Customize your JAMA Network experience by selecting one or more topics from the list below.
Daviglus ML, Liu K, Yan LL, et al. Relation of Body Mass Index in Young Adulthood and Middle Age to Medicare Expenditures in Older Age. JAMA. 2004;292(22):2743–2749. doi:10.1001/jama.292.22.2743
Author Affiliations: Department of Preventive
Medicine (Drs Daviglus, Liu, Yan, Pirzada, Wang, Dyer, Greenland, and Stamler
and Mr Garside), Department of Medicine, Division of Geriatrics (Drs Daviglus
and Liu), Division of Cardiology (Dr Greenland), and the Institute for Health
Services Research and Policy Studies (Dr Manheim), Feinberg School of Medicine,
Northwestern University, Chicago, Ill; and Harris School of Public Policy
Studies, the University of Chicago, Chicago (Dr Manning).
Context Increasing prevalence of overweight/obesity and rapid aging of the US
population have raised concerns of increasing health care costs, with important
implications for Medicare. However, little is known about the impact of body
mass index (BMI) earlier in life on Medicare expenditures (cardiovascular
disease [CVD]–related, diabetes-related, and total) in older age.
Objective To examine relationships of BMI in young adulthood and middle age to
subsequent health care expenditures at ages 65 years and older.
Design, Setting, and Participants Medicare data (1984-2002) were linked with baseline data from the Chicago
Heart Association Detection Project in Industry (CHA) (1967-1973) for 9978
men (mean age, 46.0 years) and 7623 women (mean age, 48.4 years) (baseline
overall age range, 33 to 64 years) who were free of coronary heart disease,
diabetes, and major electrocardiographic abnormalities, were not underweight
(BMI <18.5), and were Medicare-eligible (≥65 years) for at least 2 years
during 1984-2002. Participants were classified by their baseline BMI as nonoverweight
(BMI, 18.5-24.9), overweight (25.0-29.9), obese (30.0-34.9), and severely
Main Outcome Measures Cardiovascular disease–related, diabetes-related, and total average
annual Medicare charges, and cumulative Medicare charges from age 65 years
to death or to age 83 years.
Results In multivariate analyses, average annual and cumulative Medicare charges
(CVD-related, diabetes-related, and total) were significantly higher by higher
baseline BMI for both men and women. Thus, with adjustment for baseline age,
race, education, and smoking, total average annual charges for nonoverweight,
overweight, obese, and severely obese women were, respectively, $6224, $7653,
$9612, and $12 342 (P<.001 for trend); corresponding
total cumulative charges were $76 866, $100 959, $125 470,
and $174 752 (P<.001 for trend). For nonoverweight,
overweight, obese, and severely obese men, total average annual charges were,
respectively, $7205, $8390, $10 128, and $13 674 (P<.001 for trend). Corresponding total cumulative charges were $100 431,
$109 098, $119 318, and $176 947 (P<.001
Conclusion Overweight/obesity in young adulthood and middle age has long-term adverse
consequences for health care costs in older age.
Obesity has been recognized as a major risk factor for coronary heart
disease (CHD)1 and is associated with increased
risk of hypertension, dyslipidemia, diabetes, certain cancers, and other disorders.2-8 Despite
declines in prevalence of other major CHD or cardiovascular disease (CVD)
risk factors such as hypertension, hypercholesterolemia, and smoking,9 the prevalence of overweight (body mass index [BMI,
calculated as weight in kilograms divided by height in meters squared], 25.0-29.9)
and obesity (BMI ≥30.0) has increased markedly during the last few decades
across all age, sex, socioeconomic, and ethnic groups in the United States
and in other countries.10-12 Currently,
approximately 130 million US adults are overweight or obese.13
At the same time, the US population is aging rapidly. It is estimated
that the proportion of US adults aged 65 years and older will increase from
about 12% currently to 20% by 2050.14,15 The
aging population has important implications for expenditures by Medicare—the
single largest source of health care spending in the United States. The combination
of escalating obesity and the increasing population of older individuals is
of concern to health care professionals, policy makers, and the US public.
Little is known about the relation of weight to long-term expenditures
for medical care. The few existing prospective studies on BMI and health care
costs are limited to short-term follow-up.16,17 The
impact of BMI in young adulthood and middle age on future Medicare expenditures
(CVD-related, diabetes-related, and total; annual and cumulative from age
65 years to death or to attainment of advanced age) has not been addressed.
This report examines these issues using data from participants of the Chicago
Heart Association Detection Project in Industry (CHA).
