Appendix 2: Cost-effectiveness modeling methods A model was developed to evaluate the cost-effectiveness of interventions aiming to reduce overweight and obesity. The model evaluates intervention cost-effectiveness over the lifetime of the Australian population aged 20 and over in a baseline year of 2003. This year was chosen to enable the use of information on disease epidemiology from the Australian Burden of Disease study. 1 Information on intervention costs and effects from the overweight and obesity intervention literature were used, and Australian data on the costs of disease treatment. 2 The model has been built in Excel (Microsoft Office 2003) and uses the add-in tool Ersatz (EpiGear, Version 1.0) for uncertainty analysis. Life table analysis In a proportional multi-state life table 3 (Figure I), disability-adjusted life years (DALYs) averted by an intervention are evaluated as the difference in health-adjusted years of life lived between an Australian population that mimics the Australian population in terms of BMI, and an identical population that receives the intervention. The health-adjusted years of life lived by each population are calculated by dividing each population into five-year age group cohorts (from age 15-19 to age 95+), and simulating each cohort in a life table until everyone has either died or reached 100 years of age. Years of life lived by each cohort are adjusted at each age for time spent in poor health ( disability ) due to disease or injury, using disability estimates for the Australian population. 1 For modeled disease, the disability adjustment is calculated as the number of prevalent years lived with disability (pyld) per prevalent case of disease. The age- and sex-specific probability of health loss due to disability from all other causes (i.e. those not specifically modeled) is calculated as the number of pyld per capita for these conditions. 1
m x - m x + q x l x L x e x w x - w x + Lw x HALE x Life table i x p x m x i x p x m x w w Disease process 1 Disease process 2 3 4 5 Risk factor Figure I: Schematic of a proportional multi-state life table, showing the interaction between disease parameters and life table parameters, where x is age, i is incidence, p is prevalence, m is mortality, w is disability-adjustment, q is probability of dying, l is number of survivors, L is life years, Lw is disability-adjusted life years and HALE is health-adjusted life expectancy, and where - denotes a parameter that specifically excludes modeled diseases, and + denotes a parameter for all diseases (i.e. including modeled diseases). Disease models Overweight and obesity increase risk of stroke, ischemic heart disease, hypertensive heart disease, diabetes mellitus, osteoarthritis, post-menopausal breast cancer, colon cancer, endometrial cancer and kidney cancer. 4 Each of these diseases is modeled explicitly using a set of differential equations that describe the transition of people between four states (healthy, diseased, dead from the disease, and dead from all other causes), with transition of people between the four states based on rates of mortality, incidence, case fatality and remission (Figure II). 5 Epidemiological data inputs to the disease models are derived from the Australian Burden of Disease study, 1 with the aid of DISMOD-II 5 to derive data not explicitly reported (e.g. case fatality and prevalence from incidence and mortality rates). Intervention effect on body mass index The health benefits of an intervention are modeled as a reduction in incidence of each obesityrelated disease. Intervention combinations are assumed to have a multiplicative effect (Equation 1). 2
Healthy Mortality (other) Dead (other) Incidence Remission Mortality (other) Diseased Case fatality Dead (disease) Figure II: Each disease is modeled by a conceptual model with four states (healthy, diseased, dead from the disease, and dead from all other causes) and transition hazards between states of incidence, remission, case fatality and mortality from all other causes (after Barendregt et al. 5 ) ( 1 - PIF )( 1 - PIF ) K( 1 - PIF ) I = I (1) 1 2 n where: I is the incidence of disease (e.g. ischemic heart disease) in the population (by age and sex); I is the new incidence of disease (e.g. ischemic heart disease) when an intervention is implemented; and PIF i is the potential impact fraction (PIF) for interventions 1 to n. Each potential impact fraction (PIF) is derived from three key parameters: the population prevalence of overweight and obesity in Australia, the relative risks of obesity-related diseases, and the change in body mass due to an intervention. The prevalence of overweight and obesity is derived from the distribution of BMI in the Australian population, which is determined from the Australian Diabetes, Obesity and Lifestyle Study (AusDiab). 6 Body mass index is modeled by sex and age-group (20-24.9, 25-29.9,, 80+) as a continuous lognormal distribution, and subsequently divided into three categories (Figure III): normal- or underweight (BMI <25), overweight (BMI 25-30), obese (BMI >30). This re-categorisation is necessary to enable to differentially target these different groups. Relative risks of disease per unit increase in BMI (Table I) are drawn from metaanalyses carried out for the global Comparative Quantification of Health Risks 4 for all diseases except diabetes, for which the relative risks came from the Asia Pacific Cohort collaboration. 7 3
Table I: Relative risks of disease per 1 unit increase of BMI 4, 7 Age Males Females Colorectal cancer <35 1 1 35+ 1.03 (1.01-1.05) 1.03 (1.01-1.05) Breast cancer <35-1 35+ - 1.03 (1.02-1.04) Endometrial cancer <35-1.10 (1.07-1.14) 35+ - 1.10 (1.07-1.14) Kidney cancer <35 1.06 (1.03-1.08) 1.06 (1.03-1.08) 35+ 1.06 (1.03-1.08) 1.06 (1.03-1.08) Osteoarthritis <35 1.04 (1.03-1.06) 1.04 (1.03-1.06) 35+ 1.04 (1.03-1.06) 1.04 (1.03-1.06) Ischemic heart disease <35 1 1 35-44 1.12 (1.05-1.19) 1.12 (1.05-1.19) 45-59 1.10 (1.08-1.14) 1.10 (1.08-1.14) 60-69 1.06 (1.03-1.08) 1.06 (1.03-1.08) 70-79 1.04 (1.02-1.06) 1.04 (1.02-1.06) 80+ 1.02 (1.00-1.05) 1.02 (1.00-1.05) Hypertensive heart disease <45 1 1 45-59 1.09 (1.03-1.14) 1.09 (1.03-1.14) 60-69 1.16 (1.05-1.27) 1.16 (1.05-1.27) 70-79 1.12 (1.04-1.21) 1.12 (1.04-1.21) 80+ 1.06 (1.02-1.11) 1.06 (1.02-1.11) Stroke <35 1 1 35-44 1.14 (1.05-1.23) 1.14 (1.05-1.23) 45-59 1.10 (1.03-1.16) 1.10 (1.03-1.16) 60-69 1.08 (1.03-1.13) 1.08 (1.03-1.13) 70-79 1.05 (1.02-1.09) 1.05 (1.02-1.09) 80+ 1.03 (1.01-1.05) 1.03 (1.01-1.05) Type II Diabetes <35 1 1 35-44 1.19 (1.06-1.32) 1.19 (1.06-1.32) 45-69 1.14 (1.05-1.23) 1.14 (1.05-1.23) 70+ 1.10 (1.03-1.16) 1.10 (1.03-1.16) NB. Values shown are the mean and 95% confidence intervals. 4
Figure III: Prevalence of overweight and obesity in Australia (derived from the Australian Diabetes, Obesity and Lifestyle Study (AusDiab) 6 ) 100% Males 80% Prevalence 60% 40% 20% 0% 20-24 25-29 30-34 35-39 40-44 45-49 Age 50-54 55-59 60-64 65-69 70-74 75-79 80+ 100% BMI <25 Females BMI 25-30 BMI >30 80% 60% 40% 20% 0% 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80+ Prevalence Age BMI <25 BMI 25-30 BMI >30 The intervention impact on disease is quantified by a modified version of the potential impact fraction (PIF). In our model we use a version of the PIF in which the relative risk of disease changes under the influence of changes in BMI, while the prevalence of the BMI categories (normal weight, overweight, obesity) is held constant (Equation 2). 8 The change in relative risk (from RR to RR ) is based on the change in average BMI in that category and the relative risks of disease per unit increase in BMI. 5
n pi RRi pi RR' i i = 1 i = 1 PIF = (2) n p RR i i = 1 n i where: p i is the proportion of the population in BMI-category i; RR i is the relative risk of disease associated with BMI-category i; and RR' i is the relative risk of disease associated with BMI-category i after an intervention is implemented in the population. Trends in body mass index Over time, the Australian population is becoming heavier. This increase in body mass occurred among all age groups. We modeled the trend in BMI with a formula from Haby & Marwick. 9 They used multiple linear regression analyses of log-transformed BMI (ln(bmi)) data to determine the independent effects of age and year of birth on ln(bmi) for males and females separately. The data came from 11 cross-sectional Australian surveys with measured height and weight. In our model, this trend has been set to apply 20 years into the future, and is applied to both the intervention and the reference population. The intervention effect is superimposed on this trend. Adding trend analysis to the model increases the effectiveness of the interventions and makes weight-lowering interventions slightly more cost-effective than if these trends were ignored. Disease costs Cost offsets, due to reduced rates of obesity-related diseases, are evaluated using disease treatment costs from the Australian Institute of Health and Welfare Disease Costs and Impacts Study 2001 (Table III). The average cost per incident case (colon cancer, breast cancer, endometrial cancer and kidney cancer) or prevalent case (ischemic heart disease, stroke, hypertensive heart disease and diabetes) is derived for each modeled disease, based on rates of disease in 2001 1, and adjusted to the year 2003 using the Australian Health Price Index. 10 Health care costs for all other diseases and injuries, which will accrue in added years of life, are also derived from the Disease Costs and Impacts Study data. 2 These costs are derived as a cost per person, by age and sex, excluding costs of the obesity-related diseases explicitly modeled. 6
Table III: Cost per prevalent or incident case of disease. Endometrial Cancer* Ischemic Heart Disease** Hypertensive Heart Disease** Age Colon Cancer* Breast Cancer* Kidney Cancer* Stroke** Type II Diabetes** Osteoarthritis** All other*** Males <55 $17,490 $16,298 $2,962 $2,228 $13,103 $504 $4,431 $1,555 55 64 $17,657 $16,751 $1,988 $4,942 $24,408 $660 $4,431 $2,828 65 74 $18,164 $14,748 $1,664 $9,529 $15,048 $763 $4,431 $4,731 75-84 $18,037 $14,526 $1,512 $12,856 $8,167 $639 $4,431 $7,945 85+ $19,288 $7,372 $1,394 $16,301 $1,723 $594 $4,431 $13,061 Females <55 $17,136 $12,424 $10,665 $15,505 $1,832 $1,161 $22,097 $506 $4,431 $2,009 55 64 $16,349 $10,493 $9,902 $16,363 $1,520 $2,090 $32,044 $759 $4,431 $3,225 65 74 $17,238 $11,609 $14,419 $17,133 $1,595 $5,106 $20,357 $839 $4,431 $4,829 75-84 $17,360 $12,706 $10,497 $17,198 $1,564 $13,137 $9,624 $745 $4,431 $8,197 85+ $16,545 $12,520 $13,402 $12,192 $1,670 $19,679 $1,695 $429 $4,431 $15,078 * Cost per incident case of disease. ** Cost per prevalent case of disease. *** Cost per person. NB. Costs are in Australian dollars, adjusted to the year 2003. References 1. Begg S, Vos T, Barker B, Stevenson C, Stanley L, Lopez AD. The Burden of Disease and Injury in Australia, AIHW: Canberra, 2007. 2. Australian Institute of Health and Welfare. Disease costs and impacts study data. Australian Institute of Health and Welfare: Canberra, 2001. 3. Barendregt JJ, Van Oortmarssen GJ, Van Hout BA, Van Den Bosch JM, Bonneux L. Coping with multiple morbidity in a life table. Math Popul Stud 1998; 7(1): 29-49, 109. 4. James WP, Jackson-Leach R, Ni Mhurchu C, Kalamara E, Shayeghi M, Rigby NJ et al. Overweight and obesity (high body mass index). In: Ezzati M, Lopez AD, Rodgers A, Murray CJL (eds). Comparative quantification of health risks, vol. 1. World Health Organization: Geneva, 2004, pp 497-596. 5. Barendregt JJ, Van Oortmarssen GJ, Vos T, Murray CJ. A generic model for the assessment of disease epidemiology: the computational basis of DisMod II. Popul Health Metr 2003; 1(1): 4. 7
6. Cameron AJ, Welborn TA, Zimmet PZ, Dunstan DW, Owen N, Salmon J et al. Overweight and obesity in Australia: the 1999-2000 Australian Diabetes, Obesity and Lifestyle Study (AusDiab). Med J Aust 2003; 178(9): 427-32. 7. Ni Mhurchu C, Parag V, Nakamura M, Patel A, Rodgers A, Lam TH. Body mass index and risk of diabetes mellitus in the Asia-Pacific region. Asia Pac J Clin Nutr 2006; 15(2): 127-33. 8. Barendregt JJ, Veerman JL. Categorical versus continuous risk factors, and the calculation of potential impact fractions. J Epidemiol Community Health 2009. 9. Haby M, Markwick A. Future prevalence of overweight and obesity in Australian children and adolescents, 2005-2025. Public Health Branch, Victorian Government Department of Human Services: Melbourne, 2008. 10. Australian Institute of Health and Welfare. Health Expenditure Australia, Australian Institute of Health and Welfare: Canberra, 2008. 8