- Email: [email protected]
A Simulation of Affordability and Effectiveness of Childhood Obesity Interventions
A Simulation of Affordability and Effectiveness of Childhood Obesity Interventions Sai Ma, PhD; Kevin D. Frick, PhD From the Department of Population, Family and Reproductive Health (Dr Ma) and Department of Health Policy and Management (Dr Frick), Johns Hopkins Bloomberg School of Public Health, Baltimore, Md Address correspondence to Sai Ma, PhD, Johns Hopkins Bloomberg School of Public Health, Department of Population, Family and Reproductive Health, Baltimore, Maryland (e-mail: [email protected]). Received for publication August 30, 2010; accepted April 21, 2011.
ABSTRACT OBJECTIVE: This study seeks to project at what level of effectiveness and cost a population-based or targeted intervention would yield a positive net economic benefit. METHODS: Data sources include prevalence of obesity at all ages from the National Health and Nutrition Examination Survey, the persistence of obesity from childhood to adulthood from a literature review, and a cost estimate from the 2006 Medical Expenditures Panel Survey. Econometric analysis was used to estimate medical cost related to obesity. Lifetime medical cost related to obesity is calculated by race, gender, and smoking status. Simulations were conducted to estimate the break-even point for interventions that take place between ages 0 and 6 years, ages 7 and 12 years, and ages 13 to 18 years, with a range of effectiveness. RESULTS: Results of simulations reveal that, from a pure medical cost perspective, spending approximately $1.4 to $1.7 billion at present value for each birth cohort will break even if 1 percentage point reduction in obesity among children is achieved. Population-based interventions can spend up to
between $280 and $339 per child at present value if 1 percentage point reduction in obesity rate could be achieved; in contrast, should we invest in interventions that only target obese children, we can spend up to $1648 to $2735 per obese child for every 1 percentage point reduction in obesity rate. CONCLUSIONS: This study has several important policy implications; early interventions make economic sense. Targeted interventions could yield higher cost savings than populationbased interventions for young children (aged 0–6 years), whereas a population-based approach could yield greater economic net benefits for adolescents (aged 13–18 years). Our simulation shows that childhood obesity interventions, even with moderate effectiveness, would make economic sense, which should motivate policy makers to take action.
WHAT’S NEW
reductions.4 As a result, increasing attention from policy makers and researchers has been paid to early prevention and programs.5 Nearly all decisions regarding public health investments involve options of prevention and treatment. Interventions targeting obesity can aim at preventing obesity, changing the weight of those already obese, or treating emergent consequences like diabetes. Even preventive efforts can take place at multiple stages, and it remains ambiguous as to whom policies should target to maximize economic returns. To endorse interventions at the earliest ages, one needs to understand 3 critical details: 1) the persistence of childhood obesity into adulthood, 2) the degree to which interventions are likely to be adopted by the children and their families at different stages in children’s lives, and 3) the potential returns on investment. As evidence-based interventions targeting young obese children (aged 0–6 years old) are particularly scarce,5,6 few studies have estimated under what conditions interventions could be effective and affordable. A recent study by Transande (2010) using mathematical modeling found that spending $2 billion per cohort would be cost effective if a hypothetical intervention can reduce obesity
KEYWORDS: childhood obesity; intervention; investment; persistence; prevention
ACADEMIC PEDIATRICS 2011;11:342–350
Childhood obesity interventions, even with moderate effectiveness, would make economic sense. Targeted interventions could yield higher cost savings than population based interventions for young children (aged 0 -6 years), whereas a population-based approach could yield greater economic net benefits for adolescents (aged 13-18 years). THE PREVALENCE OF childhood obesity continues to increase: over the last 30 years, the obesity rate for preschool children jumped from 5% to 12%.1 In 2006, an estimated 17% of children aged 6 to 11 years and 18% of adolescents aged 12 to 19 years were obese.2 Similarly to many other conditions, obesity rates are particularly high among minorities.1 Childhood obesity has been linked to numerous adverse health consequences, including, but not limited to metabolic syndrome, cardiac disorders, and respiratory disorders.3 Further, childhood obesity has been linked to social consequences such as academic challenges related to absenteeism and sleep disorders, workforce discrimination, loss of productivity, and income ACADEMIC PEDIATRICS Copyright ª 2011 by Academic Pediatric Association
342
Volume 11, Number 4 July–August 2011
ACADEMIC PEDIATRICS
among 12-year-olds by 1 percentage point.7 Although evidence for effective preventions is lacking, Transande’s study contributes greatly to providing the economic justifications for obesity preventions. However, it did not focus on early childhood obesity, which is important both because of the prevalence and the long-term implications of early interventions. Therefore, in this study, we constructed a novel mathematical model to simulate 2 scenarios for each of the 3 age groups: 0 to 6 years, 7 to 12 years, and 13 to 18 years, in an attempt to answer 2 salient policy questions: Given lifetime medical costs attributable to childhood obesity, at what level of effectiveness and costs would an early intervention be affordable? And secondly, should such interventions aim at the entire population or just target children who are already obese?
METHODS AND PROCEDURES OVERVIEW To project the lifetime medical cost of early childhood obesity, the parameters would ideally come from a longitudinal dataset in which a cohort is followed from birth to death, and data on height, weight, and medical expenses are collected at regular intervals. Unfortunately, such data do not exist. In this study, we constructed a mathematical model to overcome the barrier of lack of longitudinal data. We first estimated each age group’s medical costs attributable to obesity, and then applied the persistence of obesity throughout the lifespan (eg, what percentage of obese young children stay obese in adulthood) to the estimated cost at each age group. All the parameters for our simulation are based on US population. Data sources are described below. MAIN SOURCES OF PARAMETERS USED FOR PROJECTIONS NATIONAL HEALTH AND NUTRITION EXAMINATION SURVEY The statistics of national prevalence of childhood obesity over time come from a nationally representative data source: the National Health and Nutrition Examination Survey (NHANES). NHANES has been considered as the gold standard for measuring obesity because each individual’s height and weight were measured with standardized protocols and calibrated equipment during a physical examination. LITERATURE REVIEW To obtain values for key parameters, especially the persistence of obesity, we first conducted a thorough literature review. Statistics for obesity projections were identified using a PubMed search of literature published from 1998 to 2008 by using the US population or a sample. Key terms include child or adolescent combined with obesity or overweight or BMI or body mass index. To supplement this search, the reference section of each retrieved paper was reviewed for additional eligible articles. In addition, we conducted hand searches of the following journals: Health Affairs, American Journal of Public Health, New England Journal of Medicine, Journal
INVEST IN CHILDHOOD OBESITY INTERVENTIONS
343
of the American Medical Association, Pediatrics, Journal of Pediatrics, Obesity Research/Reviews, The Future of Children, and the American Journal of Preventive Medicine. A key inclusion criteria is that the study had to use the current definition of childhood obesity (defined as equal to or greater than the 95th percentile of the age- and genderspecific Centers for Disease Control and Prevention charts of childhood and adolescent body mass index [BMI]) and adult obesity (defined as BMI equal to or greater than 30). MEDICAL EXPENDITURE PANEL SURVEY Data from the 2006 Medical Expenditure Panel Survey (MEPS) were used to estimate the race-specific associations of obesity with expenditures. MEPS is a nationally representative survey of noninstitutionalized Americans that includes data on total medical expenditures aggregated over a year. The expenditures include both insurance and out-of-pocket spending. The MEPS also has abundant information on people’s health conditions (such as BMI) and sociodemographic status (such as health insurance coverage, race/ethnicity, age, education, and poverty status). Our final analytic sample included 17 927 adults (aged 18 years or above) and 4158 children (aged 6–17 years). STATISTICAL ANALYSIS PROJECTION OF OBESITY PERSISTENCE To be consistent with the current definition of childhood obesity and adult obesity, we included only studies that used the same measurements. We then summarized the persistence from different studies by baseline age, follow-up age, gender, race/ethnicity, and cohort. Given an estimate of the persistence of obesity from childhood to adulthood, we calculated the percentage of obese adults who were obese young children by dividing the product of current obesity rate of young children by obesity persistence by current obesity rate of adults. PROJECTION OF OBESITY-RELATED MEDICAL COST Finkelstein and colleagues8 have conducted several studies to estimate medical cost attributable to overweight and obesity. Following their approach, we also used a standard 2-part model approach in which 1) a logistic regression was used to estimate differences in the probability of utilization by race and obesity status, and 2) a generalized linear model with a log link function and a gamma family distribution of the error term is used to estimate differences in the expenditures among those who used any care. In both models, we controlled for a set of covariates including race (non-Hispanic white, non-Hispanic black, Hispanics), poverty status (<100% federal poverty level, 100%–125%, 125%–200%, 200%–400%, and $400%), gender, age groups, health insurance (private, public, and uninsured), census region (Northeast, Midwest, South, and West), education (less than high school, high school, college, and graduate school), and marital status. Current smoking status is also controlled in the analysis. This approach is consistent with previous research that investigates obesity-related cost.9–11
344
MA AND FRICK
We then calculated age-, gender- and race/ethnicity– specific medical expenditures attributable to obesity by multiplying the predicated probability and predicted expenditures from the previous steps. Finally, we discounted the estimated medical costs at a 3% annual rate and aggregated these costs to quantify lifetime medical expenditures from the perspective of 6-, 12-, and 18-year-old children. The difference in the present value of lifetime costs between normal-weight adults and obese adults provided an estimate of the lifetime costs attributable to obesity throughout the lifespan. Since life expectancy varies greatly by smoking status, race/ethnicity, and gender, we calculated the corresponding lifetime cost for each group.12 We then calculated a weighted lifetime medical cost by multiplying lifetime cost for each gender-, race- and smoking status–specific group with the proportion of each corresponding group in the total population. In the following simulation, we applied the weighted lifetime medical cost to project cost savings of interventions. SIMULATIONS OF POSSIBLE INTERVENTIONS The goal of the simulation was to demonstrate at what level of effectiveness and cost interventions would be affordable. We used parameters estimated from the previous steps, including persistence of obesity from childhood to adulthood and lifetime medical expenditures related to obesity. We considered 2 very different intervention strategies in the simulations: population-level interventions and targeted interventions. Population level interventions are those that use the typical public health approach and are universally delivered to everyone in the intervention groups. Targeted interventions are those that target only already obese children in the age groups. We considered them separately because the 2 strategies could be very different in terms of outreach, uptake, nature of intervention, and effectiveness. We simulated targeted and population level interventions for the 3 age groups: when interventions take place at early childhood (0–6 years old), middle childhood (7–12 years), and adolescence (13–18 years). For each age group’s interventions, we then applied a range (10%–60%) of effectiveness of the interventions. The effectiveness level is defined as the percentage of children who would have been obese, transition to, and remain in a nonobese status after the intervention. To make effectiveness levels more explicit and understandable, we translated the effectiveness levels to percentage point reduction in obesity rate (Appendix). The simulation follows 30 000 000 children for all 6 birth cohorts in each of the age groups of 0 to 6 years, 7 to 12 years, or 13 to 18 years (for demonstration purposes, we just use this approximate number instead of the actual number for each birth cohort). We then calculated the per capita cost savings should interventions take place: breakeven points are calculated for each age group: early childhood (0–6 years), middle childhood (7–12 years), and adolescents (13–18 years) by aggregating obesity-related cost by life expectancy. Finally, the total cost savings are divided by the total number in the cohort (for population-based interventions)
ACADEMIC PEDIATRICS
or by the total number of obese children in each cohort (for targeted interventions).
RESULTS PREVALENCE OF CHILDHOOD OBESITY The statistics of national prevalence of childhood obesity over time come from 2 key studies based on the NHANES data.1,13 The childhood obesity rate has steadily increased over time, with Hispanic boys and black girls having the highest prevalence. Our simulation relies on the 2003 to 2006 NHANES estimates of obesity prevalence. PERSISTENCE OF OBESITY THROUGH LIFE STAGES Our literature review identified 5 studies that investigated persistence of childhood obesity throughout life stages and qualified our inclusion criteria, as shown in Table 1. Two studies used nationally representative samples,14,15 and both found obesity persistence ranged from 77% to 92%, from adolescence to adulthood. The other 3 studies used local samples and followed young children to adulthood. Two papers based on the Bogalusa Heart Study (Louisiana) found relatively high obesity persistence from early childhood to young adulthood, ranging from 71% to 93%.16,17 The paper based on Fels Longitudinal Study used a white sample from Ohio and found a much lower obesity persistence: 37% and 31% of 5-year old obese girls and boys, respectively, were found still obese at age 35.18 The probability of being obese throughout childhood remained low until the participants turned 18, and then the persistence became 77% and 68% among girls and boys, respectively. Although the Fels Longitudinal Study found a much lower persistence, we observed some consistency across all studies: the correlation between early obesity and later obesity increases as baseline age rises. The wide range of obesity persistence is probably a result of small samples. Therefore, we concluded the obesity persistence at 6 years is reasonably around 50% (ie, 50% of obese children at age 6 years will remain obese in adulthood). For the purpose of modeling, we chose to use persistence of 76% for children aged 7 to 12 years and 86% for children aged 13 to 18 years. At the next step, we calculated the proportion of obese adults who were obese young children. Taking “female” as an example: given that the current obesity rate is 12.1% for girls aged 0 to 6 years and 23.5% for female adults, if 50% of obese girls at age 6 years stay obese in adulthood, then 25.7% (ie, [12.1% 50%]/23.5%) of obese female adults were obese girls at age 6 years and 74.3% became obese at some point after age 6 years. OBESITY-RELATED MEDICAL EXPENDITURE, PER CAPITA ANNUALLY VERSUS LIFETIME In 2006, obese adults spent $1548 per capita and obese children (aged 6–17 years) spent $264 per capita more on annual medical expenditure than normal-weight adults or children. Since age, gender, and race/ethnicity all are associated significantly with care utilization and
ACADEMIC PEDIATRICS
INVEST IN CHILDHOOD OBESITY INTERVENTIONS
345
Table 1. Reviewed Studies on Persistence of Childhood Obesity* Age
Persistence of Obesity†
Baseline
Follow-up
Women, No. (%)
Men, No. (%)
2–5 6–8 9–11 12–14 15–17 6.4–12.9 6.4–12.9 13–20 13–20 13–20 5 8 11 14 17 20 16–17
22.4–32.5 22.4–32.5 22.4–32.5 22.4–32.5 22.4–32.5 21–32 21–32 19–26 19–26 19–26 35 35 35 35 35 35 37/38
15 (73) 54 (83) 72 (78) 64 (83) 31 (90) 841 (65), W‡ 477 (84), B§ 269 (91), W 199 (89), B 100 (88), Hǁ N/A (37), W N/A (46), W N/A (59), W N/A (64), W N/A (77), W N/A (99), W 16 (92)
14 (93) 32 (78) 51 (76) 59 (88) 36 (86) 691 (71), W 383 (82), B 385 (77), W 128 (84), B 109 (77), H N/A (31), W N/A (22), W N/A (28), W N/A (40), W N/A (52), W N/A (98), W 29 (80)
Data Source and Population
Study
Bogalusa Heart Study, LA
Freedman et al, 2005 USA16
Bogalusa Heart Study, LA
Freedman et al, 2005, USA17
National Longitudinal Study of Adolescent Health, Wave II & III
Gordon-Larson et al, 2004, USA14
Fels Longitudinal Study, OH
Guo et al, 2002, USA18
National Longitudinal Survey of Youth 1979
Wang et al, 2008, USA15
*Childhood obesity, BMI $ 95th percentile of Centers for Disease Control and Prevention charts; adult obesity, BMI $ 30. †Sample size. ‡W ¼ white. §B ¼ black. ǁH ¼ Hispanic.
spending, we predicted the annual medical expenditure attributable to obesity in each age-, gender- and race/ ethnicity–specific group for adults and children. We then aggregated the lifetime medical cost attributable to obesity by using smoking-, race/ethnicity– and genderspecific annual cost and life expectancy, which are summarized in Table 2. For example, when other characteristics are similar, compared with a white, nonobese, female nonsmoker, the additional average lifetime medical cost of a white, obese, female nonsmoker is $40 874. The difference in lifetime medical cost between obese and nonobese people is the largest for white, female nonsmoker (due to high life expectancy and high annual cost attributable to obesity), and the smallest for black, male smokers (due to low life expectancy and low annual cost attributable to obesity). SIMULATIONS Table 2 presents the parameters used in the simulations, including the prevalence of obesity, the proportion of obese adults who were obese children, lifetime medical expenditures attributable to obesity, life expectancy, and the discount rate. Based on the parameters we chose to use, we first calculated the breakeven point at the birth cohort level. Interventions that can achieve a 1 percentage point reduction in the prevalence of obesity would result in approximately $1.7 billion cost-savings if interventions target children aged 0 to 6 years, or $1.4 billion if targeting children aged 7 to 12 years, or nearly $1.7 billion if targeting children aged 13 to 18 years. In other words, spending $1.4 to $1.7 billion in present value in combating childhood obesity could be cost-saving if 1 percentage point
reduction in obesity rate among children is achieved (Table 3). Table 3 also shows the per capita cost savings for both population-based interventions and targeted interventions. It shows that population-based interventions can spend up to $339 per child among those aged 0 to 6 years, $280 per child among those aged 7 to 12 years, and $339 per child among those aged 13 to 18 years for every 1 percentage point reduction in obesity rate. In other words, population-based interventions targeting children aged 13 to 18 years can spend up to $339 per child to breakeven, and any cost smaller than $339 per person will yield positive returns. In contrast, if we invest in interventions that target only obese children in each age group, we can afford to spend a higher amount on each child. For every 1 percentage point reduction in obesity rate, we can spend up to $2735 per obese child for those aged 0 to 6 years, or $1648 for those aged 7 to 12 years, or $1924 for those aged 13 to 18 years to yield positive economic returns. The Figure shows the relationships between effectiveness levels and break-even point for interventions in the 6 scenarios. Finally, to reflect the possibility of shortened life expectancy due to obesity and potential cost savings from lost life expectancy19 in our initial analysis, we did additional sensitivity analysis by excluding obesity-related medical costs after certain ages. The average difference in life expectancy by obesity status seems to be 5 years.19 So we start the sensitivity analysis by subtracting the life expectancy and its corresponding obesity-related medical costs from our estimated lifetime costs in 5-year intervals, until we reach life expectancy of age 65. In other words, the lowest lifetime costs are estimated for life expectancy of 65.
346
MA AND FRICK
ACADEMIC PEDIATRICS
Table 2. Parameters Used in Modeling* Parameter
Base Model Value
Current prevalence of obesity Age 0–6 years 12.4% Age 7–12 years 17.0% Age 13–18 years 17.6% Persistence of obesity at age 6 years 50.0% Male 26.4% obese male adults were obese children at age 6 years Female 25.7% obese female adults were obese children at age 6 years Persistence of obesity at age 12 years 76% Male 56.5% obese male adults were obese children at age 12 years Female 51.1% obese female adults were obese children at age 12 years Persistence of obesity at age 18 years 86% Male 64.7% obese male adults were obese children at age 18 years Female 61.5% obese female adults were obese children at age 18 years Lifetime medical expenditures attributable to obesity (life expectancy at age 18 years) Nonsmoking white female $40 874 (85) Nonsmoking white male $32 321 (81) Nonsmoking black female $37 032 (81) Nonsmoking black male $25 960 (75) Nonsmoking Hispanic female $30 765 (83) Nonsmoking Hispanic male $23 178 (78) Smoking white female $37 264 (78) Smoking white male $28 682 (72) Smoking black female $33 782 (75) Smoking black male $22 594 (67) Smoking Hispanic female $27 588 (76) Smoking Hispanic male $19 114 (69) Discounting rate 3% *We failed to find life expectancy of Hispanics in the United States by smoking status. Therefore, we estimated life expectancy of Hispanics by smoking status by multiplying proportion of Hispanics, gender ratio, smoking-status ratio, and the ratio of life expectance of smoker to nonsmoker in other racial groups. The reported lifetime expenditures related to obesity were discounted at 3% annually. All dollar amounts are the value of the US dollar in 2006.
DISCUSSION AND CONCLUSION Based on nationally representative data on prevalence and medical expenditures and small-scale longitudinal studies on persistence of obesity, our projection clearly shows that a large fraction of obesity-related medical expenditures can be attributable to childhood obesity because a sizeable proportion of obese adults were obese children. If we could prevent children from becoming obese or reverse already obese children to normal weight, we could save a considerable amount on future medical expenditures. Therefore, interventions during childhood are critical to the success in reducing obesity prevalence and decreasing obesity-related expenditures. The economic analysis finds that in 2006, each obese adult spent $1548 and each obese child (aged 6–17 years) spent $264 more on annual medical care than a normalweight adult or child. This estimate is consistent with previous studies.20 The lifetime medical cost varies greatly by race, gender, and smoking status, with white obese female nonsmokers incurring the most ($40 874 per person) and Hispanic obese male smokers incurring the least ($19 114 per person). Smokers incur lower medical expenditures related to obesity because they have shorter life expectancy. The contradiction between the goal of saving medical costs and the goal of extending life expectancy is a classic paradox for studies that use medical cost saving to justify interventions. To overcome this paradox, one can argue that reducing medical cost should not be
the sole objective of a health care system. In analysis, loss of productivity and quality of life can be factored in to reflect the additional loss due to such diseases. In our paper, since we chose only to estimate medical cost attributable to obesity, on which we have relatively solid data, the approach we chose to address the above-mentioned problem is to apply the weighted medical cost attributable to obesity for everyone, given the current smoking rate in each group. In other words, should smoking rates continue the current trend of declining, people will live longer; therefore, higher medical costs attributable to obesity will be expected, and our estimates of cost-savings likely represent a lower bound. It is interesting to compare our results with Trasande’s7 finding. His study reported $2 billion spending per cohort could be cost effective (based on a ratio of $50 000/ QALY [quality adjusted life year], and we do not consider QALYs) if 1 percentage point reduction in obesity rate could be achieved for 12 years olds. He showed a much lower level of saved medical expenditures for a 12-yearold cohort, approximately $260 million, but this was only to age 55. Our result shows that $1.4 billion spending could result in a positive net benefit for each birth cohort among the ages 7 to 12 years. Our seemingly higher estimates are the results of using life expectancy rather than by using an end point of age 55. One could argue that a disproportionately large amount of medical expenditures incur at the end of life, and it is questionable to link such extreme future medical cost to
Effectiveness Levels, %
$1 692 853 182 $339 $1924 $1 400 910 514 $380 $1648
5.9 5.4 4.9 4.4 3.9 3.4 3.0 2.5 2.0 1.5 1.0
$8 405 463 082 $7 705 007 825 $7 004 552 569 $6 304 097 312 $5 603 642 055 $4 903 186 798 $4 202 731 541 $3 502 276 284 $2 801 821 027 $2 101 365 771 $1 400 910 514
7.6 6.9 6.3 5.7 5.0 4.4 3.8 3.2 2.5 1.9 1.3
$12 696 398 864 $11 638 365 625 $10 580 332 387 $9 522 299 148 $8 464 265 909 $7 406 232 671 $6 348 199 432 $5 290 166 193 $4 232 132 955 $3 174 099 716 $2 116 066 477
INVEST IN CHILDHOOD OBESITY INTERVENTIONS
60 2.2 $3 730 706 117 55 2.0 $3 419 813 941 50 1.8 $3 108 921 764 45 1.6 $2 798 029 588 40 1.5 $2 487 137 411 35 1.3 $2 176 245 235 30 1.1 $1 865 353 059 25 0.9 $1 554 460 882 20 0.7 $1 243 568 706 15 0.5 $932 676 529 10 0.4 $621 784 353 Breakeven point for one percentage point reduction in obesity rate Total $1 695 775 508 Per capita, population-based interventions $339 Per capita, targeted interventions $2735
Total Cost Savings Percentage Point Reduction Among Obese Children Total Cost Savings Percentage Point Reduction Among Obese Children Total Cost Savings
Interventions Taking Place Among 7–12 Years Old Interventions Taking Place Among 0–6 Years Old
Percentage Point Reduction Among Obese Children
Table 3. Simulation Results of Breakeven Points for Percentage Point Reduction in Childhood Obesity Rate
Interventions Taking Place Among 13–18 Years Old
ACADEMIC PEDIATRICS
347
early childhood. Applying a 3% annual discount rate when aggregating lifetime medical costs places a lower weight on future costs, so that these costs have a smaller influence than they would if we counted all costs equally. Additionally, current life expectancy is well above age 55, even for those who are obese or are smokers. Not considering life beyond 55 will significantly underestimate the total impact of obesity on medical expenditures, potential savings of effective preventions, and its impact on Medicare. Finally, we conducted a sensitivity analysis by excluding obesityrelated medical costs after certain ages. If we excluded the last 5 years obesity-related medical costs, the per person lifetime obesity-related medical costs will be reduced by 4.8% to 8.9%, depending on race and gender. If we choose to use age 65 as the stopping point for everyone, the per person lifetime obesity-related medical costs will be reduced by 20.1% to 24.2%, depending on race and gender (details are available upon request). That means the estimated breakeven points for intervention spending (Table 3) could be decreased by 20% to 25%, as a more conservative estimate. If we apply 75% of previously estimated cost savings, the cost savings for 1 percentage point reduction in childhood obesity rate will now be between $1.07 billion and $1.28 billion. Per capita breakeven points for population-based interventions will range from $213 to $252, and per capita breakeven points for targeted interventions will range from $1255 to $2062. The results of sensitivity analysis show that even if we use a more conservative approach to account for future medical expenditures, early interventions still make economic sense. Although our estimated medical cost savings are higher than Trasande’s7 estimates, our estimation could still be at a lower bound for several reasons. First, only the medical costs related to obesity are calculated in this study. The figures would be larger if productivity loss and restricted activity were considered, although there is no direct way of quantifying such items. Second, this simulation does not take into account that interventions may have “spillover” effects, meaning that an early intervention not only may reduce the possibility of currently obese children growing to be obese adults but also may reduce the possibility of currently nonobese children becoming obese adults. Should we have empirical evidence on such preventive effects, the per capita cost saving could be higher. POLICY IMPLICATIONS The result of simulations shows that reasonably priced early interventions on obesity make economic sense; as a society we can afford spending a significant amount, up to $1.4 to $1.7 billion, to produce a 1 percentage point reduction in childhood obesity prevalence and obtain positive economic returns. This could have an enormous impact on health cost control. Our simulations also demonstrate the similarities in breakeven points among age groups: early populationbased interventions can spend up to $339 for each child
348
MA AND FRICK
aged 0 to 6 years, compared with $280 for each child among 7- to 12-year-olds, and $339 for each child aged 13 to 18 years. In contrast, early targeted interventions can spend up to $2735 for each child aged 0 to 6 years, compared with $1648 for each child aged 7 to 12 years, and $1924 for each child aged 13 to 18 years. This result is intuitive; per capita breakeven point is higher for adolescents (aged 13–18 years) than 6- to 11-year-olds because obese adolescents are much more likely to stay obese as adults. However, the per capita breakeven point per se should not be the only measure for allocating resources. One also needs to consider the relative difficulty of achieving 1 percentage point reduction in obesity rate in each age group, which remains unknown from current literature. As shown in both Table 3 and the Figure, although for population-based interventions, per capita breakeven point for every percentage point reduction is about the same in each age group, to reach 1 percentage point reduction requires a higher effectiveness level of the intervention in the younger age group. For example, for every 2 percentage point reduction in obesity, a 55% effectiveness level is needed among the 0- to 6-years age group, 20% effectiveness level is needed for those aged 7 to 12 years, and approximately 17% is needed for those aged 13 to 18 years. This implies that interventions need to be more effective for younger children than those targeting older children in order to achieve the same economic returns. The high cost savings of targeted interventions and needed higher effectiveness of interventions for children aged 0 to 6 years implies that providing targeted approaches perhaps makes more economic sense than providing population-based interventions. Aside from
ACADEMIC PEDIATRICS
economic returns, there are several conceptual reasons: for those who are obese from a very young age, their genetic and biologic endowment might be different from the rest of the population21; therefore, a universal approach may not work for them. Limited research has found engaging parents is one single important effective factor among early interventions,22 which again requires intensive and customized interventions. Additionally, empirical evidence suggests preventions targeting high-risk children, such as children with obese parents or from disadvantaged backgrounds, could achieve better results than those offering service to the whole population of children.23 In contrast, a population-based approach could be more applicable for older adolescents because they have a much higher obesity rate (18%), and there are problems such as stigma and feasibility imbedded in targeted interventions. With this group, population-based interventions could yield positive economic benefits if they are low cost; examples may include disclosing calorie content and nutrient composition on restaurant menus, providing only healthy foods in school cafeterias and vending machines, or mandating physical activities at school. Finally, policies that are universal to all children and parents could potentially have high returns (because adults have a higher prevalence of obesity and bear nearer-term adverse consequences), and could have potentially higher effectiveness (because parental involvement is key to childhood obesity interventions). We believe such universal policies, when proved to be effective, should be promoted, since policies that make healthy dietary and activity choices of a whole family easier are likely to achieve the favorable economic returns.24
Figure. Per capita breakeven point for simulated interventions, for each birth cohort.
ACADEMIC PEDIATRICS
LIMITATIONS Our study is constrained by several limitations. First, there are limited data from longitudinal studies on the persistence of obesity. We were only able to identify 5 high-quality studies, 2 of which used nationally representative datasets. Second, the simulations rely on several assumptions; we could not specify obesity persistence by race and gender because of the limitation of existing data. Blacks seem to have a higher obesity persistence,14,17 which suggests that effective programs targeted toward black children may yield even higher savings. Additionally, the estimation of lifetime medical cost attributable to obesity is an aggregation of costs from different birth cohorts, because there are no longitudinal data that allow us to conduct direct estimation. Finally, we could not adjust our estimates of lifetime medical cost attributable to obesity due to the lack of obesity-specific survival rate. It has been estimated that severe obesity could shorten one’s life expectancy by 2 to 10 years,25 to up to 5 to 20 years,26 which in turn reduces total lifetime medical cost. (Although we need to point out that the 20 years loss in life expectancy is estimated for extremely obese [BMI > 45] young black males, which are rare cases.) Realizing that lost life expectancy due to obesity could result in reduced medical expenditures,19 we conducted sensitivity analysis by excluding obesity-related medical costs after certain ages. The results of the sensitivity analysis show that even if we use a more conservative approach of accounting for future medical expenditures, early interventions still make economic sense. We hope that our findings of the significance of early childhood obesity will motivate policy makers to take action. Our simulation shows that childhood obesity interventions, even with moderate effectiveness, would make economic sense.
ACKNOWLEDGMENTS This study was funded by the Pew Charitable Trusts (Ma & Frick) and the Zanvyl and Isabelle Krieger Fund (Guyer). The authors thank Dr Eric Finkelstein and his colleagues for generously sharing their working papers. The authors also thank Dr Elizabeth Hair, Dr Bobbi Wolfe, and 2 anonymous reviewers for their constructive comments and suggestions, and Ms Allison Roeser for her assistance with the literature review.
REFERENCES 1. Ogden CL, Carroll MD, Flegal KM. High body mass index for age among US children and adolescents, 2003–2006. JAMA. 2008;299: 2401–2405. 2. Ogden CL, Carroll MD, Curtin LR, et al. Prevalence of overweight and obesity in the United States, 1999–2004. JAMA. 2006;295: 1549–1555. 3. Choudhary AK, Donnelly LF, Racadio JM, Strife JL. Diseases associated with childhood obesity. AJR Am J Roentgenol. 2007;188: 1118–1130. 4. Desjardins E, Schwartz AL. Collaborating to combat childhood obesity. Health Aff (Millwood). 2007;26:567–571.
INVEST IN CHILDHOOD OBESITY INTERVENTIONS
349
5. Bluford DA, Sherry B, Scanlon KS. Interventions to prevent or treat obesity in preschool children: a review of evaluated programs. Obesity. 2007;15:1356–1372. 6. Guyer B, Ma S, Grason H, et al. Early childhood health promotion and its life-course health consequences. Acad Pediatr. 2009;9:142–149. e1–e71. 7. Trasande L. How much should we invest in preventing childhood obesity? Health Aff (Millwood). 2010;29:372–378. 8. Finkelstein E, Trogdon JG, Brown DS, et al. The lifetime medical cost burden of overweight and obesity: implications for obesity prevention. Obesity. 2008;16:1843–1848. 9. Finkelstein E, Fiebelkorn I, Wang G. National medical spending attributable to overweight and obesity: how much and who’s paying? Health Aff (Millwood). 2003;(suppl web exclusives):219–226. 10. Finkelstein E, Fiebelkorn I, Wang G. State-level estimates of annual medical expenditures attributable to obesity. Obes Res. 2004;12: 18–24. 11. Finkelstein EA, Trogdon JG. Public health interventions for addressing childhood overweight: analysis of the business case. Am J Public Health. 2008;98:411–415. 12. Finklestein EA, Brown DS, Wrage LA, et al. Individual and aggregate years of life lost associated with overweight and obesity. Obesity. 2010;18:333–339. 13. Hedley AA, Ogden CL, Johnson CL, et al. Prevalence of overweight and obesity among US children, adolescents, and adults, 1999–2002. JAMA. 2004;291:2847–2850. 14. Gordon-Larsen P, Adair LS, Nelson MC, Popkin BM. Five-year obesity incidence in the transition period between adolescence and adulthood: the National Longitudinal Study of Adolescent Health. Am J Clin Nutr. 2004;80:569–575. 15. Wang LY, Chyen D, Lee S, Lowry R. The association between body mass index in adolescence and obesity in adulthood. Journal of Adolesc Health. 2008;42:512–518. 16. Freedman DS, Khan LK, Serdula MK, et al. The relation of childhood BMI to adult adiposity: the Bogalusa Heart Study. Pediatrics. 2005; 115:22–27. 17. Freedman DS, Khan LK, Serdula MK, et al. Racial differences in the tracking of childhood BMI to adulthood. Obes Res. 2005;13:928–935. 18. Guo SS, Wu W, Chumlea WC, Roche AF. Predicting overweight and obesity in adulthood from body mass. Am J Clin Nutr. 2002;76: 653–658. 19. van Baal PH, Polder JJ, de Wit GA, et al. Lifetime medical costs of obesity: prevention no cure for increasing health expenditure. PLoS Med. 2008;5:e29. 20. Finkelstein EA, Trogdon JG, Cohen JW, Dietz W. Annual medical spending attributable to obesity: payer-and service-specific estimates. Health Aff (Millwood). 2009;28:w822–831. 21. Bor J. The science of childhood obesity. Health Aff (Millwood). 2010; 29:393–397. 22. Olstad DL, McCargar L. Prevention of overweight and obesity in children under the age of 6 years. Appl Physiol Nutr Metab. 2009;34: 551–570. 23. Heitmann BL, Koplan J, Lissner L. Childhood obesity: successes and failures of preventive interventions. Nutr Rev. 2009;67(suppl 1): S89–S93. 24. Frieden TR, Dietz W, Collins J. Reducing childhood obesity through policy change: acting now to prevent obesity. Health Aff (Millwood). 2010;29:357–363. 25. Prospective Studies Collaboration, Whitlock G, Lewington S, Sherliker P, et al. Body-mass index and cause-specific mortality in 900 000 adults: collaborative analyses of 57 prospective studies. Lancet. 2009;373:1083–1096. 26. Fontaine KR, Redden DT, Wang C, et al. Years of life lost due to obesity. JAMA. 2003;289:187–193.
MA AND FRICK
350
APPENDIX EFFECTIVENESS LEVEL
We used the following formula to calculate the maximum reduction in probability of obese children becoming obese adults (Y):
ACADEMIC PEDIATRICS
of 100%. If an intervention is only 20% effective, that means we could reduce the probability of obese children becoming obese adults by 30% 20% ¼ 6%. To make effectiveness levels more understandable, we translate the effectiveness levels to reduction in probability of obese children becoming obese adults, and percentage point change among obese children at age 6 years (Appendix Table):
Y ¼ P ðORa ORc PÞ =ð100% ORc Þ
Where,
Y ¼ maximum reduction in probability of obese children becoming obese adults. P ¼ persistence of obese children remain obese as adults (eg, 50%, 60%). ORa ¼ current obesity rate for adults. ORc ¼ current obesity rate for children. Here is a concrete example: the current obesity rate (ORa) is 24% among adult men, and obesity rate (ORc) is 12.8% for boys; if we apply obesity persistence of 50%, then we get the following: Y ¼ 0:5 ð0:24 0:128 0:5Þ=ð1 0:128Þ ¼ 0:30 This means that we can maximally reduce the probability of obese children becoming obese adults by 30%, or 3.8% of all obese boys at age 6 years will not become obese adults (12.8% 30%). This is assuming the maximum reduction, in other words, at effectiveness level
Appendix Table. Effective Level vs. Percentage Point Reduction in Obesity Rate
Effectiveness Level, % 100 95 90 85 80 75 70 65 60 55 50 45 40 35 30 25 20 15 10
Reduction In Probability of Obese Children Becoming Obese Adults Reduction, %
Percentage Point Reduction In Obesity Rate Among Children
Male
Female
Male
Female
29.6 28.1 26.6 25.1 23.7 22.2 20.7 19.2 17.8 16.3 14.8 13.3 11.8 10.4 8.9 7.4 5.9 4.4 3.0
30.1 28.6 27.1 25.6 24.1 22.6 21.1 19.6 18.1 16.6 15.1 13.6 12.1 10.6 9.0 7.5 6.0 4.5 3.0
3.8 3.6 3.4 3.2 3.0 2.8 2.7 2.5 2.3 2.1 1.9 1.7 1.5 1.3 1.1 0.9 0.8 0.6 0.4
3.6 3.5 3.3 3.1 2.9 2.7 2.6 2.4 2.2 2.0 1.8 1.6 1.5 1.3 1.1 0.9 0.7 0.5 0.4
ABSTRACT OBJECTIVE: This study seeks to project at what level of effectiveness and cost a population-based or targeted intervention would yield a positive net economic benefit. METHODS: Data sources include prevalence of obesity at all ages from the National Health and Nutrition Examination Survey, the persistence of obesity from childhood to adulthood from a literature review, and a cost estimate from the 2006 Medical Expenditures Panel Survey. Econometric analysis was used to estimate medical cost related to obesity. Lifetime medical cost related to obesity is calculated by race, gender, and smoking status. Simulations were conducted to estimate the break-even point for interventions that take place between ages 0 and 6 years, ages 7 and 12 years, and ages 13 to 18 years, with a range of effectiveness. RESULTS: Results of simulations reveal that, from a pure medical cost perspective, spending approximately $1.4 to $1.7 billion at present value for each birth cohort will break even if 1 percentage point reduction in obesity among children is achieved. Population-based interventions can spend up to
between $280 and $339 per child at present value if 1 percentage point reduction in obesity rate could be achieved; in contrast, should we invest in interventions that only target obese children, we can spend up to $1648 to $2735 per obese child for every 1 percentage point reduction in obesity rate. CONCLUSIONS: This study has several important policy implications; early interventions make economic sense. Targeted interventions could yield higher cost savings than populationbased interventions for young children (aged 0–6 years), whereas a population-based approach could yield greater economic net benefits for adolescents (aged 13–18 years). Our simulation shows that childhood obesity interventions, even with moderate effectiveness, would make economic sense, which should motivate policy makers to take action.
WHAT’S NEW
reductions.4 As a result, increasing attention from policy makers and researchers has been paid to early prevention and programs.5 Nearly all decisions regarding public health investments involve options of prevention and treatment. Interventions targeting obesity can aim at preventing obesity, changing the weight of those already obese, or treating emergent consequences like diabetes. Even preventive efforts can take place at multiple stages, and it remains ambiguous as to whom policies should target to maximize economic returns. To endorse interventions at the earliest ages, one needs to understand 3 critical details: 1) the persistence of childhood obesity into adulthood, 2) the degree to which interventions are likely to be adopted by the children and their families at different stages in children’s lives, and 3) the potential returns on investment. As evidence-based interventions targeting young obese children (aged 0–6 years old) are particularly scarce,5,6 few studies have estimated under what conditions interventions could be effective and affordable. A recent study by Transande (2010) using mathematical modeling found that spending $2 billion per cohort would be cost effective if a hypothetical intervention can reduce obesity
KEYWORDS: childhood obesity; intervention; investment; persistence; prevention
ACADEMIC PEDIATRICS 2011;11:342–350
Childhood obesity interventions, even with moderate effectiveness, would make economic sense. Targeted interventions could yield higher cost savings than population based interventions for young children (aged 0 -6 years), whereas a population-based approach could yield greater economic net benefits for adolescents (aged 13-18 years). THE PREVALENCE OF childhood obesity continues to increase: over the last 30 years, the obesity rate for preschool children jumped from 5% to 12%.1 In 2006, an estimated 17% of children aged 6 to 11 years and 18% of adolescents aged 12 to 19 years were obese.2 Similarly to many other conditions, obesity rates are particularly high among minorities.1 Childhood obesity has been linked to numerous adverse health consequences, including, but not limited to metabolic syndrome, cardiac disorders, and respiratory disorders.3 Further, childhood obesity has been linked to social consequences such as academic challenges related to absenteeism and sleep disorders, workforce discrimination, loss of productivity, and income ACADEMIC PEDIATRICS Copyright ª 2011 by Academic Pediatric Association
342
Volume 11, Number 4 July–August 2011
ACADEMIC PEDIATRICS
among 12-year-olds by 1 percentage point.7 Although evidence for effective preventions is lacking, Transande’s study contributes greatly to providing the economic justifications for obesity preventions. However, it did not focus on early childhood obesity, which is important both because of the prevalence and the long-term implications of early interventions. Therefore, in this study, we constructed a novel mathematical model to simulate 2 scenarios for each of the 3 age groups: 0 to 6 years, 7 to 12 years, and 13 to 18 years, in an attempt to answer 2 salient policy questions: Given lifetime medical costs attributable to childhood obesity, at what level of effectiveness and costs would an early intervention be affordable? And secondly, should such interventions aim at the entire population or just target children who are already obese?
METHODS AND PROCEDURES OVERVIEW To project the lifetime medical cost of early childhood obesity, the parameters would ideally come from a longitudinal dataset in which a cohort is followed from birth to death, and data on height, weight, and medical expenses are collected at regular intervals. Unfortunately, such data do not exist. In this study, we constructed a mathematical model to overcome the barrier of lack of longitudinal data. We first estimated each age group’s medical costs attributable to obesity, and then applied the persistence of obesity throughout the lifespan (eg, what percentage of obese young children stay obese in adulthood) to the estimated cost at each age group. All the parameters for our simulation are based on US population. Data sources are described below. MAIN SOURCES OF PARAMETERS USED FOR PROJECTIONS NATIONAL HEALTH AND NUTRITION EXAMINATION SURVEY The statistics of national prevalence of childhood obesity over time come from a nationally representative data source: the National Health and Nutrition Examination Survey (NHANES). NHANES has been considered as the gold standard for measuring obesity because each individual’s height and weight were measured with standardized protocols and calibrated equipment during a physical examination. LITERATURE REVIEW To obtain values for key parameters, especially the persistence of obesity, we first conducted a thorough literature review. Statistics for obesity projections were identified using a PubMed search of literature published from 1998 to 2008 by using the US population or a sample. Key terms include child or adolescent combined with obesity or overweight or BMI or body mass index. To supplement this search, the reference section of each retrieved paper was reviewed for additional eligible articles. In addition, we conducted hand searches of the following journals: Health Affairs, American Journal of Public Health, New England Journal of Medicine, Journal
INVEST IN CHILDHOOD OBESITY INTERVENTIONS
343
of the American Medical Association, Pediatrics, Journal of Pediatrics, Obesity Research/Reviews, The Future of Children, and the American Journal of Preventive Medicine. A key inclusion criteria is that the study had to use the current definition of childhood obesity (defined as equal to or greater than the 95th percentile of the age- and genderspecific Centers for Disease Control and Prevention charts of childhood and adolescent body mass index [BMI]) and adult obesity (defined as BMI equal to or greater than 30). MEDICAL EXPENDITURE PANEL SURVEY Data from the 2006 Medical Expenditure Panel Survey (MEPS) were used to estimate the race-specific associations of obesity with expenditures. MEPS is a nationally representative survey of noninstitutionalized Americans that includes data on total medical expenditures aggregated over a year. The expenditures include both insurance and out-of-pocket spending. The MEPS also has abundant information on people’s health conditions (such as BMI) and sociodemographic status (such as health insurance coverage, race/ethnicity, age, education, and poverty status). Our final analytic sample included 17 927 adults (aged 18 years or above) and 4158 children (aged 6–17 years). STATISTICAL ANALYSIS PROJECTION OF OBESITY PERSISTENCE To be consistent with the current definition of childhood obesity and adult obesity, we included only studies that used the same measurements. We then summarized the persistence from different studies by baseline age, follow-up age, gender, race/ethnicity, and cohort. Given an estimate of the persistence of obesity from childhood to adulthood, we calculated the percentage of obese adults who were obese young children by dividing the product of current obesity rate of young children by obesity persistence by current obesity rate of adults. PROJECTION OF OBESITY-RELATED MEDICAL COST Finkelstein and colleagues8 have conducted several studies to estimate medical cost attributable to overweight and obesity. Following their approach, we also used a standard 2-part model approach in which 1) a logistic regression was used to estimate differences in the probability of utilization by race and obesity status, and 2) a generalized linear model with a log link function and a gamma family distribution of the error term is used to estimate differences in the expenditures among those who used any care. In both models, we controlled for a set of covariates including race (non-Hispanic white, non-Hispanic black, Hispanics), poverty status (<100% federal poverty level, 100%–125%, 125%–200%, 200%–400%, and $400%), gender, age groups, health insurance (private, public, and uninsured), census region (Northeast, Midwest, South, and West), education (less than high school, high school, college, and graduate school), and marital status. Current smoking status is also controlled in the analysis. This approach is consistent with previous research that investigates obesity-related cost.9–11
344
MA AND FRICK
We then calculated age-, gender- and race/ethnicity– specific medical expenditures attributable to obesity by multiplying the predicated probability and predicted expenditures from the previous steps. Finally, we discounted the estimated medical costs at a 3% annual rate and aggregated these costs to quantify lifetime medical expenditures from the perspective of 6-, 12-, and 18-year-old children. The difference in the present value of lifetime costs between normal-weight adults and obese adults provided an estimate of the lifetime costs attributable to obesity throughout the lifespan. Since life expectancy varies greatly by smoking status, race/ethnicity, and gender, we calculated the corresponding lifetime cost for each group.12 We then calculated a weighted lifetime medical cost by multiplying lifetime cost for each gender-, race- and smoking status–specific group with the proportion of each corresponding group in the total population. In the following simulation, we applied the weighted lifetime medical cost to project cost savings of interventions. SIMULATIONS OF POSSIBLE INTERVENTIONS The goal of the simulation was to demonstrate at what level of effectiveness and cost interventions would be affordable. We used parameters estimated from the previous steps, including persistence of obesity from childhood to adulthood and lifetime medical expenditures related to obesity. We considered 2 very different intervention strategies in the simulations: population-level interventions and targeted interventions. Population level interventions are those that use the typical public health approach and are universally delivered to everyone in the intervention groups. Targeted interventions are those that target only already obese children in the age groups. We considered them separately because the 2 strategies could be very different in terms of outreach, uptake, nature of intervention, and effectiveness. We simulated targeted and population level interventions for the 3 age groups: when interventions take place at early childhood (0–6 years old), middle childhood (7–12 years), and adolescence (13–18 years). For each age group’s interventions, we then applied a range (10%–60%) of effectiveness of the interventions. The effectiveness level is defined as the percentage of children who would have been obese, transition to, and remain in a nonobese status after the intervention. To make effectiveness levels more explicit and understandable, we translated the effectiveness levels to percentage point reduction in obesity rate (Appendix). The simulation follows 30 000 000 children for all 6 birth cohorts in each of the age groups of 0 to 6 years, 7 to 12 years, or 13 to 18 years (for demonstration purposes, we just use this approximate number instead of the actual number for each birth cohort). We then calculated the per capita cost savings should interventions take place: breakeven points are calculated for each age group: early childhood (0–6 years), middle childhood (7–12 years), and adolescents (13–18 years) by aggregating obesity-related cost by life expectancy. Finally, the total cost savings are divided by the total number in the cohort (for population-based interventions)
ACADEMIC PEDIATRICS
or by the total number of obese children in each cohort (for targeted interventions).
RESULTS PREVALENCE OF CHILDHOOD OBESITY The statistics of national prevalence of childhood obesity over time come from 2 key studies based on the NHANES data.1,13 The childhood obesity rate has steadily increased over time, with Hispanic boys and black girls having the highest prevalence. Our simulation relies on the 2003 to 2006 NHANES estimates of obesity prevalence. PERSISTENCE OF OBESITY THROUGH LIFE STAGES Our literature review identified 5 studies that investigated persistence of childhood obesity throughout life stages and qualified our inclusion criteria, as shown in Table 1. Two studies used nationally representative samples,14,15 and both found obesity persistence ranged from 77% to 92%, from adolescence to adulthood. The other 3 studies used local samples and followed young children to adulthood. Two papers based on the Bogalusa Heart Study (Louisiana) found relatively high obesity persistence from early childhood to young adulthood, ranging from 71% to 93%.16,17 The paper based on Fels Longitudinal Study used a white sample from Ohio and found a much lower obesity persistence: 37% and 31% of 5-year old obese girls and boys, respectively, were found still obese at age 35.18 The probability of being obese throughout childhood remained low until the participants turned 18, and then the persistence became 77% and 68% among girls and boys, respectively. Although the Fels Longitudinal Study found a much lower persistence, we observed some consistency across all studies: the correlation between early obesity and later obesity increases as baseline age rises. The wide range of obesity persistence is probably a result of small samples. Therefore, we concluded the obesity persistence at 6 years is reasonably around 50% (ie, 50% of obese children at age 6 years will remain obese in adulthood). For the purpose of modeling, we chose to use persistence of 76% for children aged 7 to 12 years and 86% for children aged 13 to 18 years. At the next step, we calculated the proportion of obese adults who were obese young children. Taking “female” as an example: given that the current obesity rate is 12.1% for girls aged 0 to 6 years and 23.5% for female adults, if 50% of obese girls at age 6 years stay obese in adulthood, then 25.7% (ie, [12.1% 50%]/23.5%) of obese female adults were obese girls at age 6 years and 74.3% became obese at some point after age 6 years. OBESITY-RELATED MEDICAL EXPENDITURE, PER CAPITA ANNUALLY VERSUS LIFETIME In 2006, obese adults spent $1548 per capita and obese children (aged 6–17 years) spent $264 per capita more on annual medical expenditure than normal-weight adults or children. Since age, gender, and race/ethnicity all are associated significantly with care utilization and
ACADEMIC PEDIATRICS
INVEST IN CHILDHOOD OBESITY INTERVENTIONS
345
Table 1. Reviewed Studies on Persistence of Childhood Obesity* Age
Persistence of Obesity†
Baseline
Follow-up
Women, No. (%)
Men, No. (%)
2–5 6–8 9–11 12–14 15–17 6.4–12.9 6.4–12.9 13–20 13–20 13–20 5 8 11 14 17 20 16–17
22.4–32.5 22.4–32.5 22.4–32.5 22.4–32.5 22.4–32.5 21–32 21–32 19–26 19–26 19–26 35 35 35 35 35 35 37/38
15 (73) 54 (83) 72 (78) 64 (83) 31 (90) 841 (65), W‡ 477 (84), B§ 269 (91), W 199 (89), B 100 (88), Hǁ N/A (37), W N/A (46), W N/A (59), W N/A (64), W N/A (77), W N/A (99), W 16 (92)
14 (93) 32 (78) 51 (76) 59 (88) 36 (86) 691 (71), W 383 (82), B 385 (77), W 128 (84), B 109 (77), H N/A (31), W N/A (22), W N/A (28), W N/A (40), W N/A (52), W N/A (98), W 29 (80)
Data Source and Population
Study
Bogalusa Heart Study, LA
Freedman et al, 2005 USA16
Bogalusa Heart Study, LA
Freedman et al, 2005, USA17
National Longitudinal Study of Adolescent Health, Wave II & III
Gordon-Larson et al, 2004, USA14
Fels Longitudinal Study, OH
Guo et al, 2002, USA18
National Longitudinal Survey of Youth 1979
Wang et al, 2008, USA15
*Childhood obesity, BMI $ 95th percentile of Centers for Disease Control and Prevention charts; adult obesity, BMI $ 30. †Sample size. ‡W ¼ white. §B ¼ black. ǁH ¼ Hispanic.
spending, we predicted the annual medical expenditure attributable to obesity in each age-, gender- and race/ ethnicity–specific group for adults and children. We then aggregated the lifetime medical cost attributable to obesity by using smoking-, race/ethnicity– and genderspecific annual cost and life expectancy, which are summarized in Table 2. For example, when other characteristics are similar, compared with a white, nonobese, female nonsmoker, the additional average lifetime medical cost of a white, obese, female nonsmoker is $40 874. The difference in lifetime medical cost between obese and nonobese people is the largest for white, female nonsmoker (due to high life expectancy and high annual cost attributable to obesity), and the smallest for black, male smokers (due to low life expectancy and low annual cost attributable to obesity). SIMULATIONS Table 2 presents the parameters used in the simulations, including the prevalence of obesity, the proportion of obese adults who were obese children, lifetime medical expenditures attributable to obesity, life expectancy, and the discount rate. Based on the parameters we chose to use, we first calculated the breakeven point at the birth cohort level. Interventions that can achieve a 1 percentage point reduction in the prevalence of obesity would result in approximately $1.7 billion cost-savings if interventions target children aged 0 to 6 years, or $1.4 billion if targeting children aged 7 to 12 years, or nearly $1.7 billion if targeting children aged 13 to 18 years. In other words, spending $1.4 to $1.7 billion in present value in combating childhood obesity could be cost-saving if 1 percentage point
reduction in obesity rate among children is achieved (Table 3). Table 3 also shows the per capita cost savings for both population-based interventions and targeted interventions. It shows that population-based interventions can spend up to $339 per child among those aged 0 to 6 years, $280 per child among those aged 7 to 12 years, and $339 per child among those aged 13 to 18 years for every 1 percentage point reduction in obesity rate. In other words, population-based interventions targeting children aged 13 to 18 years can spend up to $339 per child to breakeven, and any cost smaller than $339 per person will yield positive returns. In contrast, if we invest in interventions that target only obese children in each age group, we can afford to spend a higher amount on each child. For every 1 percentage point reduction in obesity rate, we can spend up to $2735 per obese child for those aged 0 to 6 years, or $1648 for those aged 7 to 12 years, or $1924 for those aged 13 to 18 years to yield positive economic returns. The Figure shows the relationships between effectiveness levels and break-even point for interventions in the 6 scenarios. Finally, to reflect the possibility of shortened life expectancy due to obesity and potential cost savings from lost life expectancy19 in our initial analysis, we did additional sensitivity analysis by excluding obesity-related medical costs after certain ages. The average difference in life expectancy by obesity status seems to be 5 years.19 So we start the sensitivity analysis by subtracting the life expectancy and its corresponding obesity-related medical costs from our estimated lifetime costs in 5-year intervals, until we reach life expectancy of age 65. In other words, the lowest lifetime costs are estimated for life expectancy of 65.
346
MA AND FRICK
ACADEMIC PEDIATRICS
Table 2. Parameters Used in Modeling* Parameter
Base Model Value
Current prevalence of obesity Age 0–6 years 12.4% Age 7–12 years 17.0% Age 13–18 years 17.6% Persistence of obesity at age 6 years 50.0% Male 26.4% obese male adults were obese children at age 6 years Female 25.7% obese female adults were obese children at age 6 years Persistence of obesity at age 12 years 76% Male 56.5% obese male adults were obese children at age 12 years Female 51.1% obese female adults were obese children at age 12 years Persistence of obesity at age 18 years 86% Male 64.7% obese male adults were obese children at age 18 years Female 61.5% obese female adults were obese children at age 18 years Lifetime medical expenditures attributable to obesity (life expectancy at age 18 years) Nonsmoking white female $40 874 (85) Nonsmoking white male $32 321 (81) Nonsmoking black female $37 032 (81) Nonsmoking black male $25 960 (75) Nonsmoking Hispanic female $30 765 (83) Nonsmoking Hispanic male $23 178 (78) Smoking white female $37 264 (78) Smoking white male $28 682 (72) Smoking black female $33 782 (75) Smoking black male $22 594 (67) Smoking Hispanic female $27 588 (76) Smoking Hispanic male $19 114 (69) Discounting rate 3% *We failed to find life expectancy of Hispanics in the United States by smoking status. Therefore, we estimated life expectancy of Hispanics by smoking status by multiplying proportion of Hispanics, gender ratio, smoking-status ratio, and the ratio of life expectance of smoker to nonsmoker in other racial groups. The reported lifetime expenditures related to obesity were discounted at 3% annually. All dollar amounts are the value of the US dollar in 2006.
DISCUSSION AND CONCLUSION Based on nationally representative data on prevalence and medical expenditures and small-scale longitudinal studies on persistence of obesity, our projection clearly shows that a large fraction of obesity-related medical expenditures can be attributable to childhood obesity because a sizeable proportion of obese adults were obese children. If we could prevent children from becoming obese or reverse already obese children to normal weight, we could save a considerable amount on future medical expenditures. Therefore, interventions during childhood are critical to the success in reducing obesity prevalence and decreasing obesity-related expenditures. The economic analysis finds that in 2006, each obese adult spent $1548 and each obese child (aged 6–17 years) spent $264 more on annual medical care than a normalweight adult or child. This estimate is consistent with previous studies.20 The lifetime medical cost varies greatly by race, gender, and smoking status, with white obese female nonsmokers incurring the most ($40 874 per person) and Hispanic obese male smokers incurring the least ($19 114 per person). Smokers incur lower medical expenditures related to obesity because they have shorter life expectancy. The contradiction between the goal of saving medical costs and the goal of extending life expectancy is a classic paradox for studies that use medical cost saving to justify interventions. To overcome this paradox, one can argue that reducing medical cost should not be
the sole objective of a health care system. In analysis, loss of productivity and quality of life can be factored in to reflect the additional loss due to such diseases. In our paper, since we chose only to estimate medical cost attributable to obesity, on which we have relatively solid data, the approach we chose to address the above-mentioned problem is to apply the weighted medical cost attributable to obesity for everyone, given the current smoking rate in each group. In other words, should smoking rates continue the current trend of declining, people will live longer; therefore, higher medical costs attributable to obesity will be expected, and our estimates of cost-savings likely represent a lower bound. It is interesting to compare our results with Trasande’s7 finding. His study reported $2 billion spending per cohort could be cost effective (based on a ratio of $50 000/ QALY [quality adjusted life year], and we do not consider QALYs) if 1 percentage point reduction in obesity rate could be achieved for 12 years olds. He showed a much lower level of saved medical expenditures for a 12-yearold cohort, approximately $260 million, but this was only to age 55. Our result shows that $1.4 billion spending could result in a positive net benefit for each birth cohort among the ages 7 to 12 years. Our seemingly higher estimates are the results of using life expectancy rather than by using an end point of age 55. One could argue that a disproportionately large amount of medical expenditures incur at the end of life, and it is questionable to link such extreme future medical cost to
Effectiveness Levels, %
$1 692 853 182 $339 $1924 $1 400 910 514 $380 $1648
5.9 5.4 4.9 4.4 3.9 3.4 3.0 2.5 2.0 1.5 1.0
$8 405 463 082 $7 705 007 825 $7 004 552 569 $6 304 097 312 $5 603 642 055 $4 903 186 798 $4 202 731 541 $3 502 276 284 $2 801 821 027 $2 101 365 771 $1 400 910 514
7.6 6.9 6.3 5.7 5.0 4.4 3.8 3.2 2.5 1.9 1.3
$12 696 398 864 $11 638 365 625 $10 580 332 387 $9 522 299 148 $8 464 265 909 $7 406 232 671 $6 348 199 432 $5 290 166 193 $4 232 132 955 $3 174 099 716 $2 116 066 477
INVEST IN CHILDHOOD OBESITY INTERVENTIONS
60 2.2 $3 730 706 117 55 2.0 $3 419 813 941 50 1.8 $3 108 921 764 45 1.6 $2 798 029 588 40 1.5 $2 487 137 411 35 1.3 $2 176 245 235 30 1.1 $1 865 353 059 25 0.9 $1 554 460 882 20 0.7 $1 243 568 706 15 0.5 $932 676 529 10 0.4 $621 784 353 Breakeven point for one percentage point reduction in obesity rate Total $1 695 775 508 Per capita, population-based interventions $339 Per capita, targeted interventions $2735
Total Cost Savings Percentage Point Reduction Among Obese Children Total Cost Savings Percentage Point Reduction Among Obese Children Total Cost Savings
Interventions Taking Place Among 7–12 Years Old Interventions Taking Place Among 0–6 Years Old
Percentage Point Reduction Among Obese Children
Table 3. Simulation Results of Breakeven Points for Percentage Point Reduction in Childhood Obesity Rate
Interventions Taking Place Among 13–18 Years Old
ACADEMIC PEDIATRICS
347
early childhood. Applying a 3% annual discount rate when aggregating lifetime medical costs places a lower weight on future costs, so that these costs have a smaller influence than they would if we counted all costs equally. Additionally, current life expectancy is well above age 55, even for those who are obese or are smokers. Not considering life beyond 55 will significantly underestimate the total impact of obesity on medical expenditures, potential savings of effective preventions, and its impact on Medicare. Finally, we conducted a sensitivity analysis by excluding obesityrelated medical costs after certain ages. If we excluded the last 5 years obesity-related medical costs, the per person lifetime obesity-related medical costs will be reduced by 4.8% to 8.9%, depending on race and gender. If we choose to use age 65 as the stopping point for everyone, the per person lifetime obesity-related medical costs will be reduced by 20.1% to 24.2%, depending on race and gender (details are available upon request). That means the estimated breakeven points for intervention spending (Table 3) could be decreased by 20% to 25%, as a more conservative estimate. If we apply 75% of previously estimated cost savings, the cost savings for 1 percentage point reduction in childhood obesity rate will now be between $1.07 billion and $1.28 billion. Per capita breakeven points for population-based interventions will range from $213 to $252, and per capita breakeven points for targeted interventions will range from $1255 to $2062. The results of sensitivity analysis show that even if we use a more conservative approach to account for future medical expenditures, early interventions still make economic sense. Although our estimated medical cost savings are higher than Trasande’s7 estimates, our estimation could still be at a lower bound for several reasons. First, only the medical costs related to obesity are calculated in this study. The figures would be larger if productivity loss and restricted activity were considered, although there is no direct way of quantifying such items. Second, this simulation does not take into account that interventions may have “spillover” effects, meaning that an early intervention not only may reduce the possibility of currently obese children growing to be obese adults but also may reduce the possibility of currently nonobese children becoming obese adults. Should we have empirical evidence on such preventive effects, the per capita cost saving could be higher. POLICY IMPLICATIONS The result of simulations shows that reasonably priced early interventions on obesity make economic sense; as a society we can afford spending a significant amount, up to $1.4 to $1.7 billion, to produce a 1 percentage point reduction in childhood obesity prevalence and obtain positive economic returns. This could have an enormous impact on health cost control. Our simulations also demonstrate the similarities in breakeven points among age groups: early populationbased interventions can spend up to $339 for each child
348
MA AND FRICK
aged 0 to 6 years, compared with $280 for each child among 7- to 12-year-olds, and $339 for each child aged 13 to 18 years. In contrast, early targeted interventions can spend up to $2735 for each child aged 0 to 6 years, compared with $1648 for each child aged 7 to 12 years, and $1924 for each child aged 13 to 18 years. This result is intuitive; per capita breakeven point is higher for adolescents (aged 13–18 years) than 6- to 11-year-olds because obese adolescents are much more likely to stay obese as adults. However, the per capita breakeven point per se should not be the only measure for allocating resources. One also needs to consider the relative difficulty of achieving 1 percentage point reduction in obesity rate in each age group, which remains unknown from current literature. As shown in both Table 3 and the Figure, although for population-based interventions, per capita breakeven point for every percentage point reduction is about the same in each age group, to reach 1 percentage point reduction requires a higher effectiveness level of the intervention in the younger age group. For example, for every 2 percentage point reduction in obesity, a 55% effectiveness level is needed among the 0- to 6-years age group, 20% effectiveness level is needed for those aged 7 to 12 years, and approximately 17% is needed for those aged 13 to 18 years. This implies that interventions need to be more effective for younger children than those targeting older children in order to achieve the same economic returns. The high cost savings of targeted interventions and needed higher effectiveness of interventions for children aged 0 to 6 years implies that providing targeted approaches perhaps makes more economic sense than providing population-based interventions. Aside from
ACADEMIC PEDIATRICS
economic returns, there are several conceptual reasons: for those who are obese from a very young age, their genetic and biologic endowment might be different from the rest of the population21; therefore, a universal approach may not work for them. Limited research has found engaging parents is one single important effective factor among early interventions,22 which again requires intensive and customized interventions. Additionally, empirical evidence suggests preventions targeting high-risk children, such as children with obese parents or from disadvantaged backgrounds, could achieve better results than those offering service to the whole population of children.23 In contrast, a population-based approach could be more applicable for older adolescents because they have a much higher obesity rate (18%), and there are problems such as stigma and feasibility imbedded in targeted interventions. With this group, population-based interventions could yield positive economic benefits if they are low cost; examples may include disclosing calorie content and nutrient composition on restaurant menus, providing only healthy foods in school cafeterias and vending machines, or mandating physical activities at school. Finally, policies that are universal to all children and parents could potentially have high returns (because adults have a higher prevalence of obesity and bear nearer-term adverse consequences), and could have potentially higher effectiveness (because parental involvement is key to childhood obesity interventions). We believe such universal policies, when proved to be effective, should be promoted, since policies that make healthy dietary and activity choices of a whole family easier are likely to achieve the favorable economic returns.24
Figure. Per capita breakeven point for simulated interventions, for each birth cohort.
ACADEMIC PEDIATRICS
LIMITATIONS Our study is constrained by several limitations. First, there are limited data from longitudinal studies on the persistence of obesity. We were only able to identify 5 high-quality studies, 2 of which used nationally representative datasets. Second, the simulations rely on several assumptions; we could not specify obesity persistence by race and gender because of the limitation of existing data. Blacks seem to have a higher obesity persistence,14,17 which suggests that effective programs targeted toward black children may yield even higher savings. Additionally, the estimation of lifetime medical cost attributable to obesity is an aggregation of costs from different birth cohorts, because there are no longitudinal data that allow us to conduct direct estimation. Finally, we could not adjust our estimates of lifetime medical cost attributable to obesity due to the lack of obesity-specific survival rate. It has been estimated that severe obesity could shorten one’s life expectancy by 2 to 10 years,25 to up to 5 to 20 years,26 which in turn reduces total lifetime medical cost. (Although we need to point out that the 20 years loss in life expectancy is estimated for extremely obese [BMI > 45] young black males, which are rare cases.) Realizing that lost life expectancy due to obesity could result in reduced medical expenditures,19 we conducted sensitivity analysis by excluding obesity-related medical costs after certain ages. The results of the sensitivity analysis show that even if we use a more conservative approach of accounting for future medical expenditures, early interventions still make economic sense. We hope that our findings of the significance of early childhood obesity will motivate policy makers to take action. Our simulation shows that childhood obesity interventions, even with moderate effectiveness, would make economic sense.
ACKNOWLEDGMENTS This study was funded by the Pew Charitable Trusts (Ma & Frick) and the Zanvyl and Isabelle Krieger Fund (Guyer). The authors thank Dr Eric Finkelstein and his colleagues for generously sharing their working papers. The authors also thank Dr Elizabeth Hair, Dr Bobbi Wolfe, and 2 anonymous reviewers for their constructive comments and suggestions, and Ms Allison Roeser for her assistance with the literature review.
REFERENCES 1. Ogden CL, Carroll MD, Flegal KM. High body mass index for age among US children and adolescents, 2003–2006. JAMA. 2008;299: 2401–2405. 2. Ogden CL, Carroll MD, Curtin LR, et al. Prevalence of overweight and obesity in the United States, 1999–2004. JAMA. 2006;295: 1549–1555. 3. Choudhary AK, Donnelly LF, Racadio JM, Strife JL. Diseases associated with childhood obesity. AJR Am J Roentgenol. 2007;188: 1118–1130. 4. Desjardins E, Schwartz AL. Collaborating to combat childhood obesity. Health Aff (Millwood). 2007;26:567–571.
INVEST IN CHILDHOOD OBESITY INTERVENTIONS
349
5. Bluford DA, Sherry B, Scanlon KS. Interventions to prevent or treat obesity in preschool children: a review of evaluated programs. Obesity. 2007;15:1356–1372. 6. Guyer B, Ma S, Grason H, et al. Early childhood health promotion and its life-course health consequences. Acad Pediatr. 2009;9:142–149. e1–e71. 7. Trasande L. How much should we invest in preventing childhood obesity? Health Aff (Millwood). 2010;29:372–378. 8. Finkelstein E, Trogdon JG, Brown DS, et al. The lifetime medical cost burden of overweight and obesity: implications for obesity prevention. Obesity. 2008;16:1843–1848. 9. Finkelstein E, Fiebelkorn I, Wang G. National medical spending attributable to overweight and obesity: how much and who’s paying? Health Aff (Millwood). 2003;(suppl web exclusives):219–226. 10. Finkelstein E, Fiebelkorn I, Wang G. State-level estimates of annual medical expenditures attributable to obesity. Obes Res. 2004;12: 18–24. 11. Finkelstein EA, Trogdon JG. Public health interventions for addressing childhood overweight: analysis of the business case. Am J Public Health. 2008;98:411–415. 12. Finklestein EA, Brown DS, Wrage LA, et al. Individual and aggregate years of life lost associated with overweight and obesity. Obesity. 2010;18:333–339. 13. Hedley AA, Ogden CL, Johnson CL, et al. Prevalence of overweight and obesity among US children, adolescents, and adults, 1999–2002. JAMA. 2004;291:2847–2850. 14. Gordon-Larsen P, Adair LS, Nelson MC, Popkin BM. Five-year obesity incidence in the transition period between adolescence and adulthood: the National Longitudinal Study of Adolescent Health. Am J Clin Nutr. 2004;80:569–575. 15. Wang LY, Chyen D, Lee S, Lowry R. The association between body mass index in adolescence and obesity in adulthood. Journal of Adolesc Health. 2008;42:512–518. 16. Freedman DS, Khan LK, Serdula MK, et al. The relation of childhood BMI to adult adiposity: the Bogalusa Heart Study. Pediatrics. 2005; 115:22–27. 17. Freedman DS, Khan LK, Serdula MK, et al. Racial differences in the tracking of childhood BMI to adulthood. Obes Res. 2005;13:928–935. 18. Guo SS, Wu W, Chumlea WC, Roche AF. Predicting overweight and obesity in adulthood from body mass. Am J Clin Nutr. 2002;76: 653–658. 19. van Baal PH, Polder JJ, de Wit GA, et al. Lifetime medical costs of obesity: prevention no cure for increasing health expenditure. PLoS Med. 2008;5:e29. 20. Finkelstein EA, Trogdon JG, Cohen JW, Dietz W. Annual medical spending attributable to obesity: payer-and service-specific estimates. Health Aff (Millwood). 2009;28:w822–831. 21. Bor J. The science of childhood obesity. Health Aff (Millwood). 2010; 29:393–397. 22. Olstad DL, McCargar L. Prevention of overweight and obesity in children under the age of 6 years. Appl Physiol Nutr Metab. 2009;34: 551–570. 23. Heitmann BL, Koplan J, Lissner L. Childhood obesity: successes and failures of preventive interventions. Nutr Rev. 2009;67(suppl 1): S89–S93. 24. Frieden TR, Dietz W, Collins J. Reducing childhood obesity through policy change: acting now to prevent obesity. Health Aff (Millwood). 2010;29:357–363. 25. Prospective Studies Collaboration, Whitlock G, Lewington S, Sherliker P, et al. Body-mass index and cause-specific mortality in 900 000 adults: collaborative analyses of 57 prospective studies. Lancet. 2009;373:1083–1096. 26. Fontaine KR, Redden DT, Wang C, et al. Years of life lost due to obesity. JAMA. 2003;289:187–193.
MA AND FRICK
350
APPENDIX EFFECTIVENESS LEVEL
We used the following formula to calculate the maximum reduction in probability of obese children becoming obese adults (Y):
ACADEMIC PEDIATRICS
of 100%. If an intervention is only 20% effective, that means we could reduce the probability of obese children becoming obese adults by 30% 20% ¼ 6%. To make effectiveness levels more understandable, we translate the effectiveness levels to reduction in probability of obese children becoming obese adults, and percentage point change among obese children at age 6 years (Appendix Table):
Y ¼ P ðORa ORc PÞ =ð100% ORc Þ
Where,
Y ¼ maximum reduction in probability of obese children becoming obese adults. P ¼ persistence of obese children remain obese as adults (eg, 50%, 60%). ORa ¼ current obesity rate for adults. ORc ¼ current obesity rate for children. Here is a concrete example: the current obesity rate (ORa) is 24% among adult men, and obesity rate (ORc) is 12.8% for boys; if we apply obesity persistence of 50%, then we get the following: Y ¼ 0:5 ð0:24 0:128 0:5Þ=ð1 0:128Þ ¼ 0:30 This means that we can maximally reduce the probability of obese children becoming obese adults by 30%, or 3.8% of all obese boys at age 6 years will not become obese adults (12.8% 30%). This is assuming the maximum reduction, in other words, at effectiveness level
Appendix Table. Effective Level vs. Percentage Point Reduction in Obesity Rate
Effectiveness Level, % 100 95 90 85 80 75 70 65 60 55 50 45 40 35 30 25 20 15 10
Reduction In Probability of Obese Children Becoming Obese Adults Reduction, %
Percentage Point Reduction In Obesity Rate Among Children
Male
Female
Male
Female
29.6 28.1 26.6 25.1 23.7 22.2 20.7 19.2 17.8 16.3 14.8 13.3 11.8 10.4 8.9 7.4 5.9 4.4 3.0
30.1 28.6 27.1 25.6 24.1 22.6 21.1 19.6 18.1 16.6 15.1 13.6 12.1 10.6 9.0 7.5 6.0 4.5 3.0
3.8 3.6 3.4 3.2 3.0 2.8 2.7 2.5 2.3 2.1 1.9 1.7 1.5 1.3 1.1 0.9 0.8 0.6 0.4
3.6 3.5 3.3 3.1 2.9 2.7 2.6 2.4 2.2 2.0 1.8 1.6 1.5 1.3 1.1 0.9 0.7 0.5 0.4