Between November 1967 and January 1973, the CHA study screened 39 522
men and women aged 18 years and older of varied ethnic and socioeconomic backgrounds
(mainly non-Hispanic white and about 10% African-American) employed at 84
Chicago-area organizations (Table 1).
Survey details have been reported.18,19 Briefly,
trained staff measured height, weight, a single casual supine blood pressure,
and levels of serum total cholesterol.20 All
measurements were collected in a standardized way. A self-administered questionnaire
was used to collect demographic data, smoking history, and information on
medical diagnoses and treatments, including those for hypertension and diabetes.
Resting electrocardiograms were classified as having major, minor only, or
no abnormalities.21,22 Race/ethnicity
categories were defined by investigators and assessed by interviewers to clarify
reasons for CVD rates being higher in blacks than in whites—a major
problem in the United States then and now. Vital status was ascertained through
2002, with mean (SD) follow-up of 32 (1.3) years. Deaths were determined by
several methods. Before 1979, deaths were determined by direct mail, telephone,
contact with employer, and matching of cohort records with Social Security
Administration files; and after 1979, by matching study records with National
Death Index records. Using these methods, only 0.23% of the total cohort (86
persons) have been untraced since baseline.
The study protocol has received periodic institutional review board
approval, and a waiver as described by the Health Insurance Portability and
Accountability Act was granted by the institutional review board prior to
commencement of the present project. Appropriate administrative and physical
safeguards were established to protect confidentiality of the data and to
prevent unauthorized use or access.
Medicare fee-for-service claims data were obtained from the Centers
for Medicare & Medicaid Services from 1984 (the first year Medicare data
were available for research use) through 2002 for participants aged 65 years
or older who were eligible for Medicare benefits. Medicare files for each
participant were cross-referenced by Social Security number, sex, and birth
date. For each medical service billed to Medicare, records include date of
service, total charges, principal diagnosis, and up to 9 other diagnoses coded
according to the International Classification of Diseases,
Ninth Revision, Clinical Modification (ICD-9-CM).23 Claims for acute inpatient (including skilled nursing
facility) and outpatient hospital-related care were available from 1984-2002;
physician visit (Part B) and durable medical equipment claims, home health
agency, and hospice claims were available from 1992-2002.
For each beneficiary, all health care charges were totaled and then
annualized by dividing the total by the number of years of Medicare coverage.
For the subgroup of individuals with data from age 65 years to death or to
attainment of age 83 years, cumulative charges were summed across all years.
For CVD-related costs, charges include those for health care services with
primary ICD-9-CM discharge diagnosis codes 390-459.
Diabetes-related costs are based on primary ICD-9-CM discharge
diagnosis code 250. To account for inflation, all charges were adjusted to
year 2002 dollars with use of the hospital and related services component
of the consumer price index (CPI).24 Since
health care charges in the United States have been escalating more rapidly
than other costs, analyses were repeated after substituting the all-item CPI
for the hospital component CPI to ensure that no bias was introduced with
adjustment for inflation.
Charges may overstate costs, but these are highly correlated, and relationships
of BMI to costs and charges are likely to be consistent.25 To
estimate costs, annual cost-to-charge ratios for hospital patient care services
obtained from the Medicare Payment Advisory Commission were applied to each
year’s Medicare charges.26 Cost-to-charge
ratios from 1984-2002 (ie, 19 ratios for 19 years) ranged from 0.800 to 0.413
(extrapolated for 1984). Sensitivity analyses substituting estimated costs
for charges were conducted to evaluate whether relationships between BMI and
Medicare charges were affected by the declining cost-to-charge ratio. In addition,
because Medicare fee-for-service claims do not consistently include charges
for beneficiaries enrolled in managed care plans (enrollment increased from
7.6% in 1991—the first-year enrollment information was available for
our cohorts—to 15.3% in 2002), supplemental analyses were performed
with exclusion of beneficiaries enrolled in managed care plans.
Of the 39 522 CHA participants, 21 253 men and women (baseline
ages <65 years) were eligible for Medicare benefits between 1984 and 2002
(ie, were not deceased before 1984 and were ≥65 years old during 1984-2002).
To increase the likelihood that participants would have incurred Medicare
charges, persons with fewer than 2 years of eligibility for Medicare coverage
(n = 968) were excluded. Of the remaining participants, 2684 were
excluded due to baseline findings: CHD (n = 236), diabetes mellitus
(n = 475), or major electrocardiographic abnormality (n = 1652);
missing data on height or weight (n = 7) or other covariates (n = 135);
or underweight (BMI <18.5; 28 men, 151 women). Thus, this report is based
on 9978 men and 7623 women. With identical inclusion and exclusion criteria,
the subcohort of participants eligible for Medicare from 1992-2002 (when charge
data for all types of claims were available) included 8857 men and 7022 women.
