ISBN 9289050425 9 7 8 9 2 8 9 0 5 0 4 2 5 Health Policy Series No. 47 www.healthobservatory.eu Health Policy Series 47 AN EVIDENCE REVIEW AND PROPOSED FRAMEWORK 47 Assessing the economic costs of unhealthy diets and low physical activity Christine Joy Candari Jonathan Cylus Ellen Nolte An evidence review and proposed framework Unhealthy diets and low physical activity contribute to many chronic diseases and disability; they are responsible for some 2 in 5 deaths worldwide and for about 30% of the global disease burden. Yet surprisingly little is known about the economic costs that these risk factors cause, both for health care and society more widely. This study pulls together the evidence about the economic burden that can be linked to unhealthy diets and low physical activity and explores • How definitions vary and why this matters • The complexity of estimating the economic burden and • How we can arrive at a better way to estimate the costs of an unhealthy diet and low physical activity, using diabetes as an example The review finds that unhealthy diets and low physical activity predict higher health care expenditure, but estimates vary greatly. Existing studies underestimate the true economic burden because most only look at the costs to the health system. Indirect costs caused by lost productivity may be about twice as high as direct health care costs, together accounting for about 0.5% of national income. The study also tests the feasibility of using a disease-based approach to estimate the costs of unhealthy diets and low physical activity in Europe, projecting the total economic burden associated with these two risk factors as manifested in new type 2 diabetes cases at €883 million in 2020 for France, Germany, Italy, Spain and the United Kingdom alone. The ‘true’ costs will be higher, as unhealthy diets and low physical activity are linked to many more diseases. The study’s findings are a step towards a better understanding of the economic burden that can be associated with two key risk factors for ill health and they will help policymakers in setting priorities and to more effectively promoting healthy diets and physical activity. The editors Christine Joy Candari was an independent consultant at the time of writing this report. She is currently Chief Consultant for Health Research, U Consult Us Inc, Manila, The Philippines Jonathan Cylus is Research Fellow, European Observatory on Health Systems and Policies, London School of Economics and Political Science Ellen Nolte is Head of London Hubs, European Observatory on Health Systems and Policies DIETS AND LOW PHYSICAL ACTIVITY ASSESSING THE ECONOMIC COSTS OF UNHEALTHY Christine Joy Candari, Jonathan Cylus, Ellen Nolte Cover_WHO_nr47_Mise en page 1 7/06/17 14:31 Page 1 Assessing the economic costs of unhealthy diets and low physical activity An evidence review and proposed framework The European Observatory on Health Systems and Policies supports and promotes evidence-based health policy-making through comprehensive and rigorous analysis of health systems in Europe. It brings together a wide range of policy-makers, academics and practitioners to analyse trends in health reform, drawing on experience from across Europe to illuminate policy issues. The Observatory is a partnership hosted by the WHO Regional Office for Europe, which includes the governments of Austria, Belgium, Finland, Ireland, Norway, Slovenia, Sweden, Switzerland, the United Kingdom, and the Veneto Region of Italy; the European Commission; the World Bank; UNCAM (French National Union of Health Insurance Funds); the London School of Economics and Political Science; and the London School of Hygiene & Tropical Medicine. The Observatory has a secretariat in Brussels and it has hubs in London (at LSE and LSHTM) and at the Technical University of Berlin. Assessing the economic costs of unhealthy diets and low physical activity Christine Joy Candari, Jonathan Cylus, Ellen Nolte Keywords: DIET – ECONOMICS SEDENTARY LIFESTYLE CHRONIC DISEASE – ECONOMICS CHRONIC DISEASE – PREVENTION AND CONTROL HEALTH CARE EVALUATION MECHANISMS DELIVERY OF HEALTH CARE © World Health Organization 2017 (acting as the host organization for, and secretariat of, the European Observatory on Health Systems and Policies) All rights reserved. The European Observatory on Health Systems and Policies welcomes requests for permission to reproduce or translate its publications, in part or in full. The designations employed and the presentation of the material in this publication do not imply the expression of any opinion whatsoever on the part of the European Observatory on Health Systems and Policies concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. Dotted lines on maps represent approximate border lines for which there may not yet be full agreement. The mention of specific companies or of certain manufacturers’ products does not imply that they are endorsed or recommended by the European Observatory on Health Systems and Policies in preference to others of a similar nature that are not mentioned. Errors and omissions excepted, the names of proprietary products are distinguished by initial capital letters. All reasonable precautions have been taken by the European Observatory on Health Systems and Policies to verify the information contained in this publication. However, the published material is being distributed without warranty of any kind, either express or implied. The responsibility for the interpretation and use of the material lies with the reader. In no event shall the European Observatory on Health Systems and Policies be liable for damages arising from its use. The views expressed by authors, editors, or expert groups do not necessarily represent the decisions or the stated policy of the European Observatory on Health Systems and Policies or any of its partners. ISBN 978 92 890 5042 5 Printed in the United Kingdom Typeset by Tetragon, London Cover design by M2M Address requests about publications to: Publications, WHO Regional Office for Europe, UN City, Marmorvej 51, DK-2100 Copenhagen Ø, Denmark. Alternatively, complete an online request form for documentation, health information, or for permission to quote or translate, on the Regional Office web site (http://www.euro.who.int/ pubrequest). Table of contents Acknowledgements vii Foreword viii List of tables, boxes and figures ix Summary xi Chapter 1 Introduction 1 Chapter 2 The economic costs of unhealthy diets and low physical activity: what does the published literature tell us? 3 2.1 Characteristics of reviewed studies 3 2.2 What the evidence tells us: the economic costs of unhealthy diets 5 2.3 What the evidence tells us: the economic costs of low physical activity 7 2.4 Review of the evidence: a summary 10 Chapter 3 Estimating the economic costs of unhealthy diets and low physical activity is complex 13 3.1 Defining the concepts: how can we understand ‘unhealthy diet’ and ‘low physical activity’? 13 3.2 Costing studies differ in key assumptions, influencing estimates for the economic burden of unhealthy diets and low physical activity 16 3.3 The nature and range of costs considered is likely to underestimate the ‘true’ economic burden of unhealthy diets and physical activity 18 3.4 Conceptual and methodological challenges of estimating the economic costs of unhealthy diets and low physical activity: a summary 19 Chapter 4 Taking available approaches to determining the economic costs of unhealthy diets and low physical activity further: a proof-of-concept approach applied to five European countries 21 4.1 Diabetes as an outcome of unhealthy diets and low physical activity 23 4.2 The principal approach used in this study to estimate the economic costs that can be associated with unhealthy diets and low physical activity 25 4.3 The estimated total economic costs of unhealthy diets and low physical activity related to diabetes and its complications 40 vi Assessing the economic costs of unhealthy diets and low physical activity Chapter 5 Discussion and conclusions 43 5.1 Limitations of the costing framework 45 5.2 Implications for future studies 47 References 49 Appendices 55 Acknowledgements This study forms part of a wider programme of work that seeks to explore in further depth the available evidence on the economic costs that can be associated with unhealthy diet and lack of physical activity in Europe, or high‐income countries more broadly, and their effects on health. A better understanding of the economic burden associated with these critical risk factors can provide important pointers to inform evidence‐based decision‐making within the European region as to the most appropriate policies to improve the health, well‐being and quality of life of the population. We gratefully acknowledge the very helpful and insightful comments provided by Nick Cavill on an earlier draft of this work. We would also like to acknowledge the very useful comments and suggestions provided by the European Commission, which further helped inform the discussions presented here. We are indebted to Charles Normand and Josep Figueras for their constructive comments and guidance that helped shape the final version of this book. We would like to thank Sarah Cook for copy-editing and Jonathan North and Caroline White for the production of this book. The authors are fully responsible for any errors. Foreword Lifestyle related health problems are now a challenge of global proportions. For example, levels of obesity have doubled since 1980, and it costs a staggering 2.8% of the world’s GDP. Europe is not an exception: half of EU adults today are either overweight or obese, and rates for children are even more worrisome: one in three. Unless we act, we will condemn a whole generation to a lifetime of poor health. With governments across Europe struggling to curb these ever-rising rates, it is more important than ever to make the economic case for investing in strategies that promote health and prevent diseases. Unhealthy diets and lack of physical activity are risk factors for developing a range of chronic diseases such as diabetes, cancer and cardiovascular disease. They not only reduce people’s quality of life and life expectancy, but also place a burden on our health systems and our economies, and on society as a whole. There is growing evidence that many prevention interventions are cost-effective. It is therefore surprising that OECD countries still spend an average of only 3% of their health care budgets on disease prevention programmes. In contrast, already some 7% of EU health budgets are spent on treating chronic diseases linked to obesity. This clearly points to a need to change mindsets. If we wish to keep our population healthy, active and productive, and our health systems strong and resilient in the long term, and reduce the increasing pressure on national health care budgets, we need to reset our thinking. We need to focus more on disease prevention and health promotion to save future expenditure on treatment and cure. This sounds self-evident – yet it is a surprisingly difficult message to get across. Underpinning policies with sound evidence is essential. I therefore welcome this report, which provides a useful contribution to the debate on the increasing importance of promotion and prevention, and I invite you to read the assessment of the evidence contained within these pages of the costs of unhealthy diets and low physical activity. Xavier Prats Monné Director General of the Directorate-General for Health and Food Safety List of tables, boxes and figures Tables Table 1 Annual economic costs of unhealthy diets as reported in the published literature 6 Table 2 Annual economic costs of low physical activity reported in the published literature 8 Table 3 Intake of major food groups associated with the lowest risks for chronic disease 13 Table 4 Population-attributable fractions used in costing studies: low physical activity 17 Table 5 The quality of diets according to the alternate healthy eating index (AHEI) 2010 27 Table 6 Assignment of AHEI scores on mean dietary intakes of AHEI food groups in France, Germany, Italy, Spain and the United Kingdom 28 Table 7 Relative risks for diabetes related to unhealthy diets and low physical activity estimated by Li et al. (2015) 29 Table 8 Estimated prevalence rates of unhealthy diets and low physical activity in the populations eventually developing diabetes and relative risks for incident diabetes in France, Germany, Italy, Spain and the United Kingdom 30 Table 9 Estimated incident diabetic cases in 2020 attributable to unhealthy diet and low physical activity patterns in 2015 32 Table 10 Estimated diabetes-related health care costs in 2020 attributable to unhealthy diets and low physical activity patterns in 2015 in France, Germany, Italy, Spain and the United Kingdom 33 Table 11 Estimated diabetic complication-related costs in 2020 attributable to unhealthy diets and low physical activity patterns in 2015 in France, Germany, Italy, Spain and the United Kingdom 34 Table 12 Estimated number of incident type 2 diabetes cases at working age (15–64 years) that can be attributed to unhealthy diets and low physical activity in France, Germany, Italy, Spain and the United Kingdom, 2020 35 Table 13 Estimated total number of productivity days lost and cost due to absenteeism and presenteeism among incident type 2 diabetes cases at working age that can be attributed to unhealthy diets and low physical activity and who are expected to be in the labour force in France, Germany, Italy, Spain and the United Kingdom, 2020 36 Table 14 Estimated total number of productivity days lost and cost due to absenteeism and presenteeism among incident type 2 diabetes cases x Assessing the economic costs of unhealthy diets and low physical activity at working age that can be attributed to unhealthy diets and low physical activity and who are expected to be outside the formal labour force in France, Germany, Italy, Spain and the United Kingdom, 2020 38 Table 15 Estimated cost of working years lost due to work disability, early retirement and premature death among incident type 2 diabetes cases at working age that can be attributed to unhealthy diets and low physical activity and who are expected to be in the formal labour force in France, Germany, Italy, Spain and the United Kingdom, 2020 39 Table 16 Estimated economic cost that can be associated with unhealthy diets and low physical activity patterns in 2015 as manifested in incident diabetes and complication in France, Germany, Italy, Spain and the United Kingdom in 2020 41 Table 17 Estimated total and per capita economic cost that can be associated with unhealthy diets and low physical activity patterns in 2015 as manifested in incident diabetes and complications in France, Germany, Italy, Spain and the United Kingdom in 2020 41 Boxes Box 1 Population-attributable fraction 5 Box 2 Examples of moderate-intensity and vigorous-intensity physical activity 15 Box 3 Conceptualizing economic costs 19 Box 4 Health risks associated with unhealthy diets and low physical activity 22 Box 5 Computing the fully-adjusted population-attributable fraction (PAF) 31 Figures Figure 1 Relationship between unhealthy diets and low physical activity with type 2 diabetes and associated costs 24 Summary Unhealthy diets and low levels of physical activity are among the main risk factors for major chronic diseases. While this is well documented, the economic burden associated with these two risk factors remains uncertain. A better understanding of the economic burden could help inform priority setting and motivate efforts to promote more effectively healthy diets and physical activity in Europe and worldwide. This volume seeks to help advance the debate through • critically reviewing the literature that has sought to estimate the economic costs associated with unhealthy diets and low physical activity and presenting the range of estimates of economic burden; • analysing the measurement, methodological and practical issues in assessing the economic burden from unhealthy diets and low physical activity; and • developing a framework for assessing costs and testing the feasibility of this approach to provide better estimates of the economic burden. The published evidence overwhelmingly shows that unhealthy diets and low physical activity are predictive of higher health care expenditure, but estimates vary greatly. We used a rapid assessment of the evidence that has been published between January 2000 and February 2016, including a total of 30 studies for detailed review. Of these studies, six addressed diet, 21 looked at physical activity, and three considered both. Over half of the studies were set in North America, with only six set in Europe. Most studies retrospectively assessed the economic burden of unhealthy diets and low physical activity. About half adopted a disease-based approach, often looking at cardiovascular disease, type 2 diabetes and selected cancers. In most cases this approach used the population-attributable fraction (PAF), which estimates the proportion of a disease that can be attributed to a particular risk factor. We found that, overall, estimates presented in the reviewed studies varied widely. Not all included studies reported national or per capita costs that can be associated with the two risk factors. Of those that did, the estimated health care costs for unhealthy diets ranged from €3.5 per capita in China to €63 in Australia and xii Assessing the economic costs of unhealthy diets and low physical activity €156 in the United Kingdom. For low physical activity, the estimated annual per capita health care costs ranged from €3 in the Czech Republic to €48 in Canada. In the United States of America alone, these ranged from less than €1 in two studies up to €185. Only four studies also considered indirect costs in their estimates for low physical activity, and these varied from €3.7 per capita in China to between €127 and €224 in Canada. It is plausible that differences in socioeconomic conditions and health care and labour costs might lead to some of the differences between cost estimates. But much of the observed variation results from differences in approaches to measurement, such as what is meant by ‘unhealthy diets’ or ‘low physical activity’; large differences in the methodological approaches chosen (for example, the use of a retrospective, disease-based approach or of a prospective approach); differences in the populations studied and underlying data that are being used; and the range and types of costs considered. Defining what constitutes an ‘unhealthy diet’ and ‘low physical activity’ remains challenging and different conceptualisations make a comparative assessment of available evidence difficult. Definitions of unhealthy diets often refer to those high in specific nutrients such as saturated fats, salts or sugars. Yet, growing evidence finds that intakes of specific foods rather than the actual nutrients are most relevant for the development of chronic disease, and studies increasingly look at recommended intakes of selected food groups, such as fruit, vegetables, nuts and seeds, whole grains, seafood and unprocessed red meats. This is based on studies that found intake levels for these food groups to be associated with coronary heart disease, stroke, type 2 diabetes and certain cancers. But even with this approach there remains the problem of how to score different intake levels. These conceptual challenges are reflected in the reviewed studies of unhealthy diets, which often focused on single food items, such as intakes of fruit and vegetables, while others considered a range of nutrients and foodstuffs including saturated- and trans-fat, fruit, vegetables and whole grains. Again others examined diets that scored low on a measure of dietary diversity, or they analysed dietary patterns according to national dietary recommendations. A small number of studies did not define what was meant by unhealthy diet. There is more consensus about what constitutes low physical activity. For measurement purposes, physical activities are classified into categories of intensity, from light to moderate to vigorous, depending on the amount of effort required to perform the activity. Guidelines by the WHO combine intensity with duration, for example recommending that adults should engage in at least 150 minutes of moderate-intensity aerobic physical activity throughout the week, in bouts of at least 10 minutes’ duration, to reduce the risk of chronic disease. Summary xiii Yet, as with unhealthy diet, reviewed studies that have assessed the costs of low physical activity used different definitions, ranging from those focusing on different thresholds of intensity of activities to those looking just at the duration. Others only considered specific types of physical activity (or lack thereof), such as leisure-time physical activity, walking, participation in an exercise programme or sedentary behaviour. Two studies did not define low physical activity. These differences in definitions and conceptualisations undermine the comparative assessment of available cost estimates that can be associated with either risk factor. Costing studies differ in analytical approaches and in the nature and scope of data used, influencing estimates for the economic burden of unhealthy diets and low physical activity. Just under half of reviewed studies used population-attributable fractions (PAFs) to assess the contribution of unhealthy diets or low physical activity to a range of diseases (or death) and this then formed the basis to estimate the economic burden. Studies analysed different numbers and combinations of diseases, most commonly coronary heart disease, stroke, type 2 diabetes and colorectal and breast cancer. They also varied in the population-attributable fractions used. They commonly applied the PAFs to contemporary prevalence of unhealthy diets or low physical activity and then estimated the related disease costs incurred in the same year. Such an approach overlooks the time lag between exposure to the risk factor and the development of disease. There was equally great variation in the range of costs being considered by individual studies. The majority only looked at the direct health care costs, and this is likely to greatly underestimate the true economic burden that can be associated with unhealthy diets or low physical activity. For example, three Canadian studies that had analysed the costs associated with low physical activity found indirect costs caused by lost productivity to be about twice as high as direct health care costs, together accounting for between 0.4% and 0.6% of gross domestic product (GDP). We developed a framework for estimating the economic costs of unhealthy diets and low physical activity using a disease-based approach, and applied it to estimate the economic burden of type 2 diabetes associated with these two risk factors. Building on the insights from the critical appraisal of the literature and the review of measurement and methodological challenges, we developed a framework for assessing the costs of unhealthy diets and low physical activity. We adopted a disease-based approach, which incorporates a time perspective to account for the xiv Assessing the economic costs of unhealthy diets and low physical activity natural progression of disease. We further estimated incidence rather than prevalence, included costs related to the disease and its complications, and considered indirect costs of productivity losses as a consequence of absenteeism, presenteeism, work disability, early retirement and premature mortality. The framework was tested by assessing the costs of type 2 diabetes. Diabetes has been associated with a high individual, social and economic burden and related expenditure was estimated to account for some 9% of total health care expenditure in the European region in 2015. There is the specific advantage of using diabetes as the disease for testing the framework in that there is a well-established causal pathway from these risk factors and disease. We projected the total economic costs that can be associated with unhealthy diets and low physical activity in 2015 as manifested in incident type 2 diabetes cases in 2020 in five large European countries to be €883 million. Using this approach, we projected the total economic cost of diabetes that can be associated with unhealthy diets and low physical activity. The approach takes the patterns of diet and activity levels in 2015 and projects incident diabetes cases for the year 2020. The estimated direct and indirect costs associated with these cases ranged from €82.4 million in Spain to €266.7 million in Germany. This equates to a per capita cost of €1.77 in Spain to €3.29 in Germany. Relating costs more specifically to the population projected to develop diabetes in 2020 as a consequence of unhealthy diets and low physical activity in 2015, the United Kingdom showed the highest amounts, at €18 953, closely followed by Germany and France, while Italy had the lowest cost, at just over €10 720. The total cost in the five high-income countries studied (France, Germany, Italy, Spain and the United Kingdom) was projected to amount to about €883 million in 2020. The populations in the five countries studied account for almost two thirds of the total population in the European Union (EU-28). This would imply a total EU cost of around €1.3 billion, but care must be taken in any extrapolation given differences in population characteristics, costs of care and value of lost productivity. While these estimates of the economic costs are substantial, they represent only a small proportion of health care expenditure and a very small proportion of GDP. Even on the higher estimates in the sensitivity analysis it is likely that the burden of disease associated with unhealthy diets and low physical activity as measured by poor health and shortened life will be at least as important as the financial costs of additional health care and lost productivity. It is difficult to compare the findings of the analyses presented here with estimates published elsewhere since only diabetes costs are estimated. The principal analytical steps used in our analysis are similar to those in the recent Lancet Physical Activity 2016 Series for low physical activity. Where our model differs is that we Summary xv only considered the costs of new cases, which can be causally linked to the risk factor, and we take account of the expected time lag between exposure to the risk factor (unhealthy diets, low physical activity) and the development of the disease, and of complications that arise from diabetes. We also considered a wider range of indirect costs linked to lost productivity because of work absence, disability, early retirement and premature death among incident diabetes cases that can be attributed to unhealthy diets and low physical activity. While our estimates are restricted to diabetes, they provide a fuller picture of the likely future costs that can be attributed to contemporary dietary and physical activity patterns. Where do we go from here? This study has tested the feasibility of estimating the costs of unhealthy diets and low physical activity using a disease-based approach. While there are limitations, it has shown that it is broadly feasible to populate the model with data from a range of sources, and the results show a reasonable consistency across countries. While the disease burden from diabetes is not currently as large as that for, for example, ischaemic heart disease, it is a good exemplar because of the strong relationship between these lifestyle factors and the risk of diabetes. In other chronic diseases there will be additional challenges in identifying the contribution of these lifestyle factors and disease risk. Given the very wide range of estimates of costs from the studies reviewed, this may be a more promising approach. Chapter 1 Introduction Unhealthy diets and low physical activity are among the key risk factors for major chronic, non-communicable diseases such as cardiovascular diseases, cancers and diabetes. In 2015, diets that are low in fruit and vegetables or high in sugar, processed foods or sodium were estimated to directly account for 37% of all deaths and just over a quarter of the total disease burden (disability-adjusted life years, DALYs) (GBD 2015 Risk Factors Collaborators, 2016). Low levels of physical activity accounted for another 5% of all deaths and 3.4% of DALYs. Taken together, these two risk factors were thus responsible for some two in five deaths worldwide and about 30% of the global disease burden. It is against this background that several strategies have been launched at European and global levels since the early 2000s to promote healthy diets and physical activity and so reduce the related burden of ill-health (Commission of the European Communities, 2007; Council of the European Union, 2014; WHO Regional Office for Europe, 2001; WHO Regional Office for Europe, 2005; WHO Regional Office for Europe, 2015; WHO Regional Office for Europe, 2016; World Health Organization, 2004; World Health Assembly, 2013). There is an expectation that effectively promoting healthy diets and physical activity can help reduce the economic burden associated with chronic diseases, which was estimated to account for 70–80% of health care budgets, or €700 billion annually across the European Union alone (European Commission, 2014). However, while there is good evidence about the positive impacts of, for example, a healthy diet on outcomes such as major cardiovascular events (Estruch et al., 2013; Stefler et al., 2015; Tong et al., 2016; Liyanage et al., 2016) or about the association between physical activity and mortality (Samitz et al., 2011; Woodcock et al., 2011), only a small number of studies have provided robust estimates for the economic impacts on health care and the wider society that are directly related to either factor. This is in part because the relationships between unhealthy diets or low physical activity and health care costs is complex, and it is the consequences of these behaviours, for example obesity or specific lifestylerelated diseases, that lead to health care costs. However, there may be more direct effects of unhealthy diets and low physical activity on lost productivity. Available studies provide widely varying estimates, reflecting the range of assumptions and estimates that inform underlying models (Cecchini & Bull, 2015). For 2 Assessing the economic costs of unhealthy diets and low physical activity example, Scarborough et al. (2011) estimated the economic burden of ill health that could be attributed to unhealthy diet to the National Health Service in the United Kingdom in 2006–07 to be £5.8 billion (€8.5 billion) and £0.9 billion (€1.3 billion) for low physical activity. Maresova (2014) calculated the financial cost of low physical activity to public health insurance in the Czech Republic to be 0.4% of total health care cost in 2008. These figures do not take account of the wider societal costs that can be attributed to these risk factors. More recently, Ding et al. (2016) estimated that in 2013 health care costs associated with low physical activity accounted for an average of 0.6% of total health expenditure across EU Member States. This volume seeks to help provide a clearer picture of the economic costs of unhealthy diets and low physical activity in Europe, to understand the methodological and practical difficulties in assessing costs, and to provide a test case to show how costs might better be assessed. Chapter 2 provides a targeted review of the literature that has sought to estimate the economic costs associated with unhealthy diets and low physical activity. This provides estimates of the costs as reported in the different studies, and shows how the estimated costs vary with the different assumptions and methods used. Chapter 3 reviews the methodological and practical challenges in estimating the economic costs of unhealthy diets and low physical activity. It looks in detail at issues of measurement of diet and physical activity, and at how there are strong interactions between diet and physical activity in terms of risks of disease (and indeed in different elements of diet), and shows how these challenges affect the estimates of costs. Chapter 4 develops a bottom-up framework for assessing the costs of unhealthy diets and low physical activity using a disease-based approach. This is then tested out for a disease (type 2 diabetes) for which there is strong evidence that the disease is related to both diet and the level of physical activity. This provides a test of concept, and shows how evidence from a range of sources can potentially be combined to improve our understanding of the economic cost. Chapter 5 briefly summarises the findings of the work and proposes avenues for further research. Chapter 2 The economic costs of unhealthy diets and low physical activity: what does the published literature tell us? This chapter reports the findings of a targeted review of the literature that has sought to estimate the economic costs associated with unhealthy diets and/or low physical activity. No previous review has covered costs associated with both of these, and existing reviews of costs of low physical activity (Kruk, 2014; Oldridge, 2008) did not provide information on the methodological approach or a critical evaluation of reviewed studies. The aim of this review is to present the current best estimates of the economic costs, and the strengths and weaknesses of the available studies. Drawing on the principles of a rapid evidence assessment (Khangura et al., 2012), we carried out a targeted search of PubMed, the National Library of Medicine’s Medline and pre-Medline database (NCBI, 2016). We identified studies using medical subject headings (MeSH) as follows (‘/’ indicating ‘or’): ‘sedentary lifestyle/leisure activities/motor activity’ or ‘diet/food’ in combination with ‘health care costs[statistics and numerical data]/costs and cost analysis/ public health[economics]/cost of illness’. We limited our search to studies that were published between January 2000 and February 2016 and that were in the English language. We included original studies estimating the costs that can be associated with unhealthy diets or low physical activity. We did not consider studies that focused on populations with established disease (e.g. people with osteoarthritis), intervention studies or those that compared different populations with different levels of physical activity or differing dietary behaviours, except where these were quantified further. Given the overall scarcity of studies, we adopted an inclusive approach and we did not formally assess the quality of included studies. We excluded editorials, commentaries or letters. 