic, political and socio-cultural context in which consumers engage with the food system to make their decisions about acquiring, preparing and consuming food (p. 28). It has been proposed that the most important elements of food environments influencing the choices consumers make are physical and economic access to food, the way in which food is promoted and advertised and what information is provided to consumers, and food quality and safety (Caspi, Sorenson, Subramanian, & Kawachi, 2012; Hawkes et al., 2015; Swinburn et al., 2013).  Public policy is defined as “‘a set of interrelated decisions taken by a political actor or group of actors concerning the selection of goals and the means of achieving them’” (Jenkins, 1978 in Howlett, Ramesh, & Perl, 2009, p. 6). Therefore, public policy is characterised, first and foremost, as actions taken by government to address social challenges (Howlett et al., 2009; Lang et al., 2012). Public health policy, then, is action taken by government actors to address preventable morbidity and 9  mortality by addressing upstream determinants of health. Public health policy is distinct from health service/ health care policy which targets primary health care and treatment of morbidities.    Public health policy is a type of complex social intervention. Pawson (2006) describes seven characteristics of social interventions. First, interventions are the manifestation of theories in that they are created as a result of beliefs about how a certain type of action or approach will lead to desired outcomes. Next, social interventions are considered “active” in that whether they ‘work’ or not depends upon the reasoning processes of individuals who are meant to implement or who are the intended beneficiaries. The third characteristic of social interventions is that implementation chains are long and involve many stakeholders who, as they are asked to move an intervention forward, may make the decision to either sustain or undermine earlier implementation processes. These decisions can either support implementation or create opposition further down the chain. In addition, Pawson (2006) characterises these long implementation chains as non-linear and sometimes reversing directions, whereby a top-down intervention becomes bottom-up, for example, and creates feedback loops. Next, the creation and implementation of social interventions are embedded in multiple layers of social systems and, as such, it should be expected that the same intervention will achieve different levels of success and failure when introduced into different contexts. In addition, interventions and their program theories are “leaky” (p. 32). This means that the manner in which a complex intervention is actually implemented, in different contexts and by different individuals, will likely veer from the official and expected underlying program theory.  Lastly, social interventions are open systems and therefore not only are they influenced by social context but they also can change it. Changes anywhere along implementation pathways may be affirming, in that they promote further implementation efforts. They may also be defeating whereby further implementation efforts are stifled.  Public policy, and therefore public health policy, is carried out through the use of public policy instruments, or “the set of techniques by which government authorities wield their power in attempting to ensure support and effect social change” (Bemelmans-Videc, Rist, & Vedung, 1998, p. 3). Vedung (1998) devised a concise typology of the types of instruments governments use: economic means, regulations, and the provision of information and knowledge. In other words, all government policy instruments are either carrots, sticks, sermons or a combination of these. This typology is based on “the degree of constraint, or even better, degree of power, that the governing 10  body has invested in the governance attempt” (p. 35). Regulations are directives, rules, and mandates that remove the choice of whether people change their behaviour and non-compliance is typically met with sanctions, although not always. These are, in this dissertation, considered ‘strong’ interventions. Economic policy instruments involve attempting to change behaviour through changing the distribution of material resources, either money or in-kind, whereby these resources are distributed but with ‘strings attached’. People may choose not to change their behaviour and will consequently not receive resources (in the case of incentives) or will lose resources (in the case of tools like pricing disincentives). On the other hand, potential recipients may change their behaviour and receive resources or keep their existing resources. Ultimately though, potential beneficiaries are allowed to choose whether or not to change their behaviour. Lastly, the least authoritative policy instrument, ideally absent of coercion, is “moral suasion, or exhortation...through the transfer of knowledge” (p. 33). Information can consist of “objective, correct knowledge” (p. 33) and it can also involve spreading information underpinned by value judgements about right and wrong behaviours. Within each of these types of public policy instruments are a range of more specific tools that involve varying degrees of authoritative force. For example, under the umbrella of regulations, the most authoritative of the three policy instruments, there are specific types of regulatory tools that are more or less authoritative.  Sisnowski, Street, and Merlin (2017) conducted a review to explore policy success of regulatory interventions aiming to improve food environments since 2004. Reviewing evidence across menu-labelling interventions, physical access to food, subsidies for healthy food, taxation of unhealthy items, procurement standards for public institutions, and nutrition labelling, they found that most interventions show some success at achieving proximal outcomes such as changes in purchasing behaviours or increases in the availability of healthier foods and beverages. However, in their review of the literature around these types of food environment interventions, the authors found that these interventions have yet to lead to significant influence on consumption practices that might result in wider impacts on population-level health status. The authors suggest, though, that because the food environment interventions reviewed in their study were implemented at all shows that they are feasible and acceptable policies. Rather than discarding interventions because expectations of policymakers (like lower prevalence of overweight and obesity) have not yet been met, they recommend policymakers “pursue the example of these jurisdictions as necessary building blocks for more stringent and comprehensive nutrition policy and obesity prevention regimes” (p. 12) (e.g., not 11  only regulating that menus must contain nutrition information but also regulating menu item ingredients through policies like trans fat bans). Also, regardless of the evidence-base not showing improved population-level health outcomes as a result of regulatory food environment interventions, the precautionary principle can be invoked to support the continued practice of ensuring food environments consist of accurate information for consumers, affordable options, and an array of items that are known to be nutritious.  Many food systems and health scholars still believe that not only will health be improved by making the healthy choice the easy choice, but interventions addressing higher order levels of food environments (e.g., public sector institutions) through structured demand processes can help generate larger food systems change (L. M. Ashe & Sonnino, 2012; Nehring, Miranda, & Howe, 2017; Renting & Wiskerke, 2010; Rojas et al., 2011; Sonnino, Torres, & Schneider, 2014) as opposed to, for example, interventions addressing household food environments. Specifically, they argue that the procurement resources of public sector institutions can create sufficient demand from industry for a certain type of product, so much so that industry changes the way it conducts business and thus the larger food system is changed. Numerous examples of interventions targeting public sector food environments are reported in the literature, including in recreation centers (Olstad, Poirier, Naylor, Shearer, & Kirk, 2015; Vander Wekken, Sorensen, Meldrum, & Naylor, 2012), hospitals (J. Miller, Lee, Obersky, & Edwards, 2014), public health facilities (Kuo et al., 2016), universities (Friedmann, 2007), and schools (Harris, Lott, Lakins, Bowden, & Kimmons, 2012; Otsuki & Arce, 2007; Sumberg & Sabates-Wheeler, 2011).   2.2 Individual behaviour change and ecological health interventions addressing nutrition Public policy scholars have proposed that government decisions around what type of public policy instrument to use are related to a governing body’s policy style, or the political culture and the values of state (Howlett et al., 2009; Linder & Peters, 1989). By this, I mean that the existing values underpinning government decisions and actions, and how they frame a social problem, influence what policy instruments are used. Consequently, the use of different policy instruments to address what are essentially the same challenges is often observed in different nation-states with different ideologies (Bemelmans-Videc et al., 1998). Public health interventions are a clear example of how the framing of a public health problem by a governing body leads to the choice of certain 12  intervention methods over others. Here, I discuss two main approaches typically used for addressing public health challenges related to nutrition: individual behaviour change and ecological approaches.  The first approach, behaviour change, aims to promote the prevention of diet-related morbidity and mortality by targeting the behaviours of individuals, or changing their food choices. Raine (2005) outlines a number of determinants of individual food choices that may be targeted including, food preferences, nutrition knowledge, perceptions of healthy eating, and psychological factors. Prevention efforts at the level of individual behaviour change often take the form of education campaigns whereby the intervention consists of providing information about nutrition to convince people to change their behaviour. Some nutrition-related health promotion initiatives to change individual behaviour include, for example, education, training, and home visits targeting parents to support the nutritional health of their children (Golley, Hendrie, Slater, & Corsini, 2011) or use of mobile technology to track daily nutrition activities (Bacigalupo et al., 2013). These types of interventions are characterised by the promotion of individuals to take responsibility for their health behaviours and are typically founded upon health behaviour theories like, for example, the health belief model (Glanz & Rimer, 1997).    The health belief model (HBM) is a middle range theory (MRT), a type of theory that is broad enough to be used in a variety of different contexts but is specific enough in the articulation of its constructs (Pawson, 2007). The HBM proposes there are a set of beliefs, founded on psychological theories, that an individual can hold that will predict or explain their health-related behaviours whether preventive, detrimental, or curative (Maiman & Becker, 1974; Rosenstock, Stretcher, & Becker, 1988). These beliefs, initially, included perceived susceptibility and severity, perceived benefits of taking action, perceived barriers to taking action (Janz & Becker, 1984). Perceived susceptibility and severity related to how vulnerable a person believed they were to a health condition and beliefs about how serious the consequences are of contracting an illness or leaving it untreated (Janz & Becker, 1984). Perceived benefits relate to how much an individual believes the proposed intervention will help prevent or cure their illness (Janz & Becker, 1984). Finally, perceived barriers relates to an individual’s cost-benefit analysis of deciding to take action to prevent or treat illness. Rosenstock (1974) proposed that the “combined levels of susceptibility and severity provided the energy or force to act and the perception of benefits (less barriers) provided a preferred path of action” (p. 332). Later the dimension, “cue to action”, was added to the HBM and this 13  referred to any sort of internal cue, like a symptom, or an external cue, like mass media communications (Janz & Becker, 1984). Rosenstock et al. (1988) proposed incorporating one more dimension, the confidence in one’s ability to take action, for a total of six components of the HBM. Not only has the HBM been used to help predict and explain individual health behaviours, it has also been used to provide a conceptual scaffold to influence health behaviours (Rosenstock et al., 1988). In particular, it has been used to inform health behaviour change interventions by gearing messages (cues to action) towards individuals that attempt to increase their perceived susceptibility, severity, benefits of taking the proposed action, and their sense of self-efficacy.  The health belief model is one example of a behaviour change theory used to inform health behaviour change interventions (Glanz & Rimer, 1997). This individualistic approach to improve health maintains a focus on personal responsibility and is embedded in neo-liberal ideology that seeks to move the responsibility of care from the public domain (welfare) to the personal domain (self-help) (Guthman, 2009). In response, a growing body of literature critiques individual behaviour change interventions for both their neglect of structural influences on health behaviours and their limited success to achieve population level improvements to health (Garner & Wooley, 1991; Health Canada, 2013; Hill & Peters, 1998; Vargas-Garcia et al., 2017).  Along these lines, a second approach to improving health, the ecological approach, uses an ecological theory of health. Similar to the HBM, this is also a middle range theory. Interventions underpinned by an ecological theory of health aim to prevent diet-related morbidities and mortalities through initiatives that change the context in which people make decisions about their health. This is because it is now widely acknowledged that determinants of individual food and beverage consumption practices are complex and include both behavioural and structural factors (Glanz & Bishop, 2010; Glanz, Rimer, & Viswanath, 2008; Health Canada, 2013; Hill & Peters, 1998; Kumanyika, 2017; Penney, Almiron-Roig, Shearer, McIsaac, & Kirk, 2014):  …the choices people make in relation to food are, to a large extent, dictated by the choices they have. While individual characteristics, such as taste preference, and innate appetite and satiety responses, are clearly important, other factors external to the individual, such as food cost and availability, are increasingly being recognised as important determinants of dietary intake. (Penney et al., 2014, p. 226) 14   In the 1980s, calls for public health approaches to become more ecologically-leaning led to the development of an ecological model of health (World Health Organization [WHO], 2017b). This view of how health status is created and perpetuated focuses on both ecological factors influencing health outcomes in terms of the natural environment and also in terms of social, cultural, and economic patterns of societies (Kickbusch, 1989; McLeroy, Bibeau, Steckler, & Glanz, 1988), also known as the socioecological factors (WHO, 2017b).   Cohen, Scribner, and Farley (2000) built on this conceptualisation by connecting ecological factors with individual behaviour change by proposing a structural model of health behaviour and expanding upon this model’s implications for public health interventions. Cohen et al.'s (2000) model consists of four broad factors: (1) availability and accessibility of consumer products, (2) physical structures (or physical characteristics of products), (3) social structures and policies, and (4) media and cultural messages. The difference between the implications of this model for interventions, and the interventions implied by the individualistic view of health, is that interventions that aim to address any of the above four factors are not targeted at specific individuals but rather aim to influence the health behaviours of entire populations (Cohen et al., 2000). Some examples of nutrition interventions reflecting this structural model of health behaviour include: recommendations that schools rearrange the offerings in a school cafeteria so the healthier choices become default choices for students (Thaler & Sunstein, 2009), nutrition labeling regulations (Baltas, 2001), or providing food pricing subsidies for store owners in geographically hard-to-reach areas such as the Canadian north (Government of Canada, 2014). These types of interventions assume that dietary behaviour change potential is maximised when people are both individually motivated to make changes to their diets and the environment provides a context, through policies, social norms and social support, in which healthy choices can be made (Glanz et al., 2008). In other words these interventions aim to, as the mantra has become, make the healthy choice the easy choice (M. Ashe, Graff, & Spector, 2011; Kahn-Marshall & Gallant, 2012; Koelen & Lindstrom, 2005; Volpp & Asch, 2017).  Story, Kaphingst, Robinson-O’Brien and Glanz (2008) proposed a specific ecological framework for conceptualising the complexity of factors that determine what people eat. These include individual factors, the social environment, macro-level food environments, and the physical food environment. 15  Individual factors may be biological, cognitive frames of references, skills, and/or socio-economic demographics. The social, or intrapersonal, environment refers the food environments created by family, friends, and peers (Raine, 2005; Story et al., 2008). Macro-level food environments are systematically created through social and cultural norms, the activities of the food and beverage industry, economic and political systems, to name a few.  Regarding the physical environment, it is sometimes equated with geographic food access but also with whether the kinds of foods that are available (if there is food available) are affordable and nutritious (Health Canada, 2013). Countless interventions aimed at addressing physical food environments are reported on in the literature, in homes (DeSmet, Liu, De Bourdeaudhuij, Baranowski, & Thompson, 2017), workplaces (Schwarte et al., 2010), hospitals (Moran, Krepp, Curtis, & Lederer, 2016), neighbourhood recreational centers (Olstad et al., 2015), restaurants (Sisnowski et al., 2017), convenience and corner stores (Paek et al., 2014), and supermarkets (Adam & Jensen, 2016). 2.3 School food and beverage environment interventions Schools have become a popular institutional physical food environment in which to intervene. For one, this is because of the possibility for institutional and thus systems change (Morgan & Sonnino, 2010; Rojas et al., 2011). However, this is also because of concern for the increased incidence of obesity and overweight in children over the past decades in a number of western contexts (Rokholm, Baker, & Sorensen, 2010) and the predicted health care costs and loss of economic productivity this will create (Wang, Beydoun, Liang, Caballero, & Kumanyika, 2008; Withrow & Alter, 2011). School food environments, in many western nations, have evolved in a manner that reflects the evolution of the modern food system in the last sixty years, a system that has led to a food environment in which processed, convenient (Morgan & Sonnino, 2010; Poppendieck, 2010), addictive (Gearhardt, Grilo, DiLeone, Brownell, & Potenza, 2011; Moss, 2013), and low-nutrient energy dense foods (Poppendieck, 2010) are common. One American academic has even referred to public schools in the United States as “a 7-11 with books” (Kelly Brownell in Soechtig,2014). Moreover, educational systems, for better or worse, have long been effective sites of social and ideological perpetuation (Illich, 1971) or change (Deal, 1985) through the influencing of a young, captive audience. It is also widely accepted that undernourished students are more likely to underperform academically (Boschloo et al., 2012; Nyaradi et al., 2014; Olson, 1999; St. Leger, Kolbe, Lee, McCall, & Young, 2007; Taras, 2005) and exhibit behavioural challenges (Olson, 1999) so it is in the best interests of children and school-level staff that nutritious options be available at school. Lastly, students spend a large portion of their waking lives in school, during which time they 16  consume upwards of 35% of their daily calories (Briefel, Wilson, & Gleason, 2009; Tugault-Lafleur, Black, & Barr, 2016). Given all this, policy and programmatic ecological interventions to promote salutogenic school food and beverage environments have been recommended as strategies to improve students’ individual consumption practices (Bundy et al., 2009; CDC, 2012; Public Health Agency of Canada [PHAC], 2013; World Health Organization [WHO], 2008; World Health Organization [WHO], United Nations Children’s Fund [UNICEF], & United Nations Educational, Scientific and Cultural Organization [UNESCO], 1991). 2.3.1 School food and beverage sales standards One popular public health approach, and important component of the food policy family, is the introduction of nutritional criteria for food provided in and by schools during the school day (i.e., not food brought from home) (Hawkes et al., 2015; Lang et al., 2012). Depending on the jurisdiction, nutritional criteria have been applied to subsidised school meal programs (e.g., ActNowBC & the Government of British Columbia, n.d.; United States Department of Agriculture & Food and Nutrition Service, 2016), to food and beverages being sold during the school day and school events (e.g., CDC, 2012; Chumley, 2014; Dick et al., 2012), or to both (e.g., Adamson et al., 2013; Government of Nova Scotia, Education and Early Childhood Development, 2017).   Interventions to regulate what types of foods and beverages are available for purchase during the school day, the focus of this dissertation, have been launched in numerous jurisdictions around the world (Hawkes et al., 2015), including most of the Canadian provinces and territories (Holmes, 2016).  Food and beverage sales in Canadian schools are common. Canada does not have a national school meal program and therefore foods and beverages offered in cafeterias are typically items for sale. Some jurisdictions and/or schools do provide subsidised meals for vulnerable students but this is not the focus of this research. Rather, the focus of this research is on food sold in schools; items sold in cafeterias as well as the fundraising revenue generated from vending machines, food and beverages sold at sports days or special events, or other hot lunch fundraisers. These sales venues are important sources of income to fund activities outside of conventional educational activities (Downs et al., 2012; Masse et al., 2013; Morin, Demers, Gray-Donald, & Mongeau, 2012; Vine & Elliott, 2014). At the heart of the issue with the common food and beverage sales observed in Canadian schools is that items that have been selected to sell to students have often been high in sodium, carbohydrates, and fats (herein referred to as “salt, sugar, and fat”). This is true for cafeterias, 17  vending machines, or snack shops aiming to sustain themselves economically or whether foods and beverages are sold for fundraising activities (Winson, 2008).   The strength of school food and beverage sales policies or guidelines present in Canadian and other international jurisdictions range from a softer paternalism consisting of the language of ‘recommended’ or ‘encouraged’ (e.g., Larson et al., 2016; Quintanilha et al., 2013), to harder forms of paternalism such as mandates or legislations (Dick et al., 2012; Government of Ontario, 2017; Government of British Columbia, 2013; Sanchez-Vaznaugh et al., 2015). In British Columbia, the backdrop of this research, the school food and beverage sales policy is a government mandate but is not accompanied by official accountability mechanisms, making it an example of lex imperfecta, a regulatory policy tool not accompanied by sanctions (Vedung, 1998).   Regardless of the prevalence of the use of school food and beverage sales policies, the evidence from different jurisdictions is not unanimous as to whether the school food and beverage sales environment even influences children’s consumption practices or health outcomes, like overweight and obesity. Terry-McElrath, O’Malley, Delva, and Johnston (2009) used data from a nationally representative sample and found no significant association between availability of unhealthy items in vending machines and overweight or obesity in middle or high schools. Similarly, Van Hook and Altman (2012) found no significant associations between availability of foods and beverages for sale and BMI. A BC-based study did, on the other hand, find a significant association between availability of  SSBs for sale and moderate to high consumption and their consumption was positively associated with obesity (but not overweight) (Masse, de Niet-Fitzgerald, Watts, Naylor, & Saewyc, 2014). In the US, Nanney et al. (2016) also found that availability of SSBs in schools was associated with increased consumption and availability of unhealthy snacks for sale was associated with a small significant increase in BMI. Another older study compared nutrient intakes between two groups of US-students enroled in the National School Lunch Program, those students who did not purchase additional food and beverage options and those who did purchase additional items. They found the latter consumed significantly more calories, more fat, and lower intakes of a macronutrients (Templeton, Marlette, & Panemangalore, 2005). Vericker (2013) conducted a longitudinal analysis to explore whether having access to foods and beverage sales in schools was associated with adolescent consumption practices. They found that, on average, having access to food and beverage sales was not associated with consumption of SSBs. However, when access to 18  food and beverage sales was gained by 8th grade males, consumption increased significantly by 20.7% compared with males who have no access to sales. Low-income youth who gained access to SSB sales in 8th grade consumed SSBs over 60% more times per week than their other 8th grade counterparts. In a review of US studies, Larson and Story (2010) conclude that students have better diets when unhealthy foods and beverages are not available for sale.   This array of findings regarding the general influence of availability of foods and beverages for sale on consumption and health outcomes, presented in the previous paragraph, is reflected in findings from other specific evaluations assessing whether the introduction of school food and beverage sales policies change consumption practices or influence health outcomes. An evaluation from the Canadian province of Nova Scotia showed an overall reduction in caloric intake and reductions in the consumption of LNED foods between 2003 and 2011, which coincided with the enactment of a provincial policy in 2006 (Fung et al., 2013). A US-based systematic review conducted by (Chriqui et al., 2014) found that evidence associating the enactment of food sales policies with improved consumption practices of students was mixed. Six of the 11 studies reviewed found student consumption practices improved while five showed a mix of expected, unexpected, and non-significant results. This same study also reviewed evidence for associations between policy and BMIs or weight status. Again, they found mixed results, with two of the four evaluations reviewed finding that enactment of policy was associated with lower BMIs and two others finding a mix of expected, unexpected and non-significant results (Chriqui et al., 2014). Finally, a more current California-based study found that while, on average, BMIs decreased in the post-policy period, only students attending schools in high-income neighbourhoods experienced this benefit while students attending schools in poorer neighbourhoods had BMIs that remained the same (Sanchez-Vaznaugh et al., 2015). 2.4 Implementation evaluations of school food and beverage sales policies in BC and beyond The above section presents findings similar to those found in Sisnowski et al.’s (2017) review of food environment interventions presented earlier: there is limited evidence to show if and how food environment change, as a result of policy, influence specific consumption or health outcomes. This has also been found in evaluations from BC as well as from jurisdictions with food and beverage sales policies.   19  As BC’s school food and beverage sales policy was first being rolled out in 2005, the British Columbia Ministry of Education (BCMoEd) conducted a baseline assessment of availability and types of food and beverage sales that were occurring (British Columbia Ministry of Education [BCMoED] & British Columbia Ministry of Health [BCMoH], 2005). A follow-up study conducted in 2008 showed that progress had been made regarding elimination of SSBs from vending machines, but the presence of unhealthy foods in vending machines remained prevalent (BCMoEd & BCMoH, 2008). A later study by Watts et al. (2014) found that administrators reported significantly lower odds of having a number of important unhealthy food ‘culprits’ available in 2011/12 compared with 2007/08, and significant improvements were reported by elementary administrators regarding availability of fruits and vegetables. This was in contrast to middle and high school administrators who reported no change in availability of fruits and vegetables. In general, by 2011/12, 66% of middle and high school principals reported full implementation of the policy in vending machines, and only 45%, 36%, 10%, and 8% reported snack bars, cafeterias, fundraising events, and special events, respectively, were fully compliant.   As is the case with BC-based evaluations, literature from other jurisdictions generally shows that the existence of a school food and beverage sales policy contributes to improved school food environments. Yet in no study can it be found that compliance is occurring 100% of the time in all places. Some studies have explored these differences in compliance between schools with different geographies (e.g., rural vs urban vs suburban). Stakeholders from urban schools report significantly higher levels of implementation success, or compliance, than rural schools (Dick et al., 2012; Hills et al., 2015; Pettigrew, Pescud, & Donovan, 2012a). Urban and suburban schools offer more fresh fruits and vegetables for sale compared with rural schools (Nanney et al., 2013). Other studies have compared implementation between school types (elementary vs. secondary, for example, or public vs. private schools). One such study found that secondary schools have higher levels of compliance than middle schools (Gorski et al., 2016). Other studies have found primary schools with higher levels of compliance than secondary schools (Dick et al., 2012; Kubik et al., 2010) and public schools with higher levels of compliance than private schools (Hills et al., 2015). The influence of school size has also been explored, where small schools have shown lower levels of compliance than large schools (Hills et al., 2015; Kubik et al., 2010). The influence of school socio-demographics and economic status has also been explored. One US study found that schools with low proportion of students enrolled in the subsidised lunch program (i.e., wealthier student population) were 20  significantly more likely than both those with middle and high enrolment (i.e., poorer student populations) to have LNED foods available for sale (Nanney et al., 2013).  Other studies have found a significant positive correlation between the socioeconomic status (SES) of schools and level of compliance with their nutrition sales standards (Hills et al., 2015; Nanney et al., 2013; Ohri-Vachaspati et al., 2012; Taber et al., 2015).  A number of the above studies have also found non-compliant food to be widely available even after implementation efforts are made (Gorski et al., 2016; Kubik et al., 2013; Kubik, Davey, MacLehose, Coombes, & Nanney, 2015; Samuels, Hutchinson, Craypo, Barry, & Bullock, 2010; Whatley Blum et al., 2007), with fluctuating levels of compliance from year to year (Gorski et al., 2016).  The level of compliance to a policy is related to whether or not expected implementers choose to take action to implement. The quantitative implementation studies measuring compliance are limited in how much they can reveal about why implementation of school food and beverage sales policies happens (or does not happen) the way it does and how this may be contributing to differential levels of implementation success. For example, studies show that for the most part, a higher level of policy compliance is occurring in urban areas compared with rural areas, meaning there is more successful implementation in urban areas. However, this is not the case 100% of the time and this heterogeneity is a useful starting point for inquiry as to why this is the case and to contribute to a deeper understanding of the social processes related to implementation of this intervention. To use a clarifying metaphor, observing one black swan is enough to disprove the theory that all swans are white and to lead to asking: what is occurring that most of the time swans are white but sometimes they are black (Tilley, 2017)? This observation, articulation, and exploration of heterogeneous outcome patterns can provide more knowledge about a phenomenon (i.e., the inner workings of an intervention).   Only one in-depth qualitative study has been conducted in BC to contribute to explaining the varying levels of compliance with BC’s school food and beverage sales policy (“Guidelines”). Masse et al. (2013) conducted research in 2010/11 to identify barriers and facilitators to implementation across a variety of stakeholders. The policy was perceived to be compatible with schools’ or teachers’ expectations of what the school learning environment should provide, but respondents felt the complexity of the Guidelines led to issues of clarity about their scope (i.e., the foods and beverages to which they apply). Furthermore, struggles to maintain profit margins, as well as the disconnected 21  beliefs about school food between parents and teachers were also perceived barriers to implementation. There was also tension around whether it was families’ responsibility to ensure healthy options or whether it was the schools’ responsibility to not profit from the sales of unhealthy foods for children. Finally, lack of capacity to implement was also reported as a barrier to implementation. Regarding facilitators, Masse et al. (2013) found that one administrator reported the top-down mandate helped them achieve their goals. In addition, the existence of supportive implementation tools from the provincial government, having existing nutrition programming in schools, access to dietitians, and local suppliers that complied with the Guidelines were all specifically reported facilitators that do not appear to be articulated in any other studies of school food and beverage sales policy implementation.   Qualitative work from other western jurisdictions identified a number of similar barriers and facilitators to implementation. One barrier highlighted, similar to the BC experience, is a lack of clarity about the scope of the intervention on the part of expected implementers and insufficient communication about the policies to stakeholders (Government of Manitoba, 2015; MacLellan et al., 2010; Vine & Elliott, 2014). Also reported in other studies is the struggle to maintain profit margins (Dick et al., 2012; Han-Markey et al., 2012; Holthe, Larsen, & Samdal, 2010; Pettigrew, Pescud, & Donovan, 2012b). The paternalistic nature of school food and beverage sales policies was also identified as a barrier to implementation (MacLellan et al., 2010; Nollen et al., 2007; Pettigrew, Donovan, Jalleh, & Pescud, 2013; Pettigrew et al., 2012b; Vine & Elliott, 2014), similar to concerns highlighted in BC around whose responsibility it is to ensure healthy food for children (Masse et al., 2013).  And lastly, generic ‘lack of capacity’ is cited as a key implementation barrier reported across many other contexts besides BC (Dick et al., 2012; Holthe et al., 2010; Kubik, Lytle, & Farbakhsh, 2011; MacLellan et al., 2010; Mincher, Symons, & Thompson, 2012; Nollen et al., 2007; Pettigrew et al., 2013, 2012b; Vine & Elliott, 2014). Regarding facilitators, as Masse et al. (2013) found in BC, another study from Australia identified that the top-down pressure helps support implementation efforts (Pettigrew et al., 2012a). This is in contrast with stakeholders from these and other studies reporting concerns surrounding the paternalistic nature of school food and beverage policies. Lastly, having someone in the school with a high level of interest to implement was reported as bolstering efforts in the Canadian provinces of Alberta (Downs et al., 2012) and Ontario (Vine & Elliott, 2014). Having this kind of ‘champion’ in a school was not specifically identified in the BC-based study.  22  2.5 Potential for inequitable implementation of school food and beverage sales policies Regulation of school food and beverage sales environments, the specific focus of this dissertation, is a type of public health intervention focused on preventing diet-related morbidity. Public health interventions, however, have the potential to be inequitably implemented (MacDonald et al., 2016). This means that, for any type of public health intervention, there is the potential that those contexts in which there are already sufficient resources to implement and where diet-related health indicators may already be relatively good, will do so and their health statuses, more generally, will be improved. On the other hand, populations in those contexts where there are insufficient resources to implement, and where health indicators may be relatively worse, will not benefit as a result of the existence of the intervention. Therefore, the gap between these two contexts may be increased.   Bryant and Raphael (2010) proposed that the most effective way to foster health equity, or reduce the health status gap between members of a population, is to ensure access to social determinants of health, like healthy food environments. In jurisdictions that have adopted school food and beverage sales policies, if they are implemented successfully in all schools, there should be an equal distribution of opportunities to maximise healthy behaviours. However, as demonstrated in the previous section, existing literature from many western contexts show varying levels of success in achieving the key implementation outcome, namely that of increasing healthy options for sale in schools and decreasing unhealthy ones (Chriqui, Turner, Taber, & Chaloupka, 2013; Dick et al., 2012; Kubik et al., 2013; Watts et al., 2014; Whatley et al., 2011). Heterogeneous outcome patterns point to the potential for this public health intervention to contribute to health inequities. Referring back to the implementation research in the previous section, compliance appears to be influenced by contexts that have been identified as important social determinants of health, such as geography and SES (Commission on Social Determinants of Health [CSDH], 2008). Poor levels of implementation and compliance in contexts that already experience barriers to maximising opportunities for health, while higher levels of implementation and compliance occur in contexts that are already health promoting, could exacerbate these inequities. Unearthing why implementation of public health policy, like school food and beverage sales policies, happens sometimes in some places and not others can help inform efforts to either support implementation in under-resourced contexts and/or consider different approaches that may better fit these contexts. Implementation research is useful for understanding implementation processes in-depth to help guide more equitable implementation. 23  2.6 Implementation research and the realist approach It is well-acknowledged that few policies and programs in the ‘real world’ are fully implemented as intended (Damschroder et al., 2009; Durlak & DuPre, 2008; Helfrich et al., 2010) and uptake of health promoting interventions has often proved challenging and slow. As Pressman and Wildavsky (1973) pointed out over forty years ago:  The view from the top is exhilarating. Divorced from problems of implementation, federal bureau heads, leaders of international agencies, and prime ministers…think great thoughts together. But they have trouble imagining the sequence of events that will bring their ideas into fruition. Other men, they believe, will tread the path once they have so brightly lit the way. (p.136)  As a result, public health policy often fails to achieve consistent and sustainable desired short and long-term outcomes, which may not only waste valuable resources but also potentially exacerbate health inequities. Pressman and Wildavsky (1973) were the first implementation researchers to theorise that the inevitable array of heterogeneous intervention outcomes, often disappointing, is due to the complexity of implementation. Leow, Macdonald, Hay and McCuaig (2014) describe this gap between policy intent and implementation as “lost in translation as [policies] migrated from the recontextualising field to the reproduction field” (p. 1008).  If processes of implementation are indeed the crux at which the level of success of an intervention will be determined, there is a strong rationale for the role implementation research can play in helping narrow the gap between the intention of decision-makers and the actions of implementers ‘on-the-ground’. Implementation research is the scientific inquiry into implementation-related concerns. Peters, Tran, and Adam (2013) define implementation research specifically in terms of health policy and systems research as:  …being concerned with the study of clinical and public health policies, programmes, and practices, with the basic intent being to understand not only what is and isn’t working, but how and why implementation is going right or wrong…(p. 27)  24  Peters et al. (2013) present a variety of possible approaches to conducting implementation research. These include: pragmatic or explanatory controlled trials to focus specifically on intervention effects, effectiveness implementation hybrid trials to assess both the effectiveness of the intervention and of implementation strategy used, quality improvement studies that typically work on a plan-do-act-study cycle, participatory action approaches promote community-led interventions, or a mixed-methods approach can be used to understand multiple perspectives and different types of pathways and outcomes. While Peters et al. (2013) include “realist review” as a potential approach, I propose that a realist evaluation approach can also be used to conduct implementation research to explore social processes that influence how and if expected implementers engage in implementation activities.  Realist evaluation is part of the family of theory-based approaches to evaluation which begin with the clarification of theories about how an intervention, in a given context, is expected to lead to outcomes (Westhorp, 2014). The realist approach is useful for evaluating complex social interventions, like public health interventions, that have multiple stakeholder types, involve complex tasks, have various outcomes, contain feedback loops, and are constantly adapting to the context (Craig, Dieppe, & Macintyre, 2008). Realist evaluators assume all social interventions have some level of complexity as they are, in themselves, complex programs launched in complex social contexts to be implemented and taken up by complex individuals (Pawson, 2013). It is this complexity that leads to interventions ‘working’ differently in different settings. To use an illustrative example from BC in the literature presented earlier in this chapter, BC school food and beverage sales policy ‘worked’ to successfully remove SSBs from most vending machines in schools yet were less successful in removing unhealthy foods from vending machines (BCMoED & BCMoH, 2008). In New South Wales, Australia, the introduction of a mandatory healthy school canteen policy ‘worked’ to improve the nutritional quality of what was available for purchase  more effectively in urban areas compared with rural areas (Hills et al., 2015).  A realist evaluation perspective holds that these differences in outcomes in different settings in the examples above (i.e., SSB vending machines versus food vending machines, urban versus rural) are because the mechanisms that are needed for successful implementation are triggered to different extents in these different circumstances (Jagosh et al., 2015)—A mechanism, according to realist evaluation methodology, is conceptualised as the cognitive responses of stakeholders when they are 25  introduced to the intervention resources (Pawson, Greenhalgh, Harvey, & Walshe, 2005). In a realist approach to evaluation, identifying the role of context in shaping the cognitive responses of stakeholders is a key focus of analysis and is a means by which to conceptualise complex social processes associated with implementation.  