The subcohort with available data for cumulative charges from age 65 years
to death or to attainment of age 83 years included 2616 men and 2056 women.
Participants were grouped as nonoverweight (BMI, 18.5-24.9 [the reference
group]), overweight (BMI, 25.0-29.9), obese (BMI, 30.0-34.9), and severely
obese (BMI ≥35).2
Results are presented for men and women separately. Baseline characteristics
were compared across the 4 BMI groups; χ2 (for categorical
variables) or F tests (for continuous variables)
were used to assess statistical significance. Average annual and cumulative
charges were computed by sex and BMI categories in 3 general linear models:
model 1, adjusted for baseline age and race; model 2, additionally adjusted
for education (years) and smoking (cigarettes/d); and model 3, variables in
model 2 plus minor electrocardiographic abnormalities, vital status (indicator
for death during 1984-2002), and variables (serum cholesterol levels, systolic
blood pressure) potentially in causal pathways between BMI and development
of conditions such as that generate charges. These risk factors were included
in model 3 to assess independent effect of BMI on health care costs, because
numerous studies have demonstrated that the relationship of BMI to mortality
is attenuated but not eliminated with adjustment for risk factors potentially
intermediate in the causal pathway,6,27,28 and
because obesity was recently reclassified as a major modifiable risk factor
A modified Cox regression technique was used to test for statistical
significance of associations between baseline BMI and subsequent Medicare
charges.29 Linear trends across the 4 BMI groups
were tested using the significance level for coefficients for BMI as a continuous
variable in age-adjusted and multivariate-adjusted Cox regressions. All analyses
used SAS version 8.02 (SAS Institute Inc, Cary, NC). P<.05
was used to determine statistical significance.
For the cohort of participants aged 33 to 64 years at baseline with
at least 2 years of Medicare eligibility (1984-2002), mean baseline age was
46.0 years for men (n = 9978) and 48.4 years for women (n = 7623).
A majority of men were overweight (55.3%) or obese (14.3%) at baseline. Men
with higher BMI tended to be older and less educated and had higher prevalence
of minor electrocardiographic abnormalities and lower prevalence of smoking
(Table 1). Average blood pressure and
cholesterol levels were higher with higher BMI. Prevalence of overweight (30.3%)
and obesity (8.7%) in women was lower than in men. Relationships of BMI categories
with other characteristics were in general similar for women and men.
Among men, a significant positive relation was observed between BMI
and inpatient and outpatient hospital-related Medicare charges (1984-2002)
(Table 2). Age- and race-adjusted CVD-related,
diabetes-related, and total charges were significantly higher for overweight
and obese men. Total charges for severely obese men were $6192 more (84% higher)
than for nonoverweight men (model 1). Patterns of association remained similar
with multivariate adjustment (models 2 and 3); differences across BMI groups
decreased only slightly (all P values <.001 for
Women had lower Medicare inpatient and outpatient CVD-related, diabetes-related,
and total charges than men, but had larger proportional differences across
BMI groups (Table 2). Thus, total age-
and race-adjusted charges for severely obese women were $5618 more (88% higher)
than those for nonoverweight women. In model 2, adjusted also for education
and smoking, the graded relationship between BMI and charges among women remained
virtually unaltered and highly significant. Additional adjustments for risk
factors potentially in the causal pathways (model 3) only slightly attenuated
the association. Age- and race-adjusted diabetes-related charges were also
significantly higher among overweight and obese women compared with nonoverweight
women. These associations persisted with multivariate adjustment (models 2
and 3) (all differences P<.001).
For subgroups with Medicare data from age 65 years to death or to attainment
of age 83 years (2616 men, 2056 women), cumulative CVD-related, diabetes-related,
and total Medicare charges, adjusted for age, race, education, and smoking
were higher with higher BMI, significant for most of the comparisons, and
with P<.001 for trend in both sexes (Table 3). Results were only slightly attenuated,
with additional adjustments for risk factors possibly in the causal pathways
(model 3, data not shown).
Because baseline age and years of Medicare eligibility were highly correlated
(Pearson correlation coefficient, 0.56 for men and 0.58 for women), these
2 variables were not included in the same models. In analyses adjusted for
years of Medicare eligibility, results for both sexes were similar to those
with adjustment for baseline age (P<.001 for trend)
(data not shown). All tabulated analyses included participants who survived
through 2002 as well as those who died during the study period, adjusted for
vital status. Analyses among survivors (6383 men, 5154 women) showed similar
differences across BMI groups. For example, model 3–adjusted average
annual total charges (1984-2002) for nonoverweight vs severely obese participants
were $3995 vs $8228 (men) and $3562 vs $7206 (women).
