Australian Research Council Linkage Grant Application

• Chief Investigators and Industry Partner Investigators
• Australian Research Council Linkage Grant Application

Project Title

Maximising the Effectiveness of Public Health Policies: The Case of Smoke-Free Policies.

Project Aims and Background

At the very core of policy innovation is the growing debate regarding the best way to conceptualise the relationships between research, policy settings that have been informed by the research and subsequent outcomes. This is particularly evident with respect to the implementation of tobacco control measures, such as smoke-free environments; a growing issue faced by governments around the world. The Framework Convention on Tobacco Control, an international treaty for tobacco regulation and control, is likely to fuel the need for a more consistent approach to the formulation and implementation of smoke-free policies. However, any approach that seeks to maximise the effectiveness of such policies must be able to accommodate the existing diversity in countries’ approaches to smoke-free environments.

There have been numerous attempts to describe and understand the effects of such policies using linear, regression-based models. However, these assume a linear sequence between research evidence outlining the need for action, the optimal forms of action, policy development, and the subsequent effects of policies. Interestingly, even simple analyses have already demonstrated that smoke-free rules feed back in a non-linear way on attitudes to smoke-free places in a so-called virtuous cycle. Similarly, when evaluating behavioural changes of smokers in response to policy initiatives, such relations are often non-linear. As noted by Coghill and Petrovic-Lazarevic, significant behavioural changes may be a consequence of what appear to be marginal increments in social/health policy.

Further, regression approaches do not adequately address the potential influence of variables that cannot be directly measured by such models. As such, current models run the risk of ignoring variables that may well influence the successful uptake of a particular policy. Furthermore, early work was restricted to local policies and tended to be based on detailed case studies of policies enacted in individual countries.

In addition, existing approaches have tended to focus on individual, independent decision makers. As such, they do not address the interrelatedness of policy developers, decision makers and the targets of their policies. For example, the targets of policy do not automatically comply and qualitative research suggests that areas of low compliance with smoke-free rules may be attributable to poor managerial practices rather than simple resistance from smokers. A combination of local rules with varied exemptions, poor communication about the value of having smoke-free places and a culture of resistance to authority may all contribute to non-compliance. There is, however, no empirical evidence of what the relative contribution of these factors is, or the various ways in which they may interact.

Decision-making processes associated with the development of any government policy, including tobacco control policies, are primarily supported by the supply of research data and information to decision makers. However, the collection and dissemination of data and information cannot, on its own, facilitate the development of effective tobacco control policies. Differences in the effectiveness of such policies are a reflection of how well research data and information are translated into relevant policies and decisions. Therefore, the fundamental research question is how to maximise the quality of this translation process. We believe that our approach will provide both a theoretical and methodological framework for answering this question.

This project will develop and implement an innovative methodology for evaluating and refining public health policies. The primary focus will be on formulating an approach that will assist in the efficacious translation of science into policy, and policy into practice, while accommodating national differences. Particular attention will be directed towards examining the factors that influence the effectiveness of smoke-free policies across different countries.

Our Industry Partner, The Cancer Council of Victoria (CCV), is a key collaborator in the International Tobacco Control Survey (ITC) survey, which provides the most detailed data set on the impacts of tobacco control policies world-wide, and also provides access to the team of researchers (ITC network) who are all well placed to further identify the nature and effects of the policies. The ITC study, which commenced late in 2002, has recruited cohorts of smokers from the US, UK, Canada and Australia and is surveying them annually. It uses self-report variables from smokers as predictors of the impacts of diverse tobacco control policies, including smoke-free policies. The findings of the ITC project are now beginning to appear in the international literature. Over the period of the study, there have been marked changes in smoke-free policies, and more are scheduled. In Australia, alone, most States and Territories are introducing smoke-free conditions in bars and clubs over the next two years. Since its inception, the research team, in conjunction with local researchers, has instituted parallel surveys in three other countries (Ireland, Thailand, Malaysia), with planning under-way for two more countries (Korea and China) to become involved. The ITC survey measures changes over time within countries, and differences between countries. Coupled with country-level data collected by the ITC team, or collectable by them for this project, this will provide us with the broadest range of measures to define the outcomes of smoke-free policies, as well as measures of many predictor variables.

The CCV is interested by the potential of a range of new (for them) theoretical frameworks and their associated quantitative techniques for carrying out more sophisticated analyses of the ITC data. The theoretical frameworks include Actor Network Theory (ANT), the theory of Complex Adaptive Systems with its Agent Based Models (ABM), and the theory of System Dynamics with its Equation Based Modelling (EBM). ANT is an approach that focuses on innovation, particularly science based innovation, as a process that emerges from dynamic networks of heterogeneous co-determinants. The ANT will be used to define the heterogeneous co-determinants of the outcomes of smoke-free policies. We will then use ABM, rather than EBM, to model the dynamic interactions between these co-determinants. The rationale for using the ABM is that it affords advantages over the EBM when modelling processes that are essentially determined by social, rather technical factors. Moreover, the Monash University team, comprising Petrovic-Lazarevic, Coghill, and Yeh, have recognised expertise in the application of ABM in areas including adolescent smoking, higher education, airline safety and decision analysis.

The ITC survey and ITC network of tobacco control experts are resources for both existing measures of many key variables, and for the identification and measurement of the other variables that ABM models can use. Some of these may come from the empirical data already collected, some will be available from other sources, at least in some countries, and for others we will rely on expert judgements and ratings – something ITC network makes us uniquely placed to collect.

Despite common access to research data, countries differ widely in their progress towards implementing smoke-free environments. This suggests that policy outcomes are not based solely on research evidence, but rather are influenced by a complex configuration of factors, including existing regulations, media strategies, and institutional arrangements. Actor Network Theory (ANT) has several distinctive attributes: First, the theory can accommodate heterogeneous variables. It assumes that the effectiveness of a policy, like smoke-free policies, is driven by the interaction of both human factors (smokers, politicians, non-smokers, managers, scientists) and non-human factors (regulations, research data, settings, advertisements, penalties). Second, the theory facilitates the translation of research data and policies into practice. That is why ANT is also known as the “sociology of translation”. Fundamental to the ANT is the notion that interested parties tend to translate research data and policies to meet their own goals. Failure to align tobacco control policies with the goals of actors such as politicians, on the one hand, or smokers, on the other, weakens the tobacco control actor network relative to its competitors. Third, ANT, like the theory of Complex Adaptive Systems, positions non-linear, discontinuous and emergent change as the key signal of the success or failure of a policy innovation. When policy innovations succeed, the signal is a sudden and irreversible reconfiguration of the field within which the innovation is applied. For example, when smoking switches from being socially normative to non-normative; gradual quantitative change suddenly produces an emergent qualitative change that is not amenable to linear analysis. This is one of the key links between ANT and ABM, because both treat non-linearity and emergence as intrinsic properties of substantive change.

ANT provides an innovative, robust theoretical framework for identifying the kind and range of variables that need to be considered in relation to tobacco control policy, particularly those pertaining to smoke free environments. Then the ITC data and the judgments of ITC network will enable the measurement of these variables. Then ABM will provide the ability to model the dynamic interactions between research and practice.

The research objectives of the project are to:

• Identify the heterogeneous variables that potentially co-determine smoke-free policy outcomes across a range of countries with different policies, and different outcomes.
• Measure identified predictor variables in multiple contexts and over time.
• Develop dynamic models of how the predictor variables interact to produce policy outcomes over time.
• Test and evaluate the models on new countries, and/or datasets and/or outcomes.

Significance and Innovation

Promoting good health and well being for all Australians is one of the Australian Government’s National Research Priorities. Tobacco smoking is not only the largest single preventable cause of death in Australia, but smokers who do not die of tobacco related causes will often suffer chronic disease that requires long-term hospitalisation, and/or early retirement from work, due to permanent disability. The project will contribute directly to this National Research Priority. It will do this by using an innovative theoretical and methodological framework to improve the effectiveness of public health policies designed to minimise exposure to environmental tobacco smoke.

The theoretical framework and methodology to be developed will be unique. The framework will integrate the theory of Complex Adaptive Systems with ANT using data from ITC study, international social research, and the judgements of tobacco control experts. Based on an extensive literature review, such a complex combination of theoretical approaches does not appear to have been used in any field, let alone tobacco control, although individual components have been. In addition, directly combining research findings with the insights of tobacco control experts to develop a comprehensive and dynamic model of the impacts of tobacco control policies has not, to our knowledge, been previously attempted.

In order to develop such models, we need an example of tobacco control outcomes where the results are both tangible and characterised by sufficient variation to highlight the differences between high quality and low quality research-practice relationships. The development of smoke-free places provides such an example, with the advantage that smoke-free places are relevant to two of the three targets of government policy in tobacco control. The relevant targets are health of smokers (by reducing smoking) and rights of non-smokers (by preventing smoking in their presence).

The seven countries involved in the ITC survey are all at different stages in the development of smoke-free environments, employ different policy development processes and are characterised by widely different national cultures. In addition, they use different management strategies and loci of control with respect to tobacco control in general, and smoke-free environments, in particular. Moreover, they have significantly different smoking rates, smoking styles and institutional arrangements. National attitudes to smoking as well as to regulation and smoke-free environments are also different. All potentially relevant variables will be identified by integrating the findings of the ITC survey with expert knowledge of the situations within these countries. This study therefore affords a unique opportunity to develop a comprehensive ABM of the research-practice relationship in the tobacco control arena. The ensuing model will be capable of explaining the different kinds of outcomes achieved in the seven countries surveyed. This outcome is achievable because of the close links between Borland, the CCV and the ITC study. In addition, the track record and experience of the CIs in modelling will enable the development of the proposed model.

The most significant issue in policy analysis, especially in relation to public health, concerns how to deal with the ill-defined problems of human and non-human interaction in complex systems. The notion of complexity, especially its emphasis on emergence, and the dynamics of change, is appealing to both researchers and policy makers. One of the significant contributions of this project will be to demonstrate that effectiveness of ABMs is maximised when researchers take the step of first accurately defining the key variables and relationships to be included in the model. The identification of a set of key variables and their relationships will be achieved by synthesising the inputs from tobacco control experts with the ITC findings, using ANT.

The significance of this approach will be amplified by the ability of ABMs to test the different outcomes achieved when specific variables are changed. This will enable researchers, policy developers and decision makers to analyse what-if situations. The model will also highlight the differences between a policy development focus (what should be changed if all causal variables could be changed) and a management focus (what is the most effective way to maximise the effectiveness of smoke free policies if certain key variables such as lack of a national compliance strategy, cannot be changed).

It is intended that the smoke-free places models to be developed will eventually contribute to a range of tobacco-related outcomes and, as such, will represent a true innovation. The implications of the results will also be applicable, at least in part, to other domains where policy is designed to affect people’s behaviour, especially in the field of Public Health. It is important to stress that, in general, dynamic models have rarely been applied in such policy areas because of the difficulty of quantifying the predictor variables. The resources available through our industry partner give us a unique opportunity to comprehensively test models in the policy domain.

The utility of many attempts to use such models in the past have been constrained by only having measures for a sub-set of those variables that are believed to be important, thus reducing the predictive capacity of the models, and more generally, their utility. From the perspective of Monash University, the project provides an opportunity to further develop the application of ABMs to the policy implementation process, to apply a range of ABMs in areas where they have not yet been applied, to build and test models on better quality data and, in all instances, communicate the new knowledge.

In addition to these specific contributions, the research project will:

• Make a tangible contribution to National and International programs aimed at reducing the scourge of smoking. Any innovative contributions in this area should have the most positive impacts on public health costs and our ability to switch resources into other high priority areas like the diseases of ageing or youth.
• Make a unique and significant contribution to the methodological debate about the relationship between research and practice in public health. Successful attempts to model the relationship between variables such as existing policy settings, institutional arrangements, research findings and national culture are few and far between. Existing attempts have been bedevilled by methodological uncertainty and a lack of approaches that can account for, and interact constructively with, the intrinsic complexity of the field.
• Provide a model of trans-disciplinary research managed by a team of academics and practitioners with a wide range of professional qualifications and expertise. In addition to the immediate significance of such a diverse team bringing its skills to bear on such an important international and national issue, there is the further advantage of a larger international, trans-disciplinary collective (the ITC network) helping to develop the model and responding to the emergent products. Essentially, this project will be a microcosm of the research-practice relationship.

Approach and Training

Table 1 identifies the tasks required to achieve the objectives. The potential co-determinants of policy outcomes identified by applying ANT will provide the basis for deciding which variables should be included in the dynamic models, and how they can best be measured. This will be done by integrating relevant ITC survey variables with tobacco control experts’ knowledge of the situations within the seven countries and, where necessary, with social research data in the public domain. This will be achieved through interviews, surveys and consultations with members of ITC team and, if necessary, with other domain experts. The experts will evaluate the relevance of the potential variables identified by ANT analysis, rejecting those that are deemed to be of doubtful utility, and adding any new variables that they identify.

Table 1: Research Objectives, Tasks, Timing and Research Outcomes

The variables will then be organised into clusters based on their relationship with the outcome being sought – i.e. the effective implementation of smoke-free policies. This organisation will produce the simplest model possible that, nevertheless, does justice to the intrinsic complexity of the field. The approach will also facilitate the incorporation of different measurement metrics within the one model (continuous variables, ordinal variables and nominal relationships).

Once the co-determinants of smoke-free outcomes have been organised into clusters, experts from the ITC network will firstly rate the variables that makes up each cluster in terms of the overall importance to the primary goal (effective smoke-free policies) and then, carry out a country analysis. Where variables already exist as part of the ITC survey, or where secondary data exists (for example, smoking rate, rate of uptake, proportion of “places” where restrictions already apply) a score for each country will be directly available. Where expert judgement is required, for example to classify control strategies from regulation to litigation, a Delphi approach will be applied. This will be done to minimise the variance between individual judgments. The total, integrated, results will be fed back to participants, and responses to the final configuration will be collected.

Over the last decade there have been some fruitful applications of complexity theory to the evolution of living systems in general, and social systems in particular. The most successful methodologies associated with Complex Adaptive Systems are ABMs including neural networks, genetic algorithms, and cellular automata. In order to provide support for advocacy and social regulation, models that have reliably explained several cycles of outcomes can subsequently be used to identify the conditions by which the desired outcomes are most likely to emerge.

The process for producing the relevant ABM will be as follows. Firstly the draft ABM will be trained on data from Waves 1, 2 and 3 for the four English speaking countries, and on the variables derived from the tobacco control experts’ judgements. The model will then be extended cross culturally on the first two Waves of the data for Thailand and Malaysia, and Wave 3 of data from Ireland. The full model will then be fed-back to the domain experts for sanction and final specification. Secondly the predictive power of ABM will be tested on data from Waves 4 (2006), 5 (2007), and 6 (2008) for the English speaking countries and Ireland, and on Wave 3 for Thailand and Malaysia. Again, the model will then be fed-back to domain experts for sanction and final specification.

Further refinement will most likely involve extension to a country with a Confucian culture, such as China, or Korea, both of which are planned to have Wave 1 data by 2006/7, and the addition of new questions to the ITC survey to measure variables that have emerged as particularly sensitive when represented by expert judgement or secondary data. Such new variables will be added to the models as measurement data becomes available.

One of the key signals of complexity is that interactions between individuals produce innovative and emergent outcomes. In this research we are seeking to produce two kinds of innovative, emergent outcomes. The first emergent outcome will be ANT map of the key variable clusters and the relationships between them. Participants in the production of this map constitute key nodes in tobacco research and policy networks whose individual perceptions, knowledge, beliefs and judgements will interact to produce the emergent map. By organising and classifying variables into clusters, in a common space, based on their relevance to the production of smoke-free places, we will, in fact, be developing an emergent framework for understanding and implementing a more effective relationship between research, policy, institutional arrangements and management in the tobacco control arena.

The second emergent outcome consists of ABM output with its dynamic representation of how the inter-related variables produce results over time with respect to smoke-free places. However, more importantly, these emergent outcomes can themselves be fed back to tobacco experts to modify the individual assumptions that went into the production of the map and the models in the first place. Therefore, consistent with the requirements of complex systems, we can employ ANT and ABM in an evolutionary process. The goal of this process will be to demonstrate that ANT analysis and the dynamic ABM produce learning – an emergent process – in the domain experts. This will be in addition to meeting the criteria for successful prediction when tested on new Waves of data, and on new countries.

Having developed a methodology, the process could be further tested and refined by extending the work to other critical areas. In particular, given the experience of Chief Investigators, the methodology could be extended to the issue of youth smoking. Finally, having demonstrated its ability to both predict and produce outcomes across a range of countries and tobacco related issues, we will be in a position to assert that the methodology is potentially applicable across a wide range of public health and related fields.

Collaborating Organisation Commitment and Collaboration

The CCV is an internationally and nationally respected organisation, best known for its leadership in the provision of, and advocacy for, research based cancer control strategies. In particular it has a strong international reputation for both the quality of its research in tobacco control it produces, and the success of the campaigns it has run (through Quit), or to which it has contributed. The Quit program is a joint initiative of the Council, VicHealth, Victorian Department of Human Services and the National Heart Foundation. Its steering committee is also formally constituted as an advisory committee to the Minister on tobacco control issues. The current CCV director, Prof David Hill, was the Chair of the National Expert Advisory Committee on Tobacco, and is a long term researcher and leader in tobacco control. The CCV employs people who are recognised internationally for their research achievements, and for their knowledge of the key issues in public health in general, and cancer control in particular. Staff have won grants from the most prestigious national and international organisations, including the National Cancer Institute of the US, Robert Wood Johnson Foundation, Canadian Institute of Health Research, Cancer Research UK, National Health and Medical Research Council (Australia).

The CCV is intrigued by the potential of alternative conceptual frameworks to add new insights that might improve the quality and utility of its applied research. It is vitally interested in the outcomes of the study and is committed to building those aspects of the new methods that prove fruitful into its continuing work.

CCV experts are members of programme committees of many prestigious national and international conferences in cancer control, addiction and the social sciences. They also have an international reputation in tobacco control, and the potential to influence all aspects of government tobacco regulation including smoke-free places and regulation of tobacco pricing, promotion, distribution and product attributes. Their research achievements are reported in highly ranked international and national journals.
The CCV Tobacco Control Unit has significantly contributed to the ITC study, not only technically, but most importantly through the tobacco control knowledge of its experts. The Tobacco Control Unit staff have over thirty years of theoretical and practical involvement in the relationship between research and policy, with respect to both cancer and tobacco control. Their research was the first to provide longitudinal evidence on the effects of workplace smoking bans on cigarette consumption. Their current work, using ITC data, is extending the evidence to restrictions on smoking in recreational settings, and self-imposed restrictions in smokers’ homes. Tobacco Control Unit experts at the Council advise both state and federal government on tobacco control issues, including smoke-free places.

The commitment of CCV to this project involves active participation in the design and management of the research, in-kind provision of access to ITC data, provision of staff time to extract and analyse data, and assistance in the creation of new measures for currently unmeasured factors. CCV will also provide access to national and international tobacco control domain experts. In terms of financial commitment, CCV will allocate $20,000 per annum to cover direct and indirect costs of creating and testing ABM, and in-kind contribution of a conservatively estimated $50,000.

Further in-kind contributions will consist of establishment of the web-based survey site, running the web-based surveys, and the provision of teleconference facilities.

As a consequence of its participation in this project CCV will acquire significantly enhanced capacity in the development of tobacco research methodologies, tobacco control policies and strategies, and understanding of how to best to manage the research-practice cycle. This enhanced capacity will be reflected in the delivery of more effective methodologies, policies and practices in the tobacco control arena. CCV is highly committed to this project, which is pivotal to the long-term relationship with the research team at Monash University.

National Benefits

The national benefits are extensive. Firstly there is the direct contribution to the National Research Priorities, through the project’s focus on smoking. We expect the project to identify new strategies for improving the effectiveness of tobacco control policies. These have the potential to save numerous lives, and improve quality of life for non-smokers who will be protected from unwanted exposures to environmental tobacco smoke, as well as for smokers who will be protected from some of the adverse effects of their addiction. In doing this, the project will also have an extremely significant impact on health expenditures, given the costs of tobacco induced cancers, and the economic effects of the chronic illnesses that smoking causes. Secondly there is the Intellectual Property that will accrue to Australia as a consequence of the innovative theoretical and methodological framework, in a field that is of interest to all countries. Thirdly the project offers a longer term potential to make a significant contribution to the more general field of public health in Australia. Tobacco control is but one, albeit critical, component of the public health agenda. To the degree that capacity in this specific field can be enhanced theoretically, methodologically and practically, benefits are sure to flow on to other areas of public health with similar configurations of actors and issues. These other important areas include nutrition, cardio-vascular health, illicit drug taking, gambling, and to the degree that lifetime choices affect the ageing process, some of the costs of ageing. In an increasingly threatening global context, modelling and understanding how to best organise and integrate policy research, policy development and policy outcomes offers the potential to improve our collective ability to manage Australia’s future.

Communication of Results

Publishing: Results will be reported to CCV and at least ten papers will be published in highly ranked scholarly journals, both national and international, in tobacco control, public health, complex systems, science studies, and public policy such as Tobacco Control, Knowledge Based Systems, Addiction, American Journal of Public Health, Social Science and Medicine, Australian Journal of Public Administration, Public Policy, Australia and New Zealand Health Policy.