Realist evaluation is a theory-centric approach because, as a member of the theory-driven evaluation family (Chen, 1990; Pawson, 2013), not only does it begin with a program theory, it has aims of refining program theories. A program theory is a plausible explanation of how an intervention is supposed to work (Bickman, 1987) and is used as an inquiry framework for an evaluation (Lipsey & Pollard, 1989).  Regarding the realist evaluation aim of refining program theories, this refers to taking a program theory as a starting point and to continue to refine it through research to better reflect the reality of implementation of complex interventions. As stated earlier, from a broad perspective, there are actually very few types of social interventions, which I referred to earlier: carrots, sticks, and sermons (Bemelmans-Videc et al., 1998). It is thought that in refining the program theories for these limited types of interventions, devising middle-range program theories (e.g. health belief model or ecological theory of health behaviour) that transcend the highly specific program theories of day-to-day implementation activities, lessons emerge about how interventions that appear to work well in some contexts can be more effectively transferred to a different one (Jagosh, 2017a). A more in-depth explanation of realist ontology and methodology is presented in Chapter 3. 2.7 Purpose of the research As reviewed in Chapter 1 and the earlier sections of this chapter, the food environments that have evolved as a result of the modern food system contribute to increasing rates of nutrition-related morbidity and inequities in access to nutritious food (IPES-Food, 2017). As a result of this complex problem, multiple potentially transformative interventions are needed and school food environments have become a popular institutional-level target (CDC, 2012; Rojas et al., 2011; Winson, 2008). In a number of jurisdictions around the world, including most of the provinces and territories of Canada, one popular policy tool being adopted is establishment of nutritional standards for foods and beverages being sold in schools (Antilla et al., 2015; Dick et al., 2012; (GoO) Government of Ontario, 2017; Samuels et al., 2010). The success of implementation of these interventions, measured in the literature as adherence to the nutrition recommendations, has been found to be highly variable in many jurisdictions and similar barriers and facilitators to implementation have 26  been identified.  Studies aiming to help explain why implementation sometimes happens or sometimes does not happen tend to identify discrete facilitators and barriers to implementation (MacLellan et al., 2010; Masse et al., 2013; McDonnell, Probart, Weirich, Hartman, & Bailey-Davis, 2006; Pettigrew et al., 2012b; Vine & Elliott, 2014) but usually neglect to illuminate the complex relationships between these facilitators and barriers, context, and implementation-related outcomes. While providing an important foundation of insight from which to build understanding, neglecting to explicate links between facilitators, barriers and implementation-related outcomes means the ability to explain why implementation is happening the way it is limited. What is needed is a nuanced understanding of “barriers” and “facilitators” as interactions between contexts and individual implementers’ responses to an intervention that leads them to make decisions about whether to take action to implement policy or not.   The purpose of this dissertation work, then, is to contribute to existing literature exploring the social processes associated with food environment public health interventions. I provide specific analysis of implementation of BC’s school food and beverage sales policy and, in so doing, support a more nuanced understanding of this complex intervention. A realist approach to implementation research provides a logic of inquiry which helps articulate social processes and explain how they relate to implementation outcomes.   The specific objectives of this research are to: (1) Develop a retrospective BC-specific logic model and program theory to help inform the development of a broad program theory for school food and beverage sales environment interventions to articulate expected implementation processes and how they are expected to lead to the desired outcomes; (2) Explore whether this program theory reflects the reality of implementation processes in different real-world contexts  in multiple school districts of BC; (3) Suggest how the program theory for this popular public health intervention might be refined and used for informing how it may be launched in different contexts.  Achieving these objectives will provide empirical, theoretical, and methodological contributions. The empirical contribution will consist of offering stakeholders a nuanced understanding of how and why implementation of the school food and beverage sales policy is happening the way it is in 27  British Columbia. This can help identify practical leverage points at which implementation can be effectively supported and improved in both BC and other jurisdictions with similar school food and beverage sales policies. The theoretical contributions will include development of the first articulated program theory for school food and beverage sales policies. In addition, the use of a realist evaluation approach to conducting this implementation research requires that once highly specific empirical findings are gathered from specific settings, they are federated into a more broad program theory that is relevant for not just a specific intervention in a specific place but for a wider family of interventions in various contexts. These broader program theories can then be used to guide further research questions or even guide thinking about transferring interventions from setting to setting. Finally, this study provides a methodological contribution in that it can be included in the growing body of realist evaluation literature in which actual application of realist methodology to evaluate interventions is still relatively nascent. Moreover, it is the first example of using a realist evaluation to explore the implementation of any type of food environment intervention and gives an idea as to the level of nuanced insight about these complex interventions that can be obtained. 28  Chapter 3~ Realist evaluation methodology   A number of studies from British Columbia and other jurisdictions have explored implementation of school food and beverage sales policies (Dick et al., 2012; Kubik et al., 2010; McDonnell et al., 2006; Vine & Elliott, 2014; Watts et al., 2014). Most have focused on population-level impacts (e.g., food and beverage availability across a province or state) (Samuels et al., 2010; Watts et al., 2014) and all have found that levels of compliance to school food and beverage sales interventions vary across time and contexts.  According to proponents of a realist approach to evaluation, this outcome pattern, whereby no intervention works 100% of the time in all places for all people, is due to both the diversity and complexity of social systems into which interventions are launched. With a constant focus on the relationships between contexts, mechanisms, and outcomes, the realist approach to evaluation helps unpack implementation processes from the “black box” (Astbury & Leeuw, 2010), the opaque space to which implementation processes are typically relegated in conventional input-output evaluations (Chen, 1990).   This chapter will present the philosophical underpinnings of the realist approach, including dimensions of realist ontology (Bhaskar, 2008; Jagosh, 2016; Pawson, 2013; Sayer, 2000) and the realist view of causation. Furthermore, it will describe how these philosophical underpinnings are connected to realist methodology in praxis, including a broad overview of the research process used for this work. Because this dissertation is publication-based, the methods and data collection strategies will only be summarised in this ‘higher-level’ chapter. More detailed descriptions are included in each journal-manuscript style paper in Chapters 4, 5, and 6. In addition, because these findings chapters were prepared as manuscripts that do not require the level of detail about researcher positionality, this is also presented at the end of this chapter. 3.1 Realist philosophy of science  The on-going ‘paradigm wars’ (Gage, 1989; Given, 2017) between different ontological and epistemological perspectives most often encompass debates between those who fall into a positivist camp and those who fall into a constructionist camp (Johnson & Onwuegbuzie, 2004). These ‘camps’ have been formed around paradigms, or philosophies of science, which shape how a scientist approaches the activity of scientific inquiry (i.e., how they view the world, the questions they ask about the world, and how they approach obtaining knowledge about the world).  29  While of course researchers are a heterogeneous crowd with beliefs along the spectrum of ontological philosophies (Edwards, Ashmore, & Potter, 1995), ontological positivists (now often called post-positivists but still influenced by much of the values of classic positivism (Maxwell, 2017)) typically view only that which can be observed as ‘real’ (Comtean positivism).  For example, while they might not view the existence of a god as real, they may view the physiological effects that religious beliefs have in humans as real because they can see the effects in brain scans. They are epistemological empiricists in that they believe gaining valid knowledge about reality can only be done through sensory experience, through direct observation and measurement (Johnson & Onwuegbuzie, 2004). They tend towards also being reductionist in their belief that a phenomenon can be broken down to its constituent parts, each part can be measured, and claims about validity can be made (Shadish, Cook, & Campbell, 2002). Lastly, adherents to this paradigm place a high value on research objectivity and believe that this objectivity is possible with the right set of inquiry techniques and strategies (Maxwell, 2012), such as the use of blinded experiments.   Ontological social constructionists, in contrast, posit that reality is comprised of mental constructions rather than one absolute reality (Guba & Lincoln, 1994). Furthermore, they tend towards a constructivist/interpretivist epistemology that holds that knowledge about phenomena is constructed by individuals and therefore a wide variety of interpretations of reality can exist (Morcol, 2001). Gaining knowledge about these various interpretations of reality, then, involves exploring different interpretive communities’ constructions of reality and an understanding of the social conditioning that shapes perceptions of these different communities’ (Gordon, 2009). The possibility of multiple constructed realities is dependent on social conditions and value-laden perceptions makes it almost impossible, if not fully impossible, to make generalisable claims about social phenomenon across time and social contexts (Mayan, 2009). Furthermore, epistemological constructivists take a more holistic perspective of phenomena whereby no phenomenon can be reduced to its constituent parts and explored without the loss of the credibility of interpretations of data. They also approach research in full acknowledgment that objectivity is impossible as even the researcher is a member of an interpretive community that may or may not be different from respondents’ communities (Day, 2012; Lincoln & Guba, 1985).  Ultimately, Maxwell (2017) proposes that the ongoing split between these two paradigms occurs because they typically present themselves as antithetical to one another, with incommensurable 30  beliefs, rather than engaging in the creation of a more ‘neutral language’ to increase commensurability. The creation of a language that more fluidly crosses paradigms will be, undoubtedly, a necessary evolution as researchers aim to contribute to addressing globally important challenges of epic proportions of complexity never before observed. It has been suggested that a realist philosophy of science can help with this necessary evolution by providing an ontological perspective that is inclusive of both paradigms (Cupchik, 2001; Maxwell, 2017). Realist philosophy of science emerged from the work of Roy Bhaskar (2008), Margaret Archer (1995) and Andrew Sayer (2000) and continues to be developed in terms of the implications of this particular philosophy of science for scientific inquiry by numerous scholars (e.g., (Greenhalgh, Wong, Westhorp, & Pawson, 2011; Jagosh, 2016; Maxwell, 2012; Pawson & Tilley, 1997). The following section presents elements of realist philosophy of science that are foundational for this dissertation. The dimensions of realist ontology, the realist conceptualisation of generative causation, and the realist conceptualisation of validity will be discussed.  Where appropriate, the compatibility of realism with a constructivist epistemology will be highlighted.  3.1.1 Realist ontology Mind-independent reality Realist ontology takes as its starting point the idea that a mind-independent reality exists (Bhaskar, 2008), or that reality exists even outside of mental constructions or perceptions of it. This may appear to be in contrast to a social constructionist ontology held by most qualitative researchers but it has been argued that:  Many (if not most, I suspect) qualitative researchers have a common-sense realist ontology, that is, they take seriously the existence of things, events, structures, people, meanings, and so forth in the environment as independent in some way from their experience with them. (Schwandt, 1997, p. 134)  Realists hold that mental constructions or perceptions of the world are also part of reality (Maxwell & Mittapalli, 2010) and that the real world consists of not just physical matter but also of less concrete dimensions such as emotions, motivations, culture, etc. (Maxwell, 2012). Therefore, Maxwell and Mittapalli (2010) argue that realist ontology is actually commensurate with a constructivist epistemology given it is almost impossible to measure many of these parts of reality. Instead, interpretations of phenomena or events relayed through narratives may be the only way to 31  understand less tangible dimensions of reality.2 Maxwell (2017) proposes that humans can (and should) test constructed ideas (our own and others’) about both the observable and unobservable features of reality. To test constructions of reality, it is entirely permissible within a realist philosophy of science to pragmatically draw upon any strategies available, constructivist and/or empirical, appropriate for the inquiry. Therefore, rather than a relativist stance that everyone’s constructions of reality are ‘truths’, realism allows space to come to the possible conclusion that any one particular construction of reality may, in fact, be incomplete or even wrong (Haig & Evers, 2016) and that there are parts of reality that may never be fully known.   Implicit in the belief in a mind-independent reality is the idea that there is an absolute reality, or one singular true reality (Bhaskar, 2008; Jagosh, 2016). At the same time, scientific realism acknowledges that, while it may be possible to come close to apprehending knowledge about this absolute reality, human minds are incapable of having full knowledge about it in every given moment of time (Jagosh, 2017a) (i.e., constructions of reality will always hold the potential to be fallible). For example, in this moment I am aware that I am typing this paragraph, but I have not been concerned with paying attention to my inhalations and exhalations, or why my fingers can type on a keyboard without looking at the keys, or the sound of espresso being made or even that I am sitting in a café at all. However, now once the existence of these phenomena is acknowledged, my level of awareness begins to include many dimensions of reality in this given moment. This provides an optimistic perspective that the capability of expanding levels of awareness, and knowledge of, the complexity of reality does indeed exist.   While the realist understanding of an absolute reality may seem at odds with social constructionist ontology, scholars have articulated otherwise (Cheek & Gough, 2005; Sayer, 1997).  Sayer (1997) distinguishes between strong and weak social constructionists. Weak social constructionists hold that it is only the knowledge about reality that is socially constructed and that there is a ‘real world out there’. Strong social constructionists, on the other hand, hold that reality itself only consists of social constructions and therefore knowledge about the world is infallible. However, if this were truly the case, Sayer (1997) argues, it would follow that every event or phenomenon should occur exactly as expected because the world is nothing but social constructions. However, this is not the case and it is                                                           2 Realist ontology is also commensurate with an empiricist epistemology but accepts the possibility that a reality exists that is not yet, or perhaps cannot easily be, supported by empirical evidence (Maxwell, 2017). 32  this inevitable and objective fact that lead realists to hold that there is a real world ‘out there’ (Barth, 1987, p. 87 in Maxwell, 2012):   …for critical realists… the world is relatively inaccessible precisely because it is not reducible to, or a construction of, our concepts of it. If our knowledge were infallible, if things always worked out as we expected, there would be no reason for supposing that there was a reality which was not merely our construction. (Sayer 1997, p. 465)  Sayer’s (1997) weak social constructionism, then, is not an ontological perspective that challenges the existence of a real world. It is a perspective capable of identifying that real objects and phenomena exist independent of the researcher regardless of the belief that knowledge and meaning of real objects or phenomena are socially created.  Socially contingent reality Related to the realist belief of an absolute reality, is the belief that, while an absolute reality exists, it is not possible to achieve universal truths about reality (i.e., universal laws) because real events can have different (but real) outcomes in different (but real) contexts (Bhaskar, 2008; Jagosh, 2016; Sayer, 2000). Within a positivist paradigm, the development of universal truths about an event is based on Hume’s regularity principle (Maxwell, 2012); because an event occurs regularly in a closed system it can be said to be a universal law (Bhaskar, 2008). Realists posit that while of course there are indeed regularities that occur and statements about their reliability can be made, these regularities are always susceptible to be influenced by the context. So rather than call some events ‘regularities’ when they are not actually truly regularities because of their contextually contingent nature, realists have coined the term ‘demi-regularities’ to better reflect the reality of outcome patterns (Jagosh et al., 2012; Lawson, 1997). Pawson and Tilley (1997) use the example of gun power to illustrate the context-contingent nature of regularities. When fire is applied to gun powder, it does not ALWAYS ignite. A chemist may explain that the cause of gun powder igniting when fire is applied is due to the chemical composition of the powder: a combination of potassium nitrate, charcoal and sulfur whose respective molecular structures, in the presence of flame, produces an exothermic reaction. In a closed-system, controlled setting the gunpowder may fire 100% of the time leading an investigator to claim that gunpowder will always ignite in the presence of flame. However, if the powder is damp or there is insufficient oxygen present or not enough powder, there 33  likely will not be an explosion (p.58). These different outcomes that may occur, depending on the context, are not events occurring in different realities, but rather all outcome possibilities are encompassed within the real.  Multiple levels of reality It is also assumed, within realist ontology, that there are multiple levels of reality, or that reality has ‘ontological depth’ (Bhaskar, 2008; Jagosh, 2017a).  Jagosh (2016) illustrates this ontological depth using an ice berg metaphor. The tip of the iceberg represents the concrete, observable and relatively easily measured reality. As such, many people may perceive this level of reality or experience it in a similar way and a certain level of consensus may be achievable. Drawing on the previous examples above, this level of reality may be equivalent to me knowing that I am in café and that I can hear the sounds of espresso being made. Others can also see me and others in the café can also hear the sound of espresso being made. It is likely that I and many observers can come to consensus on these facts about reality.  However, underneath the surface of the water is the body of the iceberg. This is the next level of reality that is not easily observable and not easily measured but it is still ‘real’. Back in the café, while people may see me in there working on a laptop, they likely are not aware that I am working on writing my dissertation or what might have driven me to go work in a café in the first place and why I do not have an office to work in or even the means by which I arrived at this particular café and why I chose this one. With the right set of inquiry tools it is possible to know more about this level of reality but with potentially less ability to achieve consensus. For example, this reality consists of common social science conceptualisations such as ‘trust’ (Capra, Lanier, & Meer, 2008) or ‘happiness’ (Thin, 2012), which researchers make attempts to measure using a variety of established tools around which much disputation about validity can occur.   Lastly, the most profound level of reality is, in keeping with the iceberg metaphor, the water surrounding the iceberg. Jagosh (2016) likens this to the dimensions of reality that are lying dormant and have not yet manifested but are still considered ‘real’. One example of these “latent potentialities” (Jagosh 2016) are emotions. At a given moment in time an individual may not be exhibiting signs of anger but that does not mean that the emotion of anger is not a real phenomenon. It continuously lies dormant, in a sense. Given the right context, anger may be triggered and may emerge as a phenomenon that can be easily observed. Obtaining knowledge about this invisible level 34  of reality, the latent potentialities driving us to make decisions about how to act, is extremely challenging. 3.1.2 Generative versus successive causation The realist conceptualisation of causation is also an important dimension of the realist philosophy of science. Scientific realism subscribes to the philosophical concept of generative causation which is distinguishable from successive causation (Bhaskar, 2008; Jagosh, 2016). The concept of successive causation adopted by more positivist researchers is the idea that causation can be determined through observing an outcome occur regularly in a closed system into which a ‘treatment’ or ‘variable’ is introduced (connected to Hume’s regularity principle) (Maxwell, 2017; Sayer, 2000). A successionist inquiry question might appear as: does the introduction of x lead to y? On the other hand, the conceptualisation of causation as ‘generative’ within a realist philosophy of science is the view of causation being determinable through the exploration of why an outcome occurs, rather than simply if an outcome occurs: what is it about x that leads to y?  Blossfeld (2009) explains a generative understanding as per David Cox (cited in Blossfield, 2009):   …it is crucial to the claim of a causal link that there is an elaboration of an underlying, generative process existing in time and space. A causal association between X and Y must be considered as being produced by a process and is created by some (substantive) mechanism.  (p. 87-88)   Successionist causation is the foundation of randomised control trials (RCTs), the long-accepted gold standard of evidence. RCTs are typically viewed as more valid and capable of determining causation than all other types of scientific inquiry. This may be, Maxwell (2017) proposes, why many constructivist-leaning researchers, who typically do not use RCTs in their research, are uncomfortable with making statements about causality of a phenomenon. The generative conceptualisation of causation arguably aligns more with the in-depth insight-seeking goals as to why or how a phenomenon is occurring (the general domain of inquiry of qualitative researchers), but it also aligns with quantitative inquiry. Ray Pawson has written and discussed extensively the argument for the commensurability of generative causation for inquiries like clinical trials which are typically solely underpinned by a successionist view of causation (e.g.,  Pawson, 2013, 2016). He argues that clinical trials are not, in reality, simply “treatmentoutcome” processes as per the successionist view. They should also be considered as a social intervention whereby there are long 35  and complex implementation chains influenced by macro, meso, and micro contextual factors. For example, during an RCT it can be determined whether taking a particular drug leads to the alleviation of a disease, an obviously useful successionist inquiry. However, it is true that taking the drug will not always lead to the same outcome and it is not until generative causation is explored that an explanation for why the drug works only sometimes can be determined. In this specific RCT example, this might mean exploration of the underlying physiological mechanisms of the drug in different bodily contexts or perhaps the greater social contexts of the individuals taking the drug and the influence of these complex social environments on the drug’s mechanisms (Pawson, 2016). 3.1.3 Realist conception of validity Validity refers to the approximate truth of an inference (Shadish et al., 2002). As understood within the positivist paradigm, validity of conclusions or inferences reached through research have often depended on the assumption that there is a singular, knowable reality. As mentioned briefly above, judgments about validity are typically based on the methods used, thus creating a hegemonic hierarchy of evidence, with randomised control trials being considered the ‘gold standard’ that most successfully generate knowledge about this singular reality (Maxwell, 2012). This understanding of validity has largely been rejected by qualitative researchers who propose alternative concepts more suited to a constructivist paradigm that assumes multiple realities (Mayan, 2009), or multiple interpretations of reality (Cheek & Gough, 2005). Lincoln and Guba (1985) proposed criteria for the alternative concept of trustworthiness in qualitative work. These include credibility, transferability, dependability, and confirmability which, like quantitative conceptions of validity, also rely on a set of techniques and procedures, including but not limited to audit trails, memos, data triangulation, etc. (Mayan, 2009; Polit & Beck, 2004; Rothe, 2000; Tuckett, 2005). Seale (1999) critiqued this alternative qualitative understanding of validity, trustworthiness, as research accounts that “do no more than represent a sophisticated but temporary consensus of views about what is considered to be true” (p.46) without acknowledgment of the possibility of anything beyond these views.  Maxwell (2012) presents further critiques of both the positivist and constructivist understandings of validity and the belief that validity or trustworthiness can be ‘bought’ using methods or techniques, and that no technique or method will guarantee validity of a causal inference (Porter, 2007; Seale, 1999; Shadish et al., 2002). A realist approach to validity relates to the conclusions made by researchers rather than validity being related to the method itself (Maxwell, 2012). In other words, 36  the validity “of an account is inherent, not in the procedures used to produce and validate it, but in its relationship to those things that it is intended to be an account of” (Maxwell, 2012, p. 130).  What is problematic about taking validity to mean, basically, how close research findings are to absolute reality, is that realists adhere to the assumption that all knowledge produced through research is only constructed ideas about a ‘true’ absolute reality that it can never be fully grasped. This is the reason knowledge constructions about this true reality are inevitably fallible. Therefore, in terms of validity, how is it possible to have any ability to ‘check’ inferences against the reality of the phenomena under study if reality can never be fully known? (Maxwell, 2012). Maxwell (2012) argues that this does not mean that this unknowable reality does not have a role at all in realist judgments of validity and reminds the reader of grade five science classes and the scientific method. A key focus of the scientific method is that conclusions are tested against existing and potential evidence with the aim of exploring alternative explanations, which are the realist conception of a validity threat. With that being said, an important activity of the realist researcher is to evaluate the conclusions that emerge from the data for alternative explanations, asking, “how might I be wrong about this?”. Alternative explanations can be sought from data (our own or others’) that does not fit our conclusions and then the importance of this difference can be questioned. Maxwell (2017) also supports the ideal of research being a social process and, as such, the researcher can ask others, “where might I be off base?”.  Another important dimension of the realist approach is the value placed on inference as a contribution to the process of doing science. Even when there is limited empirical evidence that has emerged from data collection and analysis to support an inference, the inference is not immediately disqualified as invalid (Jagosh, 2017c). Further testing of initially not-well-supported inferences, as a key component of the scientific method, is an important part of the theory-building process. Weak inferences (i.e., inferences not supported by a large quantity of empirical data, either one’s own or others’) about generative causation that emerge from realist research are encouraged in the presentation of findings as long as there is transparency about the strength of evidence underlying it. 3.2 Realist methodology and the realist evaluation cycle Pawson and Tilley (1997) were the first to articulate the implications of a realist philosophy of science for evaluation research and called it, creatively, realistic evaluation (alternatively, “realist evaluation”). This theory-driven approach to evaluation (Chen, 1990; Pawson & Tilley, 1997; Weiss, 37  1997; Westhorp, 2014) has become increasingly popular for assessing the implementation of complex social interventions, particularly in health services and public health research (Lacouture, Breton, Guichard, & Ridde, 2015; Porter, 2007). This has been said to be due to the explanatory appeal of its broad line of inquiry cum mantra: What works, for whom and why? The main concern of a realist approach to evaluation is with the complexity of interventions whose outcomes are socially contingent (i.e., how interventions manifest, in-practice, and lead to outcomes is influenced by various dimensions of context) and with understanding how contexts and underlying mechanisms of an intervention lead to outcomes, whether desired or not, intended or unintended (Pawson, 2013).  Furthermore, while the output of realist evaluation can be highly useable for informing a specific intervention, the goals of realist evaluation are also theoretical and might include questioning the integrity of a program theory, adjudicating between rival program theories, or considering the same program theory in comparative settings (Jagosh, 2017a). 3.2.1 Realist research cycle Developing program theory As stated above, the goal of realist evaluation is first and foremost theoretical; the aim is to test and refine program theories. To do this, the first step in the realist research cycle (as seen in Figure 3.1) is to begin with a hypothesised program theory (Pawson, 2013; Westhorp, 2014). Basically, realist evaluation begins and ends with theory (Vareilles, Pommier, Kane, Pictet, & Marchal, 2015).               38   Figure 3.1 Realist evaluation cycle (Vareilles et al., 2015) Figure 3.1 above shows the ‘beginning’ of the realist evaluation cycle to include both middle range and program theory.  Robert Merton (1949) coined the term “middle range theory” to describe the level of abstraction at which a theory operates and defined them as:  …theories that lie between the minor but necessary working hypotheses that evolve in abundance during day-to-day research and the all-inclusive systematic efforts to develop a unified theory that will explain all the observed uniformities of social behaviour, social organisation, and social change. (p. 448)  While program theories can be specific hypotheses about the mechanisms of a specific intervention, or about specific intervention activities, MRTs are a type of theory that is a slightly higher level of abstraction than specific program theories and are a lower level of abstraction than grand, general Middle range theory 39  theories of social systems. The health belief model and the ecological theory of health behaviours, outlined in Chapter 2, are examples of MRTs, being concrete enough to do empirical testing while being broad enough to apply to different types of interventions in different contexts. The grand sweeping theories that were being proposed by sociologists in Merton’s time, Merton (1949) argues, were the result of sociologists trying to devise social theories that mirrored the theories of the physical sciences (e.g., laws of thermodynamics). Merton (1957) believed broad sweeping theories derived from observations of social processes, such as, for example, dominant class theory or the theory of conflict, were not useful because of the inability to scientifically derive knowledge about them; one variable seeking to explain all social processes was considered to lead to “epistemological obscurantism” (Boudon, 1991, p. 520). In other words, these grand theories were too vague to use to underpin scientific investigation and therefore would be unable to support the accumulation of knowledge in the field of sociology (Merton, 1949).   Realist evaluation ultimately aims to contribute to knowledge accumulation around broader program theories about whole families of interventions, transferable between contexts (Jagosh et al., 2015). It takes these broad program theories as its starting point in the realist research cycle shown in Figure 3.1. However, if no broad program theories exist in the literature pertaining to an intervention of interest, there is a need to explore real-world examples of interventions, using either primary and/or secondary data collection, to first hypothesise program theories that can then be tested in an evaluation. As indicated in Figure 3.1, MRT can interact with program theory. MRT can be devised from the federation of empirical findings from the experiences of specific interventions in specific contexts, from primary and/or secondary data collection. In addition, MRTs in the extant literature can also be drawn upon to shape the development and refinement of program theories, providing what Jagosh (2017c) has called “conceptual currency”.  Ideally, program theory is developed and articulated prior to the launch of an intervention. One way to do this is by using an integrative approach that integrates both stakeholder perspectives and existing social science perspectives to articulate a program theory (Chen, 1990). For this approach, evaluators use stakeholder input, through interviews and document analysis, to refine what they expect is going to happen to lead to desired intervention outcomes. In a “dual theorising process” (Chen, 1990, p. 71), it is recommended that evaluators draw upon their own expertise, knowledge and understanding to build on stakeholder ideas of the program theory. Discussions around the 40  hypothesised program theory between the evaluator and the stakeholders can help to verify that the hypothesised program theory reflects the intent of the stakeholders while at the same time being more strongly situated in existing social theory. However, often the development of an a priori program theory does not happen due to resource constraints, such as funding or lack of knowledge and skills (Lobo, Petrich, & Burns, 2014; McLaughlin & Jordan, 1999). In these cases, prior to evaluation, it is acceptable to develop retrospective program theory (Chen, 1990).  Designing the study Following the development of a hypothesised program theory, an appropriate study design should be developed that allows for testing the different elements of the program theory (Vareilles et al., 2015). It is important to remember that realist evaluation is not a method, but rather a framing perspective for evaluations (Westhorp, 2014). In fact, realist evaluation is method neutral (Pawson & Tilley, 1997; Westhorp, 2014). Therefore, when it comes to data collection, strategies are viewed as tools that can be used to, pragmatically, complete a task. This is in contrast to a paradigm perspective that will often equate a set of tools with a paradigm and, taking it one step further, equate the use of certain tools within a paradigm as a means to ensure validity (where some tools are viewed as superior to others for obtaining ‘true’ knowledge) (Maxwell, 2012).    Data analysis Data analysis within a realist approach largely depends on generating explanations for phenomena. The realist conceptualisation of causation as generative has an important influence on data analysis. To reiterate, generative causation seeks to explain outcomes in terms of what ‘generates’ them, or the underlying mechanisms. There are ‘three planks’ of a generative explanation: outcome patterns, mechanisms, and context (Pawson, 2008). First, generative social research takes as its starting point observed outcome patterns, or ‘demi-regularities’ (Lawson, 1997), and devises theories that explain both the occurrence of an outcome and the absence of an outcome (Jagosh, 2016). Outcomes in realist evaluation can be any effect and observable reality resulting from any part of the intervention, including activities related to implementation activities. They can be intended, unintended, desirable or undesirable, proximal, intermediate, or final (Jagosh et al., 2014). Outcome mapping and outcome measurement, according to Pawson (2013), have become “decidedly inelegant” (p. 39). He posits that “one yardstick twice applied” (p. 39), as is the practice of pre- and post-intervention evaluations, neglects the more realistic messiness of the complexity of links in between inputs, activities and outcomes. It is inevitable that with a complex intervention involving a diverse array of actors and a 41  diverse array of activities, often tailored specifically in different contexts, potential evaluable outcomes will also be diverse and numerous (Craig et al., 2006; Pawson, 2013).  The second plank of generative causation, mechanisms, are the ‘generative’ force that leads to an outcome (Jagosh, 2016; Pawson, 2013). From the realist evaluator’s perspective, programs do not ‘work’ but it is the complex individuals, within a complex social system into which a complex intervention is inserted, who ultimately ‘make decisions’ (sub-consciously or consciously) that determine the success of failure of an intervention (Pawson, 2014). All interventions provide resources, whether concrete or non-concrete, and it is the reaction to these resources (Jagosh et al., 2014) that occurs in the minds of intervention stakeholders that are considered the mechanisms of an intervention; the ‘engine’, that brings about an intervention effect, or drives individual behaviour change (Pawson & Tilley, 1997). This is often conceptualized in realist evaluation circles as: intervention resources + reasoning = mechanism (Jagosh et al., 2014).  In generative social research, while the inquiry begins with the observation of outcome patterns, theorising about choices and reasoning of individuals that leads to those outcome patterns (i.e., the mechanism) is the beginning of explaining the outcome.   The last plank of generative explanation, context, are the numerous and multi-layers of pre-existing factors that influence the triggering of a mechanism, or influence the response of a stakeholder to the intervention resources. As discussed previously, realist ontology posits that all phenomena (including intervention processes) are socially contingent and this is why there are no true regularities but rather demi-regularities or, rather, variable intervention outcome patterns observed across social contexts. Interventions are complex processes introduced into complex social contexts to be implemented and taken up by complex individuals. Contexts are what Jagosh et al. (2014) have called the “‘backdrop of programs and research” (p. 134). The realist evaluation perspective assumes that when interventions are introduced into fertile pre-existing contexts, latent mechanisms can be triggered that lead people to decide to take the desired action. Vice versa, when interventions are introduced into infertile contexts, it is expected that the desired mechanisms will not be triggered and that other less-desirable mechanisms, or no mechanisms at all (or both), will be triggered. As such, uncovering plausible theories about what constitute fertile and unfertile contexts to contribute to guiding the launch of an intervention is one of the key goals of realist evaluation.  Furthermore, contexts triggering mechanisms do not necessarily act abruptly but more like a “dimmer switch” 42  where mechanisms are triggered and outcomes are produced in varying levels depending on the individual and other dimensions of their contexts (Jagosh, 2017b). Realists acknowledge that making decisions around what facets of the environment are relevant when evaluating an intervention is at the “heart of the realist struggle” (Jagosh, 2017b) and that there are no established rules as to what is considered to be an important dimension of context.  Out of these three planks of generative causation emerges the hallmark heuristic tool of data analysis in realist evaluation: the Context (C) + Mechanism (M)  Outcome (O) configurations (CMOcs). CMOcs can be used in the praxis of qualitative analysis as both a connecting strategy and categorising strategy. Connecting strategies are based on the concept of contiguity which means related in time or place as opposed to being related by similarity, the basis of categorising strategies (Maxwell, 2012). Categorising strategies “fragment the data into discrete units” (Maxwell, 2012, p. 111) through processes like coding. These units of data are labeled and grouped, examined and compared. Connecting strategies for analysis are more narrative and contextual approaches that aim to identify relationships in the narratives while only maintaining information that is the most relevant to the relationship. Categorising data can be useful to break the data “out of the fixed contextual frameworks”, or to decontextualise the data, and allows the researcher to “see other relationships that they had been blind to” (Maxwell, 2012, p. 117). Maxwell (2012) also cautions against a sole focus on connecting strategies as it can lead to an “imprisonment in the story” (p.117).  Ultimately, Maxwell (2012) recommends an integrated approach, using both categorising and connecting strategies. One way to do this is to conduct linked CMO coding (Jackson & Kolla, 2012; Puton, Vogel, & Lloyd, 2016). Here, narratives are coded with CMO configurations, either open-coded, coded using codes developed during the creation of the initial program theory, or both. Codes might look like: C1-M1-O1, C2-M2-O2, C1-M2-O3, etc., and provide generative explanations for specific phenomena in the narratives. This helps to ensure that the relationships being described in the narrative are maintained rather than each C, or M, or O being discretely coded, detached from their relationship to one another.   Sometimes in the narratives of participants, they only identify a dimension of context and its relationship to an outcome (a C-O) or a mechanism and its relation to an outcome (an M-O). During analysis, it is acceptable to make plausible inferences to fill in these blanks to make a complete 43  CMOc (i.e., inferring a plausible M in the former or a plausible C in the latter) using existing theories (Pawson, 2013).   Categorising strategies become useful following the use of linked CMO coding as a connecting strategy. Linked CMO codes can then be categorised by, for example, what particular outcome the linked CMO code is related to. Whether the linked CMO code contains an outcome that is intended or unintended does not matter. All the linked CMO codes related with the same outcome can be grouped together. This is functional categorising that helps organise thought processes as well as see if there are similar (or different) mechanisms and contexts at work relating to the same outcome category. Alternatively, perhaps it is found that the same context and mechanism may lead to a number of different outcomes.  Guiding questions can be asked, for example:  Which of these outcomes are more prevalent?  When considering outcome patterns, are there similar contexts and mechanisms related to these outcomes?  Which outcomes are specific to just one CMO configuration? And why is this special? Does it matter? Synthesis and theory refinement After analysis, the synthesis process can consist of within- and cross-case analysis of the CMOcs. To do this, CMOcs are compared with the CMOcs derived from the initial program theory, which can then be accepted, refined, or rejected as part of the refined specific program theory. From data analysis, new CMOcs  not originally hypothesised can also be created so as to develop plausible inferences about how and why implementation strategies work or do not work (Vareilles et al., 2015). This can be done within each specific case and then across the cases to guide the refinement of the broader program theory for a family of interventions; these broader program theories will include important contexts and mechanisms impacting implementation that likely need to be considered while remaining broad enough to potentially transfer an intervention to a different context.  As stated previously, realist evaluation begins and ends with a broad program theory around families of interventions. After conducting data collection, analysis, and synthesis from an evaluation of a specific intervention, the realist evaluator aims to refine the hypothesised program theory. These 44  refined program theories are meant to provide insight into what considerations around contexts and mechanisms might need to occur when transferring an intervention to a different context (Vareilles et al., 2015). As per Figure 3.1 and 3.2, the “end” of the realist cycle again includes the interaction of program theory and middle range theory. Here, as stated in the previous section, middle range theories can be devised from program theories. This process of devising MRTs is the act of federating empirical findings, or rather creating statements that organise a set of hypotheses emerging out of segregated empirical events or observations (Boudon, 1991). More poetically, Liehr and Smith (1999) consider the development of MRTs to be the metaphorical spinning of one single thicker thread from many smaller threads. Bringing this metaphor back to the literal world of realist evaluation, this is the process of making inferences, from the empirical data gathered during an evaluation, that are a level of abstraction up from the program activity-specific inferences made about the specific intervention. There are a number of ways by which to develop an MRT from an array of empirical, segregated data. These include inductive theory building through research, deductive theory building from the grand sweeping theories described earlier, deriving theories from other disciplines, and/or synthesising theories from other disciplines (Pawson, 2000; Roy, 2014). In addition, as presented in the previous section, MRTs already in existence, such as the health belief model or the ecological theory of health behaviour, can be drawn upon and used to help provide added explanatory power and plausibility to the findings and refined program theory (Jagosh, 2017b).  Ultimately, MRTs emerge quite naturally through the use of the context + mechanism  outcome heuristic to analyse data (Pawson, 2000). This is because the first step in MRT development is to inter-relate the segregated bits of empirical evidence gathered through data collection in theoretical statements about causation (Roy, 2014). This is precisely what the use of CMO configurations, and conducting linked CMO coding, aid realist evaluators to do when analysing data.  The refined broad program theory at this point is not ‘finished’ nor will it ever be, according to the realist perspective that holds that, while a close understanding of absolute reality is attainable, a complete understanding will never be obtainable. The realist cycle is a ‘cycle’ because this new refined program theory can then be subjected again to investigation in other contexts to see if it holds water. The more water it holds, the closer it comes to reflecting reality. This is the essence of the scientific method: observed patterns of outcomes lead to theory and this leads to testing to a refined theory to more testing to an even more refined theory. 45  3.2.2 The realist research cycle for the implementation evaluation in this dissertation  Figure 3.2 presents the overall research cycle that was undertaken for this realist evaluation of the implementation of school food and beverage sales policies using the case of British Columbia, Canada.     Figure 3.2 The research cycle for this research (adapted from Vareilles et al., 2015) Chapter 4 presents the first phase of my research which was the development of the retrospective logic model which is highly specific to BC’s school food and beverage sales policy. Following this, a BC-specific program theory is also presented, derived from the specific intervention activities outlined in the logic model and already existing social theories. Finally, based on the specific BC logic model and program theory, I propose an initial broader program theory for the family of school 46  food and beverage environment interventions which is the focus of program theory refinement. This was used to devise candidate CMOcs. To develop the logic model and program theory, data collection strategies that align with a constructivist epistemology were used and included observations at public events related to the BC Guidelines, document analysis, and semi-structured interviews with those who have been involved in the creation and development of the BC Guidelines over the years (See the methods section in Chapter 4 for more details). I purposively identified relevant stakeholders and approached them about participation. Snowball recruitment was also used whereby participants were asked if they had ideas of other relevant stakeholders I might contact. For the analysis, an interpretive policy analysis lens was used.  This lens emphasises accessing local knowledge and is an alternative to conventional policy analysis that has been criticised for neglecting and sometimes devaluing the influence of stakeholder’s knowledge and specific context (Yanow, 2000). Interpretive policy analysis is founded on the presupposition that “we live in a social world that is characterised by the possibilities of multiple interpretations… as living requires sense-making, and sense-making entails interpretation, so too does policy analysis ( Yanow, 2000, p.5)”.  Throughout the process, I took into consideration that even my own interpretations of policymakers’ interpretations are subject to interpretation (Hay, 2011). Relating this back to realist ontology, all interpretations are considered to be an act of ‘theorising’ about what the absolute reality might be, with all interpretations being candidates for scrutiny and testing.  In the second phase of the dissertation research, I selected a qualitative, multiple case study approach to explore the social processes associated with implementation of the Guidelines in different BC contexts and consider how they aligned with the hypothesised broad program theory developed in phase 1. Having different contexts in which to explore how and why implementation was happening in different contexts was a means to further assess how valid my hypothesised broad program theory might be in different contexts. Using multiple cases allowed me to also delineate clear boundaries around the contexts of each case and to be able to compare findings within and across cases (Yin, 2009). I took school districts as a case unit. Districts were purposively selected. I aimed to include one district from each of BC’s five Regional Health Authority (RHA) catchment areas. Because the health system in BC is decentralised to the five RHAs, each RHA is organised differently which has implications for the way in which health sector staff engage with schools or school districts regarding implementing the policy. This means that a school district within one RHA will have different dimensions of context than another located in another RHA. I also wanted to have both 47  urban and rural district cases to consider how this contextual difference may influence implementation.  However, I was only able to obtain participation from school districts in four of the five RHAs due to challenges in recruiting interested school districts. Of the five districts in which data collection occurred, two were rural. The challenges I experienced working with school districts and schools are presented in more detail in Chapter 7 (see Appendix A for more in-depth descriptions of districts, including relevant contextual information).   Within each school district, attempts were made to gather data from relevant stakeholders (i.e., people who were purposively identified in the logic model and program theory as expected implementers) at the school district offices, school-level stakeholders, vendors, parents, and school food staff. I used semi-structured interviews to gather data from district-staff and vendors.  In consideration for their time during the school day, short questionnaires were used to engage and gather data from expected implementers in schools. Interview guidelines and questionnaires were both developed based on the program theory developed in phase 1 (Chapter 4). Also, because the interviews used in phase1 to develop the program theory were so rich, describing not only how the policy had been developed or the expectations around implementation activities, but also insight into the social processes of present-day implementation, data from these were also drawn upon in phase 2. I recruited participants using a semi-convenience approach, whereby I collected data from any individual who a) fell into one of the relevant stakeholder groups (listed above) and b) who was willing to participate. I also used a snowball approach and asked participants if they knew of anyone else who might be able to provide relevant information who I might contact or with whom they were willing to share the online questionnaire.  For the analysis in phase 2, I used the linked CMO coding and categorising approach described earlier in this chapter. In circumstances where chunks of narratives only contained a C-O or an M-O, I selected to use an existing middle-range implementation theory, contextual interaction theory (CIT), as a heuristic tool for inferring plausible contexts or mechanisms in the cases where Cs or Ms were absent. CIT is a relatively nascent implementation theory that emerged from systems theory in the late 1990s (Bressers, 2007; O’Toole, 2004; Owens, 2008). The theory’s main assumptions are that policy processes, including implementation, are processes undertaken by individuals in a constant dynamic interaction with their contexts. How implementers might respond to an intervention depends on three broad individual level characteristics: their motivations, their 48  knowledge, and power (Owens, 2008; Owens & Bressers, 2013). In simple terms, motivations are meant to encompass an implementer’s own goals and values, external motivating pressures, and perception of self-efficacy. Knowledge (or cognitions as it has also been called) refers to an implementer’s interpretations, their frames of reference, and their experience of reality. The characteristic termed ‘power’ is meant to encompass not only the attribution of power by others onto an actor but also the accessible and available resources (Bressers & Lulofs, 2010). Additionally, dimensions of context have an influence on how implementers engage with implementation (or not) only in so far as  contextual factors impact these three broad individual-level characteristics (Owens, 2008; Owens & Bressers, 2013). These three core characteristics dynamically influence one another whereby the characteristics of the implementers shape the processes of implementation but are also, in turn, influenced by their experiences with the process. This means the characteristics of actors can continually and gradually change throughout implementation processes. The external contextual factors that are posited to impact the motivations, cognitions, and power of implementers range from the specific implementation context to the wider, structural context (Bressers & Lulofs, 2010; Owens, 2008). I used these three dimensions, knowledge, motivation, and power to hypothesis Ms missing from participants knowledge and drew upon CIT’s extensive framework of contexts to hypothesis Cs that were missing from them. This implementation theory is an appropriate heuristic tool to do this as it aligns with the realist perspective in both its consideration of complexity and its focus on the interactions between contexts and how individual implementers might choose to respond to an intervention.  I then conducted a within-case and cross-case comparison to see if there were certain phenomena that may be occurring in just one district or if there were phenomena that were occurring across all the districts. This provided insight into what dimensions of context, specific to one district or common across districts, were related to social processes of implementation. It emerged during the analysis and interpretation that different social processes were occurring at district levels compared with the school levels. As a result, I decided to separate the analysis and presentation of findings by school district level and school level. This provided a clearer view of the similarities and difference of the implementation-related social processes occurring at both levels.   In the final concluding chapter, I round off the realist research cycle by critically reflecting on the differences between my hypothesised broad program theory and what was found to be actually 49  occurring in districts and schools. In addition, I build upon the hypothesised program theory, by considering my findings and existing literature, and offer a refined broad program theory for school food and beverage environment interventions. As per the realist research cycle, this can act as the starting point for guiding a realist synthesis of existing literature to further help refine the program theory or even subsequent realist evaluations in other contexts, perhaps for different types of school food and beverage environment interventions, to further test the refined program theory.  Ethical approval was obtained from the University of British Columbia’s Human Research Ethics Board. A detailed description of the ethical considerations for this study is included in Appendix B. 3.3 Locating the researcher Given that realist philosophy of science posits that all reality is socially contingent, that I am a part of a social reality (Haig & Evers, 2016), and that constructivist epistemology depends on the interpretations of not only the research participants but also on the interpretations of the researcher, it is important to explicitly situate my positionality. First, I am an outsider relative to any of the stakeholders associated with BC’s school food and beverage sales policy. I am not a parent with a child in school or any other member of a public school or district community. I am also not a part of the government health sector, a politician, or a participant in the school food and beverage industry. I do not have children (nor wish to) and as such will never see direct benefits to my child from changing school food environments. I am a single, middle-class, privileged, mostly urban, white Canadian studying in a doctoral degree program that is underpinned by a critical perspective on the current global food system. Related to my doctoral program, I view the BC Guidelines as an intervention that is, while specifically meant to be implemented in school settings, but one example of an intervention that has the potential to contribute to changing the food system writ large, moving it towards becoming more socially, economically, and environmentally sustainable and health-promoting. Moreover, I tend towards a balanced and moderate perspective of food consumption where my preference is for whole foods and less consumption of animal products but where ‘junk food’ treats happen. I am a democratic socialist and therefore also hold the belief that the food system would benefit from less concentrated power in food production, less corporate influence in nutrition and food policies, and more stringent government interventions when it comes to informational (e.g., transparent labeling, effective nutrition education), ethical (e.g., patents, farm labour), agricultural, environmental and economic policies related to food production, sales, and access. However, I also agree with Golden, McLeroy, Green, Earp, and Lieberman (2015) that 50  individuals still do have some role in deciding what to purchase and consume. There is still a need, though, to further uncover the social processes involved in individuals’ unhealthy consumption choices, whether economic, political, environmental, or even psychosocial so we/they can be supported to have access to healthier choices and to make healthier choices.51  Chapter 4~ Development of a retrospective logic model and program theory for British Columbia’s school food and beverage sales standards  4.1 Introduction and background School food environments can influence what children eat during the school day, as up to 35% of daily calories are consumed in the school setting  (Briefel et al., 2009; Tugault-Lafleur et al., 2016). Policy and programs promoting a salutogenic school environment have been found to be somewhat effective strategies for improving student eating habits and may be useful tools for improving general public health. They can promote increased consumption of fruits and vegetables (Driessen, Cameron, Thornton, Lai, & Barnett, 2014; Larson & Story, 2010) and decreased consumption of low-nutrient energy dense foods (Mullally et al., 2010; Wordell, Daratha, Mandal, Bindler, & Nicholson-Butkus, 2012), as well as decreased body mass indices (Sanchez-Vaznaugh, Sanchez, Baek, & Crawford, 2010; Taber, Chriqui, Perna, Powell, & Chaloupka, 2012).  Internationally, implementing nutritional guidelines for items sold to students is an increasingly popular school food and beverage environment intervention (Hawkes et al., 2015), including in almost all Canadian provinces and territories (Holmes, 2016).   As Canada is a federalised system, there is no national-level entity directing provincial ministries of education. The development of policies to be implemented in individual schools, then, is done at the provincial level, sometimes at the school district level, and/or sometimes at the school level. As a result, each province and territory, and sometimes individual districts or schools, make decisions about food and nutrition policies, including whether to have them or not. If a province or district does decide to develop food and beverage sales policies, they are each developed separately from other provinces or districts. While it is possible that content of provincial or district policies may be influenced by policies in other jurisdictions, what is observed across Canada is disparate initiatives being developed and implemented differently in each province.  Upon winning its bid to host the 2010 Olympics, the BC government publicly committed to making the province the healthiest jurisdiction to ever host them (Government of British Columbia, 2008), with improving school food and beverage environments an important supporting goal. In 2005, the British Columbia (BC) Ministries of Health (BCMoH) and Education (BCMoED) partnered to develop the Guidelines for Food & Beverage Sales in BC Schools (“the Guidelines”) (BCMoH & 52  BCMoEd, 2005). These Guidelines, while largely created by BCMoH policymakers, were meant to be implemented by the education sector.  The Guidelines were made mandatory in 2008 but no plans were made for official enforcement. Revised in 2013, the Guidelines are nutrient-based criteria aiming to reduce the amount of salt, sugar, and fat in items sold to students in all venues in schools, including vending machines, cafeterias, fundraisers, sports days, and school stores (Government of British Columbia, 2013).   In 2008, the BCMoEd and BCMoH conducted an evaluation to assess whether any change had occurred in food and beverage availability since the roll out of the Guidelines three years earlier (BCMoEd & BCMoH, 2008). Progress was found regarding elimination of sugar-sweetened beverages from vending machines while snack machine compliance remained challenged. Over a quarter of all schools had snack machines stocked with 60 to 80% of “not recommended” options, while a further 25% had snack machines stocked with 80-100%  “not recommended” foods”. Watts et al (2014) conducted a subsequent evaluation of the changes in school food environments between 2007 and 2012 (including vending machines products). They found that in 2012, schools had significantly lower odds of having a number of important unhealthy food ‘culprits’ available compared with 2007, signifying progress towards  reducing the presence of unhealthy items in schools in BC. The availability of some foods typically thought of as being unhealthy (pizza, hot dogs, and hamburgers) however, remained unchanged. The authors postulated that this may be due to products being reformulated to comply with the Guidelines. Masse et al. (2013) conducted implementation research in the 2011/12 school year that sought to explore the barriers and facilitators to implementing the Guidelines and perceived implementation. They found elementary school respondents perceived implementation to be lower in elementary schools than middle and high schools and fewer principals perceived the Guidelines were being met than other teacher/school informants. This suggests that there have been persistent challenges with implementation and thus further work is needed to explore the current state of implementation of the Guidelines across different jurisdictions in BC.   To develop the lines of inquiry for an evaluation of an intervention, whether evaluating process outcomes related to implementation or overall outcomes, it is helpful to use an articulated theory of change (Chen, 1990; McLaughlin & Jordan, 1999). Theories of change help to articulate components of an intervention and how these components are expected to relate to one another (Cooksy, Gill, & 53  Kelly, 2001). Ideally, a theory of change would be developed during creation of an intervention to guide future evaluations, helping clarify the evaluable components and expected implementation processes (Chen, 1990). The benefits of this activity reported in other studies include having the tools to devise sound monitoring systems to improve the functioning of an intervention (Li, Yao, & Gu, 2014), as well as to develop intervention strategies that are “culturally sensible and locally relevant” (Perez-Rodrigo et al., 2005, p. 67). However, sometimes development of a theory of change prior to the launch of an intervention does not happen in practice due to resource constraints (Lobo et al., 2014; McLaughlin & Jordan, 1999), necessitating development of a retrospective theory of change (Chen, Cato, & Rainford, 1998) to aid in framing an evaluation. Logic models and underlying program theories, as components of a theory of change, are widely acknowledged among health promotion professionals as important program planning and evaluation tools (Chen, 1990; Chen et al., 1998; McLaughlin & Jordan, 1999; Renger & Titcomb, 2002; Weiss, 1995). Logic models are useful for graphically demonstrating the flow of logic between the components of an intervention, from program inputs and activities to the anticipated short and long term impacts (Chen et al., 1998; Renger & Titcomb, 2002). Identifying program outputs helps to further provide an explicit picture of the links between program components and outcomes. However, using only a logic model to articulate a theory of change, whereby elements of an intervention are diagrammatically catalogued with minimal information on the links between elements, may result in oversimplification of an intervention (Bickman, 2000). Weiss (1995) advocates for not only capturing the components of an intervention in the logic model but also for paying special attention to the hypothesised links between the components; in other words, articulating what has been termed a program theory (McLaughlin & Jordan, 1999). Bickman (2000) defined program theory as “a plausible and sensible model of how a program is supposed to work” (p.5). Hypotheses about how intervention activities are expected to lead to outputs and outputs to outcomes are made by drawing upon existing social theory about human behaviour (Weiss, 2000).  This grounds the links between intervention components in theory and can help establish how to approach the complex exploration of implementation processes.  There is a small body of literature that has presented theories of change for other types of school food and beverage environments. One study presented a logic model of an intervention in rural China that aimed to provide schools kitchen equipment and teacher training in health and nutrition, to be able to provide training to students (Li et al., 2014). Ratcliffe (2012) published a sample theory 54  of change for farm-to-school programs. Another theory of change describing a school-based fruit and vegetable promotion program in multiple European countries has also been published in which the intervention strategies consisted of a classroom component, a school food environment component, and a family component (Perez-Rodrigo et al., 2005). Passmore and Donovan (2014), in an evaluation of the UK’s Health for Life in Primary Schools, essentially end up presenting a version of a detailed logic model. This program consisted of four interrelated strands including healthy eating, cooking skills, growing food, physical activity, and family involvement. Finally, Sumberg and Sabates-Wheeler (2011) propose a theory of change for home-grown school feeding programs which focus on the creation of institutional demand for locally and regionally grown products to not only improve the nutrition of students but to support local agriculture. However, there is yet to be an articulated theory of change for school food and beverage sales standards published in the literature.   To fill this gap, this chapter develops a retrospective logic model program theory for BC’s Guidelines. From these, a program theory is devised that is relevant for school food and beverage sales interventions more generally. In this chapter I propose some foundational assumptions and beliefs underpinning not just this specific BC-based policy for improving school food and beverage sales environment but also for this type of intervention more generally as it has been adopted in a number of other jurisdictions. I also discuss the implications of these proposed assumptions and beliefs for developing an evaluation framework for exploring implementation processes of this type of intervention. 4.2 Methods I developed a retrospective logic model and program theory using an interpretivist and social constructivist perspective.  Qualitative data collection strategies included: (1) Document analysis; (2) Participant observation; and (3) Semi-structured interviews with stakeholders involved in the creation and evolution of the Guidelines (i.e., health sector stakeholders). These data collection strategies iteratively explored: the discursive and historical context around the creation and implementation of the Guidelines; the inputs that contributed to their creation and development and continue to contribute to their implementation; the implementation activities that have occurred to-date; and the expected outcomes.  Sixty-five documents were reviewed that fell into one of 10 categories listed in Table 3.1 below:  55   Table 4.1 Documents reviewed   British Columbia context  School District* context  DOCUMENT TYPE pre-2005 2005-2013 post-2013   pre-2005 2005-2013 post-2013 TOTAL Memos       2 2           Implementation tool   4   4  1 9           Reports  2 5 1  6 2 1 17           Event documents  2 3    3 8           Correspondence   1 1  2   4           Presentations   2    1 2 5           Evaluation tools   1  1   2           Policies** 2 2 1     5           Newspaper articles      4   4           Transcripts of government sessions  7  1     8 TOTAL 11 16 8  17 3 9 64**  * These documents were mainly from two school districts (situated within two different Regional Health Authorities) that have been relatively active in implementation and thus had numerous documents to explore. Elaboration on this decision is provided in the strengths and limitations section. ** One federal-level policy document from the 2005-2013 range of years was used, making the total 65  Documents were collected iteratively from a number of sources including materials provided at public events, online searches from open sources, searches conducted in the BC legislative archives, and directly from or on the advice of interview respondents. The documents were selected for their relevance to: a) the Guidelines themselves and b) the context of school food environments and childhood obesity in BC to garner an understanding of how this issue and intervention have been framed both in the years prior to the intervention and after its implementation. Documents were 56  categorised as “pre-2005” (before the first set of Guidelines), “2005-2013” (implementation), and “post-2013” (implementation following the release of the most current version in 2013).   I attended four public events that provided resources and information to support implementation of the Guidelines (to parents, teachers, and school food staff). This was done to gain an understanding of how the Guidelines are being ‘marketed’ to implementers by health care staff and other advocacy groups, the discourses used, and to collect information about implementation activities. I also participated in informal discussions, and created field notes to record observations.   The purpose of the document analysis and participant observations were to: 1) identify the ‘on-paper’ values and beliefs of policymakers embedded in the Guidelines from pre-2005 to post-2013; 2) elicit information regarding the political, economic, and structural context in which the Guidelines were developed; and 3) inform semi-structured interviews with stakeholders involved in the development and evolution of the Guidelines.  Purposive, snowball recruitment was used to garner participation for semi-structured interviews with key stakeholders. In total, I interviewed 27 policy and decision-makers and implementers from the health sector at the provincial and regional levels, ranging from those involved in school food and beverage environment work before 2005 to those who currently work with the Guidelines. While all of the ‘implementers’ interviewed were dietitians from across the province, many of them had also played a role in the evolution of the Guidelines through stakeholder consultation processes over the years in addition to having a history of working closely with the education sector. An interview was also conducted with the only decision-maker from the education sector that had involvement with the creation of the Guidelines. As stated earlier, for the most part, the Guidelines were developed by the health sector and, as such, it was expected they would be able to provide the best insight into how they expected the Guidelines to work. Six semi-structured interviews were conducted with stakeholders from the private sector responsible for providing food and beverages in some schools and districts. The interviews lasted from 45 to 90 minutes and focused on the perceived key mechanisms respondents expected would play a role in achieving the desired outcomes of the Guidelines (Appendix C). A focus on the 2013 Guidelines in the interview process helped to avoid lapses in recall, capture perspectives of stakeholders who are still active, and more accurately reflect the current state. This is important because the nature and magnitude of intervention activities over 57  the past decade has been inconsistent and variable. With permission from the respondents, all interviews were digitally audio-recorded and transcribed verbatim.   All data were analysed using the same approach. Analysis was ongoing and iterative throughout data collection. The coding process, using Atlas ti software (Atlas.ti, 2015), was both deductive and inductive. Basic components of a logic model were used as a priori codes, including: 1) problem statement, 2) goals and objectives, 3) inputs, 4) specific activities, 5) expected outputs, 6) short, medium, and long-term intended outcomes.  Inductive codes were developed guided by broad questions, as suggested by Rapley (2007), such as:   What is the rhetorical work of the text and how does it seek to persuade?   What understanding of the world emerges from the content? A retrospective, BC-specific logic model and program theory was created from this analysis. This process involved first creating a BC-specific logic model that provided a visual image of all the specific components the Guidelines as an intervention. The logic model emerged through creation of memos for each a priori code corresponding to a logic model component (i.e., inputs, activities, etc.). Within each memo, related coded excerpts were included and a timeline was created that displayed the evolution of that component based on the coded excerpts. The coded excerpts in each memo were synthesised into broader categories. Some components remained consistent between 2005 and 2013 while others did not. Components of the intervention that no longer existed were excluded from the final logic model.    The open, inductive coding process led to the development of a series of codes related to: (1) justifications used for the creation and implementation of the Guidelines over the years; (2) discourse surrounding the Guidelines; (3) intervention activities, (4) beliefs and opinions of various policymakers and implementers about the Guidelines, and (5) how this intervention was intended to work (See Table 4.2 for examples of inductive codes that emerged from the data).      58   Table 4.2 Examples of inductive codes CODES (for example) SUBCODES JUST_JUSTIFICATION FOR INITIATIVES just_society_health   just_society_financial   just_student_physical health   just_student_cognitive/psychosocial/emotional health     RHETOR_RHETORIC    rhetor_capitalism   rhetor_harm reduction   rhetor_food vs tobacco comparison   rhetor_comprehensive school health  rhetor_students as consumers     DIEBEL_DIETITIAN BELIEFS diebel_whole foods vs nutrient based   diebel_personal history   diebel_how to engage schools   diebel_general perspective on the guidelines  After considering the diverse insights from stakeholders, in conjunction with the depiction of the programmatic components in the logic model, broad logical statements were developed to theorise how the program was expected to work (i.e., if-then statements).3 These logical statements were then situated within existing social theories. Finally, elements of the logic model and the theory-based if-then statements were combined to create the broader program theory for school food and beverage                                                           3 These refer to hypothesised context + mechanism  outcome configurations used to guide phase two data collection and analysis. I made the decision to call them ‘if-then’ statements for the purposes of this manuscript so as to focus only on the development of the broader program theory itself rather than the realist evaluation approach, which I only decided to use after data collection and analysis for this phase was almost complete. This late decision-making is not problematic, however, as it has been proposed that program theories, like the one developed in this manuscript, can be used to formulate the specific lines of inquiry for a realist evaluation (Blamey & Mackenzie, 2007). A complete table of the somewhat rough, naive, simplistic hypothesised CMOcs, derived from the broader program theory, and used to direct phase two, is in Appendix D.  That they are rough and naive is not problematic because CMOcs are not meant to be the ‘star’ of realist evaluation but are meant to be  used only as an aide memoire for looking at evidence with a generative understanding of causation (Pawson, 2008, 2013). 59  sales environment interventions more generally. This was presented to various stakeholders from the health sector to ensure plausibility and their feedback was incorporated where appropriate. 4.3 Findings The overall goals of the Guidelines are to contribute to the reduction of childhood obesity, and improve child health, development, and learning capacity by improving school food and beverage environments.  The core aspects of this intervention are laid out in an 84-page document containing nutrition information, justifications for implementation, and a variety of tools to help assess whether food and beverage items being sold in schools meet the nutrient criteria.    The 2013 Guidelines are a key component of a Comprehensive School Health (CSH) approach, as emphasised ‘on-paper’ and at public events:  CSH is a model for helping educators, health practitioners, school staff, students, parents and others...create an environment that makes their school the best place possible to learn, work and play. The Guidelines are an example of a healthy school policy that schools can use and expand upon within an overall healthy schools approach. (Government of British Columbia, 2013, p. 5)  The Joint Consortium for School Health (JCSH) developed its first CSH framework in Canada in 2008, which was subsequently adopted by BC stakeholders (Directorate of Agencies for School Health in BC [DASH], 2013). The JSCH CSH framework includes four dimensions: (1) social and physical environment; (2) healthy school policy; (3) partnerships and services; and (4) teaching and learning (Pan-Canadian Consortium for School Health [PCCSH], 2015). The CSH approach represents an ecological perspective of health promotion in that it considers not only education at the individual level, but also the social, physical, and political environment, and the need for support from the larger community. The Guidelines, as a broad intervention, address the policy dimension of CSH. However, the actual intervention inputs and activities that have occurred to-date, the most current of which are reflected in the logic model and program theory presented in the following sections, address other dimensions of the CSH framework in an interactive and dynamic manner. 60  4.3.1 A logic model for British Columbia’s school food and beverage sales policy Figure 4.1 presents a hypothesised integrative logic model specific to the BC Guidelines, which theorises and articulates inputs, activities, outputs, and anticipated outcomes that emerged from the analysis of documents, interviews, and observations at public events. 61    Figure 4.1 Retrospective logic model specifically for the 2013 Guidelines for Food and Beverage Sales in BC Schools * CSOs= Community Service Organisations; HLBC=Health Link BC, dietitian hotline62  Inputs and Activities The logic model illustrates the multi-sectoral nature of this intervention with inputs and activities expected to come from, and be undertaken by, all levels of the public health and education sectors (provincially, regionally, and locally) as well as the private sector. The BCMoH and BCMoEd, along with BC Regional Health Authorities (RHAs), and the private sector have contributed human resources towards developing the Guidelines as they evolved from 2005 and beyond. The BCMoH dietitians, in conjunction with RHA dietitians, are responsible for establishing and maintaining technological infrastructure to host implementation resources like the online database of products that have been assessed for compliance. Both have also created numerous implementation tools such as cook books with compliant foods and worksheets to use for manually assessing nutrition labels on products.  According to interview respondents involved in revisions of the Guidelines, funding was provided by the provincial government to hire consultants to facilitate the process of revising the Guidelines and to pay for design and printing of the 2013 Guidelines document and associated resources like an abridged version of the Guidelines called “Guidelines at a Glance” (Government of British Columbia, 2016). No respondents reported any additional funding specifically allotted by either the BCMoH or the BCMoED for implementation in schools outside of existing human resources.   Dissemination The BCMoED was expected to disseminate the Guidelines to school districts while districts, in turn, were expected to disseminate them to individual school communities who then further disseminate to relevant school-level stakeholders. The BCMoH does the same for its RHAs and relevant staff (RHA Public Health Dietitians, for example). It was reported that health sector stakeholders disseminate pertinent Guidelines information and resources to industry stakeholders, to school districts and individual school stake-holders. As one dietitian stated: “In our school district... they did ask us to do the presentation [about the Guidelines] to all the principals, which we did” (Dietitian, P13).  Collaboration  Collaboration is expected to occur vertically and horizontally across sectors. For example, one dietitian reported that they had suggested that a teacher responsible for the school store might work 63  with a Community Service Organisation (CSO) to improve the food offered for sale in their school store: … working with marketing teachers [in schools] and asking that rather than having the school stores sell food that doesn’t align with the Guidelines would [they] consider working with [the school market garden] and marketing products from [there]. (Dietitian, P1)  Both provincial and regional dietitians are also available to collaborate with private sector food vendors to support them in adhering to the Guidelines:  [A large food distributor] wanted to develop some menus that would meet the school food guidelines and so they asked for our support…. I provided an in-service [training workshop] to them about what the Guidelines were and then worked with them a little bit to confirm that their menus were meeting the Guidelines. (Dietitian, P3)  Assessing and selling  District and school-level implementers (district staff, principals, parents, teachers, food staff), as well as private sector vendors who might want to have their products sold in schools, are expected to use the Guidelines and the available implementation tools to assess foods and beverage offerings and make changes to ensure compliance.  While this intervention was described by participants as focused on student health, none of the intervention activities are directed toward students, but rather, activities are directed towards adults. These include assessing products for compliance or hosting events to provide information about the Guidelines to adults in decision-making roles enable them to shape the transformation of the school food environment. In the case of students who sell food in school stores for a class or for a fundraising initiative, proposed implementation was discussed more in terms of supporting teachers to implement the Guidelines by advising students appropriately rather than a direct focus on students. The expectation then is that adults are required to change their behaviours to implement the Guidelines while students are beneficiaries.     64  Outputs The logic model contains implementation output targets. Outputs are specific effects of an activity. They reflect the links between the activities and outcomes and can help with devising indicators for an implementation evaluation. From the data, the only expected output with an explicit numerical target was that at all times at least 50% of all items for sale must fall into the ‘sell most’ category with the remaining falling into the ‘sell sometimes’ category. The remainder of the outputs selected to be in the retrospective logic model are offered as potential implementation indicators, which I logically devised from the intervention activities, and include: an updated set of Guidelines and implementation tools exist that reflect up-to-date  nutrition-related evidence, new tools are created based off of stakeholder consultations, there is dissemination and use of available implementation tools, there is an increased awareness of how well schools are complying with the Guidelines, and that changes are made, if needed, to what food and beverages are sold. Regarding the vendors, an important output might be that there is an increase in compliant products available in the marketplace that schools may offer students. That there are almost no explicit implementation output targets may reflect the limited consideration given to development of implementation and evaluation strategies by policymakers. Rather, the focus has been on achieving a singular implementation process output—to change what products are sold. While this is an important output, if not the most important, it is only one output in a long implementation chain that was never articulated.   Outcomes The expected activities and outputs outlined in the logic model are projected to lead to a multi-dimensional set of outcomes, including immediate expected outcomes, such as education stakeholders perceiving implementation resources as easy to use and harmonising implicit food messaging with what is being explicitly taught in schools:  …our kids are faced with being in an environment...where there are no healthy foods available… we’re trying to change the environment… so that [it] reinforces and reflects what we, and hopefully their teachers, are going to try to teach them in terms of healthy choices... (Dietitian, P10)  The analysis also showed that intermediate outcomes relate to psychosocial, physical and cognitive outcomes for students, and long-term outcomes include overall decreases in diet-related morbidities 65  and mortalities along with cultural shifts in beliefs about school nutrition and nutrition in general. Moreover, according to those involved in creating and revising the Guidelines, this cultural change would lead to long-term demand for healthy food and beverage options outside of schools and the food industry would rise to meet this demand, effectively changing the wider food environment. 4.3.2 Program theory for British Columbia’s school food and beverage sales policy The retrospective logic model (Figure 4.1), as a reflection of health sector interviews and public documents related to the Guidelines, clearly articulates components of the intervention, helping to identify which are evaluable, but does not sufficiently capture the processes expected to occur between intervention components. For this, I developed a program theory to theorise what underlying mechanisms are expected to lead to stakeholders taking action to implement the Guidelines.  The program theory makes explicit the ‘how’ of the intervention: how are the intervention activities expected to lead to the outputs and outcomes? While it was presented above that the Guidelines are meant to be a key intervention of a CSH approach, the CSH conceptual framework is insufficient as a ‘program theory’ due to its lack of elaboration of the dynamic interactions and theoretical mechanisms of its four dimensions. Further, the Guidelines can be categorised as an ecological approach to improving nutrition in that they aim to alleviate structural barriers to creating healthy school food and beverage environments. Ecological approaches, though, still require individual implementers to actually implement; to change the behaviour of those responsible for creating school food and beverage environments. It is implementer behaviour that is expected to determine whether the larger goal of improving children’s health is achieved. The hypothesised program theory below (Figure 4.2)  reflects how this intervention, from the perspective of the health sector, has largely consisted of convincing and supporting adults to change the way they create school food environments.66    Figure 4.2 Hypothesised underlying program theory specifically for BC’s Guidelines67  The role of information and education campaigns While there was some evidence from informal conversations to suggest the existence of motivated stakeholders in schools who desired to change their school food environments but were constrained by infrastructure (e.g., lack of food preparation space and financial resources to provide healthier options), intervention activities have not typically focused on addressing these constraints. Intervention activities reported by health sector staff have revolved around information and education campaigns geared towards adults responsible for creating school food and beverage environments (i.e., school district staff, school food staff, principals, parents, and vendors) about: (1) the magnitude of the problem of unhealthy school food environments, (2) how these environments may be contributing to overweight and obese children and the lifelong impacts being overweight or obese, (3) the existence and benefits of the Guidelines and the expectations contained within, and (4) the available tools for implementation and how to use them.  First, health stakeholders aimed to provide education stakeholders with knowledge of the magnitude of the problem of childhood obesity or other diet-related morbidities to convince adults to change their practices when creating school food environments. This has been done over the years through the use of the use of epidemiological statistics. For example: “…it is estimated that the prevalence of diabetes in BC will be 1 in 10 by 2020.” (Dietitians of Canada, 2014, p. 1, Briefing for the BC Government).   Also, the health sector has made efforts to inform adults of how susceptible children are to becoming overweight or obese through the consumption of ubiquitously available sugar sweetened beverages: For example:   One can of pop contains nine teaspoons of sugars; one can of sweetened fruit beverage contains 12 teaspoons… The result? Pop consumption has increased 500 per cent in the past 50 years….teens drink twice as much pop as milk….Obesity in Canadian children has basically doubled since the early 80s. The reasons for this are multiple, and are complex, but nutrition and obesity researchers are recognizing the significant role that sweetened drinks play in the equation. (Richards, 2003, p.1, newspaper opinion piece written by Public Health Dietitian)  68  Health sector stakeholders aim to convince education stakeholders of the benefits of implementing the Guidelines with arguments such as “next to parents, schools have the most impact on shaping students’ eating habits” (Interior Health Authority & The Vancouver Foundation, 2004, p. 1) and the school food environment “contributes to children being overweight or obese”’ (British Columbia, Provincial Health Officer, 2008, p. 59, BC Government report). The benefits for students, health stakeholders have argued, are not only benefits to physical health but also cognitive and psychosocial well-being. The life-course perspective has also been promoted, providing information on how healthy behaviours developed in childhood carry over into adulthood and therefore implementing the Guidelines can lead to children growing into healthier, more successful adults:   Schools teach kids valuable lessons about healthy lifestyles. The school environment, and the foods sold in school, can offer kids a chance to act on these valuable lessons. Making wise food choices increases self-esteem and leads to further lifetime accomplishments! (Okanagan Service Area Nutritionists, 2001, p. 4, brochure for parents)  The Guidelines, as an intervention, also include supportive information and implementation resources and tools to make it easier for implementers to assess whether food and beverage offerings fit within the nutrient criteria and to provide ideas for specific actions. One example is the worksheet that helps guide school-level implementers through a series of mathematical equations to assess sugar, fat, and salt content to determine if foods they have prepared from scratch comply:  I think the addition of the checklist is really helpful. At first glance, it’s very overwhelming, but for those folks that are working in food service…once they dig into it, I think they would find that it is quite user friendly and very helpful in ensuring that the foods that are sold through food services also meet the Guidelines. (Dietitian, P10)  Ultimately, the resources and implementation tools on offer are meant to enhance the confidence of stakeholders to take action to change their school food and beverage environments.   Responsibilisation and morality: ‘just the right thing to do’  Much of the information disseminated by the health sector to education stakeholders tries to ‘convince’ them that implementing the Guidelines is a matter of doing what is best for children. 69  While the Guidelines are mandatory, they are not “supported with a compliance and enforcement program” (Government of British Columbia, 2013, p. 10). This reflects the way in which many present-day governments intervene in society through a process of responsibilisation. Opting for less heavy-handed mechanisms like mandated guidelines with no official sanctions, rather than legislation with sanctions, neo-liberal governments assume the “reflexive moral capacities of...social actors” (Shamir, 2008, p. 7) to problem-solve within their own communities. This moral reflexive capacity refers to community-members being aware of their duty and having both capacity and motivation to take action. This ideally leads people to take action without coercion (Selznick, 2002) rather than through, for example, the threat of sanctions for non-compliance. In the case of this type of intervention, this moral dimension includes the assumption that adults—including all district and school-level implementers—will be motivated to implement this intervention once they become aware of the magnitude of the health problem and the susceptibility of children through information and education campaigns.  In the Canadian context, parents are important stakeholders expected to implement the Guidelines because they are often responsible for fundraising to provide a well-rounded curricular experience for children beyond core instruction. An important implicit, basic assumption of this intervention has been that adults possess an innate motivation to take action to ensure their own children have healthy food and beverage options while in school to provide them with the best possible health outcomes in the short- and long-term. This basic assumption is reflected in the following quote from a dietitian/parent speaking to government officials:  I’ve spoken to about a thousand [student] athletes…this has quite often been as a volunteer parent, because I have three active children—…why is it important for you to choose healthier food choices? Because you want to give them the winning edge...and tips for eating for peak performance. (Legislative Assembly, Province of British Columbia, 2006, p. 154, Dietitian speaking before the Select Standing Committee on Health)  Another assumption of the Guidelines is that adults, whether district or school-level staff or parents, will exhibit pro-social behaviour. They will not only care about their own children’s ability to access healthy foods, but also they will be motivated to ensure other people’s children have the ability to access healthy options. Throughout the years, individuals and organisations from the health sector 70  have attempted to convince various stakeholders to care about other people’s children when it comes to nutrition interventions. These pleas often draw upon the shock-value of the statistics from classically defined vulnerable communities, citing statistics about overweight, obesity, diabetes, child poverty and food insecurity. For example: “...young people of aboriginal origin (off-reserve) had a significantly high combined overweight/obesity rate of 41 percent, including an obesity rate 2.5 times the national average” (Select Standing Committee on Health, 2006, p. 23, BC Government report).  While these kinds of statistics are likely used to invoke empathy (what if this was your child?), they are also used in conjunction with economic arguments; if you care about these children now there will be economic benefits in the future. This turns decision-makers, who make up government, into stakeholders with advocates hoping this will motivate them to filter more resources towards interventions:  The underlying driver [for the Guidelines]...comes down to obesity. And who cares about obesity? In the end it really comes down to health expenditures which is really where the axe tends to fall…. I think if obesity did not cost any more money [we wouldn’t do anything about it]… (Dietitian, P3)  Another moral perspective used to promote the development of the Guidelines in the early years was the argument for the vulnerability of children, in general, and their need for adult protection. In an article from a regional newspaper in 2003, as advocacy efforts for improving school food were culminating, a school trustee was quoted arguing this moral duty:  It’s ground breaking turf for the board…but the board was adamant it wanted to take a leadership stance. “It’s just the right thing to do”, said trustee [name]… “we can’t hold our children’s health to the highest bidder,” referring to the profits made by junk food sales. (Nichol, 2004, p. unknown, newspaper article)  The moralising arguments that emerged from the data reflect the foundational belief in governance through responsibilisation underpinning the Guidelines; that when citizens are convinced they have a 71  moral duty to act and solve their own problems in their communities, they do not need to be legislated to act.  4.3.3 A broad program theory: combining BC-specific logic model and program theory Figure 4.3 below combines elements of the BC-specific logic model and the hypothesised BC-specific program theory and presents a broader program theory for school food and beverage sales environment interventions, linking the different components of this type of intervention. This retrospective program theory, while developed by starting with the specific case of BC’s Guidelines, is meant to be a somewhat higher level of abstraction of how school food and beverage sales environment interventions are expected to work. Therefore, this program theory is relevant to contexts outside of BC. 72    Figure 4.3 A hypothesised broad program theory for school food and beverage sales environment interventions  73   4.4 Discussion To my knowledge, this is the first time an articulation of a program theory has been proposed for school food and beverage sales policies. The retrospective broad program theory emerged from this study highlights that the explicit objective of the intervention addresses children’s well-being and that its overarching strategy reflects an ecological approach to health promotion. However, the on-the-ground activities and outputs describe a behaviour change intervention targeting adults, motivating them to create healthy school food environments, while students are passive beneficiaries. This is distinct from other types of school nutrition interventions whose theories of change include activities targeted towards both students and adults (Li et al., 2014; Perez-Rodrigo et al., 2005; Ratcliff, 2012). In the BC context, these school food and beverage  guidelines are just one part of a complex of school nutrition interventions that not only includes health education as part of curriculum but also other unique, albeit often piece-meal and organised by CSOs, nutrition-related initiatives geared towards students (Mansfield, 2016). It is possible that the creators of the Guidelines, in not incorporating activities geared towards students, expected existing nutrition education and other private sector initiatives to be sufficient to lead students to accept food environment changes.  The intent of the intervention resources that have been, and continue to be, offered to adults to convince them to change the way they create school food environments is reminiscent of the Health Belief Model (HBM) (Alberta Health Services, 2010), comprised of six areas in which to intervene to convince people to change their behaviours: (i) perceived susceptibility; (ii) perceived severity; (iii) perceived benefits; (iv) perceived barriers; (v) cues to action; and (vi) confidence in one’s ability to take action. The intervention resources related to the Guidelines are in alignment with these dimensions and aim to motivate adults to change their behaviour (e.g., convince adults that there is a child obesity problem, convince adults their child is susceptible, etc.). This classic conceptual framework, described previously in Chapter 2, has been criticised for focusing too much on individual behaviours that impact health (i.e., in this case, the behaviours of adults in creating school food and beverage sales environments), blaming individuals for the creation of their own poor health (or in this case, the poor health of children) with little consideration for structural determinants (i.e., in this case, all the external pressures that cause adults to create food and beverage sales environments the way they do such as a lack of food preparation space or funding to provide 74  healthier options) (McLeroy et al., 1988). Regardless of the ultimate objective of this school food environment intervention—to passively improve health behaviours of students through addressing a structural barrier (i.e., food availability)—it is individual adults who must change their behaviours when deciding whether or not to take action towards changing the environment. This reinforces the necessity of not only conducting implementation research with an ecological focus (what types of foods and beverages are available for sale and how well do they comply?) but also considering the individual experiences of those responsible for implementation and the social processes underlying their behaviour change.   The larger social context influences whether and how these adults make changes to their school food environments. Responsibilisation assumes a moral imperative is sufficient to motivate people and communities to solve their own problems (Shamir, 2008). However, neglecting to consider structural constraints to behaviour change, even when a morally-based motivation exists, contributes to a lack of understanding of why and how implementation works in some contexts but not others (Pawson, 2013). This is reflected in existing evaluations of school food and beverage sales interventions, which tend to focus on quantitative measures of food and beverage availability as the implementation outcome of interest (Aarestrup et al., 2015; Caspi et al., 2012; Kubik et al., 2013; Turner, Ohri-Vachaspati, Powell, & Chaloupka, 2016; Watts et al., 2014). These studies report various levels of compliance and present potential contextual influences on compliance, including: school size  (Hills et al., 2015; Kubik et al., 2010), school types (Hills et al., 2015; Kubik et al., 2015; Kubik, Lytle, Farbakhsh, Moe, & Samuelson, 2009; Morin et al., 2012; Turner & Chaloupka, 2012), and socioeconomic demographics of schools and their communities (Kubik et al., 2011; Nanney & Glatt, 2013; Pledger, McDonald, & Cumming, 2012; Taber et al., 2015; Turner et al., 2016).  They provide important information but do not explain why implementation is successful in some of these contexts but not in others. To help explain this, a small body of qualitative literature has explored stakeholder-reported barriers and facilitators to implementing school food and beverage sales policies (e.g., MacLellan et al., 2010; Pettigrew et al., 2012b; Vine & Elliott, 2014).  The broad program theory presented here suggests outputs and short-term outcomes that could also be explored to help with understanding the different levels of compliance in different contexts observed in the findings of the above-mentioned evaluations. For example, because this is a responsibilising intervention that relies on morally inducing the motivation of people to take action 75  and address their own issues, elaboration of the motivations of implementers would be useful. This would not only provide insight into how (and if) responsibilisation processes are effective for changing school food environments, but would also contribute to the growing body of intervention literature around responsibilisation approaches being used to address social problems, including food security (Bastian & Coveney, 2013) and obesity (Share & Strain, 2008). This may be important given alternatives to a responsibilising approach, like legislating or mandating people to comply with a school food and beverage sales policy, is not necessarily effective all the time, in all contexts for ensuring sustained compliance (Gorski et al., 2016; Kubik et al., 2013; Ohri-Vachaspati et al., 2012; Samuels et al., 2010).  Not only would exploring the motivations of stakeholders be useful to help explain some of differences in compliance levels in different contexts, but so too would process evaluations that focus on awareness of and perceived usefulness of intervention resources. This could shed light on whether stakeholders are being provided intervention resources that align with their needs, an important influence on the decision to implement (Moreno & Moriano, 2016; Rashidian, Eccles, & Russell, 2008). The few studies that exist regarding school nutrition policies and stakeholder awareness center on awareness of the existence of the policy (Downs et al., 2011; MacLellan et al., 2010; Ohri-Vachaspati et al., 2012). Awareness of policy is important but so too is awareness of available implementation resources and tools, and how useful they are perceived to be (Adamson et al., 2013). This is a critical link in the journey from creation of intervention and resources to behaviour change (Bressers & Lulofs, 2010) and, as such, requires consideration.  Strengths and limitations A strength of this research is that data were collected from many of the key architects of the Guidelines within the health sector, increasing the validity of the program theory articulated here. Moreover, although this is a retrospective program theory, in contrast to one created alongside the development of the original Guidelines, it was presented to key health sector stakeholders and their feedback was incorporated into the final theory presented here. While this exercise contributed to supporting the validity of the retrospective logic model and program theory, it could also be argued that creating this broad program theory using almost only health sector stakeholder perspectives is a limitation. However, the purpose of this exercise was to develop a program theory that reflected what the creators of the Guidelines, who were mostly health sector staff, expected to happen. A 76  potential limitation is that regional documents were analysed from only two of BC’s 60 school districts. However, these two districts were known for their high level of engagement in improving school food environments and, as such, provided information into the types of implementation activities that have actually been occurring. I selected not to pursue seeking documents in any other districts because there are 60 school districts in BC and doing this extensive of a search and analysis would have not been possible within the constraints of this research. However, the purpose was not to explore specific district contexts but rather to gather an idea of activities that had been occurring throughout the years, bolstered by the other data collection strategies as well, to help shape this hypothesised broad program theory.  Conclusion and lessons learned The broad program theory devised in this paper was articulated to form the basis of a realist implementation evaluation of the BC school food and beverage sales Guidelines. A critical assessment of the process of its development, though, can be used to inform the creation of future theories of change for similar interventions. First, the program theory was based almost solely on health sector stakeholder perspectives. This coupled with the inherent potential for logic models to be oversimplifications of complex interventions (Bickman, 2000), ultimately led to the documentation of a relatively un-nuanced program theory: provide information, resources, and cues to action in order to motivate adults to change the way they create school food and beverage sales environments. An inclusive program theory development process – ideally prior to developing any particular intervention -- may facilitate the creation of a more nuanced and precise program theory that includes, for example, the identification of specific important contextual factors or theorising feedback loops and unintended consequences.  By “inclusive”, I mean inclusion of the multiple stakeholder types involved in implementing school food and beverage sales standards, such as private vendors, community-service organisations, and on-the ground education stakeholders. Kaplan and Garrett (2005) found that when logic models are developed with multiple stakeholder types, they are not only more complete but the process helped foster collaboration among the stakeholders subsequent to model development. Moreover, an inclusive process may also broaden the articulation of priorities to be included in the program theory. A more complete logic model might include a wider array of activities, process and outcome indicators relevant for different stakeholders, increasing acceptability of the intervention (Sekhon, Cartwright, & Francis, 2017). In a school food and beverage sales environment intervention, these may include indicators devised by 77  private industry or even indicators more relevant to an education system’s mandate. Including student voices in the process may also contribute to the overall acceptability of an intervention. Implementation success is not just dependent on acceptance by intervention deliverers but also of recipients (Diepeveen, Ling, Suhrcke, Roland, & Marteau, 2013).     The broad program theory presented here adds to the larger conceptualisation of not just school food and beverage environment interventions, but about criteria for evaluating food environment interventions more broadly. Developing this broad program theory provided a glimpse into the complexity of intervening in school food and beverage environments as well as insight into underlying assumptions and belief systems contributing to the decisions of intervention creators who use this specific type of approach. First, nutritional guidelines like the ones explored in this paper clearly fit within the internationally accepted CSH approach to addressing children’s health (Pan-Canadian Consortium for School Health, 2015; World Health Organization, United Nations Children’s Fund, & United Nations Educational, Scientific and Cultural Organization, 1991). Furthermore, that they are underpinned by aspirations of responsibilisation aligns with present-day hegemonic ideals of neo-liberal governance. In practice, however, the public sector continues to provide crucial implementation support for stakeholders expected to implement these nutritional guidelines.    Nutritional criteria applied to all schools in a region, province/state or country could be considered a population-level health intervention, addressing structural barriers to healthy eating and they have been found to be associated with healthier food environments and health benefits (Mullaly et al., 2010; Sanchez-Vaznaugh et al., 2010; Watts et al., 2014; Hills et al., 2015). From a broad perspective, the Guidelines are indeed one such population-level health intervention. Viewing it as a population-level health intervention has implications for evaluations, which will often attempt to measure outcomes across a large segment of the population, like measuring food and beverage availability across an entire region or country. These types of studies ultimately conclude that the intervention is only successful in some places some of the time. Evaluations such as these offer little by way of explanations for why implementation of nutrition standards is successful in one place and not another. To find these explanations requires a shift in perspective when conducting an evaluation: namely that all interventions, even population-level health interventions, must be taken from paper-to-practice by individual people in their individual circumstances via long complex 78  implementation chains. A program theory that reflects this practical reality is crucial to the development of an evaluation framework that seeks to provide an in-depth explanation as to why things are happening as they are.79  Chapter 5~ Social processes associated with district-level implementation of British Columbia’s school food and beverage sales policy: what works and why?  5.1 Introduction Interventions for improving school food environments have been reported to have numerous benefits to children’s health, including improved eating habits (CDC, 2012; PHAC, 2013; WHO, 2008) and decreased body mass indices (Sanchez-Vaznaugh et al., 2010; Taber et al., 2012). One increasingly popular approach to improving school food and beverage environments is implementation of criteria to improve nutritional quality of foods and beverages sold in schools. School food and beverage sales policies have been adopted in various jurisdictions including the United States (US), Australia, and Canada (e.g., Downs et al., 2012; Hills et al., 2015; Kubik et al., 2015; Pettigrew et al., 2012a; Phillips et al., 2010; Samuels et al., 2010; Vine & Elliott, 2014).    In Canada, health programming and public education are decentralised to the provincial level. Therefore, regulation of food available in schools is also decentralised, meaning that development and implementation of school nutrition policies is left to provincial ministries of health and education. School districts within each province may also develop policies independently. Most provincial and territorial governments have developed and implemented their own school food and beverage sales guidelines (Holmes, 2016), including the province of British Columbia (BC). The Guidelines for Food and Beverage Sales in BC Schools (“the Guidelines”) were created in 2005 by the BC Ministries of Health (BCMoH) and Education (BCMoED), with the most recent iteration launched in January 2014 (Government of British Columbia, 2013). These mandatory guidelines apply to all public schools, aiming to reduce the amount of salt, sugar, and fat in items sold to students in all food sales venues in schools. These include vending machines, cafeterias, fundraisers, sports days, and school stores. BC’s Guidelines require that at least 50% of all items for sale fall within the “sell most” category while the remaining items available for sale fall within the “sell sometimes” category. Schools are not to sell any items that fall within the “sell never” category. Previous BC-based studies have found that while school food sales environments have improved since the launch of the Guidelines, improvements have not necessarily occurred in all schools (BCMoEd & BCMoH, 2008; Watts et al., 2014). As there are 60 school districts with varying geographies and demographics in BC, the BCMoED further decentralises power to the district level, whereby districts determine how to carry out top-down directives, including implementation of 80  mandated school food and beverage sales guidelines. Districts may thus incorporate implementation processes differently, with varying levels of intra-district and inter-district compliance. Differential implementation of a public health intervention, like the Guidelines, has the potential to exacerbate health inequities if it is successfully implemented in some contexts but not others (MacDonald et al., 2016).  Research from other jurisdictions shows the important role school districts play in implementing food and beverages sales policies. In the US, studies have found district policies influence students’ consumption of sugar-sweetened beverages and significantly reduce availability of unhealthy items for sale (Chriqui et al., 2013; Kubik et al., 2013; Larson et al., 2016; G. F. Miller, Sliwa, Brener, Park, & Merlo, 2016; Mozaffarian et al., 2016). Also in the US, researchers have proposed that weak district policies and lack of accompanying resources for enforcement from the district level are potentially responsible for perpetuating poor school food and beverage environments (Belansky et al., 2010; Bergman et al., 2015).   In Canada, studies from various provinces have identified barriers and facilitators to implementing school food and beverage sales policies at the school level (e.g., Downs et al., 2012; MacLellan et al., 2010; Masse et al., 2013).  District support for implementation has been acknowledged by school-level stakeholders to be an important facilitator of implementation, including in BC (Masse et al., 2013; Mcisaac, Shearer, Veugelers, & Kirk, 2015).  However, little research has elaborated on this in the Canadian context and explored social processes associated with implementation occurring at the level of school districts to help explain how and why district practices may contribute to different levels of compliance. A nuanced understanding of this can offer leverage points that may be useful for intervening to improve implementation and, consequently, compliance.  Realist evaluation A realist approach to evaluating the implementation of complex public health interventions, like BC’s Guidelines, can begin to provide this nuanced understanding. Realist evaluation attempts to explain the inconsistent success of an intervention by assessing the relationships between contexts (C), mechanisms (M), and outcomes (O). First, context is considered the backdrop of an intervention, pre-existing dimensions of which may influence mechanisms and, therefore, outcomes.  Although the concept of context is broad and can include, for example, social or cultural norms, 81  histories of organisations or people, and governance structures, realist evaluation aims to identify what dimensions of a broad range of potential contextual factors are the most relevant for the success of a specific type of intervention. Second, mechanisms, according to realist methodology, are defined as the way in which stakeholders respond to the concrete and non-concrete resources provided by an intervention, such as funding (a concrete resource) or the fear of ill health created through a health education campaign (a non-concrete resource). Responses to intervention resources can be cognitive, emotional, or motivational, or anything else that happens in the mind of stakeholders who encounter the intervention resources. This can help explain why stakeholders make decisions about if and how they will take action. Additionally, the way individuals respond to intervention resources are influenced by pre-existing specific dimensions of context (Jagosh et al., 2012). The interaction of contexts and mechanisms leads to outcomes, which are simply defined as effects of an intervention, whether intended, unintended, short-term, intermediate, or long-term. In the case of BCs school food and beverage sales intervention, as proposed in the hypothesised program theory in the previous chapter (Figure 4.3), the intervention activities consist of dissemination of knowledge to motivate expected-implementers (mechanism) to take action to change the school food environment (outcome). Once expected implementers are motivated, the intervention provides a variety of capacity support to enable them to act (mechanism).  Realist evaluators theorise that interventions will lead to desired outcomes only when intervention resources are launched in facilitative contexts (Jagosh, 2017c; Pawson & Tilley, 1997). A facilitative context is a context that will promote stakeholders to decide to take action to implement (or take up) an intervention. If an intervention resource is offered in a facilitative context it will trigger the desired mechanism in the minds of stakeholders. It is this triggering that prompts the decision to implement (or take up) the intervention. Equipped with a deeper understanding of what is and is not a facilitative context for an intervention, decisions can be made prior to its launch as to whether it can be introduced ‘as is’ or whether other pre-intervention or coinciding activities are necessary to foster a facilitative context (Jagosh, 2017c). 5.2 Research objective This chapter aims to explore the social processes associated with district-level implementation of school food and beverage sales standards, focusing on implementation of the Province of British Columbia’s, Guidelines for Food and Beverage Sales in BC Schools. Specifically, I explore the influence of contexts on mechanisms that drive district-level stakeholders to take action to 82  implement. What dimensions of context are facilitative? What perceived intended or unintended process-outcomes are occurring as a result of the district-level processes of implementation? Are there potential implications for different types of district-level implementation processes to contribute to health inequity?4  5.3 Methods Realist evaluation begins with a program theory, a plausible explanation of how the intervention will achieve the desired outcomes (Bickman, 1987). This was articulated in an earlier phase of the research where I described the intervention using a logic model and program theory (Chapter 4). In summary, the program theory for school food and beverage sales policies, as articulated by the ‘authors’ of the intervention, is: provision of nutrition standards for foods and beverages sold to students in schools and information on why these standards are needed will motivate those responsible for creating school food environments to use the implementation tools provided to change those food environments (the implementation-related outcomes which are the focus of this research). Then schools will only sell foods and beverages that adhere to the standards. Intermediate and long term outcomes (not the focus of this implementation research) are that students will purchase and consume healthier options, learn how to make healthy choices, and ultimately have better health outcomes both in the short and long term. This logic model and program theory guided the lines of inquiry for the current investigation (Appendix D).    Data collection strategies I used a multiple case study approach to explore district-level implementation processes in different BC contexts. This allowed me to delineate clear boundaries around the contexts of each case and to be able to compare findings within and across cases (Yin, 2009). Districts were defined as the case unit. Five districts participated in the study: three urban and two rural districts. To explore how different contexts may influence implementation processes, districts were purposively selected to include two rural and three urban districts from across BC’s five decentralised Regional Health Authorities (RHAs).                                                              4 It is important to note that while this paper focuses on district-level implementation of the Guidelines, school food and beverage sales environments are also shaped at the school-level by school-level stakeholders. School-level implementation in BC is explored in Chapter 6. 83  In alignment with the realist approach, I used multiple qualitative data collection strategies outlined in Table 5.1 below: Table 5.1 Data collection strategies in each school district and regional health authority (RHA) Data collection strategies District 1 urban, RHA A  District 2 urban, RHA B  District 3 urban. RHA B  District 4 mostly rural, RHA C  District 5 mostly rural, RHA D Total Semi-structured interviews with district staff* 2 1 1 -- -- 4 Semi-structured interviews with public health sector staff** 3 2 --*** 3 5 13 Semi-structured interviews with vendors 4 2 -- -- -- 6 Event attendance**** 3 -- -- -- -- 3 Questionnaires***** 14 19 15 10 4 62 District and RHA context scan Yes Yes Yes Yes Yes -- School website scan 10 12 10 17 13 62 *One semi-structured interview was also conducted and included in the analysis with the singular person at the BCMoED who has the Guidelines as part of their portfolio  ** While only 13 semi-structured interviews with public health sector staff are outlined in the table, a total of 22 were conducted with public health sector staff across the province and are not included in the table as they fell outside of the RHAs in which the participating districts were located. While the remaining 9 were not formally included in the CMO analysis for each specific district, the broader discussions they contained about the Guidelines influenced interpretation. ***Because district 2 and 3 were in the same RHA, the interviews with health sector staff from this RHA were used for the analysis of both districts ** One event was specifically in school district one, organised by the school district in partnership with the RHA; the other two public events were organised by the same RHA but in a different school district not part of this study; the public health dietitians involved in organising these public events, however, utilize the same format and the informal discussions I heard at all the events, regardless of the district, were similar in nature. *****There were also 10 additional questionnaires on which the specific district of the respondent was not identified. Comments broadly related to Guidelines implementation influenced interpretation.  Semi-structured interviews were conducted with purposefully selected school district-level staff, public health stakeholders, and private food vendors. This provided in-depth information about relationships and interactions among stakeholders, contexts, and key activities related to Guidelines implementation. Interviews lasted from 30-90 minutes and were recorded and transcribed, with the exception of one private vendor who did not consent to recording but allowed note taking and reviewed the notes after the interview. Two iterations of interview guides were used. The first version was more exploratory where one of the goals was to garner insight for devising the initial program theory for phase 1 (Appendix C), presented in Chapter 4. The second iteration was developed to obtain more specific information about participant’s experiences with implementation as it related to the retrospective program theory in Chapter 4 (Appendix E).    84   To obtain school-level perspectives of district-level implementation, 15-20 minute structured questionnaires were administered online, via telephone, or in-person with school-level stakeholders. Four districts allowed me to contact school administrators directly. The remaining district allowed me to attend a public event where data collection was conducted in-person with primarily parent stakeholders. Administrators were also asked if they were willing to forward the online link to the questionnaire to others in their school community. As a result, questionnaires were answered largely by administrators and by a small number of other types of school-levels stakeholders (Table 5.2). The questionnaires were developed based on the retrospective logic model and program theory created in phase 1 of this research and included questions around dissemination of information, engaging in collaborative processes, and use of tools for assessing foods and beverages for compliance (see Appendix G). Table 5.2 Questionnaires completed by district and school-type  District 1 (urban) District 2 (urban) District 3 (urban) District 4 (mostly rural) District 5 (mostly rural) Unknown district  Respondent-type Elem. Sec. Elem. Sec. Elem. Sec. Elem. Sec. Elem. Sec. Elem. Sec. TOTAL Parent 12  7  2  1      22 Administrator   6 5 10 3 7 2 4  1  38 School food staff 1 1         2 1 5 Teacher            1 1 Unknown    1       3 2 6              72  I also conducted a website scan of schools, school districts and regional health authorities to obtain more context around organisational structures and school food.  Additionally, I scanned the websites of private vendors who participated in the study to garner more information about their products. These scans were an ongoing and iterative process.  Analysis Using qualitative analysis software (Atlas.ti, 2015), I adopted a linked CMO coding technique for analysis of the interview transcripts and long answer portions of the questionnaires (Jackson & Kolla, 2012; Puton, 2016). The purpose of this approach was to aid in viewing the data in configurations of context, mechanism, and outcomes to establish their relationships to one another. 85  This technique involves first identifying and labeling an outcome, expected or unexpected, in a portion of text. The text around the outcome is examined closely for the participants’ explanation of how the outcome came about, a relatively natural way of relaying a description of events for most people (i.e., in the text, a mechanism and relevant context are often mentioned as having led to the observed outcome). Once Cs, Ms, and Os are identified in a ‘chunk’ of narrative, a linked CMO code (e.g., C24-M10-O23) is created. This helps reveal linked explanations for implementation phenomena directly from the respondent rather than rely solely on the logical inferences of the researcher. However, in the cases where it was necessary to fill in the blanks because, in a given ‘chunk’ of text, only a context or mechanism was specified in relation to an outcome, plausible inferences were used to articulate full CMO relationships. For this, I underpinned my inferences with an existing implementation theory, contextual interaction theory (CIT) (Owens, 2008). CIT posits that all implementers have three broad characteristics (knowledge, motivation, and capacity) which dynamically interact with each other and with dimensions of context. These interactions and characteristics are posited to be the broader driving forces behind the decisions of implementers to take action (or not) and the type of action they take. This theory is described in more detail in Chapter 3.     The linked CMO codes were categorised based on their outcome and which component of implementation they related to as per the logic model in Chapter 4. For example, linked CMO codes containing outcomes related to the activity of assessing and selling compliant food and beverage offerings were grouped together. Memos for each component of implementation were created and the linked CMO codes, based on their outcome, were organised into each memo. This was done for each district. Contextual information garnered from the website scans was incorporated into the creation of these memos. Following this, I considered the data across the districts through comparing the memos. This helped to more clearly determine which implementation processes were specific to one context and which cut across cases. The short answer portions of the questionnaire were used to determine school-level stakeholders’ perceptions of how often they felt district-organised food complied with the Guidelines, adding another data source to support conclusions that emerged from my interpretations of respondents’ narratives. 5.4 Results The following section describes implementation processes occurring at the district level: (1) the top-down mandatory directive from the Province prompts districts to take action by incorporating 86  adherence to the Guidelines into the request for proposal (RFP) process when seeking bids from private food service providers; (2) this creates an economic incentive for private industry to comply; and (3) an informal system of regulation and enforcement helps promote compliance. Within these implementation processes, however, complexity emerges due to the influence of specific dimensions of context which results in different ways of engaging in implementation activities (e.g., assessing food and beverage sales compliance, changing what items are available for sale) among expected-implementers at the district level.  5.4.1 Mandatory mechanism: driver of demand creation The school food and beverage Guidelines, as an intervention, offers both concrete and non-concrete resources. How stakeholders respond to these ‘intervention resources’, as shaped by pre-existing contexts, constitute the mechanisms of an intervention. I found a key ‘intervention resource’ provided at the district level is the power that accompanies the BCMoED declaring them ‘mandatory’. The BCMoED, in a context of a governmental sector hierarchy, expected that districts would develop strategies to ensure the food they bring into their district is compliant. In this study, among participating districts, declaring compliance ‘mandatory’ as a top-down BCMoED directive has led to responsibility for their implementation becoming part of a new job role for specific individuals at the school district level. This was the case in all five of the districts. District-level staff did not explicitly report feeling forced or pressured outside of describing the implementation of the Guidelines as a “ministry directive” but one school principal perceived that “the government threatens the district guy” (Administrator, SD 4, questionnaire).   In the three urban districts, vending machine suppliers and private cafeteria catering companies are procured centrally. In these cases, adherence to the Guidelines is incorporated into the request for proposal (RFP) process. Urban district staff presented their role in implementation as relatively straightforward whereby delegation of actually implementing the Guidelines falls to the vendors who secure district contracts. In the two rural districts, as reported by administrators, food and beverage sales are not centrally organised and thus schools themselves are fully responsible for creating their own food and beverage sales environment. I was unable to secure interviews with the district staff in the two rural districts but considering they were not playing a role in securing food and beverage service providers, it is likely they play less of a role in implementation processes compared with the urban district staff. The rural district staff, with Guidelines implementation in their portfolio, do however reportedly engage in unofficial enforcement activities directly with 87  schools. This is presented in more detail below in a later subsection of the findings. This difference in ways of organising school food and beverage sales between rural and urban districts makes sense as stakeholders from urban areas reported access to options of food and beverage service-providers capable of serving many schools across small geographic areas. It may be that there are not vendors who can provide services to all schools in a single district spread across the long distances that characterise the two rural districts (See Appendix A for district descriptions, including relevant demographics and geography).  Table 5.3 describes the facilitative dimensions of context in which the mandatory nature of the Guidelines ‘works’ to promote demand creation whereby a school district demands potential vendors meet the Guidelines in order to obtain a district-wide contract.  Table 5.3 Describing the mandatory mechanism at the district level Context Mechanism (resource + reasoning) Outcome The education system is hierarchical  AND   Urban districts located in densely populated regions with multiple vendor options centrally organise procurement of food vendors for their schools Mandatory directive from the Ministry of Education/top-down pressure to implement the Guidelines in their district food procurement work as part of their job Districts demand potential vendors prove they will adhere to the Guidelines when applying for contracts  …everything needs to fit in the Guidelines and everyone is well versed in that now….companies won't...get in unless…everything meets the Guidelines… (District staff, SD 1, interview) The education system is hierarchical AND Rural districts do not centrally organise and schools are meant to create their own food and beverage sales environments. Mandatory directive from the Ministry of Education/top-down pressure prompts the district to take action where they can to ensure implementation. Implementation of the Guidelines becomes part of a district employee’s job where, because the district does not play a role in food and beverage sales supply procurement, they engage in other enforcement activities directly with schools. 88   5.4.2 Money mechanism: driver of supply creation Next, the demand created by urban, centrally-organised school districts, is an economic incentive to motivate industry stakeholders to create options that comply. For the purposes of this research, this economic incentive is considered an ‘intervention resource’ launched in the facilitative context of a profit-motivated, competition-based economy. This incentive was reported by vendors, school stakeholders and district staff to be effective in motivating industry stakeholders to create a compliant supply:  Last year we [started with a new vending machine supplier]...they brought their lists of potential items and what we liked about them was they had tons and tons and tons of things which fit under the healthy food guidelines…. (District staff, SD 2, interview)  Vendors fulfilling district-wide contracts are reportedly regularly offering compliant items. However, this compliance appears to be reached through two different approaches. One approach taken by vendors has been to offer reformulated compliant facsimiles of what might be thought of as conventional school food such as hamburgers, hot dogs, pizza and French fries. Some district, school, and health-sector respondents reported an overall lack of creativity and appeal in the presentation of these types of offerings, the result being that:   …the food is not great [because] much of it is bland, lacks flavour, can be visually unappealing or soggy (baked fries instead of fried)… seems relatively unhealthy even though it supposedly meets guidelines, so now it’s just not appealing to many people. (Teacher/hot lunch coordinator, SD 2, questionnaire)   One public health dietitian (RHA B) postulated the reason a vendor would opt to produce this kind of reformulated conventional cafeteria food might be due to pre-existing beliefs about school food. The two large service providers interviewed in this study (one cafeteria catering company and one vending machine company), both of whom had been in the school-food business for decades, conveyed a set of beliefs around the narrowness of children’s preferences that only included conventional school cafeteria and vending machine fare. These implied beliefs emerged through comments such as, “better to sell them a can of Coke at school than have them go across the street to 89  the 7-11 for 2-for-1 energy drinks”  and “we have to put something in the machine that will sell” (Vending machine vendor, interview).    The decision to create these often unappealing, but compliant ‘equivalents’ may also be influenced by the existence of external vendors in the vicinity of the school (e.g., fast food franchises, convenience stores). These external vendors do not have to comply with any guidelines, often offering the ‘real’ versions of conventional school food, and therefore are viewed as threats to the school-based vendors’ profits:   I hope the government doesn’t try to get any stricter because it’ll push kids off the campus even more…everything that was eliminated from schools is popping up in surrounding stores…like pizza and fries. (Cafeteria vendor, interview)  One vendor, with numerous district-wide contracts, summarised this complexity of deciding what products to sell in schools:   …And revenue loss is not related to products. Revenue loss is related to certain products in certain environments…[For example] let's use two schools…I’ll call them school A and school B…[in the case of] School A, the nearest store is 15 blocks away. [The vending machines] will survive anything…[in the case of] School B... it needs education towards helping kids make a choice on whether they want to go to the vending machines selling healthy choices or whether they want to walk across the street to the store. (Vending machine vendor, interview)  The other approach that appears to be used to comply with the Guidelines was demonstrated by another large cafeteria service provider, offering diverse and reportedly high quality, compliant, options. The staff person at one of the urban, centrally-organised districts reported recently conducting a district-wide survey which not only demonstrated changing food preferences among students towards better foods, but also provided guidance for the district’s RFP process. As a result of the survey, this district recently awarded a contract to a new company offering what the district viewed as higher quality, compliant food. The menus from school websites from this particular district confirmed that this new company was offering diverse and interesting lunch options not typically sold in school cafeterias, including curries, vegetarian stirfry, pastas, frittatas, and souvlaki. 90  While an interview with someone at this new cafeteria catering company could not be arranged, the interview with the district provided possible ideas about potential, facilitative dimensions of context that may have enabled this vendor to take a different direction with school cafeteria food.  Along with the district demand for improved food quality and diversity, it might also be important that this vendor had never worked in schools before, but rather had only operated corporate cafeterias. Not only had they not been entrenched in school food for decades but they have been providing food to adults. This may have allowed them to conceptualise cafeterias differently from the above reformulating vendor and in a way that aligned with the district’s goal of improving their schools’ cafeterias.                        91  Table 5.4 Describing the money mechanism at the district level Context Mechanism (resource + reasoning) Outcome Large private vendors have a pre-existing drive to sustain their business AND Private vendors hold beliefs about student preferences for conventional school food (e.g., Chicken nuggets, pizza, etc.) AND External vendors exist near the school and are perceived as competition, selling the kinds of items vendors believe students prefer The Guidelines are nutrient-based criteria whereby compliance is measured only in terms of salt, sugar, and fat  AND The demand by the district to comply with the mandatory Guidelines while trying to compete with external vendors for students’ business incentivises vendors   Items are compliant, reformulated versions of  conventional school food    I have heard some students say, 'I don't buy stuff in the cafeteria because it's really unhealthy'… and it's like, the cafeteria has hamburgers and French fries and I … am assuming, because I know that the vendor at the cafeteria knows about the Guidelines and there's supposed to be baked fries and there's supposed to be hamburgers that meet the Guidelines… (Dietitian, RHA B, interview) Large private vendors have a pre-existing drive to sustain their business AND The district has a desire to change cafeteria food culture AND Private vendors hold set of beliefs about food offerings in cafeterias that challenge conventional ideas of cafeteria food The demand of the district to both comply with the mandatory Guidelines and to improve cafeteria food quality incentivises the private vendor Offerings for sale in cafeterias are compliant, diverse, and well-presented …we're bringing in a new company. That [decision was] really around food quality. We had previously been with a company for about a 20 year period for our cafeterias... I wouldn’t say that we were unhappy with our previous vendor… kids were choosing not to eat that food…it wasn’t super appetizing and it wasn’t well-presented… so when [the two competing companies] came we asked them to bring their food offerings… and [the difference] was glaring in terms of the quality they were offering (District staff, SD 2, interview)  5.4.3 Monitoring mechanism: a dynamic informal process of enforcement  An important implementation process at the district level is the monitoring and enforcement of compliance. Both urban, centrally organised districts and rural de-centralised districts report engaging in informal processes of monitoring and enforcement but in slightly different roles.  In the rural districts, even though district staff are not involved with organisng district-wide procurement, 92  they engage in informal processes of school-level enforcement. For example, an administrator or teacher may file complaints about a non-compliant fundraiser that they observed occurring in another school. The district person reportedly engages in following up with the non-compliant school and exerts pressure to promote compliance. In the context of rural districts, where food services are not procured by the district, it appears that enforcement activities only occur between the district staff and school-level employees. There was no interaction reported between private vendors individual schools may decide to use and school district staff.  In urban districts, the incorporation of compliance into the district-wide RFP process theoretically helps promote adherence, as presented above. Almost universally in the three urban districts, respondents believed that, because compliance is built into the centralised procurement process, vending machines and cafeterias ‘almost always’ or ‘always’ meeting the Guidelines. The urban district staff did not report any official assessment of whether what service providers are actually selling in schools are the compliant items promised in the their contract bids. This is not surprising because even though the BCMoED  states explicitly that following the Guidelines is indeed ‘mandatory’, they “are not supported with a compliance and enforcement program” (Government of British Columbia, 2013, p.g. 10). However, these urban districts have developed informal systems of self-regulation once contracts are awarded that include a back-and-forth relationship with the private sector actors and with other school-level stakeholders:   R: If we find out that certain things aren’t great [with a cafeteria], because it’s a bottom line thing for them, the more they sell the more they make… we go back and say "hey back up the truck!” I: So you actually monitor that a little bit? R: Only because I have people tell me all the time. I: …so it's like self-monitoring basically? Self-regulation? R: [Yes] I don't [want to] be roaming around every cafeteria...no not a chance. (District staff, SD 3, interview)  District staff reported receiving complaints from private sector cafeteria staff.  For example, one type of complaint received was that non-compliant fundraising initiatives were occurring elsewhere in the school. Cafeterias are required to comply based on an official contract but also, as private 93  businesses, need to remain financially sustainable. District staff reported that they do take action to address these complaints to ensure a fair playing field for food and beverage sales.  Table 5.5 Describing the monitoring mechanism at the district level Context Mechanism (resource + reasoning) Outcome Interest groups within schools are responsible for fundraising to provide a well-rounded curricular experience AND Food and beverage sales within schools are in competition with one another Because there are no official enforcement mechanisms in place, stakeholders in competition with one another who value fairness are led to reach out to the district if they perceive non-compliant food and beverages are being sold District staff make efforts to address non-compliance within schools  ...right now I’ve got a note on my desk [from a teacher chef] about the grad group selling donuts and coffee [saying] ‘we’re supposed to follow the rules, why aren’t they?’.... because we sell almost no coffee, we’re not going to make a big issue [about the coffee]...however, the coffee AND donut fundraiser? I am going to raise the issue. (District staff, SD1, interview)  5.5 Discussion This study aimed to increase understanding of the district-level social processes related to implementation of British Columbia’s provincial school food and beverage sales policy. Specifically, I aimed to elucidate dimensions of contexts that are facilitative (or not) and what mechanisms drive district-level implementation. I found that an important context for successfully launching a mandatory set of standards is one in which there is a normative hierarchical structure. In centralised school districts, which were urban districts in this study, this engagement in implementation activities manifests as the offloading of responsibility to implement to the private sector through economic incentivisation. This structured demand in centralised districts appears to work to motivate vendors to create supply. In the two rural districts, food sales are organised at the school level and, as such, this economic demand mechanism was not a part of district-level implementation. For both the urban and rural districts in this study, while there are no official enforcement mechanisms in place, a dynamic system of informal monitoring appears to be in place, helping to promote compliance within schools.   94  First, as stated above, the acceptance of the hierarchical structure of the education system appears to be a facilitative dimension of context in which to launch mandatory nutrition standards to successfully trigger district-level stakeholders to engage in implementation activities. That top-down pressure plays a role in motivating district-level stakeholders to engage in implementation-related activities is not necessarily surprising in this context; give a mandate to the staff in a bureaucracy and they will do it (x leads to y). However, rarely are the social processes related to how mandates lead to action articulated (what is it about x that leads to y). Use of a realist approach facilitates deeper understanding of these social processes by shifting thinking about intervention resources to encompass non-concrete resources as well as more obviously concrete resources. I identified the power that accompanies a “mandate” as a non-concrete intervention resource that is meant to remove an expected-implementer’s choice of whether or not to take action to implement (Friedman, 2014). This top-down pressure prompted the integration of implementation into a job role for people at the district level which effectively removed district staffs’ choice about whether or not to engage in implementation-related activities. As a result, all five districts that participated in this study were engaging in implementation-related activities, whether ensuring compliance through district-wide contracts with private vendors or making efforts to support ‘fair’ compliance within schools through the use of more casual self-regulation measures.   I propose that this normative acceptance of the hierarchy of education system is sufficient for triggering district-level stakeholders to engage in these types of implementation-related activities even in the absence of official enforcement measures or sanctions of the kind that are used in some parts of US, for example (Gourdet et al., 2014). Friedman (2014) proposed that, despite the low acceptability of paternalistic interventions related to changing the way people eat in most western nations, when launched in an acceptable “zone of control” (p. 1743) (i.e., a ‘space’ where top-down control is normative) acceptability, and thus action, comes easier; it is acceptable that there is a high degree of government control over those employed by school districts, civil servants employed to carry out the will of a democratically elected government. These findings confirm Friedman’s (2014) hypothesis and my study adds that this is even the case when there are no official enforcement measures. Further support for this argument is provided in the Province of Alberta, which introduced youth nutrition guidelines in 2008 (Downs et al., 2011). These are voluntary standards in that school districts (or even individual schools) are allowed to decide whether they will implement them. Studies indicate adoption by school districts is not 100% (Quintanhilla et al., 2013). This is in 95  contrast to the Alberta government mandating that all school districts develop and enact policies to ensure safe schools for LGTBQ students (Mertz, 2015) where subsequently 100% of school districts adhered to this mandate (Fletcher, 2016). It would follow that in the Alberta context, where there is apparently also an accepted hierarchical power structure and mandates are proven to trigger district action, mandating Alberta school districts to adopt the nutrition guidelines may prove acceptable. This mandatory mechanism, with and without sanctions for non-compliance, requires further exploration in different contexts where, for example, there are different normative beliefs about the role of governments.   The next facilitative dimension of context is the existence of a market-based economy motivated by profit generation. In urban school districts, this is coupled with another facilitative dimension of context, namely that the district centrally organises cafeteria service providers and vending machines through centralised procurement, creating district-level demand. Private vendors, in turn, comply with nutrition standards if they want to win large, district-wide contracts. These urban districts are essentially a “customer” with large purchasing power and therefore they can effectively create in-kind financial incentives with “strings attached” (Vedung, 1998). The findings presented here suggest it is the district-demand and the profit motivation that drives them to offer compliant items. Alternatively, it is also possible that innovative vendors are not necessarily compliant because of the existence of the Guidelines but instead are influenced by other existing wellness-promoting school food initiatives. The interface between this type of policy and other interventions requires further exploration.  In general, neo-liberal discourse argues that “governments can’t do it alone” (Halpern et al., 2004, p. 3) when it comes to public services. This trend towards contracting private industry to perform what have typically been public services has been found to be useful for achieving social goals (McCrudden, 2004). These findings confirm that this also appears to be true for the goal of improving school food environments. There are numerous examples in the literature that illustrate the power of public procurement for impacting private sector operations and improving school food (Morgan & Sonnino, 2010; Sumberg & Sabates-Wheeler, 2011). Furthermore, not only can private industry be ‘convinced’ through financial incentives to help governments achieve social goals, including public health policy goals, but  the expectation of  ‘innovation’ in procurement processes can also be used to foster more effective private sector practices (Edler & Georghiou, 2007). Again, 96  it is not possible to simply create a ‘capitalist context’ into which this type of intervention might be launched to be more successful. However, these findings point to the need to consider the political economy in which it might be launched and what changes may be required before it can be successful. For example, launching an intervention to improve school food sales environments in contexts that have relatively corrupt or nepotistic contracting practices may result in poorer implementation because economic incentives to comply become less motivating. In addition, launching a school food and beverage sales intervention in a context where demand may not be strong enough to incentivise vendors to create a compliant supply, like sparsely populated rural areas,  may not be as successful as in more densely populated areas. Given the trend towards public sector contracting of private services, the processes by which economic incentivisation works (or not), to achieve public health goals, in different contexts, is also an important area for further research.   Another dimension of context that emerged from this study, the pre-existing beliefs held by private industry stakeholders, is not necessarily ‘facilitative’ but can lead to different ways of implementing nutrition standards. My findings show that some vendors choose to offer compliant foods and beverages by reformulating conventional school food to meet the salt, sugar, and fat requirements of the Guidelines. This reformulation of conventional school food was discussed in another BC-based study (Watts et al., 2014) and in Ontario (Vine & Elliott, 2014). My study identified potential reasons why vendors make the decision to take this path of implementation. It appeared that some vendors hold persistent beliefs that students will not purchase anything except ‘conventional’ school food offerings. There may also be external vendors nearby who do not have to comply with nutrition standards and are considered competition. These specific contexts may lead school vendors to feel the need to mimic conventional food to maintain customers. However, my data also suggested that reformulated foods may not always be appealing to students, the customers. This was also found in an Ontario-based study that also reported this unappealing nature of reformulated food contributed to students taking their purchasing power off-campus (Vine & Elliott, 2014). In trying to maintain customers, reformulating vendors may ultimately be losing customers and possibly contributing to students continuing to consume unhealthy food. On the other hand, a vendor may decide to create reportedly diverse and well-presented offerings that comply with the Guidelines. My findings suggest that this may be a result of the combination of an absence of pre-conceived notions about student food preferences, district-level demand that food and beverage offerings align with what the 97  students (the customers) really want, and a history of working in professional contexts that might have typically expected high quality offerings. This suggests cafeterias are not bound to perpetuating conventional ways of thinking about school food but rather could be prompted to further innovate through existing procurement processes in centrally organised school districts. Further exploration into the cognitive processes of vendors and how they make decisions about their approach school food service is required for more insight into how more innovative school-food partnerships may be fostered.  Moreover, there is a need to explore the possibility that student preferences are changing to reflect the growing ethnic diversity in Canada, particularly in urban areas (Statistics Canada, 2016), and a growing secular trend of the acknowledgement of the importance of food choices for health. Armed with this information, districts may more easily be able to select vendors who meet the preferences of the student consumers. In addition, it has been suggested that engaging students in determining what food and beverage offerings will be made available may be an effective strategy for improving their experience of school food (Cummings, Burbage, Wood, Butler, & Kuo, 2014; Samuels, Craypo, Boyle, Stone-Francisco, & Schwarte, 2006). Not only this, it has been proposed that policy can be used as a learning tool for students, encouraging engagement in the policy development process (McKenna & Brodovsky, 2016).   Theorising implications of different district-level implementation processes The differences between urban and rural districts in how they engage in implementation activities may contribute to undesirable and unintended consequences. While the normative hierarchical structure of the education sector allows for the exertion of effective top-down pressure, the BC education sector also allows significant freedom in how districts organise their day-to-day business, including their food and beverage sales environments (i.e., centralised or decentralised). In rural areas where districts are not involved in the creation of the individual school food and beverage environment as a result of being decentralised, there is less control over compliance than in urban districts. Globally, rural-urban health inequities exist where urban populations generally experience better health outcomes (CSDH, 2008), including in BC (Chasey, Pederson, & Duff, 2009). Differential implementation of a public health intervention, where urban school districts may achieve better compliance than rural districts, may exacerbate this inequity.  98  The last concern with the different ways in which nutrition standards may be implemented at the district level calls into question the more profound messages being sent to students through the type of foods and beverages available at school and the implications for later life. Because BC’s Guidelines are nutrient-based, with the parameters for compliance being focused on three nutrients, the food industry can manipulate the formulas of existing, conventional school food and beverage products to comply.  Students provided a ‘reformulated’ food environment are receiving a different set of messaging about healthy food and beverages (Titman, 1994) which could put them at a disadvantage as they make food decisions. For example, if they are served nutrient-compliant chicken nuggets and factory-processed pizza every week for 12 years, they could develop the belief that these are appropriate foods to eat on a regular basis but will not easily be able to distinguish between similar options in the general marketplace which likely will not adhere to any nutrition standards. The potential for the different ways of implementing nutrition standards at district levels to differentially maximise students’ health-promoting opportunities and capabilities should be considered by top-level decision-makers.  Implications for refining program theory The findings offer challenges to the initial program theory devised in Chapter 4. First, there was no indication from either district-level staff or vendors that they were motivated to take action to implement through a process of responsibilisation. More specifically, they did not report that they were motivated to take action to implement as a result of having received information about the severity of the problem, how susceptible children are to poor nutrition outcomes when exposed to unhealthy food and beverage environments, and why implementing nutrition standards would be beneficial. Instead, the motivation to take action to change school food and beverage environments was related to adhering to a top-down mandate as part of a job description, on the part of school district staff, and sustaining profits, on the part of vendors. This is not to say, that a process of responsibilisation did not occur in previous years whereby disseminating information to the districts and to vendors to justify mandating nutrition standards contributed to this process. It is perhaps the case that, now, the process of responsibilisation, in conjunction with global secular trend of disseminating information around the importance of healthier diets (as indicated by the wide array of television programming on the topic on numerous international networks, films, books  (e.g., British Broadcasting Channel, n.d.; Buckingham, 2017; Canadian Broadcasting Corporation, 2017; Moss, 2013; Pollan, 2007; Soechtig, 2014)), has led to a more universal understanding of why the 99  Guidelines are in place.  Second, the initial program theory did not include engagement in self-monitoring as an implementation activity promoting compliance. Lastly, the initial program theory only identified broad dimensions of context. My findings contribute to refining the program theory by identifying more specific dimensions of context related to district-level implementation, including the potential for the individual pre-existing beliefs of vendors to influence how they decide to comply with nutrition standards.  Implications for practice When considering the use of school food and beverage sales policies as a tool to improve school food environments, there is a need to consider what type of contexts will be supportive and will trigger expected implementers to actually implement. If making them mandatory rather than voluntary, policymakers can consider whether the context in which the policy will be launched is one in which official enforcement measures are needed to accompany the mandate to ensure people take action to implement or if the mandate itself is sufficient to trigger action. Also, if it is expected that the private sector will change their products to comply with the policy because of the demand from school districts, there is a need to consider whether this will be effective in regions where school districts are not procuring private sector food services centrally. Moreover, an understanding of how the pre-existing beliefs of private sector vendors may be influencing how they choose to comply with policies may help reform the process through which centralised school districts procure private sector services, opting instead for more innovative vendors. This can promote the sales of foods in their schools that are not only nutritionally compliant, but diverse and interesting to students.  Potential limitations This phase of the study has several limitations. I was not able to collect data consistently across the five district cases and used a semi-convenience approach to garnering participation, whereby questionnaire participation was open to any type of relevant stakeholder who was purposefully identified and willing to participate. From a realist approach, however, this is not considered a serious methodological flaw. While the initial analysis necessarily involved identifying what case-specific mechanisms were operating in what types of contexts, I then took a step back and set out to see which mechanisms were operating across the cases. This helps with theorising the relationship between contexts, mechanisms, and outcomes more broadly; just because each case is a different 100  ‘context’ it does not necessarily mean that cases will not share some (or many) dimensions of context. Therefore, regardless of whether it was a principal in one district discussing a particular mechanism and context interaction by sharing one story, and a district office staff in a different district discussing the same interaction through a different story, the interaction is the same, broadly speaking. Another concern is that I was unable to make claims regarding the contested concept of data saturation (O’Reilly & Parker, 2013). This contestation is elaborated on in Chapter 7. As there are 60 districts in BC (i.e., 60 different contexts) ‘true’ saturation would not have been attainable within the resource constraints of this research. However, from a realist perspective, this inability to achieve saturation does not necessarily mean that the findings are invalid, but rather that they require further exploration (Maxwell, 2012; Morse, 1995). 101  Chapter 6~ School-level implementation of the BC Guidelines  6.1 Introduction Global concern about the increased incidence of obesity and overweight in children (Rokholm et al., 2010) and the predicted health care costs and loss of economic productivity once they become adults (Wang et al., 2008; Withrow & Alter, 2011) have resulted in school food environments becoming a popular target for public health interventions. School food environments are the “the physical, economic, political and socio-cultural context in which [students] engage with the food system to make their decisions about acquiring, preparing and consuming food” (HLPE, p. 28) during the school day.  School food environments in many resource-rich countries have evolved in a manner that reflects the evolution of the modern food system in the last sixty years, commonly consisting of ultra- processed, convenient (Morgan & Sonnino, 2010; Poppendieck, 2010), addictive (Gearhardt et al., 2011; Moss, 2013), and low-nutrient, energy dense foods (Poppendieck, 2010). As a result, policy and programmatic interventions to promote salutogenic school food environments have been recommended as effective strategies to improve student consumption practices by maximising opportunities to make healthy choices (Bundy et al., 2009; CDC, 2012; PHAC, 2013; WHO, 2008; WHO et al., 1991).   Addressing the complex public health problems associated with the dominant food system requires complex interventions (Wong et al., 2016), of which school food environment interventions are an example. They often have multiple, linked components delivered to individuals or communities, perhaps involving national, regional, and/or local levels of implementation activities. Moreover, intervention success is dependent on how individuals and institutions respond to the intervention and the broader contexts in which they are launched. What may work in some contexts, for some people, may not work in another, so there is a need to deeply understand the nature of complex initiatives (The PLOS Medicine Editors, 2013; Wong et al., 2016). A variety of interventions are being used to attempt changes in the school food and beverage environments (Bontrager Yoder et al., 2014; Hawkes et al., 2015; Jaenke et al., 2012; Rojas et al., 2011; Thaler & Sunstein, 2009) and 102  understanding their inner-workings is crucial for optimising implementation and, therefore, achievement of desired proximal, intermediate, and long-term intervention outcomes.  This requires a nuanced understanding of the social processes related to implementation. Due to the complexity of these processes, few policies and programs in the ‘real world’ are fully implemented as intended (Damschroder et al., 2009; Durlak & DuPre, 2008; Helfrich et al., 2010; Pressman & Wildavsky, 1973). This is one reason public health policy may fail to achieve consistent and sustainable desired outcomes, not only wasting valuable resources but also potentially exacerbating health inequities (Macdonald et al., 2016). Implementation research can help shed light on the complex processes of implementation. Peters et al. (2013) defined health policy and systems implementation research as:  …being concerned with the study of clinical and public health policies, programmes, and practices, with the basic intent being to understand not only what is and isn’t working, but how and why implementation is going right or wrong…(p. 27)  This chapter aims to contribute to a more nuanced understanding of complex school food environment interventions through a case study of school food and beverage sales policy implementation in British Columbia (BC), Canada. In particular, I explore how key mechanisms and contexts influence how stakeholders engage in implementation activities (or not).  BC provides a fertile ground for this exploration as it was one of the first Canadian provinces to develop and launch their nutrition standards in 2005 and it has been proposed that it takes at least a decade for policy to become embedded in practice (Sabatier, 1988). Because BC’s policy is highly specific and only applies to items that are sold to students (as opposed to subsidised meal programs) it provides a clear boundary around implementation processes, useful for easing the intellectual challenges associated with exploring complex systems of policy implementation. I begin by providing background on school food and beverage sales environment policies and the quantitative and qualitative body of implementation literature revealing different levels of compliance observed in different contexts that points to ongoing implementation challenges. The study context of BC is described along with an overview of the small body of implementation research that has occurred here. I then describe the scope and methods of my research and include results of my qualitative fieldwork involving the implementation experiences of BC-based school-level stakeholders. I conclude by theorising 103  connections between the findings and existing literature and suggesting considerations for policy makers for developing or improving implementation of this type of intervention.   Implementation of nutrition standards for school food and beverage sales  Nutrition standards for foods and beverages sold in schools have been introduced in many jurisdictions, including almost all Canadian jurisdictions, at the state and district levels in the United States, Mexico, Australia, the Netherlands, Finland, Norway, and some regions of France (Antilla et al., 2015; Bonsmann, Kardakis, Wollgast, Nelson, & Caldeira, 2014; Carriere et al., 2015; CDC, 2012; Holmes, 2016; Holthe et al., 2010; Mensink, Schwinghammer, & Smeets, 2012; Monterrosa et al., 2015; Nathan et al., 2016). School food and beverage sales standards aim to increase the availability of healthy options for sale and the reduction or elimination of unhealthy ones by motivating the people who create school food and beverage sales environments to change the way they create them.  Although these policies have been shown to improve the options available for sale (Kubik et al., 2010, 2013; Larson et al., 2015; Gorski et al., 2016), the literature shows a wide variation in levels of compliance across numerous contexts. Some studies have found significantly higher levels of compliance with food and beverages sales nutrition standards in urban schools compared with rural schools (Dick et al., 2012; Hills et al., 2015; Pettigrew et al., 2012a). Comparing compliance between school types (elementary vs. secondary, for example, or public vs. private schools), one study found Massachusetts secondary schools are more compliant with nutrition standards than middle schools (Gorski et al., 2016). Other studies from Queensland, Australia and the United States have found primary schools are more compliant than secondary schools (Dick et al., 2012; Kubik et al., 2010), public schools more compliant than private schools (Hills et al., 2015), and large schools more compliant than small schools (Hills et al., 2015; Kubik et al., 2010). One US study found that schools with low proportion of students enrolled in the subsidised lunch program (i.e., wealthier student populations)  were significantly more likely to have ‘junk’ food available for sale than both schools with middle and high levels of enrolment (i.e., less wealthy schools) (Nanney et al., 2013). An Australian study found a significant positive correlation between socioeconomic status (SES) of primary schools and availability of compliant food and beverage options for sale (Hills et al., 2015). A number of these studies have also found non-compliant foods continue to be widely available even after implementation efforts are made (Gorski et al., 2016; Kubik et al., 2013, 2015; Samuels et al., 104  2010; Whatley Blum et al., 2007) and sometimes with fluctuating implementation levels from year to year (Gorski et al., 2016).  These quantitative implementation studies provide important information as to whether the introduction of food and beverage sales policies results in increased availability of health options and decreased availability of unhealthy ones (does x lead to y?). However, they are not able to provide insight about why these various levels of compliance are occurring (what is it about x that leads to y?). A number of qualitative studies exist from different jurisdictions that aim to explain why   implementation of school food and beverage sales policies happens sometimes and not others (e.g., MacLellan et al., 2010; Pettigrew et al., 2012b; Vine, Elliott, & Raine, 2014). These studies, though, tend to identify barriers and facilitators to implementation as discrete components of the complex social processes associated with implementation. What is missing from this body of work is the exploration of the relationships between implementation barriers and facilitators that lead expected implementers to decide to engage in implementation activities and the role of context in shaping these relationships. Articulating these relationships can provide deeper insight into why interventions do or do not work in different contexts, offering not only leverage points at which to intervene to improve or support implementation activities, but also insight into the nature of this popular school food and beverage environment intervention.  British Columbia context In Canada, health programming and public education, and therefore school food programming related to both curriculum and regulation of food available in schools, is decentralised to the provincial level. This leaves development and implementation of school nutrition policies to provincial ministries of health and education. School districts within each province may also develop policies independently. In BC, The Guidelines for Food and Beverage Sales in BC Schools (‘the Guidelines’) were first created in 2005 by the BC Ministries of Health (BCMoH) and Education (BCMoED), with the most recent iteration launched in January 2014. These mandatory guidelines apply to all public schools and consist of nutrient-based criteria. They aim to reduce the amount of salt, sugar, and fat in items sold to students in all food sales venues in schools. These may include vending machines, cafeterias, fundraisers, sports days, and school stores. As there are 60 school districts with varying geographies and demographics in BC, the BCMoED further decentralises power to the district level to determine how to carry out top-down directives, including 105  implementation of their mandated school food and beverage sales guidelines. This decentralisation of power to the districts means districts may incorporate implementation processes differently, with varying levels of success. The existing implementation evaluations specific to BC’s policy have found that school food and beverage sales environments improved over the years (BCMoEd & BCMoH, 2008; Watts et al., 2014), but implementation challenges continue to exist (Masse et al., 2013).  As the policy was first being rolled out in 2005, the BCMoEd conducted a baseline assessment of availability and types of food and beverage sales that were occurring (BCMoEd & BCMoH, 2005). A follow-up study conducted in 2008 showed that progress had been made regarding elimination of sugar-sweetened beverages (SSBs) from vending machines, but the presence of “not recommended” foods in snack vending machines remained prevalent (BCMoEd & BCMoH, 2008). Sixty-six percent of middle and high school principals reported full compliance with the Guidelines in vending machines.  A later study found that administrators reported significantly lower odds of having a number of important unhealthy food ‘culprits’ available in 2011/12 compared with 2007/08 (Watts et al., 2014). However, only 45%, 36%, 10%, and 8% of school snack shops, cafeterias, fundraising events, and special events, respectively, were found to be fully compliant. In addition, significant improvements were reported by elementary administrators regarding availability of fruits and vegetables, in contrast to middle and high school administrators who reported no change in this area (Watts et al., 2014).   Only one in-depth qualitative study has been conducted in BC to help explain these different levels of compliance in different places and in different venues (Masse et al., 2013). Data was collected in 2010/11 to identify barriers and facilitators to implementation across a variety of stakeholders. The Guidelines were perceived to be compatible with expectations of what the school learning environment should provide, but respondents felt the complexity of the Guidelines led to issues of clarity about their scope (i.e., the foods and beverages to which they apply). Furthermore, struggles to maintain profit margins, as well as the misaligned beliefs about school food between parents and teachers and the intent of the Guidelines were also perceived barriers to implementation. There was also tension around whether it was families’ responsibility to ensure healthy options or whether it was schools’ responsibility to not profit from the sales of unhealthy foods to children. Finally, a lack of capacity to implement was also reported as a barrier to implementation. Regarding facilitators, 106  Masse et al (2013) found that one administrator reported the top-down mandate helped them achieve their goals. Lastly, Masse et al. (2013) reported that the existence of supportive implementation tools from the provincial government, having existing nutrition programming, access to dietitians, and local suppliers that complied with the Guidelines were all specifically reported facilitators that do not appear to be articulated in other studies specific to school food and beverage sales policies. Masse et al.’s (2013) study provides a foundation of evidence from which to build a conceptualisation of the implementation experiences in the BC context. My study adds to this work by articulating underlying social processes associated with facilitators and barriers Masse et al.’s (2013) study identified.  One aspect of the complexity of this type of intervention is that it is implemented at different levels and across sectors – provinces or states, districts and schools each have different roles. For a nuanced understanding of implementation processes and the complex nature of this intervention, it is helpful to separate the experience of these different levels where different types of implementation actors and activities exist and occur.  In this chapter, I focus on school-level experience; the experience of school districts is described in the previous chapter, Chapter 5.  6.2 Methods Realist approach to implementation research  This study adopts a realist approach which is appropriate for evaluating complex public health interventions (Lacouture et al., 2015; Pawson & Tilley, 1997). One key assumption of the realist approach is that no intervention works in all places, at all times, for all people. This is because of the challenges associated with launching a complex intervention in complex social contexts. Not only this, the success of an intervention is dependent on the decision-making of individuals. It is this inevitability that accounts for the inconsistent levels of compliance observed in existing evaluations (Pawson, 2013).   Realist evaluation attempts to explain why implementation and uptake of a given intervention happens the way it does through unearthing relationships between contexts, mechanisms, and outcomes. Contexts, defined as the “back drop” for an intervention (Jagosh et al., 2015, p. 3), are considered to be pre-existing components of a social system before an intervention is rolled-out. The realist evaluator’s conceptualisation of context is broad and can include, for example, social or cultural norms, histories of organisations or people, and/or structures of governance. At the heart of 107  the struggle of realist evaluation is the identification of which specific dimensions of the broader context are the most relevant for the success of a specific type of intervention (Jagosh, 2017b). Mechanisms are defined as the interaction between intervention resources (concrete or non-concrete) and stakeholders’ response to them. A concrete resource can be considered to be something like funding. A non-concrete resource might be unintended and may emerge from a concrete resource such as the fear that may emerge from an educational campaign warning of the negative health impacts of drinking more units of alcohol per week than is recommended. Stakeholder responses can be, for example, cognitive, emotional, or motivational and help explain the decisions stakeholders make about taking action (or not). It is the dynamic interaction between how contexts shape the manner in which stakeholders respond to the resources on offer from the intervention—the mechanism—that leads to a wide variety of outcomes. Outcomes are simply defined as effects of an intervention resource, whether intended, unintended, short-term, intermediate, or long-term and can be related to the overall outcome of an intervention or to outcomes related to the processes of implementation. This study is concerned with the latter (i.e., whether people are implementing or not rather than whether the intervention is working or not).  Realist evaluation starts with a program theory, a plausible explanation of how an intervention is expected to work (Bickman, 2000). A retrospective logic model and program theory was developed in an earlier phase of this research (see Chapter 4) and guided the inquiry for the current investigation (Blamey & Mackenzie, 2007). Essentially, the program theory for this intervention, as presented in Chapter 4, is: by launching nutrition standards for foods and beverages sold to students in schools and providing information on why these standards are needed, those responsible for creating school food environments will be motivated to use the implementation tools to change how they create them (the implementation-related outcome which is the focus of this research). Then schools will only sell foods and beverages that adhere to the standards. The more intermediate and long term outcomes (not the focus of this implementation research) are that students will purchase and consume healthier options, learn how to make healthy choices, and ultimately have better health outcomes both in the short and long term. While specific contexts may mean school food and beverage sales policies will be implemented differently in different places, with different types of specific activities, this underlying, broad program theory is transferable across different contexts.   108   Data collection To investigate school-level implementation processes in different BC contexts, a multiple case study approach was used to create contextual boundaries and assist with comparing within and across them (Yin, 2009). The case unit was defined as a school district and all public kindergarten to grade 12 schools within each district were invited to participate. Defining the case at the school district level rather than the school-level was done so as to be able to explore whether and how district-level contexts influenced implementation at the school level. Five districts (out of BCs 60 school districts) participated in the study, two rural and three urban. The rationale for the selection of these five districts is discussed in the methods section of Chapter 4. These five districts collectively house roughly 21% of BC’s public school population. Each district is located within the jurisdiction of a Regional Health Authority (RHA) whose interactions with the education sector differ depending on their organisational structures.  Aligning with the realist approach, I adopted a variety of qualitative strategies outlined in Table 6.1 below:                   109   Table 6.1 Data collection strategies used by district and Regional Health Authority (RHA) Data collection strategies District 1 urban, RHA A  District 2 urban, RHA B  District 3 urban. RHA B  District 4 mostly rural RHA C  District 5 mostly rural RHA D Total Semi-structured interviews with district staff* 2 1 1 -- -- 4 Semi-structured interviews with public health sector staff** 3 2 --*** 3 5 13 Semi-structured interviews with vendors 4 2 -- -- -- 6 Event attendance**** 3 -- -- -- -- 3 Questionnaires***** 14 19 15 10 4 62 District and RHA context scan Yes Yes Yes Yes Yes -- School website scan 10 12 10 17 13 62 *One semi-structured interview was also conducted and included in the analysis with the singular person at the BCMoED who has the Guidelines as part of their portfolio  ** While only 13 semi-structured interviews with public health sector staff are outlined in the table, a total of 22 were conducted with public health sector staff across the province and are not included in the table as they fell outside of the RHAs in which the participating districts were located. While the remaining 9 were not formally included in the CMO analysis for each specific district, the broader discussions they contained about the Guidelines influenced interpretation. ***Because district 2 and 3 were in the same RHA, the interviews with health sector staff from this RHA were used for the analysis of both districts ** One event was specifically in school district one, organised by the school district in partnership with the RHA; the other two public events were organised by the same RHA but in a different school district not part of this study; the public health dietitians involved in organising these public events, however, utilize the same format and the informal discussions I heard at all the events, regardless of the district, were similar in nature. *****There were also 10 additional questionnaires on which the specific district of the respondent was not identified. Comments broadly related to Guidelines implementation influenced interpretation.  Semi-structured interviews were conducted with purposefully selected health and education stakeholders, and private vendors. This provided in-depth information on the relationships and interactions among stakeholders across all levels of implementation, contexts, and key activities related to Guidelines implementation. Interviews lasted from 30-90 minutes and were recorded and transcribed with permission of the participant. Two iterations of interview guides were used. The first version was more exploratory with the aim of developing the retrospective program theory for phase 1 (see Appendix C). The aim of the second version was to garner more specific insight about participants’ experiences with implementation as it related to the retrospective program theory (see Appendix E).   110  Short 15-20 minute questionnaires were conducted with administrators and parents. This format was selected for these respondents to increase their participation and in light of the many demands on their time. The questionnaires were administered either online, via telephone, or in-person. They were developed based on the logic model and program theory to explore a number of implementation processes, from dissemination of information to collaboration, to nutrient assessment of foods and beverages. They included both short and long answer questions (Appendix F). Questionnaire data were collected using Fluid Surveys (SurveyMonkey, 2017). Four of the five districts allowed me to contact school administrators directly, while one did not. However, in this district I was allowed to attend a public event to collect data in-person with mostly parent stakeholders. When conducting telephone questionnaires with administrators, they were asked if they would be willing to forward the online link to the questionnaire to any relevant stakeholders in their school community. The result was that questionnaires were mostly answered by administrators with a small amount of other participant types, as shown in Table 6.2 below: Table 6.2 Questionnaires completed by district and school-type  District 1 (urban) District 2 (urban) District 3 (urban) District 4 (mostly rural) District 5 (mostly rural) Unknown district  Respondent-type Elem. Sec. Elem. Sec. Elem. Sec. Elem. Sec. Elem. Sec. Elem. Sec. TOTAL Parent 12  7  2  1      22 Administrator   6 5 10 3 7 2 4  1  38 School food staff 1 1         2 1 5 Teacher            1 1 Unknown    1       3 2 6              72  Lastly, I conducted a variety of website scans exploring school websites from each district in search of mentions of food sales-related activities. Website scans of school districts and RHAs provided more context around organisational structures, explicit healthy school food policies created by school districts, and amount of information available around school food. Moreover, I scanned the websites of private vendors who participated in the study for more information about their products.  Analysis The interview transcripts and the long-answer portions of the questionnaires were open coded using qualitative analysis software (Atlas.ti, 2015). A linked CMO coding technique was used (Jackson & Kolla, 2012; Puton et al., 2016). This helped with analysing the data in context-mechanism-outcome 111  configurations (CMOcs) and making their relationships to one another more clear. This technique involves identifying an outcome described in the interview then reading for the respondent’s explanation as to how this outcome came about. When describing an event or occurrence, participants often inadvertently described relationships between contexts, mechanisms and outcomes. The Cs, Ms, and Os identified in a portion of narrative would be assigned a linked code (e.g., C24-M10-O23). However, in some cases respondents only described either a mechanism or context that related to an identified outcome and it was necessary to ‘fill in the blanks’ to articulate full CMO relationships with a plausible explanatory inference. My inferences for these circumstances were underpinned by an existing implementation theory, contextual interaction theory (CIT) (O’Toole, 2004; Owens, 2008). CIT posits that all intervention actors consist of three dimensions (knowledge, motivation, and capacity) that dynamically interact with one another as well as with components of the context. It is these interactions that influence how an actor will respond to an intervention.   Linked CMO codes were then categorised based on intervention activities identified in the development of the retrospective program theory (Chapter 4). For example, CMO codes containing outcomes related to dissemination or collaboration activities were grouped together. For each district, I then created memos for each activity-category that contained all of the category’s associated CMO codes. The information garnered from the context scans was also incorporated into the memos. These memos were used to organise the analysis and interpretation process conducted for each district, first, before considering the analysis across districts. This helped to distill how the different contexts of each district may be influencing mechanisms at the school level, leading to diverse or similar outcomes across schools. The short answers of the questionnaire were not meant to be used for quantitative analysis but rather to further support interpretations of the qualitative data (Greene, 2007).  Ethics Ethical approval for this study was obtained from the University of British Columbia’s Human Research Ethics Board and each participating district (Appendix B). Informed consent was obtained prior to data collection with the assurance of confidentiality (Appendix G and H). 112  6.3 Findings  Study participants’ observations and experiences working with school-level stakeholders and/or as school-level stakeholders provided explanatory insight about social processes influencing implementation of the BC Guidelines. Specifically, four key mechanisms were identified that interact with specific dimensions of context to trigger stakeholders to engage or not engage in implementation-related activities: (1) mandatory mechanism; (2) scofflaw mechanism;5 (3) money mechanism; (4) resource constraint mechanism. 6.3.1 Mandatory mechanism and scofflaw mechanism: Opposite responses to the same intervention resource One dimension of BC’s school food and beverage sales policy that emerged from the analysis as important is that the BCMoED declared the policy to be mandatory, albeit with no official monitoring or enforcement measures. From a realist perspective, the mandate, itself, is considered an intervention “resource” -- i.e., a dimension of the intervention that prompts people to change their behaviour and create a healthy school food and beverage sales environment. This intervention resource is distributed equally because it applies to everyone involved in creating food and beverage sales environments at the school level. Yet I found that this mandatory directive led to two different responses among expected-implementers. First, this mandate triggers some stakeholders, such as administrators, to engage in implementation-related activities, like requiring any teacher or parent who wishes to fundraise to have their ideas approved through them. These participants explained their desire to be seen to be abiding by the rules. On the other hand, for some other expected-implementers, including teachers and parents involved in fundraising, the “mandatory” resource triggers a “scofflaw” mechanism whereby forms of resistance emerge such as ignoring the mandate or finding ways to ‘skirt’ around it. From the data, I identified specific dimensions of context that could be influencing these two different responses to the mandatory nature of the intervention.  Mandatory mechanism  There were many examples of school-level stakeholders making efforts to implement the Guidelines in response to the mandatory nature of the Guidelines, pointing to this mechanism being an important driver of implementation. Administrators reported being compelled towards making                                                           5 “Scofflaw” refers to people who feel resistance to authoritative directives because they do not agree with being told what to do (especially by a government). They may respond by ignoring directives altogether or act in ways meant to openly exhibit their resistance, flouting the directive in defiance (Burns, 2011). 113  efforts to implement the Guidelines because it was a directive from the province and their school district. Four of the five school districts have compliance to the provincial Guidelines explicit in district health and wellness policies. Compliance with the Guidelines appears to have become part of school administrators’ jobs.  The influence of the mandatory directive was clear in one particular example where an administrator reported their own belief that the Guidelines are targeting the wrong audience and are just the “ministry covering their butts because of the obesity issue in Canada” (Administrator, SD 4, questionnaire). Regardless of this personal belief, this administrator reported a relatively high level of engagement in implementation activities so as to avoid being reprimanded by the district-staff, who appeared to be more authoritative around Guidelines-related issues than district-staff in other school districts.  The evidence also showed that the mandatory mechanism trickles down from the administrator, triggering some school-level stakeholders to accept the intervention and make efforts towards implementation. As one administrator stated regarding their parent advisory council (PAC), who are often involved in food-fundraising: “[The PAC] came around and realised ….that was just how it was going to be…” (Administrator, SD 1, questionnaire). The manner and magnitude in which this top-down pressure is exerted from administrators to other expected school-level implementers (parents, teachers, etc.) ranges from asking school-stakeholders at the beginning of the year to consider the Guidelines in their food fundraising efforts to outright bans of specific activities, like bake sales. Some administrators reported that they required all teacher and student fundraisers to be approved by them. Ultimately, it is likely that for a mandatory directive to trigger the expected implementers to engage in implementation activities, even in the absence of enforcement measures, it needs to be launched into the context of an organisation that is normatively hierarchical. As one administrator stated in response to what they felt triggered people to engage in implementation activities: “everyone's willingness to adhere to what the government tells them to do” (Administrator, SD 4, questionnaire).  BC is a context where the education system is relatively hierarchical, whereby districts are broadly directed by the provincial ministry, and districts direct administrators, who in turn play an important role in determining the functioning of their school.  114  There is a unique dynamic between administrators and parent advisory council members who volunteer for fundraising efforts as opposed to other school-level stakeholders. In secondary schools, in contrast to elementary schools, parents were not typically engaged in food-fundraising. In elementary schools, the volunteer parents are not under the “authority” of the school principal through employment, unlike teachers or school food staff, or through the subordinate position of students. Two important components of context were reported to influence how much pressure elementary school administrators put on PACs to comply with the Guidelines in their fundraising efforts: the quality of their relationship and the administrator’s perception of the PACs capacity. If the PAC was perceived to have sufficient capacity and there was a positive relationship with their PACs, administrators reported feeling able to exert more top-down pressure.  While the mandatory directive triggered administrators and some other school-level stakeholders to engage is implementation activities, so as to be seen as following the rules or adhering to their job role, another important contextual influence is pre-existing beliefs about food and health among school stakeholders. While this did not necessarily emerge as essential for driving engagement in implementation activities, there was some indication that the existence of “passionate parents” (Administrator, SD 1, questionnaire) or teachers, whose pre-existing beliefs aligned with an administrator’s intent, helps bolster implementation efforts.               115  Table 6.3 Context + mechanism  outcome configurations for the mandatory mechanism at the school level Context Mechanism (resource + reasoning) Outcome Example quote The education system is hierarchical  Mandatory directive from the district/top-down pressure to implement the Guidelines in schools motivates school principals, as part of their job Administrators encourage/support/mandate that school staff, students, and parent volunteers, when fundraising with food, are making efforts to meet the Guidelines Students have to come present to us what they're selling and the timelines. We tell them they need to meet the Guidelines-- actually the teacher or student always has to come to me to present the idea. (Administrator, SD 2, questionnaire) The education system is hierarchical  Enhanced efforts to implement if the beliefs of some or all of the school-level stakeholders align with the intention of the Guidelines Mandatory directive from administrator/top-down pressure motivates school-level stakeholders Efforts are made to comply by school-level implementers (e.g., alternative fundraisers or demand that private vendors offer compliant items) …We get told all the time that we better be following it and otherwise we'll get in trouble--the district guy threatens us… (Administrator, SD 4, questionnaire)   When all players are on the same page- striving for common goal-- gave me that ability to push forward with implementation. (Administrator, SD 4, questionnaire) Administrator’s quality of relationship with PAC members  School-administrator’s perception of PAC capacity Mandatory directive from the district/top-down pressure to implement the Guidelines in schools compels school principals, as part of their job Administrators encourage/support/mandate parent volunteers, when fundraising with food, to make efforts to meet the Guidelines I think that stems from having a good relationship with your parent group and your PAC. I know our PAC is fantastic and always very respectful. They always consult with me on food ideas. (Administrator, SD 1, questionnaire)  Scofflaw mechanism While the mandatory mechanism successfully drives some expected implementers to engage in implementation activities, there are other expected implementers (e.g., parents and/or teachers) among whom this mechanism is not at work. Instead, they reportedly respond differently to the mandatory directive, responding with feelings of aversion to being told what to do by the government. This drives them to ignore the Guidelines, and continue on with business-as-usual (i.e., continue selling the types of items they have always sold, many of which are categorised as unhealthy by the Guidelines), or find ways to skirt around the Guidelines. This means that 116  sometimes these expected-implementers devise means by which to technically comply but that may be counter to the intent of the Guidelines, including: continuing to sell popular items for fundraisers that do not meet the Guidelines but sell them to parents instead of students; selling items to students that technically comply due to their size but it is likely students will probably purchase more than one (e.g., donut holes, a slice of pizza); and selling an item for a fundraiser that complies, like milk, with an accompanying ‘free’ cookie that does not technically fall within the reach of the Guidelines because it is not being sold. These antithetical activities are further enabled because of the absence of official enforcement measures.  These findings align with other data I collected about perceived levels of implementation for school-level organised initiatives like food fundraisers or sporting event food sales. In the questionnaires, it was found, almost universally, that respondents perceived the food sales organised at the school level as only ‘often’, ‘sometimes’ or ‘rarely’ compliant. This is in comparison with the almost universal perception that district-organised food ‘almost always’ or ‘always’ complied with the Guidelines.  A number of important contexts influence this aversion to the mandatory nature of the Guidelines—reducing acceptability of the intervention--and the subsequent lack of engagement in implementation activities. First, is the context of needing to fundraise in schools, due to the perceived underfunding, for providing a well-rounded education experience. School website scans show that fundraising conducted by teachers is typically for supporting student clubs while PAC fundraising profits support items like educational supplies, guest speakers and performances, wellness workshops for students, playground equipment, school gardens, etc.  These expected-implementers, who engage in this important fundraising work, are then told they must change how they have been fundraising in a way they believe will decrease fundraising profits.   …it's a lot about making money. Like I know personally people on the PAC and it's like they've ordered Five Alive [a sugary drink] for the events and it was like, "well it was 25 cents a liter at Costco and how could we pass that up". And then they're selling stuff for sports day and they try to have some healthy stuff but [at the same time] they’re like "well we have to [also] have some jumbo freezies because ...we have to make some money!". (Dietitian and parent, RHA 3, interview) 117   Secondly, pre-existing beliefs around the role of government in society may not align with this type of intervention instrument (i.e., a mandate), causing a discordant aversive reaction to this “draconian imposition of someone else’s values” (Administrator, SD 1, questionnaire). Third, pre-existing beliefs about food in general and food in schools, specifically, may also not align with the content of the Guidelines. These pre-existing beliefs reported by participants, include: the belief on the part of stakeholders that children will not eat (or purchase) healthy options, that children deserve to have treats at school, that treats do not happen often at school, and that children require  larger portion sizes than those that comply with the Guidelines. Fourth, if there is inconsistent compliance of the Guidelines either between venues in a school or between schools, which was reported to often be the case, perceptions of ‘unfairness’ emerge. One dietitian spoke about experiences where parents from one school sometimes notice other schools within their district not complying and responding with the question of “ well why are we trying so hard?...” (Dietitian, RHA 2, interview). Table 6.4 Context + mechanism  outcome configurations for the opposition to implementation  Context Mechanism (resource + reasoning) Outcome Example quote Elementary school-level stakeholders are often responsible for organising external hot lunch vendors and other food sales for fundraising  AND  These school-level stakeholders hold many pre-existing beliefs about school food and food, in general that are not necessarily aligned with what the Guidelines are trying to do   AND/OR  Inconsistent implementation and enforcement in other schools is observed by school-level stakeholders The top-down mandate and knowing there are no repercussions for non-compliance leads some school-level stakeholders to have averse feelings towards the Guidelines They may choose to not make efforts to implement (Outcome A)   …and/or they may choose to ‘skirt’ around the Guidelines (Outcome B)  …then when we have sales, we have milk and cookie day--so we'll sell the milk and give the cookies out---so that is an example of when people get pushed. (Administrator, SD 4, questionnaire)  … it's too rigid. 'Therefore, you have to do this.' Then people get turned off 'cause, like, you're being pushed to do it... (Dietitian, RHA 2, interview)   118  6.3.2 Money mechanism: Demand created leads to supply creation Among those administrators, parents and teachers at the school level who are concerned with trying to comply with the Guidelines, a demand is created as a result of their efforts to comply. This demand is geared towards private food vendors and can also, from a realist perspective, be considered another “intervention resource” to which private food vendors respond.  It was reported that the demand created for compliance influenced the motivation of vendors to offer compliant food and beverage options. Unlike elementary schools, most secondary schools do not conduct hot lunch fundraisers. Therefore, secondary school-level stakeholders do not often engage with private food vendors. However, at the elementary school level, where hot lunch and other snack fundraisers might occur anywhere from once per month to once or twice per week, vendors wishing to sell their products in schools were reportedly only presenting options that comply, otherwise they risked being excluded.  Usually [hot lunch vendors] will come in and give a proposal.  [They] are generally very good at showing you how they meet the Guidelines. No one is going to bring you a product that doesn't meet the Guidelines cause you would just strike them off right away. (Administrator, SD 1, questionnaire)  It is important to acknowledge, however, this mechanism was not necessarily at work all of the time in all places. Compliance by private vendors was reported in the questionnaires to be occurring anywhere from “sometimes”, “often”, to “always”. Vending machines were largely reported as “always” compliant. Participants were more skeptical of how compliant hot lunch vendors were, with about half reporting they were always compliant and the other half reporting “rarely”, “sometimes, and “often” compliant.    Participants identified two types of hot lunch fundraiser vendors who comply with the Guidelines in two different ways. The first type of vendor, popular fast food franchises, is one in which compliance is achieved through the reformulation of products. Respondents from all the districts, website scans, and observations at public events confirm that fast food franchises are an important source used in fundraising by school-level stakeholders.  The data collected from observations and informal conversations at public events in urban areas and promotional brochures showed some franchises have reformulated some items to comply with the Guidelines. For example, in informal 119  discussions with pizza franchises attending public events related to school food, they describe creating pizzas topped with less cheese to comply with the fat criteria in the Guidelines. As one small pizza franchise states in their information sheet for school-level stakeholders responsible for organising fundraisers:  [We are] pleased to offer our Hot Lunch Program which follows the Guidelines for Food and Beverage Sales required by the BC Ministries of Health and Education. As a result, [we] are featured in the [BC Ministry of Health’s] Brand Name Food List, the list that approves ready-to-eat, packaged, and franchised foods. (Vendor information sheet provided at public event)  Data was not directly collected from rural franchises but some rural school-level respondents stated they believed their providers were reformulating their offerings. This reformulating allows PACs to provide an option for students that comply and are classically considered ‘fun foods’. This was also perceived (by parents and administrators) to help ensure students would be interested in purchasing these items, generating higher profits for fundraising and for the vendor.   Smaller hot lunch vendors that create a supply of food offerings for PAC hot lunch fundraising initiatives are a second type of vendor working at the school level. Foods provided by these vendors were reported by the vendors themselves to meet the Guidelines. In addition, these vendors promoted their offerings as being diverse, made by hand using whole foods, and as being healthier. The vendors in this study, who appear to be only active in urban areas, reported being motivated not only by a sustainable income but also their desire to contribute to a healthy, diverse food culture. Even when asked to reduce their prices by PAC members, one vendor reported they would never reduce the quality of product to maintain their bottom line but rather would renegotiate the price with them. Another vendor reported they use their philosophy about nutritious food to determine with whom they will work, gently refusing to work with PACs who do not share their whole-foods, healthy food philosophy. This suggests vendor beliefs are important in determining what kinds of offerings they provide and how these beliefs can even override profit motivations. These vendors’ compliance may not be fully related to the existence of the Guidelines, unlike some of the fast-food franchises, as their whole foods, handmade products reportedly comply without much effort. There was some evidence to suggest, though, that the existence of the mandatory nutrition standards provides vendors clout to back up their principles when dealing with school stakeholders asking for 120  “hot dog days” and “regular chocolate milk” (as opposed to the reduced sugar version) (Vendor SD 3, interview). However, regardless of receiving these types of requests from some parents involved in fundraising, they described ample demand for their whole foods, reportedly nutritious products which suggests there are many school-level stakeholders looking for healthy options. Table 6.5 Context + mechanism  outcome configurations for demand and supply creation at the school level  Context Mechanism (resource + reasoning) Outcome Example quote Franchise owners of franchises that offer conventional school  food have a pre-existing drive to sustain their business   The Guidelines are specific nutrient criteria and the demand by school stakeholders to comply with the mandatory Guidelines incentivises vendors Items are compliant but are reformulated to mimic conventional school food  We work with a local pizza place and it is one of our infrequent yellow items-- last month the PAC did a pizza day and we work with them---we work with Booster Juice-- we also work with Quiznos-- they've done their homework that we are happy to buy and sell to kids.  (Administrator, SD 5, questionnaire) Small hot lunch vendors (at least in urban areas) have a pre-existing drive to sustain their business. Small hot lunch vendors also have pre-existing beliefs supporting healthy, diverse foods and changing food culture.  These vendors are motivated by their own value system, regardless of the existence of the Guidelines, to provide healthy and delicious options and the nutrient criteria of the Guidelines provides them the information needed to align their products with expectations They create hot lunch offerings that are interesting, healthy, diverse, prepared using whole foods that comply with the Guidelines The whole idea of the [company] is that we wanted to make everything from scratch …but also, number two, to have an ethnically diverse menu….I think kids should be exposed to a wide variety of different types of foods… (Hot lunch vendor (A), SD 3, interview)  So, and my job is to provide things that are…from scratch...which removes a lot of the Guideline issues…but something that tastes good...I know that they're getting some nutrition and protein so they can learn well. (Hot lunch vendor (B), SD 3, interview)  6.3.3 Resource constraint mechanism  One other type of expected implementer at the school level is those who are motivated to change the food environment but feel they do not have the capacity to do so. Specifically, some participants reported capacity issues related to human resources, lack of ways through which to offer items that 121  comply, and the higher cost of items that comply.  Because the Guidelines do not offer intervention resources to address these capacity issues, it appears that some implementers, as a response, may be de-motivated and ultimately do not make efforts to implement.  Human resource capacity in schools, in particular elementary schools where there is a high dependence on parent volunteers, is related to the level of engagement by parents as well as the transient nature of PACs. An administrator, who told of a  high-level of engagement in implementation in a previous school with a PAC that desired change, discussed how her transfer to another school, where there was a lack of engaged parents, left her feeling that she had decreased motivation and power to engage the parents in moving towards implementation. Administrators often associated this lack of parent engagement with a low-socioeconomic status of parents and less food sales activity in their schools. Regarding the transiency of PACs, each year as children move to a different school, parents leave with no guarantee that another equally engaged parent will replace them. This perceived lack of human resources was also reflected in the aim of many parents to find the most convenient ways to do food sales.  Accessing options to meet the Guidelines was related to lack of food preparation infrastructure and, in rural areas, the lack of private vendor options. From the data, it is mostly secondary schools that have a cafeteria with food-related infrastructure. This was universal in the urban districts and varied in the two rural school districts. The lack of food preparation and food storage infrastructure, as a barrier to making efforts to implement, was echoed across all districts. Regarding the availability or existence of private vendors that could provide compliant offerings, in urban areas there appeared to be many potential options of external vendors. However, in the rural districts, there may be only more traditional fast-food franchises, or small family-owned private restaurants or, in some cases, convenience stores from which school-level stakeholders would purchase items to, in turn, sell in their schools.  We have one vendor that reaches our Guideline goals.  Just one in our community that is in our price range and desirable for offerings for the kids.  So, we choose to use other vendors and not follow the Guidelines… fully aware that we are not following them because it is just not possible to provide the food items we would like with the restrictions that have been put in place. (PAC member and dietary technologist, SD 4, questionnaire) 122   Lastly, capacity related to the cost of items that comply was reported by some administrators and parents. This is related to the overall context of the need for schools to fundraise; food fundraising input cost must be as low as possible to ensure the highest possible profit. Schools neither have sufficient financial resources to provide healthy options for free nor to subsidize them. They also reportedly do not have sufficient funds to provide well-rounded educational experiences, as reflected in the need to fundraise for items such as sports teams or clubs. Finding these funds is offloaded from government onto school communities, who seek to fundraise with offerings they believe will sell well, which requires the lowest financial input (e.g., cheap, processed items).                       123  Table 6.6 Context + mechanism outcome configurations for the resource constraint mechanism Context Mechanism (resource (or lack thereof) + reasoning) Outcome Example quote Administrator’s quality of relationship with PAC members is less than collaborative/consistent  School-administrator’s perception of PAC capacity is low Mandatory directive from the district/top-down pressure to implement the Guidelines in schools is not sufficient to trigger action as the administrator is motivated to maintain a positive relationship with parent volunteers Administrators may feel reluctant to engage with parent volunteers about compliance resulting in less efforts made to meet the Guidelines …it was easier for me to shut the canteen down than to try and make that switch in the community--I've got such a small school-- we're having more single parents or more parents working so they're not available either to come in and take this kind of thing on. (Administrator, SD 5, questionnaire) School-level stakeholders are motivated to create health maximising opportunities for children   BUT…  lack time, infrastructure, and financial capacity The Guidelines, as an intervention, does not provide financial, material, or human resources in schools nor any official enforcement mechanisms and so school-level implementers feel constrained Choosing to continue on with business-as-usual School food and beverage sales offerings may not comply with Guidelines  It’s totally efficiency because I did try to bring in… watermelon on sports day and... immediately it was like ‘I am not going to spend my time slicing off watermelon’….everyone was all up in arms about how much work this was going to be [laughs]. I was like ‘are you kidding me? It will take like 20 minutes! (Dietitian and parent, RHA 1, interview)  …it would be nice just to have fresh local healthy food. But it is also the practicality of doing that…most of the schools [in the district] do not have kitchens in their school. (Dietitian and parent, RHA 2, interview)  6.4 Discussion The findings outlined above present implementation processes, consisting of key mechanisms influenced by specific dimensions of context, that lead to expected-implementers either choosing to engage in implementation activities or not. While implementation evaluations conducted in other jurisdictions have explored barriers and facilitators that align with my findings, this is the first study to begin to explicate the specific relationships between these implementation-influencing factors and why and how they lead to decisions to engage in implementation activities or not. 124   First, the mandatory nature of the Guidelines does motivate some school-level stakeholders towards implementation. Friedman (2014) argues that the public are likely to be amenable to more authoritarian interventions, like mandates, when a) children are involved and b) when they are rolled-out in settings that are already viewed as, and accepted as, paternalistic settings. Reflecting this, studies from North American jurisdictions find parents and teachers overwhelmingly support limiting unhealthy foods in schools through policy (Kubik, Lytle, & Story, 2005; Spitters, Schwartz, & Veugelers, 2009). However, I also found that this is not necessarily the case for some when it comes to this school food and beverage sales intervention, whereby some stakeholders reported an aversion to the paternalistic nature of this mandatory intervention, which negatively affected decisions about taking action to implement. This aversion has been reported elsewhere as a barrier to implementing school food and beverage sales policies (Devi, Surender, & Rayner, 2010; MacLellan et al., 2010; Masse et al., 2013; Morgan, 2006; Vine & Elliott, 2014). The school-level stakeholders in this study who felt aversion to this intervention might be described as “scofflaws”, a term developed during the prohibition era to describe those who disagreed with the paternalistic nature of the law and, knowing it was unlikely to be enforced, continued to sell alcohol (Burns, 2011).  The term has since been used to describe anyone who flouts rules or laws, particularly when there is little chance of enforcement or repercussions. While the Guidelines in BC are not legislation, the same relationship between context and mechanisms appears to be contributing to less engaged implementation efforts, similar to what was occurring during prohibition. This reported aversion to strong interventions meant to support children’s health is disconnected from the normative ideal that adults are concerned with maximising children’s welfare. It may not necessarily be the case that implementers are averse to strong interventions to address the quality of foods and beverages for sale in schools purely on the basis of it being paternalistic (Levay, Chapman, & Seed, Forthcoming). Rather, it may be that as expected-implementers face various structural barriers to carrying out this duty imposed on them by a government mandate, they are experiencing constrained choice (Hendrickson & James, 2005, 2016; Stuart & Schewe, 2016) and loss of control, which have been found to trigger aggression, non-cooperation (Baumeister, Masicampo, & DeWall, 2009; Stuart & Schewe, 2016) and lower productivity (Iyengar & Lepper, 1999). This aversion to paternalism has been reported as an impediment to the success of various social interventions (Brooks et al., 2017; Dempers & Gott, 2017; Rhee, Zwar, & Kemp, 2012). Further research exploring the public’s aversion to the paternalistic nature of interventions meant to support social welfare can provide 125  deeper insight into why and in what contexts people do and do not support and comply with, not only paternalistic school food environment interventions, but also with paternalistic interventions more generally.  The findings also indicated that implementation efforts may be enhanced if there is a pre-existing context in which the intention of the Guidelines aligns with the food and health beliefs of school stakeholders, whether those stakeholders are parent volunteers, employed education sector staff, or private vendors. The critical role of parent and leadership support in facilitating implementation of school food and beverage sales interventions has been reported in other studies (Frerichs et al., 2016; Goldberg et al., 2009; MacLellan et al., 2010; Pettigrew et al., 2013; Vine & Elliott, 2014).  Regarding private vendors, there is a paucity of literature addressing motivations and beliefs of private vendors who are supplying school food. Scott, Hawkins, and Knai (2017) called into question the intentions of food industry actors, suggesting that reformulation of products to meet nutrient-criteria is simply a corporate political/survival strategy. However, the findings presented here suggest that, for at least some local and regional smaller private vendors, personal commitment to food and health supports their compliance with the mandatory government policy. For all stakeholder groups, further exploration of how personal commitments to food and health relate to efforts to implement school food guidelines is warranted, including examination of how that relationship varies between school volunteers, employees, and suppliers.   The need to fundraise in Canadian public schools as a barrier to implementing nutrition standards has been reported elsewhere (Downs et al., 2012; Masse et al., 2013; Vine & Elliott, 2014). The aim of fundraising is, fundamentally, to collect as much funds as possible. In the school context, when fundraising occurs, there is a need to sell products that students are willing to buy which will result in the most profit. Traditionally, this has included food and beverage items such a pizza, hot dogs, or bake sales that require a low financial input, that sell well and generate high profits. It is often perceived that nutrition standards make it difficult to have profitable fundraisers as it limits what can be sold (Downs et al., 2012; Masse et al., 2013; Vine & Elliott, 2014). However, the findings of my research suggest that this oft-reported barrier to implementation may be oversimplified and that it may be closely linked with pre-existing beliefs about both school food and food in general.  First, I found the belief in the narrowness of children’s food preferences (and therefore what they will purchase) appears to influence decisions around what foods and beverages will be sold. This has also 126  been reported in another Canadian study (MacLellan et al., 2010). My research adds to this by elucidating other beliefs that may also be influencing these decisions, including the beliefs that children are entitled to treats and that the evidence-based portion sizes recommended in the nutrition standards are insufficient. The first of these two beliefs may be founded on deeply embedded normative beliefs about a) children’s entitlement to a “joyful” (Bennet & Bergstrom, 2015, p. 27) childhood that necessarily includes treats, and b) the profound meanings of the transaction of food between parent (and perhaps other types of caregivers) and child and what provision of a treats may symbolize (Kaufman & Karpati, 2007; Maher, Fraser, & Wright, 2010). These deep belief systems may also be influencing decisions around what foods and beverages will be sold for fundraisers.   Regarding beliefs about portion sizes, the findings suggest that when parent fundraisers, in particular, select options for fundraising that meet the nutrient criteria but believe the portion sizes are insufficient, they increase portion sizes, resulting in non-compliance. One possible reason for this is that the adults involved in fundraising do not correctly perceive the appropriate portion sizes for children, believing children require more food than they physiologically do.  Emerging work, albeit with a small sample size of 52 observations, found that less than 50% of parents were able to identify appropriate portion sizes for their children (Welker, 2017). This makes sense in light of the growing portion sizes over a number of decades (Steenhuis & Poelman, 2017), a phenomenon that the National Institutes of Health has called “portion distortion” (National Heart, Lung, and Blood Institute, 2015).  Potter et al. (2017) recently found that when parents believed the ideal portion size for their child to be larger than recommended portion sizes there was a significant positive correlation with their child’s BMI.  Another possible explanation for this demand for larger, non-compliant portions is that it is an unintended consequence of BC’s Guidelines being nutrient-based criteria. As a nutrient-based criteria, the reformulation of foods and beverages is more accessible. Because vendors can reformulate to comply with the nutrient criteria for salt, sugar and fat rather than, for example, criteria based on whole foods, they can then simply produce smaller versions of the same kinds of items they were selling prior to the nutrient criteria and be in compliance. Because this study did not explore this specifically, it is impossible for me to say whether parent’s perceptions about portion size are misinformed or whether vendors’ reformulations are actually leading to smaller, calorically inadequate options, or both. Moreover, I am not able to say if those vendors who reported offering homemade options, made from whole foods that they believe students will like (e.g., homemade macaroni and cheese) are making the portions smaller so they comply. 127  Perhaps these vendors’ offerings, which are lauded as homemade and more nutritious, are actually calorically insufficient.  Lastly, some school-level stakeholders reported wanting to engage in implementation but, as a result of their context, believed they lacked the capacity to do so. This is also reported in other Canadian contexts (Government of Manitoba, 2015; MacLellan et al., 2010; Mcisaac et al., 2015; Quintanilha et al., 2013). Participants in my research discussed lack of human resource capacity to implement in terms of parents’ capacity, which relates to the dependence on volunteers for fundraising. Dependence on parent volunteers was also reported by Mclellan et al. (2010) as one reason why parents felt challenged to implement Prince Edward Island’s school nutrition standards. Lack of food-related infrastructure and financial resources to provide, not only a well-rounded curriculum, but also healthy food and beverage offerings is likely related to, what many believe to be, insufficient funding of public education (Downs et al., 2012). While a lack of available healthy options hindering capacity to implement has been reported elsewhere (Downs et al., 2012), my study also found that, in BC, this is likely particularly problematic in rural areas where there may be less access to innovative, health-conscious private vendors.    The frustration expressed by those who desire to implement the nutrition standards, but feel they cannot due to a lack of resources, may be contributing to an experience of moral distress, that occurs when a person knows the right thing to do but institutional constraints make it nearly impossible to act (Jameton, 1984). These motivated, but under-resourced, school-level stakeholders reported sometimes not making efforts to implement at all.  This is not surprising as, literature on moral distress, while specific to health care practitioners, has found that the experience of moral distress can lead to ‘burnout’ (Lins Fumis, Junqueira, Fatima Nascimento, & Vieira Jr., 2017), cynicism and low efficacy (Leiter, Frank, & Matheson, 2009). Findings of my research suggest that moral distress may be a useful conceptual tool for exploring success or failures of public health interventions, not just as a phenomenon among primary health care practitioners.    Implications for a refined program theory The findings suggest that providing information that justifies why school food and beverage sales standards are needed is not necessarily a driving force to motivate adults to change the way they create school food and beverage environments. It is possible that this intervention activity motivated 128  expected-implementers in the past, and perhaps it still is an important activity for adults who are new to creating school food and beverage environments. This requires further investigation. The initial program theory in Chapter 4 also stated that the mandatory nature of a policy would contribute to motivating implementers to implement. While this study did find this to be the case at the school level among many stakeholders, there were some stakeholders who responded to the mandate in a negative way which resulted in ignoring the policy or finding ways to ‘skirt’ around them. Individual beliefs about school food, food in general, and the role of government appear to be pre-existing contexts influencing the manner in which stakeholders respond to the mandate. Regarding vendors, the program theory in Chapter 4 did not specifically distinguish vendors from other stakeholders. However, the findings show that there is a distinct mechanism at work when it comes to vendors. All vendors have a pre-existing context of wanting to sustain their business in a market-based economy and therefore the demand created by the mandating of the policy appeared to be sufficient to motivate them to comply with the standards. However, pre-existing beliefs of the vendors also seemed to influence how they decided to comply: reformulating conventional school food or offering what was presented by vendors to be homemade, diverse options. The program theory in Chapter 4 did not include negative outcomes. My findings suggest a new dimension can be incorporated in its refinement. Specifically, even when stakeholders are motivated to ensure a healthy food and beverage environment for their school communities, the lack of perceived human resource, infrastructural, and financial capacity, can result in stakeholders feeling frustrated and little to no efforts to implement are made.  Limitations This study has several potential limitations. I was not able to consistently collect data across the five district cases due to variations in the manner in which school districts allowed me to access education stakeholders. As well, because I used a semi-convenience approach to garner participation and questionnaire participation was open to any type of relevant stakeholder who was willing to participate,  the majority of questionnaire participants were school principals, with few parents and other types of school-level stakeholders being included. As a consequence, comparing data from each type of stakeholder across cases was not possible in the conventional sense. From the realist perspective, however, this is not considered a serious methodological flaw.  The initial analysis of each participants’ responses (whether in interviews or questionnaires) involved identifying the specific contexts and mechanisms they believe to be interacting to influence specific implementation 129  activities. I then looked across the data set to examine which mechanisms were apparent across individuals’ narratives within one district as well as across the five districts. Regardless of whether it was a principal in one district discussing a particular mechanism and context interaction by sharing one story, and a district office staff in a different district discussing the same interaction through a different story, the “context + mechanism” interaction is the same, broadly speaking. An additional limitation is that I did not obtain participation from any rural vendors. This means the inferences made about private vendors and food and beverage sales in rural schools were derived from the questionnaires and the narratives of dietitians working in rural areas. The realist approach also deems this type of data as a suitable place to begin to make inferences about a phenomenon and to consider where more research is needed (Maxwell, 2012).   Finally, claims regarding the rigour of qualitative research often include reference to the concept of data saturation. This has been described as collecting data until no new insights arise (Bowen, 2008) and is a contested concept. O’Reilly and Parker (2013) identify practical weaknesses of this concept, including the reality that the potential for emerging themes is perhaps limitless in socially complex contexts (Green & Thorogood, 2004) due to the unique experience of every individual (Wray, Markovic, & Manderson, 2007). It is difficult, then, for me to make any claims about data saturation in my research because there are 60 districts in BC, with over 1550 public schools, each with numerous  individuals involved in school food and beverage sales which equates to an almost uncountable number of unique contexts. This complexity is further enhanced by there being no standardised way to implement the Guidelines. As such, ‘true’ saturation would not have been attainable within the resource constraints of this research. However, from a realist perspective, this inability to achieve saturation does not necessarily mean that the findings are invalid, but rather that they require further exploration (Maxwell, 2012; Morse, 1995).  Conclusions  This study aimed to contribute to a more nuanced understanding of complex school food environment interventions. This is important for refining program theories and advancing theories of behaviour change leading to improved food environments and public health. Using the implementation experience of school-level stakeholders in British Columbia, I identified key contexts and mechanisms that appear to be driving efforts to implement and, in contrast, lack of efforts to implement. While a number of the barriers and facilitators related to implementing this 130  type of intervention in jurisdictions around the world have been reported, this study used a realist approach to begin to provide a more nuanced explanation of why these barriers and facilitators lead to challenges and successes with implementation. Given the complexity of the public health challenge this intervention attempts to address, and the complexity of the intervention itself, it is important to arm policymakers and intervention creators with explanatory insight for improvement of implementation and effectiveness. My findings suggest it may be important to continue to advocate for, or begin advocating for, nutrition standards for school food and beverage sales being mandatory rather than weakening them out of concern for some public aversion to paternalism. Also, there may be a need for more research on the impacts of the Guidelines being nutrient-based criteria that allows for reformulation. For example, what messages are students receiving from reformulated foods and are there any long-term impacts on food preferences and decision-making around food consumption? Lastly, strategies are required for ensuring the needs of those who require additional and practical support to comply with school food and beverage standards are met, especially in under-resourced areas, like rural regions.131  Chapter 7 ~ Discussion and conclusions  This study used a realist approach to evaluate the implementation of a school food and beverage sales policy intervention.  Specifically, I explored key mechanisms and contexts driving efforts to implement a mandatory school food and beverage sales policy in British Columbia, Canada. The specific research objectives were to: • Develop a retrospective logic model and program theory to begin to articulate components of the intervention, the expected implementation processes, and how they are expected to lead to the desired outcomes; • Explore whether this program theory reflects the reality of implementation processes  in multiple school districts in BC, which each represent different real-world contexts; • Suggest how the program theory for this popular public health intervention might be refined. ‘Zooming’ in on the case of BC helps provide insight into the inner workings of not only this specific food and beverage sales policy but also of other types of complex public health interventions being adopted to address food-systems related challenges.   The first step of the realist evaluation research cycle is to hypothesise a program theory, which is typically derived from the empirical experiences of real-world interventions. Therefore, the first specific objective of this research was to develop a retrospective logic model and program theory specific to BC’s Guidelines (Chapter 4). This painted a detailed picture of what this intervention is as conceived by the policy developers and high-level implementers, providing a starting point from which to explicate links between inputs, outputs, and outcomes and, generally, how this intervention was expected to work. This was an important starting point because BC’s school food and beverage standards have essentially been, from their creation in 2005, just a dense document outlining the nutrient criteria and that also included, in the text, information ‘on’ the intervention (i.e., information justifying why nutrition standards are needed and how implementation can provide benefits for the short and long-term health outcomes of children) (Bemelmans-Videc et al.,1997). While the provincial government has, over the years, created implementation tools that can be used to assess food and beverage offerings, and determine whether they comply with the Guidelines (i.e., information ‘as’ the intervention) (Bemelmans-Videc et al., 1997), at no point were specific 132  implementation activities strategised by top-level policymakers; these decisions were left to individual Regional Health Authorities, school districts, and schools. However, policy needs to be transformed into action, somehow, to achieve objectives. Clarification of the activities undertaken, in-practice, is necessary for devising lines of inquiry for implementation evaluation. The hypothesised transformation of policy into action was what I attempted to articulate by conducting this first phase of the research.  From this retrospective broad program theory (Chapter 4), I devised a set of hypothesised context + mechanism outcome configurations (Appendix D). These hypothesised CMOcs were used as a heuristic to guide the development of the questionnaires (Appendix F) and the analysis process for phase 2 of the research (Blamey & Mackenzie, 2007). Phase 2 involved gathering data from five school districts, located in four different RHAs, to explore if the hypothesised CMOcs were relevant in practice, in different contexts. Specifically, I evaluated what intervention resources being offered trigger expected implementers at the district and school levels (Chapters 5 and 6, respectively) to: (1) make efforts to implement or (2) not make efforts to implement. Additionally, I explored which dimensions of context interact with mechanisms that may lead to these two tendencies. This approach, whereby contexts, mechanisms and outcomes are evaluated in configuration (as opposed to catalogues), helps provide some explanation of why differential implementation of this type of intervention occurs. This is important to understand because of the potential for differential implementation of public health interventions to exacerbate health inequities (MacDonald et al., 2016).   Finally, to close the loop of the realist evaluation cycle, in the following section, I provide suggestions refining the broad program theory presented in Chapter 4 (Figure 4.3). To do this, I discuss some notable differences between the original hypothesised program theory, outlined in Chapter 4, from phase 1 and the findings from phase 2.  Then I present suggestions for future research. Because the use of a realist approach tends to open the “black box” of implementation, revealing complex implementation chains and processes that are typically neglected in input-output evaluations (Astbury & Leeuw, 2010; Chen, 1990), findings from these endeavours confirm the adage: the more you know, the more you know that you do not know. Therefore, I offer examples of ‘known-unknowns’ that require more research. The chapter will conclude with a detailed discussion 133  of research limitations and implications and how the use of a realist approach is a radical one that challenges ways of thinking about normatively presented limitations of qualitative research. 7.1 Considerations for program theory refinement This next section discusses differences observed between the hypothesised program theory in phase 1 and the data collected for phase 2, what the implications are for refining the program theory (Figure 7.1), and a reflection on how these types of complex public health interventions, school food and beverage sales environment interventions, may be inequitably implemented  7.1.1 Mandatory mechanism versus hypothesised responsibilisation efforts One of the most noticeable differences between the hypothesised program theory in Chapter 4, and what is presented in the findings contained in Chapters 5 and 6, is the focus of the logic model and program theory on creation and provision of implementation tools and knowledge dissemination. This was hypothesised to be a process aimed at responsibilising adults to change the way they create food and beverage sales environments by providing information around the severity and susceptibility of the problem and why implementing the Guidelines would be beneficial for their children. The intervention also provided implementation tools to help assess which foods and beverages meet the criteria, helping to remove any perceived barriers to taking action once adults are responsibilised. This is reflected in the initial hypothesised CMOcs (Appendix D) and was subsequently reflected in the development of the phase 2 questionnaire (Appendix F), with a large proportion of questions focusing on dissemination activities and use of tools. However, once phase 2 data collection began, it became clear that the provision of information and tools were not considered important drivers of change (i.e., mechanisms) by most of the education sector stakeholders. This difference may reflect the different positionality and understanding of the role of information and tools between different stakeholders in the food and environment sector (e.g., health stakeholders who developed and refined the intervention versus education stakeholders primarily responsible for implementation).    Throughout the life of BC’s school food and beverage sales standards, one of the public policy instruments that has been used is the provision of information: information ‘on’ the intervention and information ‘as’ the intervention (Vedung, 1998). By information ‘on’ the intervention, I mean the provision of information related to operationalising the intervention. In this case, it includes providing information on the scope of the policy as well as providing the nutrient criteria and tools 134  that can be used to assess whether items school stakeholders wish to sell are compliant with the nutrient criteria. By information ‘as’ the intervention I mean the dissemination of information that aims to responsibilise adults, like information justifying why the intervention exists and why it will be beneficial for children. However, information dissemination activities, generally, have mostly been undertaken by public health sector staff rather than education stakeholders. The hypothesised program theory focused mostly on the implementation activities undertaken over the years by health stakeholders (i.e., information dissemination activities). That education stakeholders in my study did not identify these informational tools and resources as important for driving them to take action to implement differs from Masse et al.’s (2013) earlier BC-based study where education stakeholders reported this to be an important implementation facilitator. It is possible that these intervention activities are not as important as they once were for driving efforts to implement among education stakeholders. This reduced dependence on information dissemination may be due to this intervention being in place for over a decade and, therefore, the existence of the intervention has become normative. In addition, secular trends and popular media around school food and food more generally (e.g., Jamie Oliver Food Foundation, 2014; Pollan, 2007; Soechtig, 2014), may have led to reducing the necessity for responsibilising information.  In contrast to the emphasis on the provision of information to responsibilise adults and drive them to action, the phase 2 research findings point an emphasis on how the top-down pressure exerted by the BCMoED triggers expected-implementers to engage in implementation so as to fill their job role. This was mostly absent from the hypothesised logic model and program theory presented in Chapter 4. While the mandatory mechanism appeared to ‘work’ among all the district-level expected-implementers in this study, this was not the case at the school level. Instead, the mandatory mechanism fired sometimes among some school-level expected implementers, while for some stakeholders, the top-down pressure leads to the ‘scofflaw mechanism’ firing. The mandatory nature of the Guidelines as an implementation driver was originally hypothesised only as a) a cue to action motivating implementers to engage in knowledge dissemination activities and b) enhancing motivation and power of district and school-level implementers to use existing resources and tools (Appendix D). The nuances identified in the different ways that expected-implementers respond to this top-down pressure, and the way context influences their responses, was absent from the hypothesised program theory.   135  Mandates are one of the strongest types of policy instruments that a governing body may choose to utilize (Bemelmans-Videc et al., 1997), falling within the spectrum of Vedung’s (1998) typology as a ‘stick’, defined as:   …measures undertaken by governmental units to influence people by means of formulated rules and directives which mandate receivers to act in accordance with what is ordered in these rules and directives. (p. 31)  In this study, the importance of the mandatory language underpinning the Guidelines was reported despite the absence of official enforcement mechanisms or sanctions for non-compliance. While it is not a regulation coupled with penalties for non-compliance, it can be considered a set of non-sanctioned rules or, in jurisprudence terms, lex imperfecta (Bemelmans-Videc et al., 1997). This means it is a regulation that has intentionally not been coupled with threats of punishment. It has been proposed that the most important characteristic of a regulation is not necessarily its association with a negative sanction but that there is an authoritative relationship between the rule-maker and the rule-taker (Bemelmans-Videc et al., 1997; Lang et al., 2012).  While the use of strong interventions (as defined above) to achieve social goals is often criticised (Friedman, 2014; Trout, 2005; M. van den Berg, 2016), Friedman (2014) postulates they can still be launched relatively effectively in the right circumstances. In the case of school nutrition interventions, the use of strong interventions has not only been argued to be ethically justified (Asada & Garza, 2014; Crawford, Gosliner, & Kayman, 2011; Kass, Hecht, Paul, & Birnbach, 2014) but public education systems are often considered appropriate “zones of control” (Friedman, 2014, p. 1744), where a high degree of governmental control already normatively exists. Therefore, launching strong interventions to address school nutrition is possibly more palatable to stakeholders compared with, for example, interfering with private business and banning the sale of large soft drink containers by a city council (e.g., New York City’s Big Gulp ban controversy of 2013 (see Grynbaum, 2012, 2014)). These important, acceptable power dynamics that exist within education systems, and may help explain the phenomenon whereby many education stakeholders, in this study, felt compelled to comply with the mandate to improve school food and beverage sales environments even in the absence of official enforcement. They also reported going a step further to develop informal systems of enforcement and regulation. In addition, efforts to implement were reported even by more ambivalent education stakeholders as a result of top-down pressure. This suggests that lex imperfecta may be at least 136  somewhat effective in accepted hierarchical contexts for driving implementation regardless of there being no sanctions.   On the other hand, I found that the same top-down pressure that motivates implementation efforts among some people can trigger an aversive response in others particularly at the school level, and, subsequently, less effort to implement. This aversion appears to be related to pre-existing beliefs about the role of government, school food, and food in general. Moreover, where this aversion is present, the absence of formal enforcement measures enables expected implementers to either ignore the nutrition standards altogether or find loopholes such as selling foods or beverages that technically comply but that might be counter to the intent of the policy. In this study, this was mostly reported around parent and teacher fundraising efforts. This is counter to Frideman’s (2014) suggestion discussed above that strong policies launched in typically acceptable hierarchical contexts may be more acceptable and therefore more effective. This suggests that something is happening that renders the normative hierarchical experience of education sector employees a non-important context among some parents and teachers. I did not conduct any data collection with teachers and so I cannot make well-founded inferences about their decisions to skirt the standards as reported by some administrators, district staff, dietitians, and private vendors. However, there has been a long-standing power-struggle in the BC context between teachers and other education stakeholders due to labour union tensions which may be relevant here (Shaw, 2016). Some data was collected, however, with parents and it may be that the volunteer nature of their role is influential here; they are unpaid workers (therefore less accountable) aiming to fundraise to improve the curricular experience of their children and are frustrated by the government telling them they can no longer fundraise with foods and beverages that have always sold well and produced high profits. The above two possible reasons why some school-level stakeholders are triggered by this scofflaw mechanism are nascent inferences requiring further investigation. They do suggest, though, that exploring power relationships between expected implementers may be a useful framing for further exploration of implementation of complex public health interventions aiming to address food system challenges.  Ultimately, the above discussion about the mandatory mechanism and scofflaw mechanism can provide insight into what dimensions of contexts may or may not be fertile ground for a mandatory intervention to land. In decisions around what type of policy instruments may be used (i.e., carrots, 137  sticks, sermons) questions can be asked about specific dimensions of context that can guide these decisions. For example, is there an acceptance of following orders from the top-down among whoever is responsible for implementation? If so, is it really necessary to incorporate compliance provisions or can the language of “mandatory”, alone, be sufficient? If not, what might the best type of compliance provision be? Rewards and recognitions? Financial incentives or disincentives? (Gourdet, Chriqui, Piekarz, Dang, & Chaloupka, 2014). 7.1.2 Motivated by money and potential for public procurement As originally hypothesised, I found that the education sector demand for products from the private sector was promoting creation of a compliant food supply, or what I call the ‘money mechanism’. This is Vedung’s (1998) financial policy instrument, or ‘carrot’. This works by either formal incentives through request for proposal processes in districts that centrally procure cafeteria and vending machining companies, or by more casual incentives through school-level stakeholders’ explicit demand. Phase 2 data collection helped to refine the original hypothesis and identify two variations on how this outcome manifests: reformulation or innovation of the items they had on offer for sale. I also identified important specific dimensions of context that may be facilitating whether a vendor decides to comply with the nutrient criteria through reformulation or whether they decide to comply via the provision of what they perceive to be higher quality items made from whole foods. Important dimensions of context appeared to include the existing beliefs of vendors about school food, food in general, and if they feel they are in competition with other external or internal food and beverage sales venues. In fact, emergence of this important implementation mechanism, that the large portion of responsibility to implement is relegated to the private sector through official or non-official financial incentives, may help account for lack of emphasis found in the data from education stakeholders about the importance of the knowledge dissemination activities and use of implementation tools. If education stakeholders have less responsibility to actually ensure that what they are selling is compliant, then there may be less motivation to seek out information and tools to help with implementation.   In this study, while I am unable to estimate the percentage of foods and beverages available during any given school day that are within the scope of the Guidelines, it is clear that food and beverage sales in schools are common. When these sales do happen, it is often through a private vendor, whether a private cafeteria catering company, vending machine company, small hot lunch vendors, a local restaurant, or other fast food franchises. This points to the private sector having an important 138  role in the actual implementation BC’s policy. The hypothesised program theory proposed a ‘mandate’ as a cue to action to trigger adults to implement the policy, including adults who are private food vendors. It neglected to articulate the pathway for the demand creation mechanism identified at both the district and school levels in Chapters 5 and 6. The importance of the creation of institutional demand for addressing nutrition-related challenges associated with the modern food system , as governments increasingly become participants in the market as both consumer and regulator (McCRudden, 2004; Lang et al., 2012), is widely acknowledged and discussed among school food ‘revolutionaries’ (Morgan & Sonnino, 2010).  In a refined program theory of this intervention, the importance of the private sector for implementation needs to be more explicitly included.   With this trend of increasing importance of the market economy in implementation of food policy (Lang et al., 2012), there is potential to harness the regulatory power (some) governments still hold and conduct strategic procurement that demands not only compliance but also innovation (Edler & Georghiou, 2007). In this study, there was a context of a ‘school food service industry’ that consisted of vendors who were running small- and medium-sized, often family-run businesses; they are actual ‘faces’ in a community with perhaps children in schools and other people’s children as neighbours. While often reliant on large corporate national or international food suppliers in many cases, these vendors are themselves still far from the large, “big food” corporations often villainised by school food reformers (Stuckler & Nestle, 2012). Public procurement within this type of vendor-context can possibly promote local and regional economies (Morgan & Sonnino, 2010; Otsuki & Arce, 2007; Sumberg & Sabates-Wheeler, 2011).   I argued in Chapter 5 that reformulated food, while compliant, may not align with current student preferences. Furthermore, because the school food environment is part of the hidden curriculum, sending mixed messages to students may influence their food choices in the future. However, in this small-medium vendor-context, potential exists for engaging with vendors who reformulate to begin to view themselves not just as business people but as citizen-producers (to play on McRae et al.'s (2012) conception of “citizen-consumer”), contributing to short- and long-term positive social goals through innovative practices. One interesting question that emerges from this line of thinking, then, is whose job is it to strategise ways to engage, and possibly support, these smaller vendors and other food sector players in implementation? Or is it no one’s job at all? Should the free-market be left to 139  determine which businesses sink or swim, possibly leaving a vacuum in the market for “big food” to fill and with whom the possibility of citizen-focused engagement activities may be more limited?  Additionally, I found that education stakeholders are making some attempts to support the private sector through engaging in back-and-forth informal enforcement practices where vendors can report non-compliant food and beverage sales occurring in schools, with which they are in competition, to the district. This helps maintain a level playing field among the food and beverage sales venues within the schools. However, this may not always be sufficient for mitigating the presence of external vendors near schools (e.g., convenience stores, grocery stores, restaurants, etc.) who are under no obligation to comply with any nutrient criteria, particularly in more densely populated areas. Holmes (2016) discusses the possibility of redefining ‘school food environment’ to include external vendors so as to be able to justifiably impose some regulations. However, it is not likely that this would be acceptable to vendors in the vicinity of a school and perhaps the only alternative approach, in a neo-liberal system that values freedom of choice and autonomy, is voluntary change on the part of external vendors. There are examples of pilot interventions aiming to address availability of foods in small community businesses (Jaskiewicz, Dombrowski, Barnett, Mason, & Welter, 2016; Paek et al., 2014). Little exists regarding effectiveness of these types of healthy corner store programs (Jaskiewicz et al., 2016) but it may be useful to conduct a realist synthesis to explore where this type of intervention has or has not worked in various western contexts and if this is a feasible leverage point for a community-based behaviour change intervention (i.e., changing the behaviour of store-owners). 7.1.3 ‘Moral distress’ among responsibilised implementers The last area for refinement of the program theory is the incorporation of the resource constraint mechanism, which may be leading to a type of ‘moral distress’, to borrow terminology from primary healthcare. Defined in Chapter 6 as when a person knows the right thing to do but institutional constraints make it nearly impossible to act (Jameton, 1984), it appears there are school-level stakeholders, in particular parents involved in fundraising, who may be experiencing this, thus de-motivating them to implement. This means that these particular stakeholders do have an idea about what is healthy and what is not and are motivated to take action to implement; in other words, they have been responsibilised to some extent. It appears, though, that there are pre-existing contexts related to capacity and resource issues as well as competing priorities, like the need to fundraise, that have led to these responsibilised stakeholders making little or no efforts to implement. These pre-140  existing contexts were missed in the development of the hypothesised program theory and need to be explicitly incorporated into a refined program theory.    Identifying these types of pre-existing contexts provides insight into what is not working and why. These may be pre-existing contexts that could be points of leverage at which to intervene to support implementation efforts. However, this may prove challenging. Importantly, the continuing reduction of education resources and infrastructure related to food may be detrimental. Education systems require more resources to ensure a holistic education experience for children, including increased capacity to provide healthy messaging about food in both the explicit and hidden curriculum (Titman, 1994) and to alleviate fundraising pressure. However, I maintain skepticism around this recommendation coming to fruition, particularly in the Canadian context where percent of Gross Domestic Product funding for education has steadily declined for the last four decades (The World Bank, 2017b). In light of this reality, then, attention should be given to devising ways to capitalise on the moral motivation that accompanies the responsibilisation which some stakeholders experience but are in contexts with limited capacity. This refers to, for example, those parents involved in fundraising in rural schools that have neither vendors nearby that comply nor food preparation infrastructure. It is possible that in attending to these stakeholders who may be experiencing some moral distress around implementation of this intervention, providing them extra support to act on what they believe to be the ‘right thing to do’, the aversion to the use of ‘sticks’ to address food system challenges may also begin to subside (Levay et al., Forthcoming). 7.1.4 Refined program theory for school food and beverage environment interventions Figure 7.1 brings together the mechanisms that emerged from my research and incorporates them into a refined version of the hypothesised broad program theory presented in Chapter 4 (Figure 4.3). Few of the items originally hypothesised remain unchanged in Figure 7.1 but, as hypothesised, the importance of having a top-down mandate to prompt implementation efforts clearly emerged. However, what was not predicted was the opposite response to the mandate, resistance to implementation. This is now clearly articulated a mechanism that requires consideration when launching school food and beverage environment interventions. In addition, responsibilisation did not emerge as a trigger for implementation. I did not discard it entirely, though, as it may be that the process of responsibilisation is necessary and effective for raising awareness in the early years of the life of a school food and beverage policy. Therefore, I have indicated the need for more exploration around the responsibilisation component of the intervention (grey text).  Figure 7.1, unlike Figure 141  4.3, highlights the important role the private sector appears to play in implementation of nutrition standards in schools. The capacity component of the intervention presented in the original program theory was also refined. The findings of my research highlighted additional resources stakeholders may require to engage in implementation efforts, including infrastructure to support access to compliant foods and funding to support the provision of healthier options which are perceived to be (and perhaps are) more expensive (in blue text). Human resources, as a perceived needed resource, emerged as important in terms of parent engagement, engaged administrators, school staff and faculty. More exploration into the role of implementation tools to assess food and beverage items in supporting efforts to implement is still warranted. Lastly, the original program theory in Figure 4.3 only listed broad contextual dimensions that may need to be considered when launching school food and beverage environment interventions. My findings provide empirical support for what specific dimensions of these broader contexts may be the most relevant to consider when launching school food and beverage environment interventions.  The refined program theory below emphasises the importance of regulation for motivating expected implementers to take action. This challenges the health belief model underpinning the hypothesised program theory in Chapter 4 because, rather than fostering responsibilised expected implementers through provision of information on the severity and susceptibility of the health issue, people appear to be less moved by a moral duty to act and more moved by an authority mandating action. As stated above, however, it may be that in the early years of this public health intervention, or in any other type of public health intervention, providing information based on the HBM may be useful for increasing acceptability of the intervention. Another way the refined program theory challenges the HBM is that providing cues to action in the form of implementation tools may not be sufficient for triggering expected implementers to take action, even in combination with providing them information on the severity and susceptibility of the health issue. Again, the cue to action that emerged in this research as the most effective to prompt implementation was the mandate which, at the same time, can prompt resistance to the intervention among some implementers. This resistance was found to be influenced by pre-existing beliefs unrelated to the constructs of the HBM, such as the individual beliefs about the role of government in addressing this health issue. This resistance to a mandate (as a cue to action) may be particular to public health issues attributed to individual choice and behaviour that may only harm the person engaging in the harmful health behaviour (e.g., eating unhealthy foods) as opposed to harmful health behaviours that affect others (e.g., smoking). 142  Additionally, the HBM construct, ‘confidence in one’s ability to take action’, refers to an individual belief or perception and seems to neglect the real physical infrastructure and material resources that may be needed to change health behaviour. An individual may have infinite confidence in their ability to take action, but without the needed physical resources they cannot. The refined program theory challenges this construct by including the material resources and physical infrastructure that appear to be necessary for implementing school food and beverage sales environments, and likely other institutional food environments.   This refined program theory aligns with Cohen et al.’s (2000) structural model of health behaviours, underpinned by an ecological theory of health behaviour. For my research, the health behaviour under exploration was the way adults create school food environments. The refined program theory below incorporates all four of Cohen et al.’s (2000) factors that determine health behaviours. Increasing availability and accessibility of consumer products (the first of Cohen et al.’s (2000) factors) is now clearly highlighted as an important feature of school food and beverage sales environment interventions. Physical structures (the second factor), like food preparation and storage infrastructure, are now included as part of the capacity component of school food and beverage sales interventions. “Social structures and policies” and “media and cultural messages” (third and fourth factors, respectively) are now reflected in the refined program theory as dimensions of context, like the individual beliefs that appear to be most relevant for influencing whether or not adults decide to change their behaviour as a result of school food and beverage sales interventions.  Ultimately, the refined program theory below presents school food and beverage sales interventions as a complex of carrots, sticks, and sermons with elements underpinned by the HBM, an individual health behaviour theory, as well as the more structural approach to changing health behaviours.    143   Figure 7.1 Refined program theory for school food and beverage sales environment interventions  144            7.1.4 The potential for exacerbating health inequities Highlighting why and where expected implementers may or may not be making efforts to implement, inevitably resulting in different levels of implementation success in different contexts, points to potential for this type of intervention to exacerbate health inequities. This is particularly salient if implementation is less successful in contexts that have already been recognised as related to important social determinants of health. This study appears to be the first to consider health inequities related to this type of intervention. Even studies that disaggregate implementation success by variables like geography or socioeconomic status of a school (e.g., Caspi et al., 2015; Hills et al., 2015) have not addressed the possibility for differential implementation to exacerbate long-term health inequities.   My analysis revealed that district-level organised food sales was perceived to ‘almost always’ or ‘always’ be compliant. This is in contrast to school-level organised food sales which were perceived to be compliant much less often. The rural districts did not organise any food sales at the district level, therefore leaving schools to organise all and any food sales, while urban districts organised cafeteria services and vending machines services through centralised procurement. This difference means rural districts have less control over the quality and consistency of implementation in individual schools, as opposed to the urban districts, which could result in more health optimising opportunities for urban students. The importance of district control over food availability has been reported elsewhere (Larson et al., 2016; Quintanilha et al., 2013). Moreover, the implementation differentials between rural and urban schools and districts have been found in other contexts where rural schools often show less compliance (Caspi et al., 2012; Dick et al., 2012; Hills et al., 2015; Nanney et al., 2013; Pettigrew et al., 2012a; Turner et al., 2016). Globally, rural-urban health inequities exist where rural populations more often have poorer health outcomes (CSDH, 2008), including in BC (Chasey et al., 2009) and differential implementation of a public health intervention may exacerbate this inequity.   The difference between rural and urban districts in terms of availability and access to compliant options that can be sold in schools, also identified in an Alberta study (Downs et al., 2012), is another issue that could potentially exacerbate health inequities. In the urban areas in which the districts who participated in this study were located, there are a number of local and regional private 145  sector vendors in competition who can offer, and may be willing to offer, compliant options. The financial incentives for private vendors, created by either formal (RFP processes) or informal demand (parents advisory committees demand), may not work in rural areas where there is little-to- no competition or where the cost to provide district contracted services are too great for a smaller rural vendor. While data were not collected directly from small rural franchises and restaurants, there was no indication in data collected from rural school stakeholders that they were applying pressure to demand compliance. This may be due to the lack of any other options in the area around some of their schools. Therefore, there is the potential that rural students are not receiving the same health maximising opportunities as urban students. As urban students gain more health maximising opportunities as a result of more capacity to successfully implement in urban areas and rural schools continue on with business-as-usual, there may be risk of increasing health gaps.  Another area of potential concern is whether or not a school or district has an individual, or individuals, who are food and health conscious, a pre-existing context that can help bolster implementation efforts. The importance of a ‘champion’ for driving the implementation of social interventions, including implementing policies related to improving school food and beverage sales, has been highlighted in existing studies (Downs et al., 2012; MacLellan et al., 2010; McDonnell et al., 2006; Vine & Elliott, 2014). The existence of this type of stakeholder in a school community may be somewhat reliant on chance; schools without this may be at a disadvantage with regard to implementation. At the school level, an implementation study from the province of Ontario reported a quote from a participant conveying this:   …it seems to be a tiny bit hit or miss. School nutrition seems to be two kinds of important things…you need buy-in from the principal because nothing goes on in a school without the buy-in from the principal. The second big key to success is having a strong person who kind of leads it and organises it… (Vine & Elliott, 2014, p. 1296)  However, studies have also found that food and health conscious people are often associated with a well-educated, upper-middle class background (Johnston & Bauman, 2010). My findings support this existing work around the relationship between salutogenic food beliefs and practices and SES. In addition, it also begins to explain how the success of implementation of an intervention like this may be influenced by these pre-existing contexts. Administrators reported they do not put pressure 146  on parents to make efforts to implement if they perceive the parents have a low capacity, or low engagement, which is essentially the opposite of a context in which there is a “strong person” to lead and guide efforts. Furthermore, administrators also reported that lower parent engagement was often due to the school community having a lower socioeconomic status, also found in other studies (de Carvalho, 2001 cited in Smith, 2006). To my knowledge, this relationship between SES and level of parent engagement with school committees has not been highlighted in other implementation evaluations of school food and beverage sales policies. However, the specific importance of parent engagement, generally, was noted in another Canadian study (MacLellan et al., 2010). Therefore, even within a district that centrally organises cafeterias and vending machines, there is the potential for differential implementation between schools that do not have a cafeteria or vending machine (i.e., elementary schools, in the case of BC). Here, it is not unreasonable to consider two elementary schools, with no vending machine or cafeteria, in the same large urban district, one in a wealthy neighbourhood with a high-level of parent engagement and the other in a lower-income neighbourhood with lower parent engagement, and how the different capacities and levels of engagement may lead to different levels of implementation and compliance. 7.2 Known-unknowns and future research There is an oft-cited parable that illustrates the relationship between the complexity of a phenomenon, ontology and epistemology: the blind men and the elephant. This parable describes a number of blind men encountering an elephant. The elephant is, of course, large, and each blind man is asked to touch the elephant and describe what they believe an elephant to be. One blind man is touching the elephant’s trunk, another touching the tail, and yet another touching an ear, etc. Inevitably, each describe their perspective of this singular and complex creature, and come to different conclusions about what an elephant is. In phase 1, by creating the hypothesised logic model and program theory, a blurry picture of the elephant emerged. This is the first imagining of an articulation of what this school food and beverage environment intervention is and how it is actually transformed from paper-to-practice. Phase 2 began the process of clarifying and garnering a deep understanding of some dimensions of the elephant, and this is the humble goal of realist evaluation. However, as a number of implementation processes discussed in this study were removed from the “black box” (Astbury & Leeuw, 2010), much understanding was left inside of it, and a large proportion of the elephant remains undefined and blurry. In other words, there is more exploring to do. The ideas discussed above, related to suggestions for refining the program theory, are not comprehensive but can be used as stepping stones for identifying components of this intervention 147  that require further inquiry and further refinement. Table 7.1 provides examples of ideas for further exploration of some of these known unknowns:148  Table 7.1 Known-unknowns emerging from this research: suggestions for future research    Broadly     Conduct an equity analysis for this kind of intervention (could be a realist synthesis).     How is it that for some, the mandatory mechanisms work to promote implementation even at the expense of things like fundraising revenue whereas in other contexts, arguably where many schools even within a district experience similar capacity issues, the mandatory mechanism does not promote implementation efforts? What are some different contexts and mechanisms that lead to these different outcomes? Is this related to the presence of someone in the school whose beliefs align with the intentions of food systems interventions who can push implementation forward regardless of the potential for lost revenue? Or is it related to socioeconomic status of the school communities where there might be lower parent engagement?  How does the nutrient criteria itself influence the acceptability of school food and beverage sales policies? For example, if the nutrient criteria leads to reformulation and, as a result, smaller portions, are expected-implementers going to accept the intervention and make an effort to ensure compliance?     District level     Is the mandatory mechanism, that I have found to successfully promote efforts to implement by district-staff,  at work in other districts in BC who were either not approached to participate in this research or did not agree to participate? If not, then why not?      School level     Does the reported/perceived compliance of district-organised food and beverage services among school-level stakeholders align with actual/observed compliance?      How different are the implementation levels in schools that differ in size, capacity and with different access to healthy food options (e.g., rural areas)? If there are differences, what is the potential for increasing health inequities of students between, for example, rural and urban districts? This could be part of an equity analysis.     How do teachers engage with implementation outside of what the administrators reported in this study? What are their perceptions of this type of intervention and impacts on fundraising efforts?     How do administrators and/or teachers engage with older students who might use food fundraising?  How much do students know about this intervention? What are their perceptions about them?     How pervasive are the pre-existing beliefs among some stakeholders, like parents, that appear to be impeding implementation?      Private sector     Do the vendor’s beliefs about children’s food preferences, align with actual food preferences of children? Are children’s preferences different in different contexts (e.g., rural versus urban children)? What is the interaction between children’s preferences, how vendors comply, and resultant profits? How might this differ between different kinds of businesses in different contexts?  Is there a significant difference in the nutritional values between reformulated food and beverage options and the sales of foods that are diverse, whole foods, homemade options? Is one way of complying with nutrient-based standards better than the other? Are there longer term impacts on children’s ideas about what constitutes healthy foods and beverages and their consumption choices?  What role do private vendors play, and what approaches are used, in implementation of school food and beverage sales policies in rural regions?   149  7.3 Study limitations  7.3.1 Influence of researcher positionality on data collection My position as a researcher from UBC’s Faculty of Land and Food Systems could have led to a bias in the information received from participants, particularly in the case of school-level and district-office participants. They might have automatically assumed I was a dietitian evaluating them and, as such, they may have felt compelled to present the situation in their district or school as better than it is (i.e., social desirability bias). However, I believe there is limited bias in the information collected from respondents because most respondents provided information, at some point, that was counter to the expectations of the provincially mandated food and beverage Guidelines, indicating some level of candidness. I made great attempts in conversations with district and school stakeholders to refrain from judgments through the careful phrasing of questions and through exchanges of humour about school food and food in general. Ultimately, to attempt to foster some trust between myself and respondents, I approached respondents with a natural, friendly demeanour and engaged in self-disclosure to help build rapport (Dickson-Swift, James, Kippen, & Liamputtong, 2007). 7.3.2 Inconsistency of data sources across cases In an ideal situation, when conducting multiple case studies, consistent data collection would occur across all cases so as to be able to conduct cross-case analyses (Baxter & Jack, 2008; Yin, 2009). As well, it was my goal to collect data from at least one school district in each of the five Regional Health Authorities. However, gathering data from within the education sector, a sector in the Canadian context that is often reported to be overburdened (Edwardson, 2017; McElroy, 2014; Riva, 2016) can be challenging. As a result, interventions not directly related to curriculum become less of a priority and fostering interest from education stakeholders about this food and beverage sales intervention research was challenging.   For this study, I obtained approval to conduct the research in six school districts, three urban districts situated in the same RHA, one urban district in a different RHA, and two relatively rural districts situated in two different RHAs. At first, garnering approval from these school districts appeared to be a success. However, the challenges did not end here as each district’s research ethics boards decided the manner in which I could engage with school-level stakeholders. In one urban district, I was not allowed to contact schools directly but rather had to rely on an email sent from the district office to administrators, to which no one responded. Because there was no other way to garner 150  participation, I decided not to pursue further data collection in this district. In four of the districts, I was allowed to contact schools directly. I began by sending emails directly to administrators about the research so as not to disturb them during their busy days; I assumed they would be more likely to participate if they could read about the research and do the online questionnaire on their own time. Again, no administrator responded. Determined to collect some data, I then began cold-calling administrators and conducted the online questionnaire with them over the phone if they agreed to participate. This was a more successful approach to obtaining participation. While some agreed to do the questionnaire online, most wanted the questionnaire conducted with them over the phone. They most often agreed to send the online questionnaire to their parent advisory groups and any other relevant school staff. This resulted in limited response from parents and other relevant school-level stakeholders. In another urban district, I was also not allowed to contact schools directly because the district believed administrators may feel coerced to participate. However, this district allowed me to attend an event about school food, organised by the Public Health Dietitian in that regional health authority, for the parent advisory council members and vendors. Here I conducted the online-intended questionnaire with consenting participants in-person. This left me with data collected from parents but almost no data collected from administrators or school staff in this district. In one of the rural districts, while I was able to collect data directly from principals, I was unable to set up an interview with the relevant employee at the district office, which I had been successful in doing in other districts. With the last district, also a rural district, I was not only unable to set up an interview with the district office stakeholder but had a challenging time even gathering data from the schools due to what appeared to be an overburdened staff in the schools. One indicator of this that stood out was when I telephoned some schools in this district and the administrator answered the telephone, informing me part of their workday involved covering for school secretaries while the secretaries were taking breaks. Another indication of this, which I observed while conducting school website scans, was that one objective of parent advisory committee fundraising was sometimes to provide support for students with special needs.  Using a realist perspective to analyse this inconsistent data across cases helps de-problematise this potential limitation because the purpose of realist evaluation is not to see whether a specific outcome happened here or happened there. Rather, the purpose is to obtain insight about broad processes of implementation to propose new or refine existing program theories. So, regardless of whether it was a principal in one district discussing a particular mechanism or a district office staff in a different 151  district discussing the same mechanism, the mechanisms are the same and conclusions can be made about what contextual factors might actually be influencing the underlying mechanisms across context and across individual stakeholders. Essentially, the broad program theory developed here, as opposed to program theories that are highly specific to a particular intervention context, are more readily comparable across-cases regardless of data source.  7.3.3 Data saturation  The concept of data saturation as a means to establish quality in qualitative research is contested (Francis et al., 2010; Manzano, 2016; O’Reilly & Parker, 2013; Smaling, 2003; Thorne, 2011). Reaching a point in data collection that can be deemed saturated, where it appears that no new information is emerging, is not only, arguably, unattainable (Green & Thorogood, 2004) but is also often constrained by forces out of the researcher’s control (O’Reilly & Parker, 2013). From the realist perspective, the purpose is not to reach saturation to support general claims, but rather to make inferences about the wide array of possible complex processes leading to a wide array of intervention outcomes, either process outcomes or overall outcomes, which can then be further explored. If saturation is not obtained, this does not mean that the findings are necessarily invalid, but rather that the phenomenon has yet to be fully explored (Maxwell, 2012; Morse, 1995). In this study, while similar information began to emerge from within each district, different information emerged across each district due to the different contexts. Considering there are 60 districts in BC, with 60 different contexts and numerous individuals with different contexts which could be contributing to different outcomes, true saturation would not be attainable within the resource constraints of this research. 7.3.4 The hearsay problem (?) and making inferences All the participants in this study (dietitians, district staff, administrators, parents, vendors, etc.) told the story of their experience thus far with implementation. There are many instances in the dissertation where I present findings about, for example, parents or students. Often, these were based on narratives from other respondents (e.g., an administrator talking about the district staff or a dietitian talking about an administrator) rather than narratives from these stakeholders themselves.  This could be seen as problematic because it may appear that rather than presenting findings of actual events as relayed by the actual stakeholder involved, I have presented perceptions of parents by, a dietitian or administrator, for example. A particular implication of realist methodology for qualitative data collection is that: 152   …data are usefully seen, not simply as ‘texts’ to be interpreted, or as the ‘constructions’ of participants (although they are this), but as evidence for real phenomena and processes (including mental phenomena and processes) that are not available for direct observation. (Maxwell, 2012, p.103)  This type of data, then, can be used to make inferences about phenomena, which can then be tested against additional data. In some cases in the presentation of findings for this study, it was possible to make confident inferences about different types of stakeholders even in situations where little data was collected directly from that particular stakeholder. I say ‘confident’ for two reasons: (1) because in many circumstances the same narrative about a particular stakeholder or stakeholder-type, from whom little data was directly collected, was reported by multiple other different types of participants; and (2) because in the development of discussion sections of Chapters 5 and 6, it became clear that many of the inferences made were also supported by evidence from a wide variety of other western contexts, thus increasing the strength (or plausibility) of the inferences made.  There were also times during analysis and presentation of findings where inferences were made that I was less ‘confident’ about in terms the magnitude of supporting empirical evidence collected. Realist methodology creates space for this in the presentation of findings as long as there is transparency regarding the magnitude of evidence supporting these weaker inferences (Jagosh, 2017c). In classical paradigms, while plausible postulations are typically acceptable in discussion sections of research papers, not strongly supported (although plausible) inferences are often frowned upon in the presentation of findings as these inferences may be considered less valid. The realist approach allows for the possibility that less strongly supported inferences could be true but that they simply need to be further tested (Jagosh, 2016, 2017a). In this study, I made attempts to include explicit qualifiers, such as “one private catering company reported XYZ,” to provide a sense of the magnitude of data supporting an inference. This manner of thinking about data, rather than casting off potentially important evidence just because only one person said it, is useful for identifying phenomena that could be further explored for continuing to expand the picture of the reality of implementation of an intervention. 153  7.4 Implications of this study 7.4.1 Implications for research This study provides insight for researchers. As indicated by the above section regarding ideas for further research, this study begins to provide a sense of a wide range of potential future research around not only this specific type of intervention but also for conceptualising food systems-related public health interventions more broadly. For example, although numerous studies have quantified differential levels of compliance in different contexts, this is the first study, to my knowledge, that has articulated the potential for this kind of school food and beverage intervention to exacerbate health inequities and the need for an equity-focused implementation evaluation. Moreover, it contributes to the rapidly growing field of realist research as this is the first realist evaluation of a public health intervention to address food systems challenges. While in recent years there have been some realist evidence syntheses published that explore food systems interventions related to public health (Ohly, Crossland, Dykes, Lowe, & Hall-Moran, 2017; Penney, Brown, Maguire, Kuhn, & Monsivais, 2015; Sisnowski et al., 2017), this is the first to take a realist approach to evaluate a specific food environment intervention, providing an example of how realist evaluation might be used to explore the implementation of a specific food environment intervention and the types of explanatory insight a realist evaluation may provide. 7.4. Implications for practitioners  For practitioners working on-the-ground, who are involved in supporting implementation efforts (e.g., health sector staff, CSOs, or other education staff), this study begins to identify ways in which pre-existing contexts are influencing the manner in which implementers are responding to the intervention. This can offer leverage points at which to potentially intervene to better support implementation efforts on-the-ground. For example, it may be that a monitoring system is required at the school level to prompt those who are resistant to comply and that these monitoring systems could be linked with some type of incentive scheme, such as a recognition system or bonus-funding for compliance (Gourdet et al., 2014). In additions, targeted resources could be provided to schools that report a lack of food preparation and storage infrastructure as a barrier to providing compliant options. To address lack of options of private vendors in smaller communities, supporting some members of smaller communities to start small part-time businesses to earn some income and provide healthy meals for school hot lunch fundraisers might be warranted. Lastly, there is a need to address the food environments around schools, starting with softer community-based interventions 154  (i.e., sermons or carrots) to try to motivate those businesses around schools to promote healthier options. If this proves ineffective, then it may be necessary to advocate for exploring the efficacy of stronger measures like zoning bylaws, which is the realm of policymakers as opposed to practitioners. 7.4.3 Implications for policymakers These findings may also provide guidance for when policymakers seek to revise existing interventions or create new ones by providing answers to questions of what type of public policy instruments may be the most widely accepted and, subsequently, effectively implemented in what types of contexts. Understanding implementation of food policy and what works (or not), for whom, and why, is a crucial part of public policy analysis that can help guide efforts to improve intervention effectiveness (Lang et al., 2012). In addition, guiding the process through which governments make the decision to select one type of policy instrument over another is important because “while broad policy objectives are at best something to be achieved far into the future, instrument choice is real and will have immediate consequences” (Doern & Phidd, 1983, p. 111). First, policymakers may want to consider how this type of school food environment public policy instrument may or may not work in different contexts. Will a mandatory school food and beverage sales policy effectively work to trigger expected implementers to take action to comply in rural versus urban areas, for example? If the demand for compliant products is effective for driving the private sector to comply with the standards, is the demand in a rural area sufficient to drive, say, the only local restaurant in the community to provide compliant options for school hot lunch fundraisers? In addition, it may be useful to assess acceptability of potential intervention options prior to selecting a policy instrument. This can help predict the level of resistance on the part of stakeholders that may occur. Moreover, policymakers may consider conducting an equity-focused analysis to assess the potential of school food and beverage sales policies, with no enforcement measures, to be inequitably implemented. Another consideration may be to reconsider the use of a set of nutrient-based standards that easily allows for reformulation which has been indicated to lead to the creation of items that are deemed as unappetising and therefore driving students to purchase foods off-campus. Nutrient-based criteria also allows space for vendors and other education stakeholders to more easily find loopholes and ‘skirt around’ them.  It could also be problematic that reformulation may be leading to portion sizes that are calorically insufficient and therefore parents or students are purchasing multiple portions rendering the items non-compliant. In both of these possible consequences of reformulation, the policy is having the opposite effect of what is intended. 155  7.4 Conclusion This study aimed to contribute to the larger picture of implementing public health interventions aiming to address food systems issues. A deeper understanding is gained through the exploration of how different public policy instruments work (or not) given the unique public health challenges to which the modern food system has contributed. To do this, I took the case of school food and beverage sales nutrition standards, a widely adopted public health and food systems intervention. Specifically, I explored the experiences of implementation in five different school districts in British Columbia, Canada. School food and beverage sales policies have been found to be helpful for increasing healthy options and decreasing unhealthy ones for children during the school day  (Chriqui et al., 2013; Dick et al., 2012; Kubik et al., 2013; Watts et al., 2014; Whatley et al., 2011). However, changing school food environments through policy is not an easy task as schools are embedded in broader food, education, economic and political systems (Barlow & Stone, 2011; Mansfield, 2016). Still, the power of school food systems to evolve and innovate around considerations of social and physical health, economics, and sustainability at community, regional, and national levels have been observed in contexts around the world (Morgan & Sonnino, 2010).  This study adds to existing implementation research literature by providing a deeper level of insight into the complexity of a number of implementation processes related to the use of public policy instruments. First, I found that, for this complex public health intervention, all three types of public policy instruments, carrots, sticks, and sermons are being drawn upon to attempt to coordinate implementation efforts across stakeholder types. Strong, more authoritative directives appear to prompt education sector employees to make efforts to implement yet also can trigger negative responses among some stakeholders like volunteer parents. Financial incentives prompt the private sector to engage in ensuring compliance of their products, particularly in urban areas where there is a high level of market-competition. However, this might not be the case in all contexts. Finally, the very minimal reporting of the importance of knowledge dissemination and use of the existing informational tools by district and school-level stakeholders led me to conclude that this type of public policy instrument may be less important for driving implementation than perhaps they once were.  The modern food system has contributed to nutrition-related wicked problems. Public health has an important role to play in addressing these challenges. While wicked problems, characteristically, can 156  never be fully ‘solved’ but only tempered, doing this requires complex interventions. Nutrition-related public health interventions that take an individual behaviour change approach aim to shoulder the responsibility to change health behaviour solely on individuals. This approach for addressing public health challenges, in general, has resulted in less than desired population-level outcomes (Garner & Wooley, 1991; Health Canada, 2013; Hill & Peters, 1998; Vargas-Garcia et al., 2017). On the other hand, ecological approaches, like altering the food environment in which consumers engage with the food system, have also shown little impact on indicators of health, like BMIs (Sisnowski et al., 2017).  Implementation research tends to be input-output evaluations or identification of discrete barriers and facilitators to implementation, cataloguing rather than configuring explanations for why things are occurring the way they are. This can provide important foundational insights from which to build more nuanced explanations for the varying levels of implementation success and failure observed with all types of social interventions. However, these types of implementation research tend to be limited in providing a deep understanding of the inner workings of the implementation of complex public health interventions, or the social processes related to implementation. As a result of garnering insight about the social processes associated with implementation, leverage points can emerge at which to intervene to support more effective intervention design and implementation in different contexts.  Better implementation can potentially lead to a more effective tempering of the negative impacts created by the wicked problem of preventable nutrition-related morbidity and mortality. Ultimately, complex public health interventions, such as school food and beverage sales standards, that are effectively implemented could play a crucial role in transforming food systems to be more salutogenic.157  References  Aarestrup, A. K., Jorgensen, T. S., Jorgensen, S. E., Hoelscher, D. M., Due, P., & Krolner, R. (2015). Implementation of strategies to increase adolescents’ access to fruit and vegetables at school: Process evaluation findings from the Boost study. BMC Public Health, 15, 1-16. https://doi.org/10.1186/s12889-015-1399-9  ActNowBC & the Government of British Columbia. (year unknown). School meal and school nutrition program handbook. Retrieved from healthyschoolsbc.ca/program/resources/87/en/79172/School-Meal-and-School-Nutrition-Program-Handbook.pdf  [Accessed 17 July 2017]. Adam, A., & Jensen, J. D. (2016). What is the effectiveness of obesity related interventions at retail grocery stores and supermarkets? A systematic review. BMC Public Health, 16, 1-18. https://doi.org/10.1186/s12889-016-3985-x  Adamson, A., Spence, S., Reed, L., Conway, R., Palmer, A., Stewart, E., … Nelson, M. (2013). School food standards in the UK: Implementation and evaluation.  Public Health Nutrition, 16, 968–981. https://doi.org/10.1017/S1368980013000621  Afshin, A., Forouzanfar, M. H., Reitsma, M. B., Sur, P., Estep, K., Lee, A., … Murray, C. J. L. (2017). Health effects of overweight and obesity in 195 countries over 25 years. New England Journal of Medicine, 377, 13–27. https://doi.org/10.1056/NEJMoa1614362  Alberta Health Services. (2010). Health promotion and behavior change theory. Calgary, Alberta: Alberta Health Services. Retrieved from http://www.screeningforlife.ca/resources/Health%20Promotion%20Evidence/Health%20Promotion%20Evidence/Health%20Promotion%20and%20Behaviour%20Change%20Theory.pdf [Accessed 16 December 2015]. Antilla, J., Rytkonen, T., Kankaanpaa, R., Tolvanen, M., & Lahti, S. (2015). Effect of national recommendation on sweet selling as an intervention for a healthier school environment. Scandinavian Journal of Public Health, 43, 27–34. https://doi.org/10.1177/1403494814558150  Archer, M. S. (1995). Realist social theory: A morphogenetic approach. New York: Cambridge University Press. Asada, Y., & Garza, K. (2014). School food environments-- If not evidence, then ethics? Journal of Science Policy & Governance, Sept 2014, 1-3. Ashe, L. M., & Sonnino, R. (2012). At the crossroads: New paradigms of food security, public health nutrition and school food. Public Health Nutrition, 16, 1020–1027. Ashe, M., Graff, S., & Spector, C. (2011). Changing places: Policies to make a healthy choice the easy choice. Public Health, 125, 889–895. https://doi.org/ 10.1016/j.puhe.2011.04.010 Astbury, B., & Leeuw, F. L. (2010). Unpacking black boxes: Mechanisms and theory building in evaluation. American Journal of Evaluation, 31(3), 363–381. Atlas.ti. (2015). ATLAS.ti: Qualitative Analyse und Datenauswertung – ATLAS.ti Software version 7.5.9. Retrieved from http://www.atlasti.com/de/index.htmlindex.html [Accessed 18 April 2014]. Bacigalupo, R., Cudd, P., Littlewood, C., Bissell, P., Hawley, M. S., & Buckley Woods, H. (2013). Interventions employing mobile technology for overweight and obesity: An early systematic review of randomized controlled trials. Obesity Reviews, 14(4), 279–291. Baltas, G. (2001). Nutrition labelling: Issues and policies. European Journal of Marketing, 35, 708–721. https://doi.org/10.1108/03090560110388178  158  Barlow, Z., & Stone, M. K. (2011). Living systems and leadership: Cultivating conditions for institutional change. Journal of Sustainability  Education, 2, 1–29. Retreived from https://www.ecoliteracy.org/sites/default/files/uploads/shared_files/CEL_JSE_Cultivating_Leadership.pdf [Accessed 13 October 2017]. Bastian, A., & Coveney, J. (2013). The responsibilisation of food security: What is the problem represented to be? Health Sociology Review, 22(2), 162–173. Baumeister, R. F., Masicampo, E. J., & DeWall, C. N. (2009). Prosocial benefits of feeling free: Disbelief in free will increases aggression and reduces helpfulness. Personality and Social Psychology Bulletin, 35(2), 260–268. Baxter, P., & Jack, S. (2008). Qualitative case study methodology: Study design and implementation for novice researchers. The Qualitative Report, 13(4), 544–559. Bowen, G. A. (2008). Naturalistic inquiry and the saturation concept: A research note. Qualitative Research, 8(1), 137–152. British Columbia Ministry of Education & British Columbia Ministry of Health. (2005). School food sales and policies provincial report. British Columbia, Canada. Retrieved from http://www.health.gov.bc.ca/library/publications/year/2005/sales_report.pdf [Accessed 23 March 2013]. British Columbia Ministry of Education & British Columbia Ministry of Health. (2008). School food sales and policies provincial report II. British Columbia, Canad