Similar analyses for average annual charges were conducted in subcohorts
with data on all types of health care services (1992-2002). As above, BMI
was positively related to charges for CVD, diabetes, or any disease (P<.001 for trend for both sexes) (data not shown). Exclusion
of beneficiaries enrolled in managed care plans during 1992-2002 had little
impact on the observed relationship of BMI with CVD-related, diabetes-related,
and total charges (range of P values for trend, .04
Other analyses revealed no clear association between BMI and cancer-related
charges. Body mass index was directly associated with utilization of medical
care as indicated by average annual number of hospital visits and hospital
days. These positive associations were statistically significant for men but
not for women (data not shown).
In analyses adjusted for all-item CPI (instead of the hospital component
CPI), with average annual and cumulative charges lower for all BMI strata,
relationships between baseline BMI and Medicare charges were again positive
and significant (data not shown). Furthermore, with lower dollar amounts when
estimated Medicare costs were used, differences in health care costs across
BMI groups were similar to those for Medicare charges. For example, compared
with nonoverweight men, age- and race-adjusted total average annual costs
for severely obese men were 74% higher (ie, $4311 vs $7501), similar to the
difference in age- and race-adjusted total average annual charges between
these 2 groups (84%). Finally, in sensitivity analyses with inclusion of participants
with baseline CHD, diabetes, or major electrocardiographic abnormalities,
Medicare charges were higher across BMI strata, but the relationship of BMI
levels to charges were similar (P<.001 for all
Our main findings are that BMI assessed during young adulthood and middle
age was significantly and positively associated with average annual CVD-related
and total Medicare health care charges in older age as well as with CVD-related
and total cumulative charges from age 65 years to death or to age 83 years.
Results held true for both sexes, for both hospital-related and all types
of Medicare-covered services, and with adjustment for age (or years of Medicare
eligibility), race, education, smoking, cholesterol level, systolic blood
pressure, minor electrocardiographic abnormalities, and vital status.
Obesity is a major risk factor for CHD1 and
diabetes mellitus and also is associated with increased risk of stroke, cancers,
and other diseases.2-8 It
has been estimated that in the US population, more than 45% of the 9.3 million
cases of CVD30 and 280 000 deaths annually
(13.3% of all deaths)31 can be attributed to
obesity. Among young and middle-aged participants in the CHA study, BMI was
directly and independently associated with 25-year risk of CVD and total mortality.6,7 Proportions of CHA participants surviving
to at least age 65 years ranged from 80% (severely obese) to 89% (nonoverweight)
for men and from 91% (severely obese) to 95% (nonoverweight) for women; ie,
a high proportion of persons who are overweight or obese earlier in life live
to experience deleterious consequences in older age, including higher health
care costs (likely due to disease and disability). In 2003, about 7% of Medicare
expenditures were attributable to obesity.32
Most previous studies of relationships of BMI to health care costs are
cross-sectional or statistical simulations of long-term costs and have found
significant associations of BMI with health care costs.33-37 Results
from the few available prospective studies involve short-term follow-up.16,17 Among 1286 members of a large health
maintenance organization aged 35 to 64 years, total annual medical care costs
for inpatient, outpatient, and pharmacy services over the next 9 years were
10% and 36% higher among participants with BMIs of 25.0-29.9 and 30.0 or greater,
respectively, compared with those with BMI of 20.0-24.9. Cumulative health
care costs over the 9-year period were also greater with higher BMI.38 Since the average baseline age of the cohort was
only about 48 years, 9-year follow-up was insufficient to assess impact on
health care expenditures in older age. In another prospective study among
5689 enrollees (aged 40 years and older) of a Minnesota health plan, a 1-unit
increase in BMI was associated with 1.9% higher health care charges over the
next 18 months.16 Another prospective study
of 41 967 Japanese men and women (baseline ages 40-79 years) reported
that, compared with persons with BMIs of 21.0 to 22.9 (with the lowest costs),
total health care costs over 4 years were 9.8% and 22.3% higher among those
with BMIs of 25.0-29.9 and 30.0 or greater, respectively.17
These longitudinal studies, with relatively short-term follow-up, may
only partially reflect the full burden of medical expenditures associated
with overweight and obesity at younger ages. Our study on health care expenditures
by Medicare over a 19-year period is, to our knowledge, the first to show
impact of BMI measured earlier in life on health care costs in older age.