2.1 Characteristics of reviewed studies This section briefly summarises the key characteristics of reviewed studies; a detailed overview is presented in Appendix 1. 4 Assessing the economic costs of unhealthy diets and low physical activity The PubMed searches identified a total of 3661 records (diet: 2347; physical activity: 1314) and, following screening of abstracts and titles, we considered 38 studies for full-text review. Of these, 30 studies were considered eligible for inclusion in the review. Six addressed diet, 21 examined physical activity, and three considered both. Eleven of the included studies were set in the United States of America (Ackermann et al., 2003; Anderson et al., 2005; Bachmann et al., 2015; Bland et al., 2009; Carlson et al., 2015; Chevan & Roberts, 2014; Daviglus et al., 2005; Garrett et al., 2004; Martinson et al., 2003; Wang et al., 2005; Wang et al., 2004), five in Canada (Alter et al., 2012; Janssen, 2012; Katzmarzyk, 2011; Katzmarzyk et al., 2000; Krueger et al., 2015), three each in the United Kingdom (Allender et al., 2007; Rayner & Scarborough, 2005; Scarborough et al., 2011) and in Australia (Collins et al., 2011; Doidge et al., 2012; Peeters et al., 2014), two in China (Popkin et al., 2006; Zhang & Chaaban, 2012) and one each in Brazil (Codogno et al., 2015), Germany (Idler et al., 2015), Czech Republic (Maresova, 2014), Japan (Kuriyama et al., 2004) and Taiwan, China (Lo et al., 2013). One study assessed the economic costs of disease-related malnutrition in health care settings in Ireland (Rice & Normand, 2012). The majority of studies provided a retrospective assessment of the economic burden that can be associated with either unhealthy diets or low physical activity or both, while nine studies adopted a prospective approach by following a cohort of people over a defined period of time (Alter et al., 2012; Bachmann et al., 2015; Bland et al., 2009; Chevan & Roberts, 2014; Collins et al., 2011; Kuriyama et al., 2004; Lo et al., 2013; Martinson et al., 2003; Peeters et al., 2014). About half of the reviewed studies adopted a disease-based approach to estimate the economic burden that can be associated with unhealthy diets or low physical activity, most frequently cardiovascular diseases (coronary heart disease, stroke, hypertension), type 2 diabetes, and colon and female breast cancer (Allender et al., 2007; Daviglus et al., 2005; Doidge et al., 2012; Garrett et al., 2004; Janssen, 2012; Katzmarzyk, 2011; Katzmarzyk et al., 2000; Krueger et al., 2015; Maresova, 2014; Popkin et al., 2006; Rayner & Scarborough, 2005; Scarborough et al., 2011; Wang et al., 2004; Zhang & Chaaban, 2012). The disease-based approach typically, although not always (Daviglus et al., 2005; Wang et al., 2004), uses the population-attributable fraction (PAF) to quantify the contribution of the individual risk factor (unhealthy diet, low physical activity) to the burden of a given disease or death (Box 1). The remaining studies used a generic, non-disease-based approach, where unhealthy diets or low physical activity data of each individual were linked to health care cost data, regardless of the type of disease or diagnosis. Such an approach is typically followed by regression techniques to identify possible The economic costs of unhealthy diets and low physical activity 5 associations between the presence or absence of the risk factor and the magnitude of costs (Ackermann et al., 2003; Alter et al., 2012; Anderson et al., 2005; Bachmann et al., 2015; Bland et al., 2009; Carlson et al., 2015; Chevan & Roberts, 2014; Codogno et al., 2015; Collins et al., 2011; Idler et al., 2015; Kuriyama et al., 2004; Lo et al., 2013; Martinson et al., 2003; Peeters et al., 2014; Rice & Normand, 2012; Wang et al., 2005). Studies using a disease-based approach based on population-attributable fractions reported costs that can be associated with unhealthy diets and/or low physical activity in aggregate terms, for example national costs. Conversely, studies that adopted a generic approach tended to report the costs as ‘additional costs’, that is, additional to the costs a non-exposed individual would otherwise incur, in per capita terms. Two studies reported risk estimates, in this case, odds ratio, illustrating the strength of the association between unhealthy diet or low physical activity and costs (Chevan & Roberts, 2014; Codogno et al., 2015). 2.2 What the evidence tells us: the economic costs of unhealthy diets Drawing on those studies that have reported aggregate costs, the annual economic costs of unhealthy diets ranged from €1.4 billion in Australia (AU$ 2 billion) (Doidge et al., 2012) to €4.5 billion in China (US$ 4.2 billion) (Popkin et al., 2006) and €8.5–9.5 billion in the United Kingdom (£5.8–6 billion) (Rayner & Scarborough, 2005; Scarborough et al., 2011) (see Appendix 2 for conversion Box 1 Population-attributable fraction The population-attributable fraction (PAF) generally refers to the proportion of cases for a given outcome of interest that can be attributed to a given risk factor among the entire population. Specifically, the PAF is a function of the proportion of individuals in the population who are exposed to the factor of interest (Pexp), for example, unhealthy diet, and the relative risk (RR) of a particular outcome given that exposure, for example, the development of type 2 diabetes. If the exposure variable is dichotomous (i.e. the risk factor is present or absent), the mathematical formula reads: For example, if the relative risk for the effect of a given exposure on a disease outcome was approximately 5, and we can infer from a population survey that about 20% of the population was exposed to this risk factor, the proportion of all disease cases in the population that can be attributed to the risk factor is calculated as: PAF = 0.05 x (20-1) / (0.05 x (20-1) + 1) = 0.95 / 1.95 = 49%. PAF (%) = Pexp (RR−1) [Pexp (RR−1)] + 1 6 Assessing the economic costs of unhealthy diets and low physical activity rates of currencies applied) (Table 1). Taking account of the population size, the per capita annual economic costs that can be associated with unhealthy diets is estimated to range from €143 to €156 for the United Kingdom, €63 for Australia and €3.5 for China. All of these costs were health care costs only. Table 1 Annual economic costs of unhealthy diets as reported in the published literature Country Estimated annual economic costs of unhealthy diets (per capita*) Definition of unhealthy diets Perspective of cost estimation Population base Source Australia €1.4 billion (€63) Low levels of dairy consumption Direct health care costs, not specified General population Doidge et al. (2012) China €4.5 billion (€3.5) Diet high in saturated and trans-fat, low in fruit, vegetables and whole grains plus heavy alcohol drinking Direct health care costs, not specified General population Popkin et al. (2006) United Kingdom €8.5 billion (€143) Not defined Direct health care costs, not specified General population Rayner & Scarborough (2005) €9.5 billion (€156) Not defined Direct health care costs, not specified General population Scarborough et al. (2011) Note: * per capita costs calculated using United Nations population data (United Nations, 2015). Reviewed studies varied widely in their definition of unhealthy diets, often focusing on single food items, such as dairy consumption (Doidge et al., 2012) or intakes of fruit or vegetables (Bland et al., 2009; Daviglus et al., 2005), while Popkin et al. (2006) considered a range of nutrients and foodstuffs including high consumption of saturated- and trans-fat, and low consumption of fruit, vegetables and whole grains, as well as heavy alcohol use. Lo et al. (2013) examined diets that scored low on a measure of dietary diversity, while Collins et al. (2011) analysed dietary patterns according to Australian national dietary recommendations. Two studies did not define unhealthy diets specifically (Rayner & Scarborough, 2005; Scarborough et al., 2011), and Rice & Normand (2012) analysed the cost of malnutrition in health care settings. Similarly, studies also varied in relation to the data that were used to assess diet and costs. A small number of studies used national surveys of self-reported food intake (Doidge et al., 2012; Popkin et al., 2006), while others did not collect data on dietary patterns specifically but instead used readily calculated populationattributable fractions published elsewhere. For example, Rayner & Scarborough (2005) and Scarborough et al. (2011) drew on PAFs produced as part of the Global Burden of Disease studies (Ezzati et al., 2004; Murray & Lopez, 1997). The economic costs of unhealthy diets and low physical activity 7 All reviewed studies estimated disease-based direct health care expenditures and only Rice & Normand (2012) also included social care costs. All estimated costs applied to the general population; the only exception was the study by Bland et al. (2009), which estimated the costs for a sample of members of one health insurance plan in the United States (n = 7983 individuals). 2.3 What the evidence tells us: the economic costs of low physical activity Turning to the economic costs that can be associated with low physical activity, fourteen studies reported aggregate costs. These ranged from €29 million per annum in the Czech Republic (Kč 700 million) (Maresova, 2014) to €1.32–1.68 billion in the United Kingdom (£0.9–1.06 billion) (Scarborough et al., 2011; Allender et al., 2007), €1.3–7.9 billion in Canada (C$ 2.1–10.8 billion) (Katzmarzyk et al., 2000; Krueger et al., 2015; Janssen, 2012; Katzmarzyk, 2011), €1.8–4.9 billion in China (US$ 1.7–6.8 billion) (Popkin et al., 2006; Zhang & Chaaban, 2012) and €90.5 million–€57.7 billion in the United States (US$ 83.6 million–79 billion) (Garrett et al., 2004; Anderson et al., 2005; Carlson et al., 2015) (Table 2). Taking account of population, the estimated annual per capita health care costs ranged from €3 in the Czech Republic to €48 in Canada. In the United States alone, per capita health care cost estimates ranged from less than €1 in two studies up to €185. Only four studies also considered indirect costs in their estimates, and these varied from €3.7 per capita in China to between €127 and €224 in Canada. As illustrated in Table 2, and similar to studies analysing unhealthy diets, low physical activity or physical inactivity was defined differently across studies, which makes it difficult to compare estimates. Four studies defined physical inactivity as not meeting recommendations for moderate- and vigorous-intensity types of physical activity (Idler et al., 2015; Janssen, 2012; Maresova, 2014; Zhang & Chaaban, 2012), while Carlson et al. (2015) considered only moderate-intensity activity and Garrett et al. (2004) only vigorous-intensity activity. Four studies defined physical inactivity as not meeting the recommended duration (which varied across studies), regardless of intensity (Alter et al., 2012; Anderson et al., 2005; Bland et al., 2009; Martinson et al., 2003). Katzmarzyk et al. (2000), Garrett et al. (2004), Katzmarzyk (2011) and Krueger et al. (2015) only considered leisuretime physical inactivity, Popkin et al. (2006) only considered sedentary behaviour, Kuriyama et al. (2004) focused on walking and Ackerman et al. (2003) defined physical activity as participation in an exercise programme. Bachmann et al. (2015) and Peeters et al. (2014) conceptualised activity in terms of metabolic equivalents achieved while performing a physical activity, and Wang et al. (2004; 2005) used small increases in heart rate or heavy breathing induced by physical activity as a measure of activity. Allender et al. (2007) and Scarborough et al. (2011) did not define physical inactivity specifically. 8 Assessing the economic costs of unhealthy diets and low physical activity Country Annual economic costs of low physical activity (per capita*) Definition of low physical activity Perspective of cost estimation Population base Source Czech Republic €29 million (€2.8) Less than 150 minutes of moderate activity or less than 75 minutes of vigorous activity a week or less than 180 minutes of walking weekly or any combination resulting in less than 600- MET minutes a week on at least three days per week or no physical activity at all. Direct health care costs, not specified General population Maresova (2014) United Kingdom €1.32 billion (€22) Not defined Direct health care costs, not specified General population Scarborough et al. (2011) €1.68 billion (€28) Not defined Direct health care costs: inpatient and outpatient, primary care, pharmaceutical and net community care service expenditures General population Allender et al. (2007) Canada €1.3 billion (€43) Leisure-time energy expenditure of less than 12.6 kilojoules per kg bodyweight per day Direct health care costs, not specified General population Katzmarzyk et al. (2000) €4.3 billion (€127) Less than 150 minutes of moderate-vigorous physical activity per week Direct health care costs: hospital, drugs, physician care, care in other institutions, ‘additional’ direct health care expenditures Indirect productivity costs: income lost due to sickness absence, injury-related work disability and premature deaths before retirement General population Janssen (2012) €5.4 billion (€160) Leisure-time energy expenditure of <1.5 kcal per kg bodyweight per day Direct health care costs: hospital, drugs, physician care, care in other institutions, other professional fees, public health, health research and pre-payment administration costs Indirect productivity costs: income lost due to premature death and short-term and long-term disability General population Katzmarzyk (2011) €7.9 billion (€224) Average leisure-time energy expenditure of less than 1.5 kcal per kg per day over the past three months Direct health care costs: hospital, physician services, other health care professional services (excluding dental services), drugs, health research and ‘other’ health care expenditures Indirect productivity costs: income lost due to short-term disability, long-term disability and premature mortality (before retirement) General population Krueger et al. (2015) Table 2 Annual economic costs of low physical activity reported in the published literature The economic costs of unhealthy diets and low physical activity 9 Data on physical activity were typically self-reported (assessed through surveys). Janssen (2012) measured physical activity using accelerometers, and Bachmann et al. (2015) assessed cardiorespiratory fitness as a measure of habitual physical activity using a treadmill test. Four studies estimated indirect costs associated with low physical activity, in addition to health care costs (Janssen, 2012; Katzmarzyk, 2011; Krueger et al., 2015; Zheng et al., 2012). Analyses mostly applied to the general population, although six studies estimated costs for specific populations, such as members of a health plan in the United States (Ackermann et al., 2003; Anderson et al., 2005; Bachmann et al., 2015; Garrett et al., 2004; Idler et al., 2015; Wang et al., 2005). Idler et al. (2015) analysed the relationship between physical activity and health care and (parental) productivity costs among children aged 9 to 12 years. Country Annual economic costs of low physical activity (per capita*) Definition of low physical activity Perspective of cost estimation Population base Source China €1.8 billion (€1.4) Sedentary behaviour Direct health care costs, not specified General population Popkin et al. (2006) €4.9 billion (€3.7) Less than 30 minutes of moderate-intensity activity for five days per week or less than 20 minutes of vigorous-intensity activity for three days per week Direct health care costs: hospital, drugs, physician care, other professional services, public health, health research, prepayment administration costs Indirect productivity costs: income lost due to sickness absence, injury-related work disability and premature deaths before retirement General population Zhang & Chaaban (2012) The United States €90.5 million (€0.3) Less than 20 minutes of vigorous leisure-time physical activity for at least three days a week or no leisure-time physical activity at all Direct health care costs: inpatient and outpatient facility, professional, x-ray, laboratory and pharmaceuticals, including out-ofpocket payments Sample population (n = 1.5 million) Garrett et al. (2004) €212.4 million (€0.8) Less than four days of physical activity lasting 30 minutes or more or no physical activity at all Direct health care costs: hospital and professional fees, excluding pharmaceuticals Sample population (n = 200 000) Anderson et al. (2005) €26.5 billion (€98) Less than 30 minutes of activity increasing the heart rate for ≥5 times per week or less than 20 minutes of activity substantially increasing the heart rate for ≥3 times per week Direct health care costs: inpatient stays and outpatient visits, medications and home care, including out-of-pocket payments General population Wang et al. (2004) €57.7 billion (€185) Less than 150 minutes per week of moderate physical activity or none at all Direct health care costs: inpatient, and outpatient, emergency room, office-based, dental, vision and home health care, including prescription drugs General population Carlson et al. (2015) Note: * per capita costs calculated using United Nations population data (United Nations, 2015). 10 Assessing the economic costs of unhealthy diets and low physical activity Two studies reported the costs of low physical activity in combination with other risk factors. Alter et al. (2012) estimated the incremental health care costs that can be associated with obesity and additional risk factors, including low physical activity (described as sedentary) (data not shown in Table 2 as authors reported per capita costs only). They found that the cumulative additional costs attributable to overweight and obesity alone were small when compared with matched normal-weight adults. However, costs increased significantly with other risk factors. Thus, for obese individuals who were also physically inactive, health care costs exceeded those of normal-weight, healthy individuals over an 11.5-year period by around €3700 (C$ 4080; p = 0.003). Kuriyama et al. (2004) estimated that low physical activity increased monthly per capita health care costs among adults in Japan by 8%, from €172 for adults without lifestyle risk to €185, with further increases where low physical activity was combined with obesity (by 16.4%) (1995–2001). Idler et al. (2015) focused on children aged 9 to 12 years and found that low physical activity increased health care costs by €6 per child annually, but decreased the productivity costs (i.e. earnings lost due to parental absence from work) by €11 per physically inactive child. However, the relationship between physical activity and costs in this age group was not statistically significant. Similarly, Ackermann et al. (2003) found no significant difference in costs between physically active and inactive members of a health plan in the United States. Three studies analysed the comparative impact of unhealthy diets and low physical activity on cost. Popkin et al. (2006) and Scarborough et al. (2011) estimated that the cost associated with unhealthy diets exceeded that associated with low physical activity by a factor of 1.5 to 5. Bland et al. (2009) found that low physical activity but not unhealthy diet (as measured by low fruit and vegetable consumption) was significantly associated with higher short-term medical costs. The latter study collected primary data on diet and physical activity among members of a health plan in the United States to estimate medical costs, whereas both Popkin et al. (2006) and Scarborough et al. (2011) based their analyses on published data on population-attributable fractions and studies are not easily comparable. 2.4 Review of the evidence: a summary In summary, of the 30 studies reviewed, 27 found a significant association between diet and/or physical activity and costs, with unhealthy diets and low physical activity predictive of higher health care expenditure. The only exception was the study by Collins et al. (2011), which reported a healthy diet to be predictive of higher health care costs; the authors noted, however, that the findings of their study of women were likely confounded by charges incurred for routine screening services (e.g. cervical and breast cancer screening), with those The economic costs of unhealthy diets and low physical activity 11 with higher dietary index scores more likely to use these services than those with poorer scores. Three studies did not find a significant association between diet or physical activity and costs. Studies that did report costs associated with the two risk factors found the annual cost of unhealthy diets to range from €3 to €148 per capita and for low physical activity from €3 to €181 per capita. The highest health care cost estimates are equivalent to between 2% and 6% of health spending in the countries. The review shows that there is a very wide range of estimates, and these are very sensitive to the measures of diet and activity and the ways in which the studies were carried out. The next section reviews these methodological and measurement challenges in assessing the costs of unhealthy diets and low physical activity. Chapter 3 Estimating the economic costs of unhealthy diets and low physical activity is complex We have shown in the review of published studies that the estimated economic burden associated with unhealthy diets or low physical activity varies widely. Reasons for this variation include differences in the definition of what constitutes unhealthy diets or low physical activity; the methodological approach chosen (such as the method to calculate population-attributable fractions); and the range and types of costs considered. We discuss each aspect in turn. 3.1 Defining the concepts: how can we understand ‘unhealthy diet’ and ‘low physical activity’? Our evidence review illustrates that ‘unhealthy diets’ and ‘low physical activity’ have been conceptualised and interpreted differently, making a comparative assessment of available studies difficult. For example, definitions of unhealthy diets often refer to those high in specific nutrients such as saturated fats, salts or sugars, but growing evidence suggests that intakes of specific foods rather than macro- or micro-nutrients are most relevant for the development of chronic disease (Morgan, 2012; Mozaffarian et al., 2011). Table 3 Intake of major food groups associated with the lowest risks for chronic disease Food group Optimal intake levels (mean ± standard deviation) Fruits 300 ± 30 grams/day Vegetables 400 ± 40 grams/day Nuts/seeds 113.4 ± 11.3 grams/week Whole grains 100 ± 12.5 grams/day Seafood 350 ± 35 grams/week Unprocessed red meats 100 ± 10 grams/day Processed meats 0 Source: Micha et al., 2015. 14 Assessing the economic costs of unhealthy diets and low physical activity It is against this background that researchers have moved towards identifying and defining healthy diets based on recommended intakes of selected food groups. For example, Micha et al. (2015) described optimal consumption levels of selected food groups, based on probable or convincing evidence about the association of intake levels and the risk for coronary heart disease, stroke, type 2 diabetes and certain cancers. An unhealthy diet can be defined as one that does not meet the recommended intake levels of selected food groups shown in Table 3. Even with this approach there remains the problem of how to score different levels of deviation from optimal intake. Intakes of beneficial dietary factors tend to be positively correlated with each other and inversely correlated with those considered unhealthy. This correlation could lead to overestimates of the relative risk of each dietary factor and the total effect of dietary risks at the population level (GBD 2015 Risk Factors Collaborators, 2016). Instead of individually assessing the risks associated with selected food groups, an alternative approach is to examine dietary patterns. Such an approach considers the balance among all food groups, including those that are recommended for frequent consumption and those that are not. It also accommodates different eating patterns, so allowing for variation depending on cultural, ethnic or personal preferences, or the costs and availability of certain foods. Examples include the Healthy Eating Index (HEI), which measures adherence to the 2005 dietary guidelines in place in the United States, and the Alternate Healthy Eating Index (AHEI), which is based on foods and nutrients predictive of chronic disease risk (Chiuve et al., 2012). Diets which score highly on either the HEI or the AHEI were shown to be associated with a significant reduction in the risk of all-cause mortality, cardiovascular disease, cancer and type 2 diabetes by around 20%, highlighting their relevance for population health (Schwingshackl & Hoffmann, 2015). The AHEI is discussed further in the context of our proposed costing framework below. In contrast to diet, there is more consensus about what constitutes low physical activity. The World Health Organization (2010) has defined physical inactivity as the “absence of physical activity or exercise” (p. 53), and it recommends that adults meet the guidelines of at least 150 minutes of moderate-intensity or at least 75 minutes of vigorous-intensity aerobic physical activity throughout the week, in bouts of at least 10 minutes’ duration, in order to improve cardiorespiratory and muscular fitness and bone health, and to reduce the risk of chronic disease and depression (World Health Organization, 2010). Physical activity includes activities undertaken while working, playing, carrying out household chores, travelling, and leisure-time activities but is distinct from exercise. For measurement purposes, physical activities are classified into different categories, ranging from light to moderate to vigorous intensity, depending on the Estimating the economic costs of unhealthy diets and low physical activity is complex 15 amount of effort (i.e. kilocalories of energy) required to perform the activity. This effort is measured in terms of the Metabolic Equivalents (METs), which is the ratio of a person’s metabolic rate while doing the activity relative to their resting metabolic rate, and the intensity is then assessed by multiples of METs spent on a given activity (Box 2) (World Health Organization, 2004). However, as we have seen in the preceding section, studies that have assessed the costs of low physical activity vary in their use of intensity thresholds, or intensity is not taken account of altogether, which makes it difficult to interpret the evidence as highlighted. Interpretation of the evidence is further complicated where low physical activity is conceptualised as sedentary behaviour, as for example in the study by Popkin et al. (2006). Individuals that are not meeting physical activity guidelines may be wrongly classified as ‘sedentary’. Defined as “any waking behaviour with low energy expenditure (≤1.5 metabolic equivalents) while in a sitting or reclining posture” (p. 540) (Sedentary Behaviour Research Network, 2012), growing evidence suggests that prolonged sedentary time is independently associated with deleterious health outcomes independent of the level of physical activity (Biswas et al., 2015). Disentangling these relationships will be important, with for example Ekelund et al. (2016) showing that moderate levels of physical activity reduced the increased mortality risks associated with high sitting time. Box 2 Examples of moderate-intensity and vigorous-intensity physical activity One MET is equivalent to an energy consumption of 1 kilocalorie per kilogram of bodyweight per hour (i.e. energy cost of sitting quietly). Examples of moderate-intensity physical activity Equivalent to approximately 3–6 METs; requires a moderate amount of effort and noticeably accelerates the heart rate Examples of vigorous-intensity physical activity Equivalent to approximately >6 METs; requires a large amount of effort and causes rapid breathing and a substantial increase in heart rate • Brisk walking • Dancing • Gardening • Housework and domestic chores • Traditional hunting and gathering • Active involvement in games and sports with children/walking domestic animals • General building tasks (e.g. roofing, thatching, painting) • Carrying/moving moderate loads (<20kg) • Running • Walking/climbing briskly up a hill • Fast cycling • Aerobics • Fast swimming • Competitive sports and games (e.g. traditional games, football, volleyball, hockey, basketball) • Heavy shovelling or digging ditches • Carrying/moving heavy loads (>20kg) Source: World Health Organization, 2004. 16 Assessing the economic costs of unhealthy diets and low physical activity 3.2 Costing studies differ in key assumptions, influencing estimates for the economic burden of unhealthy diets and low physical activity As noted earlier, reviewed studies considered different numbers of diseases to derive estimates of the economic burden that can be associated with unhealthy diets or low physical activity. The most common conditions considered include coronary heart disease, stroke, type 2 diabetes and colorectal and breast cancer, with Garrett et al. (2004) also adding mood and anxiety disorders as directly related to individual physical activity patterns in adults. The range and combination of disease groups varied among studies, as did the main approaches to estimating costs, including the use of population-attributable fractions; consideration of lag times between exposure to a given risk factor, the development of disease and ensuing cost; and the conceptualisation of economic cost itself, which we briefly discuss here. 3.2.1 Use of population-attributable fractions Fourteen reviewed studies used population-attributable fractions (PAFs) to assess the contribution of unhealthy diets or low physical activity to a range of diseases (or death) as a basis to estimate the economic burden that can be attributed to either risk factor. As noted earlier, the PAF generally refers to the proportion of cases for a given outcome of interest that can be attributed to a given risk factor among the entire population. Calculation of PAFs commonly uses a relative risk that has been adjusted for potential confounders of the association between risk factors and outcomes, such as age or sex, and the prevalence of exposure in the population under investigation (the partially-adjusted method). However, where confounding or effect modification affects the relative risk, the estimate of the attributable fraction is potentially biased even if the relative risk has been adjusted for confounding (Benichou, 2001). This is likely to be the case for the unhealthy diet and low physical activity–disease relationships, with multiple factors, in addition to age and sex, such as family history or physiological risk factors such as weight found to confound the association (Laaksonen et al., 2009; Li et al., 2015; Montonen et al., 2005). Baliunas (2011) compared the partially-adjusted method for estimating population-attributable fractions with the fully-adjusted approach, which stratifies the relative risk according to confounder or effect modifier. The comparison was applied to mortality from lung cancer, ischaemic heart disease, chronic obstructive pulmonary disease and cerebrovascular disease related to smoking. It found that the partially-adjusted method overestimated the attributable fractions by Estimating the economic costs of unhealthy diets and low physical activity is complex 17 10%. The majority of studies reviewed in this volume have used the partiallyadjusted method, and cost estimates are therefore likely to be biased, although the direction of the bias is not clear. The use of a fully-adjusted method would reduce the risk of over- or under-estimating the ‘true’ association between risk factors and costs. In addition, population-attributable fractions used in the reviewed studies varied widely. This is illustrated further in Table 4 for PAFs used for costing studies of low physical activity. This variation reflects, to great extent, the source of PAFs and whether they were adjusted for the population under investigation. For example, Allender et al. (2007) and Scarborough et al. (2011), in their analyses of the economic burden of ill health that can be attributed to low physical activity in the United Kingdom in the mid-2000s, used PAFs that were produced in the context of the first Global Burden of Disease study for the western European population (Murray & Lopez, 1997). There is uncertainty about the degree to which these PAFs are applicable to the country they were used for (in this case the United Kingdom) as PAFs should take account of the underlying prevalence of a risk factor in the population under study. Although a sensitivity analysis by Scarborough et al. (2011) revealed little impact of the choice of PAFs on the cost estimates for low physical activity, population-specific PAFs remain preferable to increase the accuracy of estimates. 3.2.2 Consideration of time lags between risk factor exposure, disease development and associated costs Reviewed studies tended to apply population-attributable fractions using the prevalence of a given risk factor (unhealthy diets, low physical activity) in a given Table 4 Population-attributable fractions used in costing studies: low physical activity Ischaemic heart disease (%) Stroke (%) Diabetes type 2 (%) Breast cancer (%) Colon cancer (%) Katzmarzyk et al. (2000) 36 20 20 11 20 Garett et al. (2004) 31 31 18 19 31 Allender et al. (2007) 23 – 15 11 16 Katzmarzyk (2011) 18 23 20 13 17 Scarborough et al. (2011) 23 12 15 11 – Janssen (2012) 26 (m)* 27 (f) 25 (m)* 26(f) 38 (m)* 29 (f) 15 (f)* 24 (m, f)* Zhang & Chaaban (2012) 12 16 14 – – Maresova (2014) 7 3 4 3 4 Note: * m – males; f – females 18 Assessing the economic costs of unhealthy diets and low physical activity year and estimating the related disease costs incurred in the same year (Allender et al., 2007; Garrett et al., 2004; Janssen, 2012; Katzmarzyk, 2011; Katzmarzyk et al., 2000; Maresova, 2014; Popkin et al., 2006; Scarborough et al., 2011; Zhang & Chaaban, 2012). Such an approach ignores the time lag between exposure to the given risk factor and the development of disease. For example, Weyer et al. (1999) found that it takes approximately five years between the initial normal glucose-tolerant stages and the development of clinically verified type 2 diabetes. Where the development of diabetes can be linked to unhealthy diet or low physical activity at the outset, associated economic costs would thus be expected to emerge only after five years at the earliest. Therefore, accurately estimating the cost that can be attributed to unhealthy diets or low physical activity would need to take account of the time it takes for the natural progression of exposure to disease. 3.3 The nature and range of costs considered is likely to underestimate the ‘true’ economic burden of unhealthy diets and physical activity Reviewed studies tended only to consider costs that can be associated with primary disease outcomes, such as type 2 diabetes or coronary heart disease. Yet, a full costing would also need to take account of complications associated with the primary disease outcome. Considering, for example, type 2 diabetes: data from the UK National Diabetes Audit found that within a one-year followup period between 2011–12 and 2012–13, people with type 2 diabetes were significantly more likely than those without diabetes to be admitted to hospital for complications such as angina, at 135.1%, heart failure (121.1%), heart attack (87.6%) or stroke (59.1%) (Health and Social Care Information Centre, 2015). Disregarding the costs associated with the development of complications will inevitably underestimate the true economic burden that can be associated with a given risk factor, but this also assumes that it is possible to quantify the contribution of the risk factor under consideration to the observed complication. Furthermore, the majority of reviewed studies considered only the direct health care costs, further underestimating the true economic burden that can be associated with unhealthy diets or low physical activity (see also Box 3). For example, three Canadian studies that had analysed the costs associated with low physical activity found indirect costs caused by lost productivity to be about twice as high as direct health care costs, and together these accounted for between 0.4% and 0.6% of gross domestic product (GDP) (Janssen, 2012; Katzmarzyk, 2011; Krueger et al., 2015). It is difficult to generalise from these studies to other settings and it will be important to broaden existing costing estimates to also capture indirect costs in order to better understand the size of the burden that can be associated with the two risk factors. Estimating the economic costs of unhealthy diets and low physical activity is complex 19 3.4 Conceptual and methodological challenges of estimating the economic costs of unhealthy diets and low physical activity: a summary This section reviewed the measurement and methodological issues in assessing the economic burden of unhealthy diet and low levels of activity. We note that the measurement of what constitutes an ‘unhealthy’ diet is made more difficult by there being positive effects of some foods, negative effects of others and interactions between the effects of different foods. Calibrating the extent of deviation from optimal consumption and the effects of this deviation is difficult. It is also clear that the context should be taken into account in terms of other population characteristics. Box 3 Conceptualizing economic costs Costing methodology generally distinguishes direct, indirect and intangible costs, although these have been conceptualized in different ways (Johnston et al., 1999). Direct costs typically refer to costs of health care services as they relate to the prevention, diagnosis and treatment of a given condition, such as inpatient or outpatient care, rehabilitation, community health services and pharmaceuticals; direct costs may also include social care costs where relevant (Suhrcke et al., 2008). Costs considered typically include those associated with service utilization, that is the use of a particular service over time (for example, physician visits, emergency room or accident and emergency department visits, hospital (re-)admissions, length of hospital stay, number of hospital days), and the actual cost of providing a particular service (health, nursing, social care), including the costs of procedures, therapies and medications, or expenditure, that is, the amount of money paid for the services, and from fees (the amount charged), regardless of cost. There are also direct costs borne directly by people using the services; these include transportation costs, out-of-pocket payments for medications and devices, special diets and home help. Indirect costs typically refer to productivity losses to society because of ill health or its treatment (Koopmanschap et al., 1995). Commonly considered dimensions include presenteeism, absenteeism, early retirement and premature mortality. Presenteeism costs refer to the value of productivity losses accrued by employees who are present at work but are unable to work at full capacity because of illness (Johns, 2010), measured as the value of reduced work output, errors on the job and failure to meet the company’s production standards (Schultz & Edington, 2007). Absenteeism costs refer to those costs incurred because of absence from work because of ill health. Costs related to early retirement refer to potential earnings forgone by not working to the formal age of retirement, while premature mortality refers to the loss of economic output calculated as the income that individuals who die before or at a given age will lose over the period of remaining labour market participation (under or to age 65). Finally, intangible costs generally describe the psychological burden placed on patients and their carers, including pain, bereavement, anxiety and suffering (Suhrcke et al., 2008). 20 Assessing the economic costs of unhealthy diets and low physical activity While there is more consensus about the measurement of physical activity, similar issues arise in terms of the independent effects of moderate and vigorous activity and sedentary behaviour, but also the interactions between these. Studies take broader and narrower perspectives in terms of what costs are included, with some limited to formal health care costs, and others aiming to take a more societal view. While current evidence makes it difficult to make accurate comparisons, it is likely that much of the economic burden comes from non-health care costs, especially from effects on productivity, absenteeism, presenteeism and other indirect costs. Chapter 4 Taking available approaches to determining the economic costs of unhealthy diets and low physical activity further: a proofof-concept approach applied to five European countries This chapter develops a bottom-up framework for assessing the costs of unhealthy diets and low physical activity using a disease-based approach. Building on the insights from the evidence review and critical appraisal of measurement and methodological issues in the preceding sections, this section builds a framework for assessing costs of unhealthy diets and low physical activity. It then applies this framework as a proof-of-concept to demonstrate the feasibility of undertaking a comprehensive, bottom-up cost assessment that addresses some of the identified limitations of existing costing studies. The evidence review demonstrated that there are essentially two approaches to estimating the economic burden associated with unhealthy diets and low physical activity. One involves the use of a disease-based approach, based on the observation that the two risk factors have been identified to be causally related to major chronic diseases such as coronary heart disease, stroke, type 2 diabetes and selected cancers (Lee et al., 2012; Micha et al., 2015), which have been associated with considerable costs to health care systems and wider society through productivity losses (European Commission, 2014). About half of the reviewed studies adopted the disease-based approach, typically using the population-attributable fraction to quantify the contribution of the individual risk factor (unhealthy diet, low physical activity) to the burden of a given disease or death and the associated economic costs. An alternative method involves a generic, non-disease-based approach that uses individual data on unhealthy diets or low physical activity data and links these to (health care) cost data, regardless of the type of disease or diagnosis. This second approach requires availability of and access to individual-level data on dietary and physical activity patterns along with data on health care use and productivity. Such data are difficult to access even within individual country settings, let alone in multiple countries. 22 Assessing the economic costs of unhealthy diets and low physical activity We use a disease-based approach to estimate the costs that can be associated with unhealthy diets and low physical activity. Diseases associated with these lifestyle factors include coronary heart disease, stroke, type 2 diabetes and selected cancers (Box 4) (Lee et al., 2012; Micha et al., 2015). As a test of concept we apply the framework using type 2 diabetes. This was chosen because of the strong association of type 2 diabetes with either risk factor, as shown in the literature. Diabetes is already an important public health issue but perhaps more importantly, globally the number of people with diabetes has doubled during the past 20 years, making it a growing challenge for health systems (Zimmet et al., 2014). Estimates for the early 2000s place the costs for type 2 diabetes in the EU at some 7.4% of total health care expenditure, compared to 12% for cardiovascular diseases (Muka et al., 2015). More recently, expenditure on diabetes was estimated to account for some 9% of total health care expenditure in the European region in 2015 (International Diabetes Federation, 2015). Taking a disease-based approach, we used population-attributable fractions to provide population-level quantitative estimates for the economic costs that Box 4 Health risks associated with unhealthy diets and low physical activity A wide range of studies have presented convincing or probable evidence for an association between the intake of selected foods and food groups and major chronic conditions (Micha et al., 2015). For example, coronary heart disease has been linked to intake of fruit and vegetables, nuts and legumes, whole grains, fish, and red and processed meat (Afshin et al., 2014; Aune et al., 2016; Boeing et al., 2012; Micha et al., 2010; Zheng et al., 2012); stroke to fruit and vegetables, nuts and legumes, fish, and red and processed meat (Afshin et al., 2014; Boeing et al., 2012; Chowdhury et al., 2012, Micha et al., 2010; Zheng et al., 2012), type 2 diabetes to nuts and legumes, whole grains, red and processed meats (Afshin et al., 2014; Aune et al., 2016; Micha et al., 2010; Zheng et al., 2012) and selected cancers to fruit and vegetables, nuts and legumes, whole grains, fish, and red and processed meats (Aune et al., 2016; Boeing et al., 2012; Bouvard et al., 2015; World Cancer Research Fund and American Institute for Cancer Research, 2007). Based on the available evidence, Micha et al. (2015) concluded that “even modest dietary changes are associated with meaningful reductions in [cardiovascular disease] morbidity and mortality, type 2 diabetes [and] specific cancer sites” (p. 2), along with major risk factors, such as hypercholesterolaemia, hypertension and obesity. These conditions have also been linked to physical activity, with Warburton et al. (2010) demonstrating that low physical activity was associated with an increased risk for cardiovascular disease, stroke, hypertension, colon and breast cancer and type 2 diabetes. The authors further demonstrated that higher levels of physical activity reduced the risk for premature all-cause mortality. A proof-of-concept approach 23 can be associated with unhealthy diets and low physical activity. In the light of the reviewed evidence, we adapted the PAF methodology by (i) using a fulladjustment formula that takes account of confounding between unhealthy diets and low physical activity and associated diseases; (ii) incorporating a time perspective that takes account of the natural progression from risk factor exposure to the development of disease; (iii) applying the PAFs to costs based on data on incidence rather than prevalence; (iv) estimating the direct costs that can be associated with the primary outcome and those associated with complications using the annual incidence rates for complications; and (v) considering indirect costs of productivity losses as a consequence of absenteeism, presenteeism, work disability, early retirement and premature mortality. The following sections provide a sample computation of the economic costs of unhealthy diets and low physical activity in five European countries: France, Germany, Italy, Spain and the United Kingdom, using type 2 diabetes and its complications as the primary outcome. 4.1 Diabetes as an outcome of unhealthy diets and low physical activity As noted above, we chose type 2 diabetes as the primary outcome because of the convincingly strong evidence of its association with unhealthy diets and low physical activity (Afshin et al., 2014; Micha et al., 2010; Warburton et al., 2010; Zheng et al., 2012). Diabetes mellitus, commonly referred to as diabetes, is a group of metabolic diseases characterized by high levels of sugar in the blood for a prolonged period of time. The most common form of diabetes is type 2, which typically occurs in adults, although it is increasingly seen in young people, including children (Zimmet et al., 2014). It results primarily from insulin resistance, and at later stages the pancreas may also fail to produce sufficient levels of insulin. Insulin resistance in type 2 diabetes is associated mainly with high bodyweight and low physical activity. The role of genetic factors in the development of type 2 diabetes tends to be small compared to lifestyle and clinical factors such as increased bodyweight, elevated liver enzymes, current smoking status and reduced measures of insulin secretion and action (Lyssenko et al., 2008). Factors that increase the likelihood of developing diabetes include high consumption of sugar-sweetened drinks (Malik et al., 2010a; Malik et al., 2010b), of saturated and trans-fatty acids (Risérus et al., 2009) and of refined grains such as white rice (de Bakker et al., 2012), which, when metabolised to glucose, increase blood sugar levels. Low physical activity has also been linked to type 2 diabetes through causing insulin resistance (Warburton et al., 2010) (Figure 1). Figure 1 Relationship between unhealthy diets and low physical activity with type 2 diabetes and associated costs Source: authors. Decreased cellular sensitivity to insulin, decreasing uptake of glucose from the blood, insufficiently compensated for by the production of insulin (increased insulin production at the early stages) Genetic factors Lifestyle High in saturated fats, trans-fatty acids and total fat intake Fat metabolised to high levels of glucose Alters fatty acid composition of cell membranes Affects gene expression by binding to cell receptors Low physical activity Unhealthy diets Impaired glucose tolerance Decreased cellular sensitivity to insulin, decreasing uptake of glucose from the blood Decreased capillary proliferation in muscles, insulinsensitive type of muscle fibres and muscle mass Ischaemic heart disease (IHD) Retinopathy Renal failure Amputation Type 2 diabetes mellitus Fatal and non-fatal myocardial infarction (MI) Diabetes-related mortality Congestive heart failure Stroke Direct health care costs (Inpatient care, outpatient care, medications, rehabilitation, long-term care, other direct expenses) Indirect (productivity) costs (Absenteeism, presenteeism, reduced productivity among the unemployed population, work disability, early retirement and premature mortality) A proof-of-concept approach 25 4.2 The principal approach used in this study to estimate the economic costs that can be associated with unhealthy diets and low physical activity The identification of the economic costs that can be associated with unhealthy diets and low physical activity as conceptualised in this study involves five principal steps: 1. determining the prevalence of unhealthy diets and low physical activity among the general population and estimating the prevalence in populations eventually developing the disease using adjustment factors; 2. determining the population-attributable fraction, or the proportion of cases attributable to unhealthy diets and low physical activity; 3. estimating the proportion of incident diabetes cases in future year X that can be attributed to present unhealthy diets and low physical activity patterns; 4. estimating the average annual per patient health care costs that can be associated with diabetes and with diabetic complications to yield diabetes-related direct costs; and 5. estimating the indirect costs that can be associated with diabetes attributable to unhealthy diets and low physical activity. We discuss these steps in turn. 4.2.1 Determining the prevalence of unhealthy diets and low physical activity Our proposed approach requires knowledge of the prevalence of the risk factor (here: unhealthy diets, low physical activity) among populations that eventually develop the outcome (here: type 2 diabetes), rather than among the general population, in order to enhance the accuracy of the estimation of costs that can be associated with the given risk factor. For low physical activity, we drew on work by Lee et al. (2012) who calculated adjustment factors for different outcomes (coronary heart disease, type 2 diabetes, breast and colon cancer, and those who died) to identify the added extent to which low physical activity occurred in people who eventually developed the outcome in question compared to the general population. Lee et al. (2012) illustrate this with an example from the Shanghai Women’s Health Study, where the prevalence of low physical activity in all women at baseline was 45.4% compared to 51.6% among women who eventually died, yielding an adjustment factor of 1.14 (51.6/45.4 = 1.14). They then calculated such an adjustment factor for a large number of original studies, and, for type 2 diabetes, derived a factor of 1.23 after averaging estimates across studies (for comparison, adjustment factors for coronary heart disease or colon 26 Assessing the economic costs of unhealthy diets and low physical activity cancer were 1.20 and 1.22, respectively, and for breast cancer, it was 1.05). We used the adjustment factor of 1.23 to estimate the prevalence of low physical activity among people who will eventually develop diabetes in France, Germany, Italy, Spain and the United Kingdom. Data on prevalence for low physical activity in the five countries were derived from the WHO Global Health Observatory database (World Health Organization, 2016). According to this source, the proportion of the general population who were not active, or who did not meet the recommendations of at least 150 minutes per week of moderate-intensity physical activity or at least 75 minutes per week of vigorous-intensity physical activity in 2010 were: 23.8% in France, 21.1% in Germany, 33.2% in Italy, 30.5% in Spain and 37.3% in the United Kingdom. Applying the adjustment factor of 1.23 to these rates yielded estimated prevalence rates of 29.3% in France, 26% in Germany, 40.8% in Italy, 37.5% in Spain and 45.9% in the United Kingdom among populations who will eventually develop diabetes. In contrast to low physical activity, determining the prevalence rate for unhealthy diets is more complex given the limited availability of appropriate data. We drew on the alternate healthy eating index (AHEI) 2010 developed by Chiuve et al. (2012). The AHEI assesses components of the diet by assigning scores ranging from 0 (worst) to 10 (best) (Table 5); scores between 67 and 110 are rated as adherence to what has been described as a healthy diet that emphasises high intakes of whole grains, polyunsaturated fatty acids, nuts and fish, and reductions in red and processed meats, refined grains and sugar-sweetened beverages. As noted above, studies found that diets that scored highly on the AHEI were shown to be associated with a significant reduction in the risk of all-cause mortality, cardiovascular disease, cancer and type 2 diabetes by around 20% (Schwingshackl & Hoffmann, 2015). We used the European Food Safety Authority (EFSA) Database to estimate the prevalence of unhealthy diets in the five countries under study. The EFSA contains information on the mean consumption of over 1500 food items in European countries, based on data from national dietary surveys (European Food and Safety Authority, 2015). Based on these data, we derived the mean intake of AHEI food groups for each country (Appendix 3) and allocated AHEI scores as shown in Table 6. Assuming that the figures shown in Appendix 3 and Table 6 are representative of the five countries as a whole, we estimate the proportions of those following an unhealthy diet to be 44% of the general population in France, 25% in Germany, 33.9% in Italy, 34.6% in Spain and 26.5% in the United Kingdom, as assessed by scores of 43% (France), 34% (Germany), 58% (Italy), 48% (Spain) and 37% (the United Kingdom) according to the 2010 alternate healthy eating index. A proof-of-concept approach 27 Unlike for low physical activity, we were unable to identify studies that provide adjustment factors that would allow assessment of the degree to which unhealthy diet is present in cases of the outcome compared to the overall population. Jacobs et al. (2015) estimated approximate adjustment factors based on a cohort of white Americans (which formed part of a larger multi-ethnic cohort in Hawaii) with poor dietary patterns assessed using the 2010 alternate healthy eating index (31 864 individuals), of whom 7.1% eventually developed diabetes (2274 subjects). The prevalence of unhealthy diets in the study population was 61.3% Table 5 The quality of diets according to the alternate healthy eating index (AHEI) 2010 Component Criteria for minimum score (0) Criteria for maximum score (10) Serving size for food components Vegetables1 (servings per day) 0 ≥5 0.5 cup or 1 cup of green leafy vegetables Fruit2 (servings per day) 0 ≥4 1 medium piece or 0.5 cup of berries Whole grains (grams per day) Women 0 75 Men 0 90 Sugar-sweetened beverages and fruit juice (servings per day) ≥1 0 8oz. Nuts and legumes (servings per day) 0 ≥1 1oz. of nuts or 1 tablespoon (15 ml) of peanut butter Red and processed meat (servings per day) ≥1.5 0 4oz. of (red) unprocessed meat or 1.5oz. of processed meat Trans-fats (% of energy) ≥4 ≥0.5 Omega-3 fatty acids (EPA + DHA) (mg per day) 0 250 Equivalent to two 4oz servings per week Polyunsaturated fatty acids (% of energy) ≤2 ≥10 Sodium (mg per day) Women ≥3 337 <1 112 Men ≥5 271 ≤1 612 Alcohol (drinks per day) Women ≥2.5 0.5–1.5 One drink equivalent to 4oz. wine, Men 12oz. beer or 1.5oz. liquor ≥3.5 0.5–2.0 TOTAL 0 110 Source: adapted from Chiuve et al., 2012. Notes: 1 Excludes potatoes (including french fries) because they are not associated with lower risk of chronic disease in epidemiologic studies; 2 whole fruits only. 28 Assessing the economic costs of unhealthy diets and low physical activity Dietary component AHEI recommended intake per day for maximum score (10 points) MEAN INTAKE PER DAY AHEI SCORE France (2007) Germany (2007) Italy (2005– 6) Spain (2009) United Kingdom (2008) France (2007) Germany (2007) Italy (2005–6) Spain (2009) United Kingdom (2008) Vegetables ≥5 cups green leafy vegetables 0.09 cups 0.12 cups 0.20 cups 0.18 cups 0.03 cups 0 0 0 0 0 Fruit ≥453.44g berries 10.97g 13.66g 2.89g 6.35g 10.81g 0 0 0 0 0 Whole grains ≥82.5g * 12.2g 0.49g 35.29g 6.84g 2.80g 1 0 4 0 0 Sugar-sweetened beverages and fruit juice ≤0.79oz. 4.19oz. 12.05oz. 1.98oz. 4.61oz. 7.91oz. 5 0 8 5 1 Nuts ≥1oz. 0.04oz. 0.10oz. 0.04oz. 0.07oz. 0.04oz. 0 1 0 0 0 Processed meat ≤0.224oz. 1.33oz. 1.76oz. 1.05oz. 1.73oz. 1.06oz. 5 3 6 3 6 Trans-fat ≤0.5% of energy intake (equivalent to ≤1.39g for a 2 500kcal diet per day) 63.5g 68.57g 30.23g 48.31g 33.48g 0 0 0 0 0 Long-chain (n-3) fats (EPA + DHA) ≥0.25g 21.39g 13.76g 31.05g 57.31g 21.1g 10 10 10 10 10 PUFA ≥10% of energy intake (equivalent to ≥27.78g for a 2 500kcal diet per day) 10.92g 2.93g 36.63g 32.6g 1.46g 3 0 10 10 0 Sodium ≤1.36g * 1.5g 0.01g 0.01g 0.17g 0.08g 9 10 10 10 10 Alcohol ≤7oz. wine, ≤21oz. beer, ≤2.6oz. liquor * 2.75oz. 6.33oz. 2.50oz. 3.12oz. 6.43oz. 10 10 10 10 10 TOTAL 43 34 58 48 37 Table 6 Assignment of AHEI scores on mean dietary intakes of AHEI food groups in France, Germany, Italy, Spain and the United Kingdom Note: * Median of recommended intakes for men and women A proof-of-concept approach 29 and among those eventually developing diabetes, the proportion of those with unhealthy diets was 67.1%; this equates to an adjustment factor of 1.09, which we used in our study. 4.2.2 Determining the population-attributable fraction, or the proportion of cases attributable to unhealthy diets and low physical activity Determining the population-attributable fraction first requires assessment of the relative risks for developing diabetes that can be associated with unhealthy diets and low physical activity. We drew on work by Li et al. (2015) who prospectively assessed the joint association of birth weight and five behavioural risk factors (smoking status, daily alcohol consumption, body mass index, dietary patterns and physical activity) in adulthood with incident type 2 diabetes based on a cohort of almost 150 000 male and female health professionals (median age 45 years), who were followed up for a period of 20–30 years (median follow-up: 24 years) (see also Appendix 4 for a justification of using the study by Li et al.). During the follow-up period, 11 709 new cases of type 2 diabetes were reported, equating to an incidence rate of 7.8%. Table 7 shows the adjusted relative risks for different levels of exposure to unhealthy diets or low physical activity as estimated by Li et al. (2015). Table 7 Relative risks for diabetes related to unhealthy diets and low physical activity estimated by Li et al. (2015) Risk factor Risk factor category AHEI score (≥67% classified as healthy diet) Level of exposure 0–22% 23–44% 45–66% ≥67% Relative risk 1.15 1.06 1.02 1.00 (ref) Low physical activity (moderate-to-vigorous intensity physical activity in hours/week) Level of exposure 0 0.01–1.0 1.0–3.5 ≥3.5 Relative risk 1.28 1.19 1.03 1.00 (ref) Source: Li et al., 2015. Table 8 provides a summary overview of the various estimates for the prevalence of unhealthy diets and low physical activity in the populations eventually developing diabetes in France, Germany, Italy, Spain and the United Kingdom described above and the associated relative risks for incident diabetes that can be associated with these risk factors as derived from Li et al. (2015). As the data on low physical activity among the general population derived from the WHO Global Health Observatory Database described above do not quantify the average weekly hours of physical activity performed by individuals, we used data from the 2013 Eurobarometer survey (TNS Opinion & Social, 2014). It assessed the 30 Assessing the economic costs of unhealthy diets and low physical activity frequency and duration of sport and physical activity among populations in the EU and quantified the proportions of the survey population who performed moderate- and vigorous-intensity physical activity. This showed that the majority of inactive individuals in the five countries studied performed physical activity for 0.01 to 1 hour per week, which gives a relative risk for the development of diabetes that can be associated with low physical activity of 1.19 (Table 7). Table 8 Estimated prevalence rates of unhealthy diets and low physical activity in the populations eventually developing diabetes and relative risks for incident diabetes in France, Germany, Italy, Spain and the United Kingdom Prevalence in the population eventually developing diabetes Relative risks for incident diabetes associated with unhealthy diets and low physical activity Unhealthy diets (%) Low physical activity (%) Unhealthy diets Low physical activity France 48.0 29.3 1.06 1.19 Germany 27.3 26.0 1.06 1.19 Italy 37.0 40.8 1.02 1.19 Spain 37.7 37.5 1.02 1.19 United Kingdom 28.9 45.9 1.06 1.19 Estimates shown in Table 8 allow for the computation of the fully-adjusted population-attributable fraction for each setting as illustrated in Box 5. 4.2.3 Estimating the proportion of incident diabetes cases in future year X that can be attributed to present unhealthy diet and low physical activity patterns We have noted that there is typically a time lag between exposure to a risk factor and the development of subsequent disease that can be (part-)attributed to the risk factor. In the case of type 2 diabetes, Weyer et al. (1999) analysed the major metabolic abnormalities occurring in the development of diabetes, finding that insulin action and secretion significantly decrease early in the development of diabetes, during the transition from normal glucose tolerance to impaired glucose tolerance, and that these early changes are preceded by increases in bodyweight. As impaired glucose tolerance progresses to established diabetes, insulin action and secretion further deteriorate, accompanied by an increase in the endogenous glucose output, and these changes are associated with further increases in bodyweight. Weyer et al. observed an average interval of 1.8 (± 0.8) years between normal glucose tolerance and impaired glucose tolerance, while the interval between impaired glucose tolerance and established diabetes was 3.3 (± 1.4) years. On average, then, it would require some 5.1 (± 1.4) years for normal glucose tolerance to develop into type 2 diabetes. A proof-of-concept approach 31 Weyer et al. (1999) further noted that the early deteriorations in insulin secretion and action are distinct from and additive to abnormalities that overweight and obese individuals may experience while normal glucose-tolerant. With the progression from impaired glucose tolerance to diabetes, the defects worsen in parallel with an increase in endogenous glucose output, implying that being overweight or obese prior to the observed stages does not alter the interval between normal glucose tolerance to impaired glucose tolerance and diabetes and that an interval of about five years is therefore consistent. Based on these observations, in this study we assume a time lag of five years between exposure to the risk factors (unhealthy diets, low physical activity) and the eventual development of type 2 diabetes, starting from a 2015 population Box 5 Computing the fully-adjusted population-attributable fraction (PAF) The computation of the fully-adjusted population-attributable fraction (PAF) follows the following principal formula: in which: Pd: prevalence of the risk factor in the population eventually developing the disease RR: The relative risk for a certain disease associated with the risk factor compared to absence of risk factor, adjusted for confounding variables Applying this formula to the prevalence and relative risk estimates for each factor in each country yields the country- and risk-factor-specific PAFs as follows: Unhealthy diet Low physical activity France = 3% 0.48 × (1.06−1) 1.06 PAF (%) = = 5% 0.293 × (1.19−1) 1.19 PAF (%) = Germany = 2% 0.273 × (1.06−1) 1.06 PAF (%) = = 4% 0.26 × (1.19−1) 1.19 PAF (%) = Italy = 1% 0.37 × (1.02−1) 1.02 PAF (%) = = 7% 0.408 × (1.19−1) 1.19 PAF (%) = Spain = 1% 0.377 × (1.02−1) 1.02 PAF (%) = = 6% 0.375 × (1.19−1) 1.19 PAF (%) = United Kingdom = 2% 0.289 × (1.06−1) 1.06 PAF (%) = = 7% 0.459 × (1.19−1) 1.19 PAF (%) = Pd (RRadj−1) RRadj PAF (%) = 32 Assessing the economic costs of unhealthy diets and low physical activity that is normal glucose tolerant. We drew on estimates of cumulative incidence of diabetes in the 53 countries of the WHO European region from 2010 to 2020 by Webber et al. (2014) to estimate the annual incidence of diabetes in 2020 in the five countries under study. We then applied the country-specific PAFs for unhealthy diets and low physical activity (Box 5) to the estimated 2020 incidence of diabetes, which gives the total number of incident cases in 2020 that can be attributed to 2015 levels of unhealthy diet and low physical activity (Table 9). Table 9 Estimated incident diabetic cases in 2020 attributable to unhealthy diet and low physical activity patterns in 2015 Country Risk factor Populationattributable fraction (%) Projected incidence in 2020 Number of incident diabetes cases attributable to risk factor France Unhealthy diets 3 151 590 4 548 Low physical activity 5 7 580 TOTAL 12 128 Germany Unhealthy diets 2 242 571 4 851 Low physical activity 4 9 703 TOTAL 14 554 Italy Unhealthy diets 1 151 306 1 513 Low physical activity 7 10 591 TOTAL 12 104 Spain Unhealthy diets 1 106 929 1 069 Low physical activity 6 6 416 TOTAL 7 485 United Kingdom Unhealthy diets 2 118 393 2 368 Low physical activity 7 8 288 TOTAL 10 656 4.2.4 Estimating the average annual per patient health care costs that can be associated with diabetes and with diabetic complications to yield diabetes-related direct costs To estimate the direct health care costs associated with the incident diabetes cases attributable to unhealthy diets and low physical activity, we used data from the 2015 Diabetes Atlas, which provides estimates for the average annual per patient diabetes cost in the studied countries in 2015 (International Diabetes Federation, 2015). We adjusted these figures by the average annual growth rate in per capita health spending during 2005 to 2013 in each of the five countries (OECD, 2015). This yielded estimated per patient costs A proof-of-concept approach 33 in 2020 to range from €3314 in Spain to €6810 in France. Based on these figures, the direct diabetes-related health care costs that can be associated with unhealthy diets and low physical activity in the five countries in 2020 are estimated to range from €24 805 290 in Spain to €96 638 560 in Germany (Table 10). Table 10 Estimated diabetes-related health care costs in 2020 attributable to unhealthy diets and low physical activity patterns in 2015 in France, Germany, Italy, Spain and the United Kingdom Country Risk factor Number of incident cases in 2020 attributable to the risk factor in 2015 Estimated per patient diabetes cost in 2020 (€) * Estimated total health care cost in 2020 (€) France Unhealthy diets 4 548 6 810 30 971 880 Low physical activity 7 580 6 810 51 619 800 TOTAL 12 128 82 591 680 Germany Unhealthy diets 4 851 6 640 32 210 640 Low physical activity 9 703 6 640 64 427 920 TOTAL 14 554 96 638 560 Italy Unhealthy diets 1 513 3 935 5 953 655 Low physical activity 10 591 3 935 41 675 585 TOTAL 12 104 47 629 240 Spain Unhealthy diets 1 069 3 314 3 542 666 Low physical activity 6 416 3 314 21 262 624 TOTAL 7 485 24 805 290 United Kingdom Unhealthy diets 2 368 5 292 12 531 456 Low physical activity 8 288 5 292 43 860 096 TOTAL 10 656 56 391 552 Source: International Diabetes Federation, 2015, adjusted for average annual growth rate in per capita health spending during 2005 to 2013 at 1.2% in France, 2.4% in Germany, 0.55% in Italy, 0.95% in Spain and 1.75% in the United Kingdom (OECD, 2015). We further estimated the direct health care costs that can be attributed to diabetic complications arising from these attributable cases. We drew on Hayes et al. (2013) who reported annual diabetes complications incidence rates based on the UK Prevention Diabetes Study Outcomes Model 2 (UKPDS-OM2) to estimate the number of incident cases of complications in 2020. Specifically, Hayes et al. derived models to predict the annual risk and incidence of a range of outcomes of diabetes, including myocardial infarction, stroke, congestive heart failure, ischaemic heart disease, amputation, blindness, renal failure and ulcer (Appendix 5). We first applied these incidence rates to the cases attributable to unhealthy diets and low physical activity in each country to estimate the number 34 Assessing the economic costs of unhealthy diets and low physical activity of incident cases of complications arising in 2020. We then applied the average annual per patient cost for each complication as derived from the published evidence to estimate the direct health care costs that can be associated with these complications (Table 11). Appendix 6 provides a detailed breakdown by country of the estimated number of incident cases of diabetes-related complications and related costs in 2020. Table 11 Estimated diabetic complication-related costs in 2020 attributable to unhealthy diets and low physical activity patterns in 2015 in France, Germany, Italy, Spain and the United Kingdom Estimated number of incident cases of diabetes-related complications in 2020 attributable to the risk factor Estimated complication-related total health care cost in 2020 (€) Unhealthy diets Low physical activity Unhealthy diets Low physical activity Total cost France 181 302 2 646 685 4 411 142 7 057 827 Germany 193 386 3 910 441 7 821 688 11 732 129 Italy 60 422 492 429 3 447 004 3 939 433 Spain 43 255 516 414 3 099 453 3 615 867 United Kingdom 94 330 1 874 299 6 560 047 8 434 347 Accordingly, we estimate that the total number of incident cases of complications arising from diabetes cases that can be attributed to unhealthy diets and low physical activity in 2020 would be 483 in France, 579 in Germany, 482 in Italy, 298 in Spain and 424 in the United Kingdom. The associated direct health care costs ranged from some €3.6 million in Spain to €11.7 million in Germany. Taken together, the total direct health care costs linked to incident diabetes and its complications in 2020 that can be attributed to unhealthy diets and low physical activity in 2015 are estimated to be €89.6 million in France, €108.4 million in Germany, €51.6 million in Italy, €28.4 million in Spain and €64.8 million in the United Kingdom (see also below for a detailed breakdown of numbers). 4.2.5 Estimating the indirect costs that can be associated with diabetes attributable to unhealthy diets and low physical activity We considered two principal categories of indirect costs that can be associated with diabetes attributable to unhealthy diets and low physical activity: first, productivity loss in the workplace due to absenteeism and presenteeism, including among the unemployed population; second, productivity forgone because of work disability, early retirement or premature death. We first identified the proportion of diabetes cases in 2020 attributable to unhealthy diets and low A proof-of-concept approach 35 physical activity who are of working age, using the average sex-specific proportions of incident type 2 diabetes cases in the United Kingdom for the period 1991–2010 as provided by Holden et al. (2013) (Appendix 7). We then computed the percentage of annual incident type 2 diabetes cases by five-year age group and sex in the United Kingdom for the period 1991–2010 (Appendix 8) and applied these to the total number of diabetes cases attributable to unhealthy diets and low physical activity in each country, again by sex and five-year age group (Appendix 9). This allowed us to compute the total number of incident type 2 diabetes cases at working age (15–64 years) that can be attributed to unhealthy diets and low physical activity in each country by 2020 (Table 12). Table 12 Estimated number of incident type 2 diabetes cases at working age (15–64 years) that can be attributed to unhealthy diets and low physical activity in France, Germany, Italy, Spain and the United Kingdom, 2020 Unhealthy diets Low physical activity Total France 1 517 2 528 4 045 Germany 1 618 3 236 4 854 Italy 505 3 532 4 037 Spain 357 2 140 2 497 United Kingdom 790 2 764 3 554 We then computed the proportion of those of working age who are expected to be participating in the labour force. In the United Kingdom, labour force data for 2013 and 2014 suggest that the proportion of working age people with diabetes was 71.3% compared with 72.3% (2013) and 73.5% (2014) in the general population (Department for Work and Pensions, 2015). Based on these observations and in the absence of labour force participation data specific to people living with diabetes in the other countries studied, we assumed labour participation rates in the diabetic population to be similar to the general population, using annual national labour force participation rates for each of the five countries for the period 2000 to 2015 by five-year age group (OECD, 2016). We acknowledge that this is a conservative estimate that likely overestimates the ‘true’ proportion of people with diabetes who are in gainful employment across the five countries. Using this assumption, we computed the number of incident cases in 2020 attributable to unhealthy diets and low physical activity at working age who are expected to be in or out of the labour force by five-year age group (Appendix 10). Based on these data, we first estimated the indirect costs that can be linked to incident cases of diabetes that can be attributed to unhealthy diets and low physical activity. We drew on a systematic review by Breton et al. (2015), which 36 Assessing the economic costs of unhealthy diets and low physical activity reported that people with diabetes lose an average of 11.9 days of productivity per year due to the disease. This included absenteeism (work time lost) and presenteeism (work time impaired). Using this figure, we estimated the total number of productivity days lost due to absenteeism and presenteeism in the five countries. We then applied this estimate to data on average daily salary in 2020, using hourly labour cost data collected by Eurostat (Eurostat, 2015), adjusted to 2020 figures (based on the average rate of hourly salary increase every five years in the five countries from 2000 to 2015). Assuming an eight-hour working day, we estimate the total cost of productivity lost due to absenteeism and presenteeism that can be attributed to diabetes because of unhealthy diets and low physical activity in 2020 to range from €2.7 million in Spain to €9.7 million in Germany (Table 13). Table 13 Estimated total number of productivity days lost and cost due to absenteeism and presenteeism among incident type 2 diabetes cases at working age that can be attributed to unhealthy diets and low physical activity and who are expected to be in the labour force in France, Germany, Italy, Spain and the United Kingdom, 2020 Country Risk factor Estimated number of incident type 2 diabetes cases at working age attributable to the risk factor (in the labour force) Estimated total number of days of productivity lost in 2020 Average daily wage in 2020 (€) Estimated cost of productivity lost in 2020 (€) France Unhealthy diets 893 10 621 224.64 2 385 959 Low physical activity 1 488 17 702 224.64 3 976 599 TOTAL 6 362 558 Germany Unhealthy diets 1 105 13 144 246.96 3 246 097 Low physical activity 2 209 26 291 246.96 6 492 864 TOTAL 9 738 961 Italy Unhealthy diets 265 3 149 152.00 478 643 Low physical activity 1 852 22 043 152.00 3 350 501 TOTAL 3 829 144 Spain Unhealthy diets 216 2 574 150.16 386 576 Low physical activity 1 298 15 451 150.16 2 320 178 TOTAL 2 706 754 United Kingdom Unhealthy diets 545 6 488 220.16 1 428 341 Low physical activity 1 907 22 699 220.16 4 997 459 TOTAL 6 425 800 A proof-of-concept approach 37 We further estimated the costs of productivity lost that can be linked to individuals with diabetes who are of working age but are outside the formal labour force. Productivity losses among this population relate to lost unpaid contributions to national productivity such as time spent providing child care, household activities and other activities such as volunteering in the community (American Diabetes Association, 2013). To determine the value of such losses, we again applied Breton et al.’s estimate of 11.9 productivity days lost per year per diabetic person to the number of cases attributable to unhealthy diets and low physical activity who are of working age but are outside the labour force. This gives the total number of productivity days lost per year among the diabetic population outside the formal labour force, to which we then applied the average minimum daily wage in 2020 as derived from Eurostat (2015) for the period 2000 to 2015, adjusted to 2020 figures. Again assuming an eight-hour working day, we estimate the total cost of productivity lost due to absenteeism and presenteeism that can be attributed to diabetes because of unhealthy diets and low physical activity among those outside the formal labour force in 2020 to range from €0.4 million in Spain to €1.8 million in Italy (Table 14). Secondly, we computed the indirect costs that can be attributed to work disability, early retirement and premature deaths. We drew on work by Herquelot et al. (2011), which prospectively assessed the impact of diabetes on the risks of work disability, early retirement and premature death in a cohort of 20 625 employees of a national gas and electricity company in France, among whom 2.4% (506 individuals) had developed diabetes. They estimated the number of working years lost among diabetic persons aged 35–60 years to be 1.1 years per person. Of these, 0.09 years were attributed to work disability, 0.7 years to early retirement and 0.28 years to premature death. We applied these estimates to the number of diabetes cases attributable to unhealthy diets and low physical activity for those aged 35–60 who are expected to participate in the labour force. This yielded an estimate of the total number of working years lost due to work disability, early retirement and premature death because of unhealthy diet- and physical inactivity-related diabetes (Appendix 11). We then combined the total number of working years lost with the average annual salary in 2020, which we estimated using the average 2015 annual salary data collected by Eurostat (Eurostat, 2015), adjusted to 2020 values (based on the average rate of salary increase every five years from 2000 to 2015) (Table 15). 38 Assessing the economic costs of unhealthy diets and low physical activity Table 14 Estimated total number of productivity days lost and cost due to absenteeism and presenteeism among incident type 2 diabetes cases at working age that can be attributed to unhealthy diets and low physical activity and who are expected to be outside the formal labour force in France, Germany, Italy, Spain and the United Kingdom, 2020 Country Risk factor Estimated number of incident type 2 diabetes cases at working age attributable to the risk factor (outside the formal labour force) Estimated total number of days of productivity lost in 2020 Minimum daily wage in 2020 (€) Estimated cost of productivity lost in 2020 (€) France Unhealthy diets 624 7 429 75.68 562 252 Low physical activity 1 041 12 382 75.68 937 086 TOTAL 1 499 338 Germany Unhealthy diets 513 6 109 70.80 432 513 Low physical activity 1 027 12 219 70.80 865 116 TOTAL 1 297 629 Italy Unhealthy diets 240 2 856 77.20 220 482 Low physical activity 1 680 19 992 77.20 1 543 377 TOTAL 1 763 859 Spain Unhealthy diets 140 1 668 33.92 56 590 Low physical activity 841 10 013 33.92 339 646 TOTAL 396 236 United Kingdom Unhealthy diets 245 2 911 92.88 270 339 Low physical activity 856 10 182 92.88 945 746 TOTAL 1 216 085 A proof-of-concept approach 39 Table 15 Estimated cost of working years lost due to work disability, early retirement and premature death among incident type 2 diabetes cases at working age that can be attributed to unhealthy diets and low physical activity and who are expected to be in the formal labour force in France, Germany, Italy, Spain and the United Kingdom, 2020 Country Risk factor to which cases are attributable Average annual salary in 2020 (in €)* Estimated costs of disability in 2020 (in €) Estimated costs of early retirement in 2020 (€) Estimated costs of premature death in 2020 (€) France Unhealthy diets 41 131.99 3 304 085 25 698 440 10 279 376 Low physical activity 41 131.99 5 506 808 42 830 733 17 132 293 TOTAL 8 810 894 68 529 172 27 411 669 Germany Unhealthy diets 41 523.63 4 127 868 32 105 639 12 842 256 Low physical activity 41 523.63 8 256 587 64 217 897 25 687 159 TOTAL 12 384 455 96 323 537 38 529 415 Italy Unhealthy diets 32 021.69 762 619 5 931 485 2 372 594 Low physical activity 32 021.69 5 338 336 41 520 393 16 608 157 TOTAL 6 100 956 47 451 878 18 980 751 Spain Unhealthy diets 31 377.75 610 939 4 751 747 1 900 699 Low physical activity 31 377.75 3 666 776 28 519 369 11 407 748 TOTAL 4 277 715 33 271 116 13 308 446 United Kingdom Unhealthy diets 49 359.15 2 421 903 18 837 024 7 534 810 Low physical activity 49 359.15 8 473 719 65 906 700 26 362 680 TOTAL 10 895 622 84 743 725 33 897 490 Note: * estimated from 2015 average annual salary adjusted using the average percentage of increase in salary every five years from 2000 to 2015 as derived from OECD (2016). 40 Assessing the economic costs of unhealthy diets and low physical activity 4.3 The estimated total economic costs of unhealthy diets and low physical activity related to diabetes and its complications Table 16 summarises the total economic cost of diabetes in the five countries in 2020 that can be associated with unhealthy diets and low physical activity patterns in 2015. Costs associated with low physical activity tended to be higher than those associated with unhealthy diets in all countries, although differences varied, ranging from some 67% higher in France to a factor of six in Spain and seven in Italy. Indirect costs that can be associated with either risk factor were higher than direct health care costs in all countries, by between 25% in France and up to 100% in the United Kingdom. The differences in total cost across countries reflect, to a great extent, differences in population sizes, health care costs and labour costs. When related to the population projected to develop diabetes in 2020 as a consequence of unhealthy diets and low physical activity in 2015, the United Kingdom showed the highest cost, at €18 953, closely followed by Germany and France, while Italy had the lowest cost, at just over €10 720 (Table 17). 4.3.1 Sensitivity analysis We carried out a limited set of sensitivity analyses in order to better understand the likely range of cost estimates provided here. We repeated the above analyses using the lower and upper values (i.e. 95% confidence intervals) for those parameters where they were available. This was the case for the prevalence of low physical activity in the general population and the adjustment factor for low physical activity. Applying the lower value of these two parameters decreased the total estimated costs by between 27% and 63%. The highest impact was on estimates for Italy, reducing the total costs by 63% (from €146 million to €54.3 million), followed by Spain at 59% (€94.1 million to €39.0 million), Germany at 58% (€266.7 million to €112.0 million), France at 29% (€151.7 million to €107.4 million) and the United Kingdom at 27% (€202.0 million to €147.8 million). Using the higher values of the same parameters increased the total costs by between 26% and 160%. The greatest effect was seen for Germany, where the total costs more than doubled (from €266.7 million to €694.1 million). A doubling of costs was also observed for Spain (€94.1 million to €212.3 million) and Italy (€146 million to €307.9 million). In France estimated costs rose by 29% (€151.7 million to €196.2 million) and in the United Kingdom by 26% (€202 million to €254.3 million). A proof-of-concept approach 41 Table 16 Estimated economic cost that can be associated with unhealthy diets and low physical activity patterns in 2015 as manifested in incident diabetes and complication in France, Germany, Italy, Spain and the United Kingdom in 2020 Country Risk factor Direct health care cost (€) Indirect cost (€) Total cost (€) France Unhealthy diets 33 618 565 42 230 112 75 848 677 Low physical activity 56 030 942 70 385 985 126 416 927 TOTAL 202 265 604 Germany Unhealthy diets 36 121 081 52 774 976 88 896 057 Low physical activity 72 249 608 105 530 564 177 780 172 TOTAL 266 676 229 Italy Unhealthy diets 6 446 084 9 765 824 16 211 908 Low physical activity 45 122 589 68 415 902 113 538 491 TOTAL 129 750 398 Spain Unhealthy diets 4 059 080 7 706 874 11 765 955 Low physical activity 24 362 077 46 253 942 70 616 019 TOTAL 82 381 974 United Kingdom Unhealthy diets 14 405 755 30 455 391 44 861 146 Low physical activity 50 420 143 106 686 304 157 106 447 TOTAL 201 967 593 Table 17 Estimated total and per capita economic cost that can be associated with unhealthy diets and low physical activity patterns in 2015 as manifested in incident diabetes and complications in France, Germany, Italy, Spain and the United Kingdom in 2020 Country Total cost Cost per head of total estimated population in 2020 (€) Cost per incident case of type 2 diabetes in 2020 (€) Cost per incident case of type 2 diabetes attributable to unhealthy diets and low physical activity in 2020 (€) France 202 265 604 2.97 1 334 16 678 Germany 266 676 229 3.29 1 099 18 323 Italy 129 750 398 2.14 858 10 720 Spain 82 381 974 1.77 770 11 006 United Kingdom 201 967 593 3.01 1 706 18 953 Chapter 5 Discussion and conclusions This study sought to contribute to a better understanding of the economic burden that can be associated with unhealthy diets and low levels of physical activity in order to help inform priority setting and motivate efforts to promote more effectively healthy diets and physical activity in Europe and worldwide. We did so through critically reviewing the available evidence on the economic costs associated with unhealthy diets and low physical activity; discussing the measurement, methodological and practical issues for estimating the economic burden from unhealthy diets and low physical activity; and developing a framework for assessing costs and testing the feasibility of this approach to provide better estimates of the economic burden. We showed that the majority of reviewed studies found a significant association between diet and/or physical activity and costs, with unhealthy diets and low physical activity predictive of higher health care expenditure. Studies that did report costs that can be associated with the two risk factors estimated the annual cost of unhealthy diets to range from €3 to €148 per capita and for low physical activity from €3 to €181 per capita. The highest health care cost estimates were equivalent to between 2% and 6% of health spending in the countries. We noted that there is a very wide range of estimates, and these are very sensitive to the measures of diet and activity and the ways in which the studies were carried out. Costing studies differ widely in their analytical approaches and in the nature and scope of data used, influencing estimates for the economic burden of unhealthy diets and low physical activity. Particular challenges arise from measuring unhealthy diets given the different effects of foods and the interactions between these effects. Calibrating the extent of deviation from optimal consumption and the effects of this deviation is difficult. It is also clear that the context should be taken into account in terms of other population characteristics. While there is more consensus about the measurement of physical activity, similar issues arise in terms of the independent effects of moderate and vigorous activity and sedentary behaviour, but also the interactions between these. Further, studies take broader and narrower perspectives in terms of what costs are included, with some limited to formal health care costs, and others aiming to take a more societal view. While current evidence makes it difficult to make accurate comparisons, much 44 Assessing the economic costs of unhealthy diets and low physical activity of the economic burden is likely to come from non-health care costs, especially from effects on productivity, absenteeism, presenteeism and other indirect costs. Based on a critical appraisal of existing approaches, we developed a framework for estimating the economic costs of unhealthy diets and low physical activity using a disease-based approach, with type 2 diabetes mellitus chosen as a disease for which both are risk factors. The aim was to demonstrate the feasibility of undertaking a comprehensive, disease-based, bottom-up cost assessment drawing on the best available data as identified from a rapid review of the published evidence that addresses some of the limitations of existing costing studies. Our choice of type 2 diabetes as the exemplar outcome was motivated by its consistently strong association with either risk factor as shown in the literature. Using this approach, we projected the total economic costs that can be associated with unhealthy diets and low physical activity patterns in 2015 as manifested in incident diabetes cases in 2020 to range from €82.4 million in Spain to €266.7 million in Germany. This equates to a per capita cost of €1.77 in Spain to €3.29 in Germany. Relating costs more specifically to the population projected to develop diabetes in 2020 as a consequence of unhealthy diets and low physical activity in 2015, the United Kingdom showed the highest cost, at €18 953, closely followed by Germany and France, while Italy had the lowest cost, at just over €10 720. The total cost in the five high-income countries studied (France, Germany, Italy, Spain and the United Kingdom) was estimated to amount to about €883 million. The populations in the five countries studied account for almost two thirds of the total population in the European Union (EU-28), which would imply a total EU cost of around €1.3 billion, but care must be taken in any extrapolation given differences in population characteristics, costs of care and value of lost productivity. While these estimates of the economic costs are substantial, they represent only a small proportion of health care expenditure and a very small proportion of GDP. Even on the higher estimates in the sensitivity analysis it is likely that the burden of disease associated with unhealthy diets and low physical activity as measured by poor health and shortened life will be at least as important as the financial costs of additional health care and lost productivity. It is difficult to compare the findings of the analyses presented here with estimates published elsewhere since only diabetes costs are estimated. Scarborough et al. (2011) calculated the cost of unhealthy diets and low physical activity in the United Kingdom in 2006–07 to be €9.8 billion (€8.5 billion and €1.3 billion for unhealthy diets and low physical activity, respectively). Ding et al. (2016), in their recent assessment of the global economic costs that can be associated with low physical activity, provided estimates of direct and indirect costs ranging from $(Int) 1.4 billion in Italy to $(Int) 2.6 billion in Germany. Again, data Discussion and conclusions 45 are difficult to compare as analyses by Ding et al. considered a wider range of disease outcomes (coronary heart disease, stroke, type 2 diabetes, breast cancer, colorectal cancer), and the cost estimates are not easily transferable. The principal analytical steps employed in the present analysis are similar to those used by Ding et al. (2016) for low physical activity in that we calculated country-specific adjusted population-attributable fractions based on available prevalence data and relative disease risks causally linked to either risk factor in order to estimate the total number of cases for the outcome (here: diabetes) in each country. Ding et al. used the same definition of low physical activity that we used, and drew on the same data sources for prevalence of low physical activity and estimated the same relative risk-adjusted population-attributable fractions that we calculated for our analysis. We also estimated health care costs as well as indirect costs that can be associated with disease developed as a consequence of the risk factor and similar to Ding et al. we drew on the latest estimates of diabetes-related health care costs provided by the International Diabetes Federation (International Diabetes Federation, 2015). Where our model differs is that we only considered the costs of incident cases, that is, new cases, which can be causally linked to the risk factor, whereas Ding et al. calculated costs on the basis of prevalence data. Also, our approach takes account of the expected time lag between exposure to the risk factor (unhealthy diets, low physical activity) and development of the disease and complications. Further, we considered a wider range of indirect costs linked to lost productivity because of work absence, disability, early retirement and premature death among incident diabetes cases that can be attributed to unhealthy diets and low physical activity. Conversely, Ding et al. only considered lost productivity that can be associated with premature death. We therefore believe that our estimates provide a fuller picture of the likely future costs that can be attributed to contemporary dietary and physical activity patterns. 5.1 Limitations of the costing framework As noted, the costing model as proposed here presents a ‘proof of concept’ approach, drawing on the best available data as identified from a rapid review of the published evidence and providing point estimates only (although we present data from a limited sensitivity analysis). Clearly, there is uncertainty associated with each input parameter, namely prevalence rates of unhealthy diets and low physical activity in the general population, adjustment factors, relative risks and average per patient disease and productivity costs, which will all impact on the estimated effect size of predicted incident diabetes cases and cost estimates. A fully costed model would consider ranges as input parameters as a reflection of variation at baseline, using for example probabilistic modelling such as Monte 46 Assessing the economic costs of unhealthy diets and low physical activity Carlo simulation, and also employ sensitivity analyses in order to better understand the influence of the various input parameters on cost estimates. A major challenge presents the availability of suitable data on the prevalence of unhealthy diets and low physical activity that are comparable across countries and over time. For example, the prevalence data for low physical activity used for the five countries were obtained from the WHO Global Health Observatory Database, with 2010 prevalence rates as the latest available data. The level of exposure among physically inactive individuals was not specified so data from a more recent Eurobarometer survey conducted in 2013 was used to assess the level of exposure. In order to arrive at comparable estimates there is therefore a need for more detailed national prevalence data on physical activity in each country, specifying the type, duration, frequency and intensity to identify the extent of low physical activity and levels of exposure to risk. We defined a given dietary pattern as unhealthy, based on a score of <67% on the 2010 alternate healthy eating index (AHEI) and we used dietary data compiled by the European Food Safety Authority, which draws on national surveys that are not directly comparable in relation to assessment methods and data collection instruments. The EFSA database also only provides aggregate-level data of mean consumption of certain food groups (European Food and Safety Authority, 2015). In the absence of more detailed data sets, we assumed that mean intakes as presented in the EFSA database are nationally representative of a relatively homogenous dietary pattern in each of the five countries. But this is not necessarily the case and in order to arrive at more precise estimates more recent, individual-level data on mean intake of the different AHEI food-groups in different countries in Europe would be needed. We considered a window period of five years from the occurrence of unhealthy diets and low physical activity to the development of diabetes and associated costs incurred. The five-year lag reflects the average latency period and it will vary according to individual risk profile, with those with other high-risk characteristics, such as genetic predisposition, likely to develop the condition more quickly. If the costing model was applied more widely this would require a systematic assessment of the evidence of the range of the latency period, and its incorporation in the form of a sensitivity analysis. Likewise, we used average annual per patient costs for diabetes and its complications in previous years, which we adjusted to reflect more closely 2020 prices. However, average costs are not sensitive to variations in individual patterns of health care utilisation and a fully costed model would ideally derive a unique set of average per patient costs for each country that take account of patient characteristics such as age, sex, ethnicity, socioeconomic status and disease severity, which may influence patterns of use. A particular challenge relates to estimating Discussion and conclusions 47 the indirect costs that can be associated with diabetes as a manifestation of unhealthy diets and low physical activity, and within the scope of this study we applied very crude assumptions, which would need to be revisited for a fully costed model. Any future modelling exercise would also need to take account of the long-term care costs associated with diabetes which are estimated to be substantial, but which we have been unable to address in this work. Finally, the costing framework as presented here uses only one outcome, namely type 2 diabetes, as a manifestation of exposure to unhealthy diets and low physical activity. Yet, unhealthy diets and low physical activity are associated with a range of other conditions of ill health as highlighted in earlier parts of this volume and a fully costed model would incorporate these also, along with their sequelae, guided by the strength of evidence of the association. Ding et al. (2016), in their analysis of the global economic cost of low physical activity, disaggregated cost figures by disease category, and diabetes cases that can be attributed to low physical activity accounted for a large majority of the estimated total health care costs. Specifically, according to Ding et al., for the five countries considered in the present study, diabetes cases accounted for approximately half (Italy) up to 85% (Spain) of the estimated total health care costs. This would mean that estimates for health care costs provided in the present study are likely to underestimate the ‘true’ health care costs that can be associated with low physical activity-related diabetes cases by at least one third. However, since Ding et al. (2016) calculated costs for prevalent cases it is very difficult to generalise from their estimates. 5.2 Implications for future studies This study has tested the feasibility of estimating the costs of unhealthy diets and low physical activity using a disease-based approach. While there are limitations, it has shown that it is broadly feasible to populate the model with data from a range of sources, and the results show a reasonable consistency across countries. While the disease burden from diabetes is not currently as large as that for, for example, ischaemic heart disease, it is a good exemplar because of the strong relationship between these lifestyle factors and the risk of diabetes. In other chronic diseases there will be additional challenges in identifying the contribution of these lifestyle factors and disease risk. Given the very wide range of estimates of costs from the studies reviewed, this may be a more promising approach. References Ackermann R, Cheadle A & Sandhu N (2003). Community exercise program use and changes in healthcare costs for older adults. Am J Prev Med, 25:232–7. Afshin A, Micha R, Khatibzadeh S & Mozaffarian D (2014). Consumption of nuts and legumes and risk of incident ischemic heart disease, stroke, and diabetes: a systematic review and meta-analysis. Am J Clin Nutr, 100:278–88. Al Tunaji H, Davis J, Dawn M & Khan K (2014). Population-attributable fraction of type 2 diabetes due to physical inactivity in adults: a systematic review. BMC Public Health, 14:469. Allender S, Foster C, Scarborough P & Rayner M (2007). The burden of physical activity-related ill health in the UK. J Epidemiol Community Health, 61:344–48. Alter D, Wijeysundera H & Franklin B (2012). Obesity, lifestyle risk-factors and health service outcomes among healthy middle-aged adults in Canada. BMC Health Serv Res, 12:238. American Diabetes Association (2013). Economic costs of diabetes in the U.S. in 2012. Diabetes Care, 36:1033–46. Anderson L, Martinson B, Crain A, Pronk N, Whitebird R, O’Connor P & Fine L (2005). Health care charges associated with physical inactivity, overweight and obesity. Prev Chronic Dis, 2:A09. Aune D, Keum N, Giovannucci E, Fadnes L, Boffetta P, Greenwood D, Tonstad S, Vatten L, Riboli E & Norat T (2016). Whole grain consumption and risk of cardiovascular disease, cancer, and all cause and cause specific mortality: systematic review and dose-response meta-analysis of prospective studies. BMJ, 353:i2716. Bachmann J, Defina L, Franzini L, Gao A, Leonard D, Cooper K, Berry J & Willis B (2015). Cardiorespiratory fitness in middle age and health care costs in later life. J Am Coll Cardiol, 66:1876–85. Baliunas D (2011). A comparison of two methods of adjusted attributable fraction estimation as applied to the four major smoking related causes of death in Canada in 2005. Ph.D. University of Toronto. URL: https://tspace.library.utoronto.ca/bitstream/1807/32061/1/ Baliunas_Dalia_O_201111_PhD_thesis.pdf (accessed 23 October 2016). Benichou J (2001). A review of adjusted estimators of attributable risk. Stat Methods Med Res, 10:195–216. Biswas A, Oh P, Faulkner G, Bajaj R, Silver M, Mitchell M, et al. (2015). Sedentary time and its association with risk for disease incidence, mortality, and hospitalization in adults: a systematic review and meta-analysis. Ann Intern Med, 162:123–32. Bland P, An L, Foldes S, Garrett N & Alesci N (2009). Modifiable health behaviours and shortterm medical costs among health plan members. Am J Health Promot, 23:265–73. Boeing H, Bechthold A, Bub A, Ellinger S, Haller D, Kroke A, Leschik-Bonnet E, Müller M, Oberritter H, Schulze M, Stehle P & Watzl B (2012). Critical review: vegetables and fruit in the prevention of chronic diseases. Eur J Nutr, 51:637–63. Bouvard V, Loomis D, Guyton K, Grosse Y, Ghissassi F, Benbrahim-Tallaa L, Guha N, Mattock H, Straif K & International Agency For Research On Cancer Monograph Working Group (2015). Carcinogenicity of consumption of red and processed meat. Lancet Oncol, 16:1599–600. Breton M, Guénette L, Amiche M, Kayibanda J, Grégoire J & Moisan J (2015). Burden of diabetes on the ability to work: A systematic review. Diabetes Care, 36:740–9. 50 Assessing the economic costs of unhealthy diets and low physical activity Carlson S, Fulton J, Pratt M, Yang Z & Adams E (2015). Inadequate physical activity and health care expenditures in the United States. Prog Cardiovasc Dis, 57:315–23. Cecchini M & Bull F (2015). Promoting physical activity. In: McDaid D, Sassi F & Merkur S (eds). Promoting health, preventing disease. The economic case. Maidenhead: Open University Press. Chevan J & Roberts D (2014). No short-term savings in health care expenditures for physically active adults. Prev Med, 63:1–5. Chiuve S, Fung T, Rimm E, Hu F, Mccullough M, Wang M, Stampfer M & Willett W (2012). Alternative dietary indices both strongly predict risk of chronic disease. J Nutr, 142:1009–18. Chowdhury R, Stevens S, Gorman D, Pan A, Warnakula S, Chowdhury S, Ward H, Johnson L, Crowe F, Hu F & Franco O (2012). Association between fish consumption, long chain omega 3 fatty acids, and risk of cerebrovascular disease: systematic review and meta-analysis. BMJ, 345:e6698. Codogno J, Turi B, Kemper H, Fernandes R, Christofaro D & Monteiro H (2015). Physical inactivity of adults and 1-year health care expenditures in Brazil. Int J Public Health, 60:309–16. Collins C, Patterson A & Fitzgerald D (2011). Higher diet quality does not predict lower Medicare costs but does predict number of claims in mid-aged Australian women. Nutrients, 3:40–8. Commission of the European Communities (2007). White Paper on A Strategy for Europe on Nutrition, Overweight and Obesity related health issues. Brussels: European Commission. Council of the European Union (2014). Council conclusions on nutrition and physical activity. URL: http://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:52014XG0708(0 1)&from=EN (accessed 23 October 2016). Daviglus M, Liu K, Pirzada A, Yan L, Garside D, Wang R, Van Horn L, Manning W, Manheim L, Dyer A, Greenland P & Stamler J (2005). Relationship of fruit and vegetable consumption in middle-aged men to medicare expenditures in older age: the Chicago Western Electric Study. J Am Diet Assoc., 105:1735–44. De Bakker D, Struijs J, Baan C, Raams J, De Wildt J, Vrijhoef H & Schut F (2012). Early results from adoption of bundled payment for diabetes care in the Netherlands show improvement in care coordination. Health Aff (Millwood), 31:426–33. Department for Work and Pensions (2015). Labour Force Survey analysis of disabled people by region and main health problem. URL: https://www.gov.uk/government/uploads/system/ uploads/attachment_data/file/406369/labour-force-survey-disabled-people.pdf (accessed 15 June 2016). Ding D, Lawson K, Kolbe-Alexander T, Finkelstein E, Katzmarzyk P, Van Mechelen W, Pratt M & Lancet Physical Activity Series 2 Executive Committee (2016). The economic burden of physical inactivity: a global analysis of major non-communicable diseases. Lancet, 388:1311–24. Doidge J, Segal L & Gospodarevskaya E (2012). Attributable risk analysis reveals potential healthcare savings from increased consumption of dairy products. J Nutr, 142:1772–80. Ekelund U, Steene-Johannessen J, Brown W, Fagerland M, Owen N, Powell K, Bauman A, Lee I, Lancet Physical Activity Series 2 Executive Committee & Lancet Sedentary Behaviour Working Group (2016). Does physical activity attenuate, or even eliminate, the detrimental association of sitting time with mortality? A harmonised meta-analysis of data from more than 1 million men and women. Lancet, 388:1302–10. Estruch R, Ros E, Salas-Salvadó J, Covas M, Corella D, Arós F, Gómez-Gracia E, et al. (2013). Primary prevention of cardiovascular disease with a Mediterranean diet. N Engl J Med, 368:1279–90. European Commission (2014). The 2014 EU Summit on Chronic Diseases. Brussels, 3 and 4 April 2014, Conference Conclusions. URL: http://ec.europa.eu/health/major_chronic_diseases/ docs/ev_20140403_mi_en.pdf (accessed 23 October 2016). European Food and Safety Authority (2015). The EFSA Comprehensive European Food Consumption Database. URL: https://www.efsa.europa.eu/en/food-consumption/ comprehensive-database (accessed 15 June 2016). References 51 Eurostat (2015). Hourly labour costs. URL: http://ec.europa.eu/eurostat/statistics-explained/ index.php/Hourly_labour_costs (accessed 16 May 2016). Ezzati M, Lopez A, Rodgers A, et al. (2004). Comparative quantification of health risks. Geneva: World Health Organization. Garrett N, Brasure M, Schmitz K, Schultz M & Huber M (2004). Physical inactivity: direct cost to a health plan. Am J Prev Med, 27:304–9. GBD 2015 Risk Factors Collaborators (2016). Global, regional, and national comparative risk assessment of 79 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015. Lancet, 388:1659–1724. Hayes A, Leal J, Gray A, Holman R & Clarke P (2013). UKPDS outcomes model 2: a new version of a model to simulate lifetime health outcomes of patients with type 2 diabetes mellitus using data from the 30 year United Kingdom Prospective Diabetes Study: UKPDS 82. Diabetologia, 56:1925–33. Health and Social Care Information Centre (2015). National Diabetes Audit 2012–2013. Report 2: Complications and mortality. URL: http://content.digital.nhs.uk/catalogue/PUB16496/ nati-diab-audi-12-13-rep2.pdf (accessed 25 October 2016). Herquelot E, Gueguen A, Bonenfant S & Dray-Spira R (2011). Impact of diabetes on work cessation: data from the GAZEL cohort study. Diabetes Care, 34:1344–9. Holden S, Barnett A, Peters J, Jenkins-Jones S, Poole C, Morgan C & Currie C (2013). The incidence of type 2 diabetes in the United Kingdom from 1991 to 2010. Diabetes Obes Metab, 15:844–52. Idler N, Teuner C, Hunger M, Holle R, Ortlieb S, Schulz H, et al. (2015). The association between physical activity and health care costs in children – results from the GINIplus and LISAplus cohort studies. BMC Public Health, 15:437. International Diabetes Federation (2015). IDF Diabetes Atlas, Seventh edition. Brussels: International Diabetes Federation. Jacobs S, Harmon B, Boushey C, Morimoto Y, Wilkens L, Le Marchand L, Kröger J, Schulze M, Kolonel L & Maskarinec G (2015). A priori-defined diet quality indexes and risk of type 2 diabetes: the Multiethnic Cohort. Diabetologia, 58:98–112. Janssen I (2012). Health care costs of physical inactivity in Canadian adults. Appl Physiol Nutr Metab, 37:803–6. Johns G (2010). Presenteeism in the workplace: a review and research agenda. J Org Behav, 32:519–42. Johnston K, Buxton M, Jones D & Fitzpatrick R (1999). Assessing the costs of healthcare technologies in clinical trials. Health Technol Assess, 3:1–76. Katzmarzyk P (2011). The Economic Costs Associated with Physical Inactivity and Obesity in Ontario. Health and Fitness Journal of Canada, 4:31–40. Katzmarzyk P, Gledhill N & Shephard R (2000). The economic burden of physical inactivity in Canada. CMAJ, 163:1435–40. Khangura S, Konnyu K, Cushman R, Grimshaw J & Moher D (2012). Evidence summaries: the evolution of a rapid review approach. Syst Rev, 1:10. Koopmanschap M, Rutten F, Van Ineveld B & Van Roijen L (1995). The friction cost method for measuring indirect costs of disease. J Health Econ, 14:171–89. Krueger H, Krueger J & Koot J (2015). Variation across Canada in the economic burden attributable to excess weight, tobacco smoking and physical inactivity. Can J Public Health, 106:e171–7. Kruk J (2014). Health and economic costs of physical inactivity. Asian Pac J Cancer Prev, 15:7499– 503. Kuriyama S, Hozawa A & Ohmori K (2004). Joint impact of health risks on health care charges: 7-year follow-up of National Health Insurance beneficiaries in Japan (the Ohsaki Study). Prev Med, 39:1194–9. 52 Assessing the economic costs of unhealthy diets and low physical activity Laaksonen M, Knekt P & Rissanen H (2009). The relative importance of modifiable potential risk factors of type 2 diabetes: a meta-analysis of two cohorts. Eur J Epidemiol, 25:115–24. Lee I, Shiroma E, Lobelo F, Puska P, Blair S & Katzmarzyk P (2012). Effect of physical inactivity on major non-communicable diseases worldwide: an analysis of burden of disease and life expectancy. Lancet, 280:219–29. Li Y, Ley S, Tobias D, Chiuve S, Vanderweele T, Rich-Edwards J, Curhan G, Willett W, Manson J, Hu F & Qi L (2015). Birth weight and later life adherence to unhealthy lifestyles in predicting type 2 diabetes: prospective cohort study. BMJ, 351:h3672. Liyanage T, Ninomiya T, Wang A, Neal B, Jun M, Wong M, Jardine M, Hillis G & Perkovic V (2016). Effects of the Mediterranean diet on cardiovascular outcomes – A systematic review and meta-analysis. PLoS One, 11:e0159252. Lo Y, Wahlqvist M, Chang Y, Kao S & Lee M (2013). Dietary diversity predicts type of medical expenditure in elders. Am J Manag Care, 19:e415–23. Lyssenko V, Jonsson A, Almgren P, Pulizzi N, Isomaa B, Tuomi T, Berglund G, Altshuler D, Nilsson P & Groop L (2008). Clinical risk factors, DNA variants, and the development of type 2 diabetes. N Engl J Med, 359:2220–32. Malik V, Popkin B, Bray G, Després J & Hu F (2010a). Sugar sweetened beverages, obesity, type 2 diabetes and cardiovascular disease risk. Circulation, 121:1356–64. Malik V, Popkin B, Bray G, Després J, Willett W & Hu F (2010b). Sugar-sweetened beverages and risk of metabolic syndrome and type 2 diabetes: A meta-analysis. Diabetes Care, 33:2477–83. Maresova K (2014). The costs of physical inactivity in the Czech Republic in 2008. J Phys Act Health, 11:489–94. Martinson B, Crain A & Pronk N (2003). Changes in physical activity and short-term changes in health care charges: a prospective cohort study of older adults. Prev Med, 37:319–26. Micha R, Wallace S & Mozaffarian D (2010). Red and processed meat consumption and risk of incident coronary heart disease, stroke, and diabetes mellitus: a systematic review and metaanalysis. Circulation, 121:2271–83. Micha R, Khatibzadeh S, Shi P, Andrews K, Engell R, Mozaffarian D & Global Burden of Diseases Nutrition and Chronic Diseases Expert Group (Nutricode) (2015). Global, regional and national consumption of major food groups in 1990 and 2010: a systematic analysis including 266 country-specific nutrition surveys worldwide. BMJ Open, 5:e008705. Montonen J, Knekt P, Härkänen T, Järvinen R, Heliövaara M, Aromaa A & Reunanen A (2005). Dietary patterns and the incidence of type 2 diabetes. Am J Epidemiol, 161:219–27. Morgan P (2012). Back to the future: the changing frontiers of nutrition research and its relationship to policy. Proc Nutr Soc, 71:190–7. Mozaffarian D, Appel L & Van Horn L (2011). Components of a cardioprotective diet: new insights. Circulation, 12:2870–91. Muka T, Imo D, Jaspers L, Colpani V, Chaker L, Van Der Lee S, Mendis S, Chowdhury R, Bramer W, Falla A, Pazoki R & Franco O (2015). The global impact of non-communicable diseases on healthcare spending and national income: a systematic review. Eur J Epidemiol, 30:251–77. Murray C & Lopez A (1997). Global mortality, disability and the contribution of risk factors: global burden of disease study. Lancet, 349:1436–42. NCBI (2016). Pubmed. URL: http://www.ncbi.nlm.nih.gov/pubmed. OECD (2015). Focus on Health Spending: OECD Health Statistics 2015. URL: http://www. oecd.org/health/health-systems/Focus-Health-Spending-2015.pdf (accessed 16 May 2016). OECD (2016). Labour force participation rate. URL: https://data.oecd.org/emp/labour-forceparticipation-rate.htm (accessed 15 June 2016). Oldridge N (2008). Economic burden of physical inactivity: healthcare costs associated with cardiovascular disease. Eur J Cardiovasc Prev Rehabil, 15:130–9. Peeters G, Mishra G, Dobson A & Brown W (2014). Health care costs associated with prolonged sitting and inactivity. Am J Prev Med, 46:265–72. References 53 Popkin B, Kim S, Rusev E, Du S & Zizza C (2006). Measuring the full-economic costs of diet, physical activity and obesity-related chronic diseases. Obes Rev, 7:271–93. Rayner M & Scarborough P (2005). The burden of food-related ill health in the UK. J Epidemiol Community Health, 59:1054–7. Rice N & Normand C (2012). The cost associated with disease-related malnutrition in Ireland. Public Health Nutr, 15:1966–72. Risérus U, Willett W & Hu F (2009). Dietary fats and prevention of type 2 diabetes. Prog Lipid Res, 48:44–51. Samitz G, Egger M & Zwahlen M (2011). Domains of physical activity and all-cause mortality: systematic review and dose-response meta-analysis of cohort studies. Int J Epidemiol, 40:1382– 400. Scarborough P, Bhatnagar P, Wickramasinghe KK, Allender S, Foster C & Rayner M (2011). The economic burden of ill health due to diet, physical inactivity, smoking, alcohol and obesity in the UK: an update to 2006–07 NHS costs. J Public Health (Oxf), 33:527–35. Schultz A & Edington D (2007). Employee health and presenteeism: a systematic review. J Occup Rehabil, 17:547–79. Schwingshackl L & Hoffmann G (2015). Diet quality as assessed by the Healthy Eating Index, the Alternate Healthy Eating Index, the Dietary Approaches to Stop Hypertension score, and health outcomes: a systematic review and meta-analysis of cohort studies. J Acad Nutr Diet, 115:780–800. Sedentary Behaviour Research Network (2012). Letter to the editor: standardized use of the terms “sedentary” and “sedentary behaviours”. Appl Physiol Nutr Metab, 37:540–2. Stefler D, Malyutina S, Kubinova R, Pajak A, Peasey A, Pikhart H, Brunner E & Bobak M (2015). Mediterranean diet score and total and cardiovascular mortality in Eastern Europe: the HAPIEE study. Eur J Nutr, [Epub ahead of print]. Suhrcke M, Fahey D & Mckee M (2008). Economic aspects of chronic disease and chronic disease management. In: Nolte E & McKee M (eds) Caring for people with chronic conditions: A health system perspective. Maidenhead: Open University Press. TNS Opinion & Social (2014). Special Eurobarometer 412. Sport and physical activity. Brussels: European Commission. Tong T, Wareham N, Khaw K, Imamura F & Forouhi N (2016). Prospective association of the Mediterranean diet with cardiovascular disease incidence and mortality and its population impact in a non-Mediterranean population: the EPIC-Norfolk study. BMC Med, 14:135–46. United Nations (2015). World population prospects, the 2015 revision. URL: https://esa.un.org/ unpd/wpp/Download/Standard/Population/ (accessed 30 November 2016). Wang F, Mcdonald T, Reffitt B & Edington D (2005). BMI, physical activity, and health care utilization/costs among Medicare retirees. Obes Res, 13:1450–7. Wang G, Pratt M & Macera C (2004). Physical activity, cardiovascular disease, and medical expenditures in U.S. adults. Ann Behav Med, 28:88–94. Warburton D, Charlesworth S, Ivey A, Nettlefold L & Bredin S (2010). A systematic review of the evidence for Canada’s Physical Activity Guidelines for Adults. Int J Behav Nutr Phys Act, 7:39. Webber L, Divajeva D, Marsh T, Mcpherson K, Brown M, Galea G & Breda J (2014). The future burden of obesity-related diseases in the 53 WHO European-Region countries and the impact of effective interventions: a modelling study. BMJ Open, 4:e004787. Weyer C, Bogardus C, Mott D & Pratley R (1999). The natural history of insulin secretory dysfunction and insulin resistance in the pathogenesis of type 2 diabetes mellitus. J Clin Invest, 104:787–94. WHO Regional Office for Europe (2001). The First Action Plan for Food and Nutrition Policy WHO European Region 2000–2005. Copenhagen: World Health Organization. WHO Regional Office for Europe (2005). Steps to health. A European framework to promote physical activity for health. Copenhagen: World Health Organization. 54 Assessing the economic costs of unhealthy diets and low physical activity WHO Regional Office for Europe (2015). European Food and Nutrition Action Plan 2015–2020. Copenhagen: World Health Organization. WHO Regional Office for Europe (2016). Physical activity strategy for the WHO European Region 2016–2025. Copenhagen: World Health Organization. Woodcock J, Franco O, Orsini N & Roberts I (2011). Non-vigorous physical activity and all-cause mortality: systematic review and meta-analysis of cohort studies. Int J Epidemiol, 40:121–38. World Cancer Research Fund & American Institute for Cancer Research (2007). Food, nutrition, physical activity, and the prevention of cancer: a global perspective. Washington DC: American Institute for Cancer Research. World Health Assembly (2013). Follow-up to the Political Declaration of the High-level Meeting of the General Assembly on the Prevention and Control of Non-communicable Diseases. WHA66.10. URL: http://apps.who.int/gb/ebwha/pdf_files/WHA66/A66_R10-en.pdf (accessed 23 October 2016). World Health Organization (2004). Global Strategy on Diet, Physical Activity and Health. Geneva: World Health Organization. World Health Organization (2010). Global recommendations on physical activity for health. Geneva: World Health Organization. World Health Organization (2016). Global Health Observatory (GHO) data. URL: http://www. who.int/gho/en/ (accessed 15 June 2016). Zhang J & Chaaban J (2012). The economic cost of physical inactivity in China. Prev Med, 56:75–8. Zheng J, Huang T, Yu Y, Hu X, Yang B & Li D (2012). Fish consumption and CHD mortality: an updated meta-analysis of seventeen cohort studies. Public Health Nutr, 15:725–37. Zimmet P, Magliano D, Herman W & Shaw J (2014). Diabetes: a 21st century challenge. Lancet Diabetes Endocrinol, 2:56–64. Appendices Appendix 1 Summary overview of key characteristics of included studies Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings DIET Daviglus et al. (2005) To examine whether fruit and vegetable intake in middle age is related to health care costs (total and diseasespecific) in old age. The United States Based on diseasespecific costs (cardiovascular disease and cancer) and overall national health insurance costs. Intake of fruit and vegetables were determined from previous dietary histories and classified into one of three groups: <14 cups per month, 14–42 cups per month and >42 cups per month. Combined fruit and vegetable intake was classified as either low, medium or high. Using Medicare claims data, individual mean annual health care costs in old age and ten-year cumulative costs prior to death were estimated. The costs were linked to previous fruit and vegetable intake and examined across three intake groups to determine possible associations. Fruit and vegetable intake: Dietary interview (using Burke’s comprehensive dietary history method) Health care costs: estimated from national health insurance claims Direct health care costs (USD) 1984–2000 Although most relationships were not statistically significant, higher fruit and vegetable intakes were generally associated with lower mean annual health care costs. Mean annual costs for cardiovascular disease for high vs low intake were US$ 3 128 vs US$ 4 223; for cancer: US$ 1 352 vs US$ 1 640 and for total costs: US$ 10 024 vs US$ 12 211. Trends were generally similar for fruit or vegetable intake alone. This trend also applied to fruit and vegetable intake and ten-year cumulative costs before death (US$ 92 757 total cumulative costs for high intake vs US$ 132 713 for low intake). Rayner & Scarborough (2005) To quantify the disease and financial burden of ill health related to food. United Kingdom Disease-based: cardiovascular disease, type 2 diabetes, other hormonal and immune system diseases, cancer, dental caries and digestive system diseases. PAFs for diseases due to risk factors published by the WHO (2004) were applied to disease DALYs to estimate the proportion of disease burdens attributable to unhealthy diets. The same PAFs were applied to NHS disease costs. PAFs: Global Burden of Disease (GBD) study (1997), WHO Comparative Quantification of Health Risks (2004) Disease DALYS: WHO Report on Diet, Nutrition and the Prevention of Chronic Diseases (2003) NHS cost data: extrapolated from estimates from an earlier study for 1992/93 Direct health care costs (GBP) 2002 Food-related disease accounted for about 10% of morbidity and mortality in the United Kingdom (as measured by DALYs). Food-related disease was associated with an annual health care cost of £6 billion, or about one third of the total NHS costs. Appendix 1 57 Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings DIET Daviglus et al. (2005) To examine whether fruit and vegetable intake in middle age is related to health care costs (total and diseasespecific) in old age. The United States Based on diseasespecific costs (cardiovascular disease and cancer) and overall national health insurance costs. Intake of fruit and vegetables were determined from previous dietary histories and classified into one of three groups: <14 cups per month, 14–42 cups per month and >42 cups per month. Combined fruit and vegetable intake was classified as either low, medium or high. Using Medicare claims data, individual mean annual health care costs in old age and ten-year cumulative costs prior to death were estimated. The costs were linked to previous fruit and vegetable intake and examined across three intake groups to determine possible associations. Fruit and vegetable intake: Dietary interview (using Burke’s comprehensive dietary history method) Health care costs: estimated from national health insurance claims Direct health care costs (USD) 1984–2000 Although most relationships were not statistically significant, higher fruit and vegetable intakes were generally associated with lower mean annual health care costs. Mean annual costs for cardiovascular disease for high vs low intake were US$ 3 128 vs US$ 4 223; for cancer: US$ 1 352 vs US$ 1 640 and for total costs: US$ 10 024 vs US$ 12 211. Trends were generally similar for fruit or vegetable intake alone. This trend also applied to fruit and vegetable intake and ten-year cumulative costs before death (US$ 92 757 total cumulative costs for high intake vs US$ 132 713 for low intake). Rayner & Scarborough (2005) To quantify the disease and financial burden of ill health related to food. United Kingdom Disease-based: cardiovascular disease, type 2 diabetes, other hormonal and immune system diseases, cancer, dental caries and digestive system diseases. PAFs for diseases due to risk factors published by the WHO (2004) were applied to disease DALYs to estimate the proportion of disease burdens attributable to unhealthy diets. The same PAFs were applied to NHS disease costs. PAFs: Global Burden of Disease (GBD) study (1997), WHO Comparative Quantification of Health Risks (2004) Disease DALYS: WHO Report on Diet, Nutrition and the Prevention of Chronic Diseases (2003) NHS cost data: extrapolated from estimates from an earlier study for 1992/93 Direct health care costs (GBP) 2002 Food-related disease accounted for about 10% of morbidity and mortality in the United Kingdom (as measured by DALYs). Food-related disease was associated with an annual health care cost of £6 billion, or about one third of the total NHS costs. 58 Assessing the economic costs of unhealthy diets and low physical activity Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings DIET Rice & Normand (2012) To establish the annual public expenditures for patients with disease-related malnutrition (DRM). Republic of Ireland Based on overall public health and social care costs. Prevalence of DRM in multiple health care settings were estimated from age-standardised comparisons between available Irish data and large-scale United Kingdom surveys. Frequency of health care utilisation among adults with DRM was estimated, to which relevant unit costs were applied to obtain DRM-related costs. Prevalence of DRM: earlier studies which used the Malnutrition Universal Screening Tool, British Association for Parenteral and Enteral Nutrition Screening Week Surveys (2007, 2008, 2010) Patient activity levels and costs: Department of Health and Children, Health Service Executive, Economic and Social Research Institute Direct health care costs (EUR) 2007 The annual public health and social care costs associated with diseaserelated malnutrition among adults was over €1.4 billion or 10% of the health care budget. The majority of the costs were from acute hospital or residential care settings (70%). The additional cost of DRM per adult patient was €5357. Collins et al. (2011) To examine whether higher diet quality was associated with lower health care claims and costs among middle-aged adults. Australia Based on national health insurance costs. Individual dietary patterns were assessed using a survey consistent with national dietary recommendations. Based on survey results, the Australian Recommended Food Score (ARFS) was calculated and grouped into quintiles (1 = highest score; 5 = lowest score). Median fiveyear cumulative Medicare costs and six-year cumulative number of Medicare claims were linked to ARFS and examined across quintiles to determine possible associations between diet quality, number of health care claims and costs. Dietary intake: Survey (using the Dietary Questionnaire for Epidemiological Studies) Number of health care claims and costs: Medicare Australia Direct health care costs (AUD) Costs: 2002–2006 Number of claims: 2002–2007 There was a statistically significant association between five-year cumulative health care costs and the ARFS, with individuals with dietary scores in the highest quintile having higher health care costs compared to those in the lowest quintile (cost difference of AU$ 110). The trend was not consistent across quintiles. Individuals in the highest quintile had a lower number of claims compared to those in the lowest quintile (significant difference of 10 claims). This trend was not consistent in the medium term. Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings DIET Doidge et al. (2012) To quantify the potential effects of increasing dairy product consumption to recommended levels in terms of population health impact and direct health care costs. Australia Disease-based: type 2 diabetes, ischaemic heart disease, stroke, osteoporosis, obesity, hypertension. PAFs for diseases related to low dairy consumption were identified based on fine levels of dairy consumption (i.e. increments of 0.1 standard serving per day), and distribution among the population and corresponding RRs. Two variant-formulas were used to calculate the PAFs, depending on data availability. Resulting PAFs were applied to disease DALYs and health care costs to determine proportions potentially avoided by increasing dairy consumption. Dairy consumption levels and population distribution: Australian National Nutrition Survey (1995) RRs: Earlier studies representing the “highest level of evidence” (as per criteria by Phillips et al. (2009)) Disease DALYs and costs: Australian Institute of Health and Welfare, earlier studies, government reports, primary analysis of publicly available databases and government reports Direct health care costs (AUD) 2010–2011 Increasing dairy consumption to recommended levels can potentially prevent 18.4% of the incident cases of obesity, 10.2% of type 2 diabetes, 5% of ischaemic heart disease, 16.2% of stroke and 8.3% of hypertension. Increasing dairy consumption can potentially save AU$ 2.0 billion in direct health care costs, while also saving an additional 75 012 DALYs. The amount comprises 1.7% of total direct health care expenditures and is comparable with total spending on public health (AU$ 2.0 billion in 2009–2010). Lo et al. (2013) To assess the relationship between dietary quality and medical care utilisation and expenditures among populations aged ≥65 years. China (Taiwan) Based on national health insurance costs. Individual dietary intakes were assessed through a 24-hour dietary recall. Dietary quality was scored from 0 points (lowest) to 6 points (highest) through the Dietary Diversity Score (DDS) method. National health insurance claims in the succeeding eight years were linked to DDS scores and examined across quartiles (DDS scores of ≤3, 4, 5, 6) to identify possible associations between diet quality and level of medical utilisation and costs. Dietary intake: National Elderly Nutrition and Health Survey (1999–2000) Medical care utilisation and costs: National health insurance (covering >99% of the population) Direct health care costs (TWD) 1999–2006 Participants with better diet quality (as indicated by higher DDS) had lower utilisation of and costs for emergency and hospitalisation services. Average annual emergency costs were NT$ 2 330 vs NT$ 1 560 for DDS of ≤3 vs DDS of 6. Hospitalisation costs were NT$ 47 600 vs NT$ 35 100. For preventive and dental services, however, higher DDS predicted greater utilisation (0.25 and 0.5 times) and costs (+NT$ 270 and +NT$ 420) compared to the lowest DDS. Overall expenditures were still lower for those with higher DDS at NT$ 64 200 (DDS of 6) vs NT$ 68 300 (DDS ≤3). Appendix 1 59 Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings DIET Doidge et al. (2012) To quantify the potential effects of increasing dairy product consumption to recommended levels in terms of population health impact and direct health care costs. Australia Disease-based: type 2 diabetes, ischaemic heart disease, stroke, osteoporosis, obesity, hypertension. PAFs for diseases related to low dairy consumption were identified based on fine levels of dairy consumption (i.e. increments of 0.1 standard serving per day), and distribution among the population and corresponding RRs. Two variant-formulas were used to calculate the PAFs, depending on data availability. Resulting PAFs were applied to disease DALYs and health care costs to determine proportions potentially avoided by increasing dairy consumption. Dairy consumption levels and population distribution: Australian National Nutrition Survey (1995) RRs: Earlier studies representing the “highest level of evidence” (as per criteria by Phillips et al. (2009)) Disease DALYs and costs: Australian Institute of Health and Welfare, earlier studies, government reports, primary analysis of publicly available databases and government reports Direct health care costs (AUD) 2010–2011 Increasing dairy consumption to recommended levels can potentially prevent 18.4% of the incident cases of obesity, 10.2% of type 2 diabetes, 5% of ischaemic heart disease, 16.2% of stroke and 8.3% of hypertension. Increasing dairy consumption can potentially save AU$ 2.0 billion in direct health care costs, while also saving an additional 75 012 DALYs. The amount comprises 1.7% of total direct health care expenditures and is comparable with total spending on public health (AU$ 2.0 billion in 2009–2010). Lo et al. (2013) To assess the relationship between dietary quality and medical care utilisation and expenditures among populations aged ≥65 years. China (Taiwan) Based on national health insurance costs. Individual dietary intakes were assessed through a 24-hour dietary recall. Dietary quality was scored from 0 points (lowest) to 6 points (highest) through the Dietary Diversity Score (DDS) method. National health insurance claims in the succeeding eight years were linked to DDS scores and examined across quartiles (DDS scores of ≤3, 4, 5, 6) to identify possible associations between diet quality and level of medical utilisation and costs. Dietary intake: National Elderly Nutrition and Health Survey (1999–2000) Medical care utilisation and costs: National health insurance (covering >99% of the population) Direct health care costs (TWD) 1999–2006 Participants with better diet quality (as indicated by higher DDS) had lower utilisation of and costs for emergency and hospitalisation services. Average annual emergency costs were NT$ 2 330 vs NT$ 1 560 for DDS of ≤3 vs DDS of 6. Hospitalisation costs were NT$ 47 600 vs NT$ 35 100. For preventive and dental services, however, higher DDS predicted greater utilisation (0.25 and 0.5 times) and costs (+NT$ 270 and +NT$ 420) compared to the lowest DDS. Overall expenditures were still lower for those with higher DDS at NT$ 64 200 (DDS of 6) vs NT$ 68 300 (DDS ≤3). 60 Assessing the economic costs of unhealthy diets and low physical activity Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Katzmarzyk et al. (2000) To estimate the mortality and economic costs of physical inactivity and the effect of a 10% reduction in inactivity levels on costs. Canada Disease-based: coronary artery disease, stroke, hypertension, colon cancer, breast cancer, type 2 diabetes mellitus, osteoporosis. PAFs were calculated based on RRs from earlier meta-analyses and large prospective, epidemiological studies. The PAFs were applied to total numbers of premature deaths per disease and disease costs. A hypothetical scenario was assumed where inactivity prevalence was reduced by 10%, under which new PAFs were calculated and applied to disease costs. Physical activity prevalence: Physical Activity Monitor Survey, Canadian Fitness and Lifestyle Research Institute (1999) RRs: Earlier meta-analyses and prospective studies Premature mortality: Statistics Canada (1995) Cost data: Canadian Health Expenditures Database, Economic Burden of Illness Canada (1993), American Heart Association (1999), earlier studies Direct health care costs (CAD) 1999 Physical inactivity was responsible for 10.3% (21 000) of deaths from all causes in 1995, through the seven diseases. Physical inactivity accounted for C$ 2.1 billion or 2.5% of total direct health care costs. The amount represented 25.5% of the total costs of treating the seven diseases. Reducing the prevalence of physical inactivity by 10% would reduce direct health care expenditures by C$ 150 million per year. Ackermann et al. (2003) To determine if changes in health care utilisation and costs for Medicareeligible enrolees of a large health maintenance organization (HMO) are related to their choice to participate in a community exercise programme offered as a health benefit. The United States Based on private health insurance costs. A retrospective, matched cohort study was conducted to determine if changes in health care costs for Medicare-eligible enrolees (aged ≥65 years) choosing to participate in the exercise programme were different from those of similar individuals who did not participate. Three enrolees who never attended the exercise programme were randomly selected as controls for each participant matching on age and gender. Changes in health care utilisation and costs were compared between the intervention and control groups. Exercise programme participation and cost data: Group Health Cooperative of Puget Sound, Decision Support System (integrates clinical information, units of service and actual costs) Direct health care costs (USD) 1997–2000 Overall, no significant differences were found in the total health care costs between exercise programme participants and non-participants. However, compared to those who did not participate, programme participants had 4.9% lower hospitalisations and displayed a stronger trend towards lower inpatient costs (i.e. US$ 708 lower). Among those who had high programme usage (i.e. ≥1 visit per week), annual total health care costs were US$ 1 057 lower and the hospitalisation risk was 7.9% lower compared to controls. These significant effects among higher users were observed at a relatively low attendance rate (average of 1.74 visits per week) and under normal daily operating conditions of the programme. Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Martinson et al. (2003) To examine the impact of changes in physical activity over the course of a single year on short-term changes in health care charges among members of a health plan aged 50 and above. The United States Based on commercial health insurance costs. Health plan members were surveyed twice for physical activity levels (in August 1995 and September 1996). Individuals were classified as active or inactive at each of the two time points. To account for alternative definitions of physical activity, five definitions of inactivity and activity were used: A: Inactive = 0 days, active = 1+ days B: Inactive = 0 days, active = 2+ days C: Inactive = 0–1 days, active = 2+ days D: Inactive = 0–1 days, active = 3+ days E: Inactive = 0–1 days, active = 4+ days A specific definition was used in categorising an individual’s physical activity changes between August 1995 and September 1996 into one of five mutually exclusive groups: (1) inactive to inactive, (2) active to active, (3) active to inactive, (4) inactive to active and (5) unclassified (falling within the “gap”” between the inactive and active categories under B, D and E). Changes in health care charges between September 1994 to August 1995 and September 1996 to August 1997 were examined across 25 physical activity change groups to determine possible associations. Physical activity levels: HealthPartners Minnesota health plan members’ survey Health care charges: commercial health insurance Direct health care costs (USD) 1994–1995 and 1996–1997 For physical inactivity/activity definitions A and B, physically active groups had larger decreases in health care charges relative to individuals who were physically inactive in both surveys. These declines ranged from US$ 1 200 to US$ 1 900. For definitions C, D and E, declines in total charges were observed primarily among individuals who moved from inactivity to activity. In particular for definition D, the decline in charges was significantly larger (in the inactive to active group) at US$ 2 200. Similar trends are seen in models using definitions C and E but the results are not significant. Garrett et al. (2004) To estimate the costs of physical inactivity among members of the Blue Cross Blue Shield health plan in Minnesota. The United States Diseasebased: heart disease, stroke, hypertension, type 2 diabetes, colon cancer, breast cancer, osteoporosis, depression, anxiety. PAFs were calculated based on RRs from earlier meta-analyses of published studies, stratified by level of exposure: inactive and irregularly active. The PAFs were applied to disease costs. Physical activity prevalence: Behavioural Risk Factors Surveillance System (2000) RRs: Earlier meta-analyses of published studies Cost data: Blue Cross and Blue Shield health plan Direct health care costs (USD) 2000 Physical inactivity was estimated to account for US$ 83.6 million of total medical expenditures among a health plan population of 1.5 million members. The cost per member was US$ 56. Appendix 1 61 Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Martinson et al. (2003) To examine the impact of changes in physical activity over the course of a single year on short-term changes in health care charges among members of a health plan aged 50 and above. The United States Based on commercial health insurance costs. Health plan members were surveyed twice for physical activity levels (in August 1995 and September 1996). Individuals were classified as active or inactive at each of the two time points. To account for alternative definitions of physical activity, five definitions of inactivity and activity were used: A: Inactive = 0 days, active = 1+ days B: Inactive = 0 days, active = 2+ days C: Inactive = 0–1 days, active = 2+ days D: Inactive = 0–1 days, active = 3+ days E: Inactive = 0–1 days, active = 4+ days A specific definition was used in categorising an individual’s physical activity changes between August 1995 and September 1996 into one of five mutually exclusive groups: (1) inactive to inactive, (2) active to active, (3) active to inactive, (4) inactive to active and (5) unclassified (falling within the “gap”” between the inactive and active categories under B, D and E). Changes in health care charges between September 1994 to August 1995 and September 1996 to August 1997 were examined across 25 physical activity change groups to determine possible associations. Physical activity levels: HealthPartners Minnesota health plan members’ survey Health care charges: commercial health insurance Direct health care costs (USD) 1994–1995 and 1996–1997 For physical inactivity/activity definitions A and B, physically active groups had larger decreases in health care charges relative to individuals who were physically inactive in both surveys. These declines ranged from US$ 1 200 to US$ 1 900. For definitions C, D and E, declines in total charges were observed primarily among individuals who moved from inactivity to activity. In particular for definition D, the decline in charges was significantly larger (in the inactive to active group) at US$ 2 200. Similar trends are seen in models using definitions C and E but the results are not significant. Garrett et al. (2004) To estimate the costs of physical inactivity among members of the Blue Cross Blue Shield health plan in Minnesota. The United States Diseasebased: heart disease, stroke, hypertension, type 2 diabetes, colon cancer, breast cancer, osteoporosis, depression, anxiety. PAFs were calculated based on RRs from earlier meta-analyses of published studies, stratified by level of exposure: inactive and irregularly active. The PAFs were applied to disease costs. Physical activity prevalence: Behavioural Risk Factors Surveillance System (2000) RRs: Earlier meta-analyses of published studies Cost data: Blue Cross and Blue Shield health plan Direct health care costs (USD) 2000 Physical inactivity was estimated to account for US$ 83.6 million of total medical expenditures among a health plan population of 1.5 million members. The cost per member was US$ 56. 62 Assessing the economic costs of unhealthy diets and low physical activity Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Kuriyama et al. (2004) To examine the joint impact of physical inactivity, including smoking and obesity, on direct health care charges. Japan Based on national health insurance costs. Risk factor status of participants in terms of BMI, physical activity and smoking status were assessed. Participants’ health insurance claims and costs were then followed-up prospectively for seven years. They were classified into eight risk groups (i.e. different combinations of the three risk factors) and examined in terms of monthly per capita health care charges to determine the cost impact of different risk factor combinations. Physical activity, BMI, smoking data: study interview Cost data: national health insurance (covers 35% of the Japanese population) Direct health care costs (USD) 1995–2001 Participants without risk (i.e. never smoking, with normal BMI and physically active) had mean monthly per capita health care charges of US$ 171.6. Compared to this group, the presence of physical inactivity alone increased per capita costs by 8% (US$ 185.3), smoking and physical inactivity by 31.4% (US$ 225.4) and obesity and physical inactivity by 16.4% (US$ 199.8). Presence of all three risk factors increased per capita costs by 42.6% (US$ 244.7). Wang et al. (2004) To estimate the costs of cardiovascular disease (CVD) associated with physical inactivity. The United States Disease-based: CVD (including coronary heart disease, hypertension, stroke and rheumatic heart disease). Individual medical expenditure data were linked to physical activity status in the previous year. Physical activity was categorised as active and inactive. Medical expenditures on CVD associated with inactivity were derived by comparing mean medical costs between population groups stratified by CVD status and physical activity status and obtaining the difference. CVD status and physical activity: National Health Interview Survey (1995) Cost data: Medical Expenditure Panel Survey (1996) Direct health care costs (USD) 1996 Among 7.3 million cases of CVD, 1.1 million or 15.3% were associated with physical inactivity. The total medical expenditure of persons with CVD was US$ 41.3 billion, of which US$ 5.4 billion (13.1%) was associated with physical inactivity. Applying the percentages to the national health and economic burden, 9.2 million CVD cases in the United States were associated with physical inactivity in 2001, costing US$ 23.7 billion in direct medical expenditures. Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Anderson et al. (2005) To estimate the proportion of total health care charges associated with physical inactivity, overweight and obesity among populations aged 40 and older. The United States Based on commercial health insurance costs. A predictive model of health care charges was developed through preliminary analysis of data from 8 000 health plan members. Five variables were assessed in association with health care charges: physical activity, BMI, chronic disease status (i.e. with diabetes and/ or hypertension or none) and smoking status. Counterfactual health care charges were also estimated by reclassifying all individuals as normal weight and physically active, leaving other characteristics unchanged. Model cells for hypothetical 200 000 health plan members were then developed and defined by status of BMI, physical activity and other co-variates. Total health care charges were estimated by multiplying the percentage of the health plan population in each cell with the predicted charges per cell and summing all cells. Counterfactual estimates were computed by using counterfactual charges. The difference between current risk profile total charges and counterfactual total charges were computed as charges associated with physical inactivity, overweight and obesity. The same calculation was performed but using national population percentages to identify national cost estimates. BMI, physical activity and other variable data: HealthPartners Minnesota health plan survey (1995) Health care charges: commercial health insurance National population percentages: National Health Interview Survey (2001) Direct health care costs (USD) 1996–1999 For a health plan with 200 000 white members aged 40 and above, total annualised health care charges of US$ 1.12 billion was estimated, 23% of which (US$ 236 million) was associated with physical inactivity and overweight or obesity. The three sub-populations with the largest charges associated with physical inactivity, overweight and obesity were: 1) men aged 50–64 without chronic disease (US$ 44.7 million) 2) men aged 65 and older with chronic disease (US$ 43.7 million) 3) men aged 40–49 with no chronic disease (US$ 41.7 million). At the national level, the percentage of national health care charges associated with physical inactivity, overweight and obesity was estimated to be 27%. Appendix 1 63 Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Anderson et al. (2005) To estimate the proportion of total health care charges associated with physical inactivity, overweight and obesity among populations aged 40 and older. The United States Based on commercial health insurance costs. A predictive model of health care charges was developed through preliminary analysis of data from 8 000 health plan members. Five variables were assessed in association with health care charges: physical activity, BMI, chronic disease status (i.e. with diabetes and/ or hypertension or none) and smoking status. Counterfactual health care charges were also estimated by reclassifying all individuals as normal weight and physically active, leaving other characteristics unchanged. Model cells for hypothetical 200 000 health plan members were then developed and defined by status of BMI, physical activity and other co-variates. Total health care charges were estimated by multiplying the percentage of the health plan population in each cell with the predicted charges per cell and summing all cells. Counterfactual estimates were computed by using counterfactual charges. The difference between current risk profile total charges and counterfactual total charges were computed as charges associated with physical inactivity, overweight and obesity. The same calculation was performed but using national population percentages to identify national cost estimates. BMI, physical activity and other variable data: HealthPartners Minnesota health plan survey (1995) Health care charges: commercial health insurance National population percentages: National Health Interview Survey (2001) Direct health care costs (USD) 1996–1999 For a health plan with 200 000 white members aged 40 and above, total annualised health care charges of US$ 1.12 billion was estimated, 23% of which (US$ 236 million) was associated with physical inactivity and overweight or obesity. The three sub-populations with the largest charges associated with physical inactivity, overweight and obesity were: 1) men aged 50–64 without chronic disease (US$ 44.7 million) 2) men aged 65 and older with chronic disease (US$ 43.7 million) 3) men aged 40–49 with no chronic disease (US$ 41.7 million). At the national level, the percentage of national health care charges associated with physical inactivity, overweight and obesity was estimated to be 27%. 64 Assessing the economic costs of unhealthy diets and low physical activity Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Wang et al. (2005) To explore the relationship of physical activity with shortterm health care costs across BMI groups among individuals aged 65 and above. The United States Based on national health insurance and indemnity/ PPO insurance costs. Physical activity of participants was assessed by the number of times per week they engage in physical activity “hard enough to induce heavy breathing and fast heart beat” for at least 20 minutes. Participants were classified into three groups: sedentary, moderately active and very active. They were also divided by BMI group (normal weight, overweight, obese). In all three BMI groups, health care costs were linked to physical activity levels to determine possible associations. Analysis was also done in specific age groups (65–69, 70–74 and 75+ years). Physical activity prevalence: study interview (using the modified Health Risk Appraisal questionnaire) Cost data: national health insurance (Medicare), indemnity/PPO plans Direct health care costs (USD) 2001–2002 Higher levels of physical activity predicted lower short-term health care costs for older individuals across the three BMI groups. Moderately active retirees had US$ 1 456, US$ 1 731 and US$ 1 177 lower total health care costs than their sedentary counterparts in the normal weight, overweight and obese groups, respectively. The very active retirees had US$ 1 823, US$ 581 and US$ 1 379 lower costs than moderately active retirees in the same BMI groupings, respectively. The same association between physical activity and short-term health care costs applies when analysis is done in specific age groups (65–69, 70–74, 75+ years). Allender et al. (2007) To estimate the health and economic burden of physical inactivity. United Kingdom Disease-based: ischaemic heart disease, ischaemic stroke, breast cancer, colon/ rectum cancer, type 2 diabetes. PAFs published by the WHO (2004) stratified by sex were applied to disease DALYs in the WHO EUR-A region to determine health burdens attributable to physical inactivity. The same PAFs were applied to NHS disease costs. Disease DALYs: WHO (2003) Cost data: extrapolated from estimates from earlier studies for 1992/93 Direct health care costs (GBP) 2002 Physical inactivity was estimated to be associated with 3% of DALYs lost. Physical inactivity was estimated to be associated with £1.06 billion in direct health care costs. Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Katzmarzyk (2011) To estimate the economic burden of physical inactivity. Canada Disease-based: coronary heart disease, stroke, hypertension, colon cancer, breast cancer, type 2 diabetes, osteoporosis. PAFs were calculated based on summary RRs derived in a previous meta-analysis of prospective longitudinal studies. The PAFs were applied to disease health care and productivity costs. Physical activity prevalence: Canadian Community Health Survey (2009) RRs: earlier meta-analysis of studies Cost data: Economic Burden of Illness in Canada (1993, 1998), Health Canada (2002), National Cancer Institute of Canada (2002), Canadian Institute for Health Information (2010), American Heart Association (2002), US Centers for Disease Prevention and Control, earlier studies Direct health care and indirect productivity costs (CAD) 2009 Physical inactivity in Ontario was associated with a cost of C$ 3.4 billion (C$ 1.02 billion in direct costs and C$ 2.34 billion in indirect costs). After extrapolation to the national situation, estimated costs reach C$ 8.6 billion (C$ 2.6 billion direct, C$ 5.95 billion indirect). Appendix 1 65 Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Katzmarzyk (2011) To estimate the economic burden of physical inactivity. Canada Disease-based: coronary heart disease, stroke, hypertension, colon cancer, breast cancer, type 2 diabetes, osteoporosis. PAFs were calculated based on summary RRs derived in a previous meta-analysis of prospective longitudinal studies. The PAFs were applied to disease health care and productivity costs. Physical activity prevalence: Canadian Community Health Survey (2009) RRs: earlier meta-analysis of studies Cost data: Economic Burden of Illness in Canada (1993, 1998), Health Canada (2002), National Cancer Institute of Canada (2002), Canadian Institute for Health Information (2010), American Heart Association (2002), US Centers for Disease Prevention and Control, earlier studies Direct health care and indirect productivity costs (CAD) 2009 Physical inactivity in Ontario was associated with a cost of C$ 3.4 billion (C$ 1.02 billion in direct costs and C$ 2.34 billion in indirect costs). After extrapolation to the national situation, estimated costs reach C$ 8.6 billion (C$ 2.6 billion direct, C$ 5.95 billion indirect). 66 Assessing the economic costs of unhealthy diets and low physical activity Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Alter et al. (2012) To estimate cumulative outcomes and costs associated with obesity alone and in combination with physical inactivity, including smoking and psychosocial distress among middle-aged adults. Canada Based on public health insurance costs. Risk factor status in terms of BMI, physical activity, psychosocial status and smoking status were assessed through a telephone survey. Individual risk factor data were linked with health care utilisation and costs and followed longitudinally for 11.5 years. Propensity-matching was done where each exposed individual was matched in 1:1 fashion to a healthy, normal weight person, based on age, gender, socioeconomic status, comorbidity and non-BMI risk factors. Differences in costs between the exposed and unexposed matched pairs were tallied and averaged throughout the follow-period and attributed to the corresponding risk factor. Risk factor data: National Population Health Survey (1994–96) Cost data: public health insurance, Canadian Institutes of Health Information, Ontario Case Costing Initiative, Ontario Drug Benefits formulary Incidence of diabetes and hypertension: Ontario diabetes and hypertension databases Mortality: registered persons database Direct health care costs (CAD) 1994/95/96 – 2005/06/07 Cumulative costs associated with obesity alone were C$ 8 294.67 per person, not significantly higher compared to costs among propensity-matched normal weight controls (C$ 7 323.59 per person). Obesity in combination with other lifestyle factors was associated with significantly higher cumulative expenditures as compared with normal-weight healthy matched controls. Excess costs were estimated to be: Overweight + physically inactive: C$ 1 095.30 Obese + physically inactive: C$ 4 079.47 Overweight + physically inactive + smoking: C$ 2 026.10 Obese + physically inactive + smoking: C$ 2 632.91 Overweight + physically inactive + distressed: C$ 1 868.57 Obese + physically inactive + distressed: C$ 7 156.94 Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Janssen (2012) To estimate the economic burden of physical inactivity. Canada Disease-based: coronary artery disease, stroke, hypertension, colon cancer, female breast cancer, type 2 diabetes, osteoporosis. PAFs were calculated based on summary RRs derived in earlier meta-analyses of RRs in prospective cohort studies, stratified by sex. PAFs were applied to disease direct and indirect costs. Physical activity prevalence: Canadian Health Measures Survey (2007–2009) RRs: earlier meta-analyses of studies Cost data: Economic Burden of Illness in Canada (2000) extrapolated to 2009 values Direct health care and indirect productivity costs (CAD) 2009 Physical inactivity costs C$ 6.8 billion (C$ 2.4 billion in direct costs and C$ 4.3 billion in indirect costs). Direct costs represent 3.8% of the overall health care costs. Zhang and Chaaban (2012) To estimate the economic burden of physical inactivity. China Disease-based: coronary heart disease, stroke, hypertension, cancer, type 2 diabetes. Impacts of physical inactivity were assessed through direct mechanisms (inactivity to NCDs) and indirect mechanisms (inactivity to overweight/obesity to NCDs). PAFs were calculated for each disease related to physical inactivity, overweight and obesity, based on RRs from earlier meta-analyses of prospective studies and single cohort studies. PAFs related to overweight and obesity were multiplied by 12% (i.e. the proportion of overweight and obesity attributable to physical inactivity). All three sets of PAFs were summed and the totals applied to disease direct and indirect costs. Physical activity, overweight, obesity prevalence: Chinese Behavioural Risk Factors Surveillance Survey (2007) RRs: Earlier meta-analyses and single cohort studies Cost data: National Health Service Survey (2003) extrapolated to 2007 Direct health care and indirect productivity costs (USD) 2007 Economic costs associated with physical inactivity were US$ 6.7 billion (US$ 3.5 billion for direct costs and US$ 3.3 billion for indirect costs). This was equivalent to 15.2% of the total costs of the five diseases, 5.3% of total NCD costs and 3.8% of total costs for all diseases. Appendix 1 67 Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Janssen (2012) To estimate the economic burden of physical inactivity. Canada Disease-based: coronary artery disease, stroke, hypertension, colon cancer, female breast cancer, type 2 diabetes, osteoporosis. PAFs were calculated based on summary RRs derived in earlier meta-analyses of RRs in prospective cohort studies, stratified by sex. PAFs were applied to disease direct and indirect costs. Physical activity prevalence: Canadian Health Measures Survey (2007–2009) RRs: earlier meta-analyses of studies Cost data: Economic Burden of Illness in Canada (2000) extrapolated to 2009 values Direct health care and indirect productivity costs (CAD) 2009 Physical inactivity costs C$ 6.8 billion (C$ 2.4 billion in direct costs and C$ 4.3 billion in indirect costs). Direct costs represent 3.8% of the overall health care costs. Zhang and Chaaban (2012) To estimate the economic burden of physical inactivity. China Disease-based: coronary heart disease, stroke, hypertension, cancer, type 2 diabetes. Impacts of physical inactivity were assessed through direct mechanisms (inactivity to NCDs) and indirect mechanisms (inactivity to overweight/obesity to NCDs). PAFs were calculated for each disease related to physical inactivity, overweight and obesity, based on RRs from earlier meta-analyses of prospective studies and single cohort studies. PAFs related to overweight and obesity were multiplied by 12% (i.e. the proportion of overweight and obesity attributable to physical inactivity). All three sets of PAFs were summed and the totals applied to disease direct and indirect costs. Physical activity, overweight, obesity prevalence: Chinese Behavioural Risk Factors Surveillance Survey (2007) RRs: Earlier meta-analyses and single cohort studies Cost data: National Health Service Survey (2003) extrapolated to 2007 Direct health care and indirect productivity costs (USD) 2007 Economic costs associated with physical inactivity were US$ 6.7 billion (US$ 3.5 billion for direct costs and US$ 3.3 billion for indirect costs). This was equivalent to 15.2% of the total costs of the five diseases, 5.3% of total NCD costs and 3.8% of total costs for all diseases. 68 Assessing the economic costs of unhealthy diets and low physical activity Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Chevan & Roberts (2014) To examine the association between leisure-time physical activity levels and short-term health care expenditures. The United States Based on shortterm public and/ or private health care costs. Physical activity was assessed by the number of minutes of vigorous- and moderate-intensity activity completed per week. By equating 2 minutes of moderate activity to 1 minute of vigorous activity, physical activity was categorised in two ways. The first categorised participants as to: (1) whether they met the physical activity guidelines for strength and aerobic activity, (2) met guidelines only for strength, (3) met guidelines only for aerobic activity and (4) not meeting guidelines for both. The second focused on aerobic activity and split up participants as completing 0, <75, 75–149, 150–299 or >300 minutes per week. The two categorisations characterised weekly leisure-time physical activity levels which were linked to individual health care expenditures to determine possible associations. Physical activity prevalence: National Health Interview Survey (2006, 2007) Cost data: Medical Expenditures Panel Survey (2007–2008 and 2008–2009) Direct health care costs (USD) 2007–2009 Whether in terms of meeting the physical activity guidelines (i.e. for strength and aerobic activity) or the number of minutes spent on aerobic activity per week, not a single level of leisure-time physical activity had a significant association with all types of short-term health care expenditures. Maresova (2014) To estimate the economic burden of physical inactivity. Czech Republic Disease-based: coronary heart disease, ischaemic stroke, diabetes type 2, female breast cancer, colon cancer. PAFs calculated from disease-specific RRs derived from WHO (2004) stratified by age group (15–69, 70–79, 80+) and physical activity level (level 1: inactive; level 2: insufficiently active). The PAFs were applied to disease costs. Mortality data: Czech Statistical Office (CZSO) Morbidity (DALYs): WHO GBD study (2004) Physical activity prevalence: Czech Republic European Health Interview Survey (EHIS) Cost data: health insurance (covering approx. 75% of health care expenditure in the Czech Republic); earlier studies Direct health care costs (CZK) 2008 Physical inactivity accounted for 2 442 of all deaths (2.3%) and 18 065 DALYs (1.24%). Physical inactivity accounted for almost CZK 700 million (~€29 million) of public health insurance expenditure (0.4% of total expenditure). Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Peeters et al. (2014) To estimate the costs associated with prolonged sitting and physical inactivity among middleaged women. Australia Based on national health insurance costs. Participants were surveyed in terms of time spent sitting, walking and in moderate and vigorous leisure-time activities. Sitting time was categorised as low, moderate and high. Physical activity was categorised as inactive, low, moderate and high. Combined sitting time and physical activity or “activity patterns” were categorised into active + low sitting time, active + high sitting time, inactive + low sitting time and inactive + high sitting time. Medicare costs averaged over the survey year ±1 year were used to calculate the annual costs. Annual median costs were linked to sitting time, physical activity and activity patterns and examined across groups to determine possible associations. Analysis was also done by BMI strata to examine for potential effect modifications of BMI. Sitting time and physical activity: Australian Longitudinal Study on Women’s Health (2001, 2005, 2007, 2010) Cost data: National health insurance Direct health care costs (AUD) 2010 The annual median costs for highly active vs inactive individuals were AU$ 689 vs AU$ 741, with a significant difference of AU$ 94. In terms of sitting time, annual median costs for people with low sitting time vs high sitting time were AU$ 671 vs AU$ 709, with a difference of AU$ 16. No statistically significant associations were found between sitting time and costs. A high sitting time did not add to inactivityassociated costs (AU$ 110 higher for inactive people with high sitting time vs active people with low sitting time). Although costs are higher for overweight and obese groups compared to normal weight, the effects of physical activity on costs were similar across BMI ranges, suggesting no BMI interaction effects. Appendix 1 69 Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Peeters et al. (2014) To estimate the costs associated with prolonged sitting and physical inactivity among middleaged women. Australia Based on national health insurance costs. Participants were surveyed in terms of time spent sitting, walking and in moderate and vigorous leisure-time activities. Sitting time was categorised as low, moderate and high. Physical activity was categorised as inactive, low, moderate and high. Combined sitting time and physical activity or “activity patterns” were categorised into active + low sitting time, active + high sitting time, inactive + low sitting time and inactive + high sitting time. Medicare costs averaged over the survey year ±1 year were used to calculate the annual costs. Annual median costs were linked to sitting time, physical activity and activity patterns and examined across groups to determine possible associations. Analysis was also done by BMI strata to examine for potential effect modifications of BMI. Sitting time and physical activity: Australian Longitudinal Study on Women’s Health (2001, 2005, 2007, 2010) Cost data: National health insurance Direct health care costs (AUD) 2010 The annual median costs for highly active vs inactive individuals were AU$ 689 vs AU$ 741, with a significant difference of AU$ 94. In terms of sitting time, annual median costs for people with low sitting time vs high sitting time were AU$ 671 vs AU$ 709, with a difference of AU$ 16. No statistically significant associations were found between sitting time and costs. A high sitting time did not add to inactivityassociated costs (AU$ 110 higher for inactive people with high sitting time vs active people with low sitting time). Although costs are higher for overweight and obese groups compared to normal weight, the effects of physical activity on costs were similar across BMI ranges, suggesting no BMI interaction effects. 70 Assessing the economic costs of unhealthy diets and low physical activity Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Bachmann et al. (2015) To evaluate the association of health care costs in later life with cardiorespiratory fitness (an objective measure of habitual physical activity) in mid-life after adjustment for cardiovascular risk factors. The United States Based on public and/or private health care costs. Participants’ cardiorespiratory fitness at mid-life (mean age: 49 years) was assessed from a treadmill test and measured in terms of maximal metabolic equivalents (METs) achieved, categorised into age- and sexspecific quintiles of fitness. The quintiles were combined into three fitness groupings, into which participants were assigned: low fit (quintile 1), moderate fit (quintiles 2 and 3) and high fit (quintiles 4 and 5). Health care costs were followed up later in old age for an average period of 6.5 years, from the date of initiating Medicare coverage until death or at the end of follow-up. Average annual health care costs were examined across fitness groupings to determine possible associations. Treadmill test data: Cooper Center Longitudinal Study Cost data: national health insurance (Medicare), Carrier, Durable Medical Equipment, Home Health Agency, Hospice, Center for Medicare and Medicaid Services Direct health care costs (USD) 1999–2009 Health care costs among those aged 65 years and above were significantly associated with cardiorespiratory fitness in midlife. Compared to men with low cardiorespiratory fitness, men with high cardiorespiratory fitness had significantly lower health care costs (i.e. US$ 12 811 vs US$ 7 569). The same trend applied to women (US$ 10 029 vs US$ 6 056). Average annual health care costs were incrementally lower per MET achieved (i.e. 6.8% and 6.7% decrease in costs per MET achieved in mid-life in men and women, respectively). The associations persisted when analysis was done between cohorts who died during the follow-up and those who survived. Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Carlson et al. (2015) To examine the association between leisure-time aerobic physical activity and health care expenditures with and without adjusting for obesity status. The United States Based on public and/or private health care costs. Physical activity was assessed through the frequency and, if applicable, the duration of leisure-time activity participated in for at least 10 minutes at a time in vigorous- and light- or moderate-intensity activities. Minutes of moderate-intensity equivalent activity were calculated by counting 1 minute of vigorousintensity activity as 2 minutes of light- or moderate-intensity activity. Participants were classified into three activity levels: 1) active, 2) insufficiently active and 3) inactive. Physical activity levels were linked to individual health care costs to determine mean and per cent differences in expenditures. The sum of the differences in costs was divided by the total predicted expenditures for all groups to determine the percentage of aggregate costs associated with inadequate levels of physical activity. Two models were used: with and without adjustment for BMI. Physical activity prevalence: National Health Interview Survey (2004–2010) Cost data: Medical Expenditure Panel Survey (2006–2011) Direct health care costs (USD) 2006–2011 Higher levels of leisure-time aerobic physical activity were significantly associated with lower health care costs. There was a mean difference of 29.9% in annual per capita costs between active and inactive groups, as inactive individuals paid an additional cost of US$ 1 437. Insufficiently active individuals paid an additional cost of US$ 713 compared to active individuals (mean difference: 15.4%). After adjusting for BMI, the means of annual expenditure and per cent differences decreased but remained significant (26.6% or US$ 1 313 between active and inactive groups and 12.1% or US$ 576 between insufficiently active and inactive groups). Before adjustment for BMI, the percentage of aggregate health care expenditures associated with inadequate levels of physical activity was 12.5% (US$ 131 billion) and remained significant at 11.1% (US$ 117 billion) after adjusting for BMI. When adults with any reported difficulty walking due to a health problem were excluded from the analysis, the percentage was 8.7% (US$ 79 billion). Appendix 1 71 Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Carlson et al. (2015) To examine the association between leisure-time aerobic physical activity and health care expenditures with and without adjusting for obesity status. The United States Based on public and/or private health care costs. Physical activity was assessed through the frequency and, if applicable, the duration of leisure-time activity participated in for at least 10 minutes at a time in vigorous- and light- or moderate-intensity activities. Minutes of moderate-intensity equivalent activity were calculated by counting 1 minute of vigorousintensity activity as 2 minutes of light- or moderate-intensity activity. Participants were classified into three activity levels: 1) active, 2) insufficiently active and 3) inactive. Physical activity levels were linked to individual health care costs to determine mean and per cent differences in expenditures. The sum of the differences in costs was divided by the total predicted expenditures for all groups to determine the percentage of aggregate costs associated with inadequate levels of physical activity. Two models were used: with and without adjustment for BMI. Physical activity prevalence: National Health Interview Survey (2004–2010) Cost data: Medical Expenditure Panel Survey (2006–2011) Direct health care costs (USD) 2006–2011 Higher levels of leisure-time aerobic physical activity were significantly associated with lower health care costs. There was a mean difference of 29.9% in annual per capita costs between active and inactive groups, as inactive individuals paid an additional cost of US$ 1 437. Insufficiently active individuals paid an additional cost of US$ 713 compared to active individuals (mean difference: 15.4%). After adjusting for BMI, the means of annual expenditure and per cent differences decreased but remained significant (26.6% or US$ 1 313 between active and inactive groups and 12.1% or US$ 576 between insufficiently active and inactive groups). Before adjustment for BMI, the percentage of aggregate health care expenditures associated with inadequate levels of physical activity was 12.5% (US$ 131 billion) and remained significant at 11.1% (US$ 117 billion) after adjusting for BMI. When adults with any reported difficulty walking due to a health problem were excluded from the analysis, the percentage was 8.7% (US$ 79 billion). 72 Assessing the economic costs of unhealthy diets and low physical activity Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Codogno et al. (2015) To analyse the association between physical inactivity in different domains and public health expenditures and identify whether clustering of physical inactivity in these domains contributes to increased health care costs. Brazil Based on public primary health care costs. Participants’ primary health care costs were estimated in the last 12 months prior to interview on physical activity. Individual habitual physical activity was assessed in three domains (i.e work, sports and leisure-time) using the Baecke Questionnaire and scored. Scores were categorised into quartiles: the bottom quartile with scores ≤25, middle quartiles with scores ≥25 but ≤75 and the high quartile with scores ≥75. Participants were analysed according to in how many domains of physical activity they were categorised in the bottom quartile (i.e. from 0 to 3 times). This frequency of being in the bottom quartile was correlated with annual expenditure data. Cost data: Basic Healthcare Unit, Sao Paolo, Brazil Physical activity: study interview Direct health care costs (USD) 2009 Lower physical activity (i.e. scores in the bottom quartile) in two or all of the domains was associated with higher expenditures (odds ratio for two domains: 1.75 and for three domains: 2.12). Those in the bottom quartile for all three domains had the highest overall expenditures (OR: 2.28). For specific domains, lower physical activity at work and in sports is associated with higher health care expenditures related to medicine discharge (OR for work: 1.58, OR for sport: 1.57). Lower physical activity in leisure-time was still associated with higher overall expenditures (OR: 1.53). Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Idler et al. (2015) To analyse the relationship between physical activity and health care and (parental) productivity costs among children aged 9 to 12 years. Germany Based on public and/or private health care and parental work absence costs. Information on children’s physical activity and frequency of health care utilisation and costs were provided by parents in an interview (using a self-administered questionnaire). Children were grouped into two categories based on weekly hours spent on moderate and vigorous physical activity: <7 hours per week and ≥7 hours per week. Direct health care and indirect productivity costs (i.e. from parental absence at work) were examined across the two physical activity groups to determine possible associations. Physical activity, frequency of health care utilisation: The German Infant Study on the Influence of Nutrition Intervention plus Air Pollution and Genetics on Allergy Development (GINIplus) study and the Influence of Lifestyle Factors on Development of the Immune System and Allergies in East and West Germany plus Air Pollution and Genetics on Allergy Development (LISAplus) study Direct medical unit costs, number of days of parental absence from work and costs per day: Working Group Methods in Health Economic Evaluation (AG MEG) national costing guide, earlier studies Direct health care and indirect productivity costs (EUR) 2007 The mean annual cost for a child with a higher level of physical activity (i.e. ≥7 hours per week) vs a child with lower activity level (i.e. <7 hours per week) were €392 vs €398 for direct costs and €138 vs €127 for indirect costs. However, there were no statistically significant associations noted between physical activity and health care utilisation and costs among children aged 9 to 12 years. Different directions of estimates were noticeable throughout the cost components. Appendix 1 73 Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Idler et al. (2015) To analyse the relationship between physical activity and health care and (parental) productivity costs among children aged 9 to 12 years. Germany Based on public and/or private health care and parental work absence costs. Information on children’s physical activity and frequency of health care utilisation and costs were provided by parents in an interview (using a self-administered questionnaire). Children were grouped into two categories based on weekly hours spent on moderate and vigorous physical activity: <7 hours per week and ≥7 hours per week. Direct health care and indirect productivity costs (i.e. from parental absence at work) were examined across the two physical activity groups to determine possible associations. Physical activity, frequency of health care utilisation: The German Infant Study on the Influence of Nutrition Intervention plus Air Pollution and Genetics on Allergy Development (GINIplus) study and the Influence of Lifestyle Factors on Development of the Immune System and Allergies in East and West Germany plus Air Pollution and Genetics on Allergy Development (LISAplus) study Direct medical unit costs, number of days of parental absence from work and costs per day: Working Group Methods in Health Economic Evaluation (AG MEG) national costing guide, earlier studies Direct health care and indirect productivity costs (EUR) 2007 The mean annual cost for a child with a higher level of physical activity (i.e. ≥7 hours per week) vs a child with lower activity level (i.e. <7 hours per week) were €392 vs €398 for direct costs and €138 vs €127 for indirect costs. However, there were no statistically significant associations noted between physical activity and health care utilisation and costs among children aged 9 to 12 years. Different directions of estimates were noticeable throughout the cost components. 74 Assessing the economic costs of unhealthy diets and low physical activity Author/s Study objective Country Principal approach Methodology Data Type of costs Base-year Principal findings PHYSICAL ACTIVITY Krueger et al. (2015) To identify the economic burden of physical inactivity and the potential reduction in economic costs if all provinces in Canada achieved a prevalence rate of physical inactivity equivalent to that of the province with the lowest rate. Canada Disease-based: malignant and other neoplasms, endocrine, nutritional and metabolic diseases, cardiovascular, respiratory infections, digestive and musculoskeletal diseases. PAFs were derived based on sex-specific RRs from an earlier study and applied to disease costs. For indirect costs of physical inactivity, the ratio of direct and indirect costs for each diagnostic category from an earlier study were applied to PAF-based direct costs to generate equivalent indirect cost data. A second set of PAFs was derived using the prevalence rate of physical inactivity in British Columbia (the province with the lowest prevalence rate) and applied to populations living in other provinces to determine potential reductions in the economic burden. Physical activity prevalence: Canadian Community Health Survey (2012) RRs: earlier study for 2001 Cost data: National Health Expenditure Database, Economic Burden of Illness in Canada (EBIC) online tool (2008) Ratio of direct to indirect costs: EBIC (1998) Direct health care and indirect productivity costs (CAD) 2013 Economic costs associated with physical inactivity in 2013 (using actual provincial prevalence rates) were C$ 10.8 billion (males: C$ 5.2 billion, females: C$ 5.6 billion; direct costs: C$ 3.3 billion, indirect costs: C$ 7.5 billion). If all provinces were to achieve the low prevalence rate of physical inactivity in British Columbia, 14% (C$ 1.5 billion) of the costs could potentially be avoided, reducing the attributable costs to C$ 9.3 billion. DIET AND PHYSICAL ACTIVITY Popkin et al. (2006) To determine the economic costs of unhealthy diets and physical inactivity through its direct effects on disease risks as well as indirect effects, through overweight and obesity. China Disease-based: coronary heart disease, type 2 diabetes, hypertension, stroke, cancers of the breast, colon, oesophagus, endometrium, lungs, stomach and bladder. PAFs related directly to unhealthy diets and physical inactivity were calculated based on RRs from earlier meta-analyses of studies and applied to disease costs. PAFs related to overweight and obesity were also calculated, of which the proportion attributable to unhealthy diets and physical inactivity was also obtained and applied to disease costs. Physical activity, overweight and obesity prevalence: China Health and Nutrition Survey RR: earlier meta-analyses of studies Cost data: National Survey of Health Services for China (1998) Direct health care costs (USD) 2000 and 2025 In the year 2000 unhealthy diets accounted for US$ 4.2 billion of direct health care costs (US$ 3.4 billion for direct effects on disease risks and US$ 0.83 billion for indirect effects through overweight and obesity). Physical inactivity accounted for US$ 1.7 billion (US$ 1.3 billion for direct effects and US$ 0.35 billion for indirect effects through overweight and obesity). Costs predicted for 2025 did not increase considerably at US$ 6.1 billion (US$ 3.9 billion for unhealthy diets and US$ 2.2 billion for physical inactivity). Appendix 2 Average annual currency exchange rates used to identify the EUR equivalent of costs in studies Cost values in non-European currencies were converted into Euros, using the average of yearly currency exchange rates from the OANDA website (https:// www.oanda.com/currency/average) during the base year/s for cost estimation in each study. The following table shows the average exchange rate used in each of the studies, according to the base year/s for cost estimation. Study Base year/s for cost estimation Original currency used Equivalent to EUR per one unit of original currency Rayner & Scarborough (2005); Allender et al. (2007) 2002 GBP 1.59 Scarborough et al. (2011) 2006–2007 GBP 1.47 Lo et al. (2013) 1999–2006 TWD 0.03 Collins et al. (2011) 2002–2006 AUD 0.59 Peeters et al. (2014) 2010 AUD 0.69 Doidge et al. (2012) 2010–2011 AUD 0.72 Daviglus et al. (2005) 1984–2000 USD 0.96 Kuriyama et al. (2004) 1995–2001 USD 1.00 Anderson et al. (2005) 1996–1999 USD 0.90 Bland et al. (2009) 1999–2000/2001 USD 1.03 Bachmann et al. (2015) 1999–2009 USD 0.87 Garrett et al. (2004); Popkin et al. (2006) 2000 USD 1.08 Wang et al. (2005) 2001–2002 USD 1.09 Carlson et al. (2015) 2006–2011 USD 0.73 Zhang & Chaaban (2012) 2007 USD 0.73 Maresova (2014); Kruk (2014) 2008 CZK 0.04 Alter et al. (2012) 1994–2007 CAD 0.91 Katzmarzyk et al. (2000) 1999 CAD 0.63 Katzmarzyk (2011); Janssen (2012) 2009 CAD 0.63 Krueger et al. (2015) 2013 CAD 0.73 Note: Average annual exchange rates from USD to EUR in 1984 to 2000 only reflect the average of rates from 1990 to 2000 as earlier figures were not available. Costs in USD reported by Martinson et al. (2003) and Wang et al. (2004) were not converted into EUR as average exchange rates during the base year/s for cost estimation were not available from the sources used. Appendix 3 Mean intake of AHEI food groups in France, Germany, Italy, Spain and the United Kingdom Country Dietary component Mean intake (grams/day) Mean intake (cups or ounces/day)d France (n = 23 048 consumers; 44% of surveyed population) 2007 Individual and National Study on Food Consumption Vegetablesa 20.99 0.09 Fruitb 10.97 0.05 Whole grains 12.2 Sugar-sweetened beverages and fruit juice 118.84 4.19 Nuts 1.13 0.04 Processed meat 37.59 1.33 Trans-Fat 63.5 Long-chain (n-3) fats (EPA + DHA) 21.39 Polyunsaturated fatty acid 10.92 Sodium 1.5 Alcoholc 78.08 2.75 Germany (n = 54 710 consumers; 25% of surveyed population) 2007 National Nutrition Survey Vegetablesa 27.31 0.12 Fruitb 13.66 0.06 Whole grains 0.49 Sugar-sweetened beverages and fruit juice 341.65 12.05 Nuts 2.95 0.10 Processed meat 49.82 1.76 Trans-Fat 68.57 Long-chain (n-3) fats (EPA + DHA) 13.67 Polyunsaturated fatty acid 2.93 Sodium 0.01 Alcoholc 52.02 1.83 Appendix 3 77 Country Dietary component Mean intake (grams/day) Mean intake (cups or ounces/day)d Italy (n = 18 035 consumers; 33.9% of surveyed population) 2005–2006 National Food Consumption Survey Vegetablesa 45.62 0.20 Fruitb 2.89 0.01 Whole grains 35.29 Sugar-sweetened beverages and fruit juice 56.03 1.98 Nuts 1.06 0.04 Processed meat 29.87 1.05 Trans-Fat 30.23 Long-chain (n-3) fats (EPA + DHA) 31.05 Polyunsaturated fatty acid 36.63 Sodium 0.01 Alcoholc 70.88 2.50 Spain (n = 6 940 consumers; 34.6% of surveyed population) 2009 Spanish Agency for Food Safety Survey Vegetablesa 40.17 0.18 Fruitb 6.35 0.03 Whole grains 6.84 Sugar-sweetened beverages and fruit juice 130.65 4.61 Nuts 1.99 0.07 Processed meat 48.97 1.73 Trans-Fat 48.31 Long-chain (n-3) fats (EPA + DHA) 57.31 Polyunsaturated fatty acid Sodium Alcoholc 1.01 United Kingdom (n = 7 046 consumers; 26.5% of surveyed population) 2008 National Diet and Nutrition Survey Vegetablesa 6.96 0.03 Fruitb 10.81 0.05 Whole grains 2.80 Sugar-sweetened beverages and fruit juice 224.12 7.91 Nuts 1.07 0.04 Processed meat 30.1 1.06 Trans-Fat 33.48 Long-chain (n-3) fats (EPA + DHA) 21.1 Polyunsaturated fatty acid 1.46 Sodium 0.08 Alcoholc 86.7 3.06 Source: European Food and Safety Authority, 2015. Notes: a Vegetables refers to green leafy vegetables; b fruit refers to berries and small fruits; c alcohol refers to either beer and beer-like beverages, wines or liquors, whichever has the highest mean intake; d converted into units used in the AHEI: 240 g = 1 cup for liquids, 226.72 g = 1 cup for solids and 28.35 g = 1 oz for both solids and liquids. Appendix 4 Quality of studies that have quantified the association between unhealthy diets and low physical activity and diabetes Quality criterion as defined by Al Tunaji et al. (2014) Quality assessment Montonen et al. (2005) Laaksonen et al. (2009) Li et al. (2015) The exposure (risk factor) is clearly defined An unhealthy diet was defined as a ‘conservative’ dietary pattern characterised by consumption of butter, potatoes and whole milk Physical inactivity was defined as less than 30 minutes of occasional or regular leisure-time exercise per day Poor diet was defined as a dietary score in the 2010 alternate healthy eating index belonging to the third to fifth quintiles (i.e. score of <67%). Physical inactivity was defined as those not meeting 150 minutes per week of moderate-intensity exercise or 75 minutes per week of vigorous-intensity exercise The exposure was measured objectively No; the exposure was measured using a one-year dietary history interview No; the exposure was self-reported in a health interview or a self-administered questionnaire No; the exposure was measured using a food frequency questionnaire and interviews on lifestyle habits and medical history The outcome (disease) was clearly defined The outcome was defined as type 2 diabetes; no diagnostic criteria were presented The outcome was defined as type 2 diabetes according to the WHO diagnostic criteria The outcome was defined as type 2 diabetes, according to national diagnostic criteria The outcome was ascertained by objective measures or, if self-reported, confirmed by other measures The outcomes were identified from a nationwide registry of patients receiving drug reimbursement (including for diabetes). Study participants were linked to this register by unique social security codes The outcomes were identified from a central register of all patients receiving drug reimbursement (including for diabetes). Study participants were linked to this register by unique social security codes The outcomes were selfreported and confirmed by a validated supplementary questionnaire (validated through hospital records) Appendix 4 79 Quality criterion as defined by Al Tunaji et al. (2014) Quality assessment Montonen et al. (2005) Laaksonen et al. (2009) Li et al. (2015) The analysis was based on raw data from a prospective or cohort study Yes Yes Yes The follow-up time was provided 23 years 10 years 20–30 years Full adjustments were made Adjustments were made for age, sex, body mass index, energy intake, smoking status, family history of diabetes, geographic area, serum cholesterol and hypertension Adjustments were made for age, sex, other lifestyle risk factors (e.g. BMI, smoking, alcohol consumption and serum vitamin D) or components of metabolic syndrome (BMI, blood pressure, serum triglyceride levels, serum HDL cholesterol, fasting glucose) Adjustments were made for sex, ethnicity (Caucasian yes/no), marriage status, living status (alone yes/no), family history of diabetes, menopausal status (pre- or post-menopausal; never, past or current menopausal hormone use), and for other lifestyle risk factors assessed in the study (i.e. BMI, smoking status, daily alcohol consumption) Appendix 5 Annual incidence rate of diabetic complications from the UKPDS Outcomes Model 2 Complication Annual incidence rate (%) First myocardial infarction 1.13 Second myocardial infarction 0.19 First stroke 0.56 Second stroke 0.09 Congestive heart failure 0.39 Ischaemic heart disease 0.83 First amputation 0.19 Second amputation 0.06 Retinopathy/Blindness 0.3 Renal failure 0.13 Ulcer 0.11 Source: Hayes et al., 2013. Appendix 6 Estimated diabetes complicationrelated costs in 2020 attributable to unhealthy diets and low physical activity in 2015 Country Complication Estimated number of incident cases of diabetes-related complications in 2020 attributable to the risk factor Estimated per patient costs in 2020 (€) Estimated diabetes complication-related costs in 2020 (€) Unhealthy diets Low physical activity Unhealthy diets Low physical activity France First myocardial infarction 51 86 19 097 981 441 1 635 734 Second myocardial infarction 9 14 19 097 165 021 275 035 First stroke 25 42 14 396 366 649 611 081 Second stroke 4 7 14 396 58 926 98 210 Congestive heart failure 18 30 4 838 85 813 143 021 Ischaemic heart disease 38 63 3 200 120 795 201 325 First amputation 9 14 39 191 338 657 564 429 Second amputation 3 5 39 191 106 944 178 241 Retinopathy/Blindness 14 23 468 6 385 10 642 Renal failure 6 10 69 186 409 055 681 759 Ulcer 5 8 1 399 6 999 11 665 Total 181 302 2 646 685 4 411 142 82 Assessing the economic costs of unhealthy diets and low physical activity Country Complication Estimated number of incident cases of diabetes-related complications in 2020 attributable to the risk factor Estimated per patient costs in 2020 (€) Estimated diabetes complication-related costs in 2020 (€) Unhealthy diets Low physical activity Unhealthy diets Low physical activity Germany First myocardial infarction 55 110 22 465 1 231 448 2 463 150 Second myocardial infarction 9 18 22 465 207 058 414 158 First stroke 27 54 29 032 788 672 1 577 506 Second stroke 4 9 29 032 126 751 253 528 Congestive heart failure 19 38 9 030 170 838 341 711 Ischaemic heart disease 40 81 5 002 201 397 402 836 First amputation 9 18 33 068 304 784 609 632 Second amputation 3 6 33 068 96 248 192 515 Retinopathy/Blindness 15 29 15 650 227 754 455 556 Renal failure 6 13 86 975 548 490 1 097 094 Ulcer 5 11 1 312 7 001 14 003 Total 193 386 3 910 441 7 821 688 Italy First myocardial infarction 17 120 12 580 215 079 1 505 553 Second myocardial infarction 3 20 12 580 36 164 253 146 First stroke 8 59 5 994 50 786 355 502 Second stroke 1 10 5 994 8 162 57 134 Congestive heart failure 6 41 3 363 19 844 138 908 Ischaemic heart disease 13 88 2 091 26 259 183 810 First amputation 3 20 9 266 26 637 186 459 Second amputation 1 6 9 266 8 412 58 882 Retinopathy/Blindness 5 32 5 021 22 790 159 532 Renal failure 2 14 39 220 77 142 539 993 Ulcer 2 12 694 1 155 8 085 Total 60 422 492 429 3 447 004 Appendix 6 83 Country Complication Estimated number of incident cases of diabetes-related complications in 2020 attributable to the risk factor Estimated per patient costs in 2020 (€) Estimated diabetes complication-related costs in 2020 (€) Unhealthy diets Low physical activity Unhealthy diets Low physical activity Spain First myocardial infarction 12 73 22 638 273 460 1 641 273 Second myocardial infarction 2 12 22 638 45 980 275 966 First stroke 6 36 5 447 32 608 195 709 Second stroke 1 6 5 447 5 241 31 453 Congestive heart failure 4 25 5 834 24 323 145 981 Ischaemic heart disease 9 53 2 592 22 998 138 031 First amputation 2 12 17 366 35 272 211 698 Second amputation 1 4 17 366 11 139 66 852 Retinopathy/Blindness 3 19 3 669 11 766 70 621 Renal failure 1 8 36 679 50 973 305 932 Ulcer 1 7 2 258 2 655 15 936 Total 43 255 516 414 3 099 453 United Kingdom First myocardial infarction 27 94 8 823 236 089 826 313 Second myocardial infarction 4 16 8 823 39 696 138 938 First stroke 13 46 5 131 68 041 238 144 Second stroke 2 7 5 131 10 935 38 273 Congestive heart failure 9 32 48 330 446 337 1 562 180 Ischaemic heart disease 20 69 39 119 768 860 2 691 012 First amputation 4 16 14 114 63 502 222 256 Second amputation 1 5 14 114 20 053 70 186 Retinopathy/Blindness 7 25 1 890 13 427 46 993 Renal failure 3 11 63 949 196 861 689 012 Ulcer 3 9 4 030 10 497 36 741 Total 94 330 1 874 299 6 560 047 Appendix 7 Number of incident type 2 diabetes cases attributable to unhealthy diets and low physical activity in each country by sex Country Number of type 2 diabetes cases attributable to the risk factor Unhealthy diets Low physical activity Males Females Total Males Females Total France 2 507 2 041 4 548 4 178 3 402 7 580 Germany 2 674 2 177 4 851 5 348 4 355 9 703 Italy 834 679 1 513 5 838 4 753 10 591 Spain 589 480 1 069 3 536 2 880 6 416 United Kingdom 1 305 1 063 2 368 4 568 3 720 8 288 Appendix 8 Average percentage of annual incident type 2 diabetes cases in the United Kingdom by sex and five-year age group, 1991–2010 Age range Average percentage of annual total incident type 2 diabetes cases Males (%) Females (%) 0–4 0.16 0.14 5–9 0.13 0.14 10–14 0.13 0.16 15–19 0.25 0.46 20–24 0.32 0.79 25–29 0.46 0.97 30–34 0.77 1.19 35–39 1.30 1.61 40–44 2.39 2.33 45–49 3.94 3.40 50–54 5.90 5.00 55–59 8.04 7.05 60–64 10.79 9.56 65–69 12.57 11.61 70–74 13.31 12.95 75–79 13.28 13.61 80–84 11.64 12.42 85–89 9.25 10.56 90+ 5.38 6.03 Appendix 9 Estimated number of diabetes cases attributable to unhealthy diets and low physical activity by sex and fiveyear age group in France, Germany, Italy, Spain and the United Kingdom Country Agegroup Number of diabetes cases attributable to the risk factor Unhealthy diets Low physical activity Males Females Total Males Females Total France 0–4 4 3 7 7 5 12 5–9 3 3 6 5 5 10 10–14 3 3 7 5 5 10 15–19 6 9 16 10 16 26 20–24 8 16 24 13 27 40 25–29 12 20 31 19 33 52 30–34 19 24 44 32 40 73 35–39 33 33 65 54 55 109 40–44 60 48 107 100 79 179 45–49 99 69 168 165 116 280 50–54 148 102 250 247 170 417 55–59 202 144 345 336 240 576 60–64 270 195 466 451 325 776 65–69 315 237 552 525 395 920 70–74 334 264 598 556 441 997 75–79 333 278 611 555 463 1 018 80–84 292 254 545 486 423 909 85–89 232 216 447 386 359 746 90+ 135 123 258 225 205 430 TOTAL 2 507 2 041 4 548 4 179 3 401 7 580 Appendix 9 87 Country Agegroup Number of diabetes cases attributable to the risk factor Unhealthy diets Low physical activity Males Females Total Males Females Total Germany 0–4 4 3 7 9 6 15 5–9 3 3 7 7 6 13 10–14 3 3 7 7 7 14 15–19 7 10 17 13 20 33 20–24 9 17 26 17 34 52 25–29 12 21 33 25 42 67 30–34 21 26 46 41 52 93 35–39 35 35 70 70 70 140 40–44 64 51 115 128 101 229 45–49 105 74 179 211 148 359 50–54 158 109 267 316 218 533 55–59 215 153 368 430 307 737 60–64 289 208 497 577 416 993 65–69 336 253 589 672 506 1 178 70–74 356 282 638 712 564 1 276 75–79 355 296 651 710 593 1 303 80–84 311 270 582 623 541 1 163 85–89 247 230 477 495 460 955 90+ 144 131 275 288 263 550 TOTAL 2 674 2 177 4 851 5 349 4 354 9 703 Italy 0–4 1 1 2 9 7 16 5–9 1 1 2 8 7 14 10–14 1 1 2 8 8 15 15–19 2 3 5 15 22 36 20–24 3 5 8 19 38 56 25–29 4 7 10 27 46 73 30–34 6 8 15 45 57 102 35–39 11 11 22 76 77 152 40–44 20 16 36 140 111 250 45–49 33 23 56 230 162 392 50–54 49 34 83 344 238 582 55–59 67 48 115 469 335 804 60–64 90 65 155 630 454 1 084 65–69 105 79 184 734 552 1 286 70–74 111 88 199 777 616 1 393 75–79 111 92 203 775 647 1 422 80–84 97 84 181 680 590 1 270 85–89 77 72 149 540 502 1 042 90+ 45 41 86 314 287 601 TOTAL 834 679 1 513 5 838 4 752 10 591 88 Assessing the economic costs of unhealthy diets and low physical activity Country Agegroup Number of diabetes cases attributable to the risk factor Unhealthy diets Low physical activity Males Females Total Males Females Total Spain 0–4 1 1 2 6 4 10 5–9 1 1 1 5 4 9 10–14 1 1 2 5 5 9 15–19 1 2 4 9 13 22 20–24 2 4 6 11 23 34 25–29 3 5 7 16 28 44 30–34 5 6 10 27 34 61 35–39 8 8 15 46 46 92 40–44 14 11 25 85 67 152 45–49 23 16 40 139 98 237 50–54 35 24 59 209 144 353 55–59 47 34 81 284 203 487 60–64 64 46 109 382 275 657 65–69 74 56 130 445 334 779 70–74 78 62 141 471 373 844 75–79 78 65 144 470 392 862 80–84 69 60 128 412 358 769 85–89 55 51 105 327 304 631 90+ 32 29 61 190 174 364 TOTAL 589 480 1 069 3 537 2 879 6 416 United Kingdom 0–4 2 1 3 7 5 13 5–9 2 1 3 6 5 11 10–14 2 2 4 6 6 12 15–19 3 5 8 11 17 29 20–24 4 8 13 15 29 44 25–29 6 10 16 21 36 57 30–34 10 13 23 35 44 79 35–39 17 17 34 59 60 119 40–44 31 25 56 109 87 196 45–49 51 36 88 180 126 306 50–54 77 53 130 270 186 456 55–59 105 75 180 367 262 630 60–64 141 102 242 493 356 849 65–69 164 123 287 574 432 1 006 70–74 174 138 311 608 482 1 090 75–79 173 145 318 607 506 1 113 80–84 152 132 284 532 462 994 85–89 121 112 233 423 393 815 90+ 70 64 134 246 224 470 TOTAL 1 305 1 063 2 368 4 569 3 719 8 288 Appendix 10 Estimated number of diabetes cases attributable to unhealthy diets and low physical activity who are expected to be in and out of the formal labour force by five-year age group Country Working age range Average annual labour force participation rate (%) Number of diabetes cases attributable to the risk factor Unhealthy diets Low physical activity Total In the labour force Out of the labour force Total In the labour force Out of the labour force France 15–19 14.94 16 2 14 26 4 22 20–24 60.90 24 15 9 40 25 16 25–29 86.52 31 27 4 52 45 7 30–34 87.68 44 38 6 73 64 9 35–39 88.98 65 58 7 109 97 12 40–44 89.58 107 96 11 179 160 19 45–49 88.59 168 149 19 280 248 32 50–54 83.87 250 210 40 417 349 67 55–59 62.14 345 215 131 576 358 218 60–64 17.69 466 82 383 776 137 639 TOTAL 1 517 893 624 2 528 1 488 1 041 Germany 15–19 30.16 17 5 12 33 10 23 20–24 70.17 26 18 8 52 36 15 25–29 81.62 33 27 6 67 55 12 30–34 86.16 46 40 6 93 80 13 35–39 87.88 70 61 8 140 123 17 40–44 89.64 115 103 12 229 206 24 45–49 88.99 179 160 20 359 319 40 50–54 85.01 267 227 40 533 453 80 55–59 74.50 368 275 94 737 549 188 60–64 38.10 497 189 307 993 378 615 TOTAL 1 618 1 105 513 3 236 2 209 1 027 90 Assessing the economic costs of unhealthy diets and low physical activity Country Working age range Average annual labour force participation rate (%) Number of diabetes cases attributable to the risk factor Unhealthy diets Low physical activity Total In the labour force Out of the labour force Total In the labour force Out of the labour force Italy 15–19 12.48 5 1 5 36 5 32 20–24 49.95 8 4 4 56 28 28 25–29 70.99 10 7 3 73 52 21 30–34 79.36 15 12 3 102 81 21 35–39 80.59 22 18 4 152 123 30 40–44 79.73 36 29 7 250 200 51 45–49 77.45 56 43 13 392 303 88 50–54 70.71 83 59 24 582 412 170 55–59 50.48 115 58 57 804 406 398 60–64 22.50 155 35 120 1 084 244 840 TOTAL 505 265 240 3 532 1 852 1 680 Spain 15–19 23.57 4 1 3 22 5 17 20–24 63.32 6 4 2 34 22 12 25–29 85.63 7 6 1 44 38 6 30–34 87.39 10 9 1 61 54 8 35–39 85.69 15 13 2 92 79 13 40–44 83.66 25 21 4 152 127 25 45–49 80.22 40 32 8 237 190 47 50–54 73.15 59 43 16 353 258 95 55–59 60.20 81 49 32 487 293 194 60–64 35.39 109 39 71 657 232 424 TOTAL 357 216 140 2 140 1 298 841 United Kingdom 15–19 52.83 8 4 4 28 15 13 20–24 74.60 13 9 3 44 33 11 25–29 84.25 16 14 3 57 48 9 30–34 84.52 23 19 4 79 67 12 35–39 84.87 34 29 5 119 101 18 40–44 86.02 56 48 8 196 168 27 45–49 85.97 88 75 12 306 263 43 50–54 83.29 130 108 22 456 379 76 55–59 71.93 180 129 50 629 452 177 60–64 44.74 242 108 134 849 380 469 TOTAL 790 545 245 2 763 1 907 856 Appendix 11 Estimated total number of working years lost due to work disability, early retirement and premature death among incident type 2 diabetes cases at working age that can be attributed to unhealthy diets and low physical activity and who are expected to be in the formal labour force in France, Germany, Italy, Spain and the United Kingdom, 2020 Country Risk factor Estimated number of incident type 2 diabetes cases aged 35–60 attributable to the risk factor (in the labour force) Total number of working years lost due to work disability early retirement premature death France Unhealthy diets 893 80 625 250 Low physical activity 1 488 134 1 041 417 Germany Unhealthy diets 1 105 99 773 309 Low physical activity 2 209 199 1 547 619 Italy Unhealthy diets 265 24 185 74 Low physical activity 1 852 167 1 297 519 Spain Unhealthy diets 216 19 151 61 Low physical activity 1 298 117 909 364 United Kingdom Unhealthy diets 545 49 382 153 Low physical activity 1 907 172 1 335 534 ISBN 9289050425 9 7 8 9 2 8 9 0 5 0 4 2 5 Health Policy Series No. 47 www.healthobservatory.eu Health Policy Series 47 AN EVIDENCE REVIEW AND PROPOSED FRAMEWORK 47 Assessing the economic costs of unhealthy diets and low physical activity Christine Joy Candari Jonathan Cylus Ellen Nolte An evidence review and proposed framework Unhealthy diets and low physical activity contribute to many chronic diseases and disability; they are responsible for some 2 in 5 deaths worldwide and for about 30% of the global disease burden. Yet surprisingly little is known about the economic costs that these risk factors cause, both for health care and society more widely. This study pulls together the evidence about the economic burden that can be linked to unhealthy diets and low physical activity and explores • How definitions vary and why this matters • The complexity of estimating the economic burden and • How we can arrive at a better way to estimate the costs of an unhealthy diet and low physical activity, using diabetes as an example The review finds that unhealthy diets and low physical activity predict higher health care expenditure, but estimates vary greatly. Existing studies underestimate the true economic burden because most only look at the costs to the health system. Indirect costs caused by lost productivity may be about twice as high as direct health care costs, together accounting for about 0.5% of national income. The study also tests the feasibility of using a disease-based approach to estimate the costs of unhealthy diets and low physical activity in Europe, projecting the total economic burden associated with these two risk factors as manifested in new type 2 diabetes cases at €883 million in 2020 for France, Germany, Italy, Spain and the United Kingdom alone. The ‘true’ costs will be higher, as unhealthy diets and low physical activity are linked to many more diseases. The study’s findings are a step towards a better understanding of the economic burden that can be associated with two key risk factors for ill health and they will help policymakers in setting priorities and to more effectively promoting healthy diets and physical activity. The authors Christine Joy Candari was an independent consultant at the time of writing this report. She is currently Chief Consultant for Health Research, U Consult Us Inc, Manila, The Philippines Jonathan Cylus is Research Fellow, European Observatory on Health Systems and Policies, London School of Economics and Political Science Ellen Nolte is Head of London Hubs, European Observatory on Health Systems and Policies DIETS AND LOW PHYSICAL ACTIVITY ASSESSING THE ECONOMIC COSTS OF UNHEALTHY Christine Joy Candari, Jonathan Cylus, Ellen Nolte Cover_WHO_nr47_Mise en page 1 13/06/17 13:21 Page 1