Our findings have important implications for future Medicare expenditures,
particularly given the continued and alarming increase in prevalence of overweight/obesity
in the United States during recent decades.10,11 Data
from the National Health and Nutrition Examination Survey (NHANES) 1999-2000
show that age-adjusted prevalences of overweight (BMI ≥25.0) and obesity
(BMI ≥30.0) among US adults are 65% and 31%, respectively, compared with
56% and 23% in 1988-1994 (NHANES III).11
Limitations of our study include a single measurement of BMI without
data on prior history of weight change or on duration of overweight and obesity,
which likely plays an important role in producing adverse effects on subsequent
Medicare expenditures. We also do not know how other related factors (eg,
physical inactivity, adverse diet) influence future health care costs. Furthermore,
although persons with CHD and diabetes at baseline were excluded from these
analyses, information was not collected to exclude those with other severe
chronic conditions at baseline that could influence health care costs. However,
the likelihood is small that participants with cancer or other severe chronic
diseases would still be alive decades later. Moreover, the CHA cohort was
derived from employed persons; thus, they were healthier than the general
population and less likely to have severe chronic diseases at baseline.
Another limitation is lack of data on use of health services not covered
by Medicare, such as, importantly, long-term nursing home care and prescription
drugs. Based on other research by our group, among CHA participants aged 65
and older, BMI levels earlier in life were directly associated with more adverse
mental, social, and physical functioning in older age and with greater use
of prescription drugs.39 Thus, potential bias
from this limitation is likely to be toward underestimation of health care
costs of high BMI levels. In addition, the use of only fee-for-service Medicare
data may lead to underestimates of actual total health care expenditures,
because health care costs incurred outside the Medicare system, mainly Health
Maintenance Organization and Veterans Administration costs, are not included.
However, Medicare is the largest single source of health care spending in
the United States. In additional analyses excluding Medicare beneficiaries
enrolled in health maintenance organizations, the positive relationships between
BMI and Medicare charges were virtually unaltered. Only a very small proportion
(<2%) of our cohort utilized Veterans Administration health care. Data
on out-of-pocket payments are also unavailable; these constitute only a small
proportion of total expenditures.
In conclusion, our findings demonstrate the adverse impact of high BMI
in young adulthood and middle age (irrespective of changes in weight that
may have occurred over the years) on future Medicare expenditures. With current
trends of increasing overweight and obesity afflicting all age groups, urgent
preventive measures are required not only to lessen the burden of disease
and disability associated with excess weight but also to contain future health
care costs incurred by the aging population. Public health efforts need to
include comprehensive national strategies and resources for primary prevention
of weight gain from early life on, with the goal to contain and end the obesity
epidemic and reduce health care costs among older persons.
Corresponding Author: Martha L. Daviglus,
MD, PhD, Department of Preventive Medicine, Feinberg School of Medicine, Northwestern
University, 680 N Lake Shore Dr, Suite 1102, Chicago, IL 60611 (firstname.lastname@example.org).
Author Contributions: Dr Daviglus had full
access to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analyses.
Study concept and design: Daviglus, Liu, Stamler.
Acquisition of data: Daviglus, Garside, Greenland.
Analysis and interpretation of data: Daviglus,
Liu, Yan, Pirzada, Manheim, Manning, Wang, Dyer, Greenland, Stamler.
Drafting of the manuscript: Daviglus, Yan,
Critical revision of the manuscript for important
intellectual content: Daviglus, Liu, Yan, Manheim, Manning, Garside,
Dyer, Greenland, Stamler.
Statistical analysis: Liu, Manheim, Manning,
Obtained funding: Daviglus, Liu, Dyer, Greenland,
Administrative, technical, or material support:
Study supervision: Daviglus, Stamler.
Funding/Support: This study was supported by
grants from the National Heart, Lung, and Blood Institute (R01 HL62684 and
HL21010), the Illinois Regional Medical Program, the Chicago Health Research
Foundation, and private donors.
Role of the Sponsors: The funding organizations
had no role in the design or conduct of the study; the collection, analysis,
or interpretation of the data; or the preparation, review, or approval of
Acknowledgment: We are indebted to the officers
and employees of the Chicago companies and organizations whose invaluable
cooperation and assistance made this study possible; to the staff members
and volunteers involved in the Chicago Heart Association Detection Project
in Industry; and to our colleagues who contributed to this endeavor. An extensive
list of colleagues is given in Cardiology. 1993;82:191-222.
Create a personal account or sign in to: