You are on page 1of 20

Reviews/Commentaries/ADA Statements


Economic Costs of Diabetes in the U.S.

in 2007

AMERICAN DIABETES ASSOCIATION revious research highlighting the
high cost of diabetes in terms of its
economic burden and cost to society
OBJECTIVE — The prevalence of diabetes continues to grow, with the number of people in has been helpful in health policy debates
the U.S. with diagnosed diabetes now reaching 17.5 million. The objectives of this study are to and decision-making (1– 4). Knowledge
quantify the economic burden of diabetes caused by increased health resource use and lost of the costs of diabetes improves under-
productivity, and to provide a detailed breakdown of the costs attributed to diabetes. standing of the importance of addressing
health care and prevention issues associ-
RESEARCH DESIGN AND METHODS — This study uses a prevalence-based ap- ated with diabetes. This 2008 update of
proach that combines the demographics of the population in 2007 with diabetes prevalence rates the 2003 American Diabetes Association
and other epidemiological data, health care costs, and economic data into a Cost of Diabetes
(ADA) study (3) encompasses the contin-
Model. Health resource use and associated medical costs are analyzed by age, sex, type of medical
condition, and health resource category. Data sources include national surveys and claims ued growth in diabetes prevalence;
databases, as well as a proprietary database that contains annual medical claims for 16.3 million changing practices, technology, and cost
people in 2006. to treat people with diabetes; and im-
provements in the data sources and meth-
RESULTS — The total estimated cost of diabetes in 2007 is $174 billion, including $116 ods to estimate the economic and social
billion in excess medical expenditures and $58 billion in reduced national productivity. Medical burden of diabetes.
costs attributed to diabetes include $27 billion for care to directly treat diabetes, $58 billion to
treat the portion of diabetes-related chronic complications that are attributed to diabetes, and
$31 billon in excess general medical costs. The largest components of medical expenditures RESEARCH DESIGN AND
attributed to diabetes are hospital inpatient care (50% of total cost), diabetes medication and METHODS — This study combines
supplies (12%), retail prescriptions to treat complications of diabetes (11%), and physician office research findings from the medical, pub-
visits (9%). People with diagnosed diabetes incur average expenditures of $11,744 per year, of lic health, and economics literature with
which $6,649 is attributed to diabetes. People with diagnosed diabetes, on average, have medical new empirical analysis to provide detailed
expenditures that are ⬃2.3 times higher than what expenditures would be in the absence of information on the cost of diabetes by de-
diabetes. For the cost categories analyzed, ⬃$1 in $5 health care dollars in the U.S. is spent caring mographic group, health care delivery
for someone with diagnosed diabetes, while ⬃$1 in $10 health care dollars is attributed to setting, medical conditions treated, and
diabetes. Indirect costs include increased absenteeism ($2.6 billion) and reduced productivity
while at work ($20.0 billion) for the employed population, reduced productivity for those not in
cost component. Major data sources ana-
the labor force ($0.8 billion), unemployment from disease-related disability ($7.9 billion), and lyzed and referenced in this article in-
lost productive capacity due to early mortality ($26.9 billion). clude the National Health Interview
Survey (NHIS), National Health and Nu-
CONCLUSIONS — The actual national burden of diabetes is likely to exceed the $174 trition Examination Survey (NHANES),
billion estimate because it omits the social cost of intangibles such as pain and suffering, care Medical Expenditure Panel Survey
provided by nonpaid caregivers, excess medical costs associated with undiagnosed diabetes, and (MEPS), Ingenix MCURE database, Na-
diabetes-attributed costs for health care expenditures categories omitted from this study. Omit- tionwide Inpatient Sample (NIS), Na-
ted from this analysis are expenditure categories such as health care system administrative costs, tional Ambulatory Medical Care Survey
over-the-counter medications, clinician training programs, and research and infrastructure de- (NAMCS), National Hospital Ambulatory
velopment. The burden of diabetes is imposed on all sectors of society— higher insurance
premiums paid by employees and employers, reduced earnings through productivity loss, and
Medical Care Survey (NHAMCS), Na-
reduced overall quality of life for people with diabetes and their families and friends. tional Nursing Home Survey (NNHS),
and National Home and Hospice Care
Diabetes Care 31:596–615, 2008 Survey (NHHCS). A brief description of
how these data sources are used and their
respective strengths and limitations for
● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
estimating the cost of diabetes is provided
This report was prepared by Tim Dall, Sarah Edge Mann, Yiduo Zhang, Jaana Martin, Yaozhu Chen, and Paul in an APPENDIX.
Hogan of The Lewin Group, Inc., Falls Church, Virginia.
Address correspondence and reprint requests to Matt Petersen, American Diabetes Association, 1701 N. The approach used to estimate medi-
Beauregard St., Alexandria, VA 22311. E-mail: cal costs largely follows the prevalence-
Abbreviations: ADA, American Diabetes Association; BLS, Bureau of Labor Statistics; CDC, Centers for based approach used in ADA’s two most
Disease Control and Prevention; CMS, Centers for Medicare and Medicaid Services; MEPS, Medical Expen- recent cost of diabetes studies (1,3). All
diture Panel Survey; NAMCS, National Ambulatory Medical Care Survey; NHAMCS, National Hospital estimates are extrapolated to the U.S.
Ambulatory Medical Care Survey; NHANES, National Health and Nutrition Examination Survey; NHHCS,
National Home and Hospice Care Survey; NHIS, National Health Interview Survey; NIS, Nationwide Inpa- population in 2007 and reported in year
tient Sample; NNHS, National Nursing Home Survey; PVFP, present value of future productivity; SSI, Social 2007 dollars. Cost data from earlier years
Security Insurance. were inflated to year 2007 dollars using
DOI: 10.2337/dc08-9017 the appropriate components of the medi-
© 2008 by the American Diabetes Association.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby
cal consumer price index (the nursing fa-
marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. cility index, the hospital inpatient index,
See accompanying editorial, p. 624. the pharmaceutical index, etc.).


American Diabetes Association

Estimating the size of the population excess medical costs and lost productivity The approach used to quantify the in-
with diabetes associated with their diabetes, but if this crease in health resource use associated
We estimate that approximately 17.5 mil- population incurs even modest disease- with diabetes was influenced by four data
lion people in the U.S. have been diag- related costs, the overall economic bur- limitations: 1) absence of a single data
nosed with diabetes. The estimated den of diabetes could be substantially source for all estimates, 2) small sample
prevalence of diagnosed diabetes is sub- higher than our estimates. size in some data sources, 3) correlation of
stantially higher than the 2002 estimate of Combining prevalence rates for the both diabetes and its comorbidities with
12.1 million from ADA’s previous cost noninstitutionalized population with es- other factors such as age and obesity, and
study, reflecting the growth and aging of timates for the long-term resident popu- 4) underreporting of diabetes as a comor-
the population, the rising prevalence of lation in nursing facilities, and applying bidity. We address these limitations and
overweight and obesity, improvements in rates to U.S. Census Bureau population their implications below.
detection, decreasing mortality, and estimates for 2007 by age-group, sex, and No single data source. There is no readily
growth in minority populations with race/ethnicity, suggests that over 24 mil- available data source that is ideal for esti-
higher rates of diabetes (5,6). Compari- lion people in the U.S. have diabetes (Ta- mating the increased use of health care
son of the 2007 estimate to the 2002 es- ble 1). Diabetes prevalence increases with
services associated with diabetes. The
timate suggests that the net number of age (type 2 diabetes). Blacks and Hispan-
MEPS is the most complete source in that
people with diagnosed diabetes is grow- ics have higher diabetes prevalence
it contains medical claims data, diabetes
ing by ⬃1 million people per year. These within each age-group, but because the
estimates are similar to those produced by age distribution of the groups is toward a status, patient demographic, and other
CDC and other published estimates (7). younger population than the age distribu- patient information such as health-
Analysis of the 2004 through 2006 tion of the nation as a whole, the higher seeking behavior. The MEPS, unfortu-
NHIS suggests that 5.6% of the noninsti- overall prevalence of these groups is nately, has insufficient sample size to
tutionalized population in the U.S. has di- masked somewhat. Approximately 2 mil- analyze health resource use and expendi-
agnosed diabetes. This estimate is based lion people with diabetes have no medical tures for many chronic complications of
on respondents answering “yes” to the insurance—with one in three of these un- diabetes by demographic group and
question of whether they have ever been insured people undiagnosed. Approxi- health care delivery setting. Also, the
told by a physician they have diabetes (ex- mately half of all people with diabetes MEPS sampling frame excludes the insti-
cluding gestational diabetes and “border- have medical insurance through the gov- tutionalized population, including long-
line” diabetes). Previous research has ernment—primarily through Medicare, term residents of nursing facilities who
found that self-report of physician’s diag- as an estimated 8.5 million people with have high prevalence of diabetes and are
nosis of diabetes is accurate (8). We com- diabetes are age 65 or older. high users of health care resources. For
bined NHIS files for years 2004 through each of the major health care delivery set-
2006 to increase sample size (n ⫽ Health resource use attributed to tings, we use a nationally representative
121,215 people) so that separate preva- diabetes data file: NAMCS for physician office vis-
lence rates could be estimated by eight People with diabetes have higher use of its, NHAMCS for hospital outpatient and
age-groups (⬍18, 18 –34, 35– 44, 45–54, hospital inpatient care, outpatient and emergency visits, NHHCS for hospice
55–59, 60 – 64, 65– 69, 70⫹), sex, and physician office visits, emergency visits, care; NIS for hospital inpatient care, and
race/ethnicity (non-Hispanic white, non- nursing facility stays, home health visits, NNHS for nursing facility care. We use
Hispanic black, non-Hispanic other, and visits with other health professionals, and MEPS to estimate health resource use and
Hispanic). Analysis of the 2004 NNHS prescription drug and medical supply use unit costs for a variety of health resources.
(n ⫽ 13,502 people) finds that ⬃24% of than their peers without diabetes. In ad- Small sample size. For some data
the estimated 1.5 million long-term resi- dition to health care services to directly sources we combined several years of data
dents of nursing facilities have been diag- treat diabetes, people with diabetes are at to increase sample size. Large samples
nosed with diabetes. We assume that the increased risk for neurological symp- were needed to produce reliable estimates
institutionalized population other than toms, peripheral vascular disease, cardio-
of health resource use and cost by age-
nursing homes (e.g., prisons) and the vascular disease, renal complications,
group and sex. (While race and ethnicity
noncivilian population have diabetes endocrine/metabolic complications, oph-
were used to estimate the number of peo-
prevalence rates similar to the noninstitu- thalmic complications, and other compli-
tionalized population controlling for de- cations (see APPENDIX for a comprehensive ple with diabetes, sample sizes were gen-
mographics. list of comorbidities) (3,9 –15). A portion erally insufficient to calculate separate
Analysis using NHANES suggests that of health care expenditures for these health resource use and unit cost patterns
2.2% of the noninstitutionalized U.S. chronic conditions, therefore, should be by race and ethnicity).
population has undiagnosed diabetes. We attributed to diabetes. We use the term Correlation of both diabetes and its co-
combined three NHANES files (1999 – “attributed” to mean the difference in morbidities with other factors. The prev-
2000, 2001–2002, and 2003–2004) to health care use for people with diabetes alence of type 2 diabetes increases with
increase sample size (n ⫽ 29,608 people) compared to what their health care use age and factors related to behavior (e.g.,
to calculate prevalence by demographic. would be in the absence of diabetes— diet and obesity), as does the risk for hy-
All cost estimates presented in this article estimating the excess health care use that pertension and other medical conditions.
are based on the population with diag- is theoretically due to (or caused by) dia- Not controlling for factors that increase
nosed diabetes. Additional research is betes and its complications. We use pri- risk of both diabetes and other medical
needed to determine the extent to which mary diagnosis code (ICD-9) to classify conditions will overestimate the health re-
people with undiagnosed diabetes incur medical claims data by medical condition. source use attributed to diabetes.


Economic costs of diabetes in the U.S. in 2007

Table 1—U.S. population (in thousands) and percent of U.S. population with diabetes, 2007

Total U.S. With diagnosed With undiagnosed

population diabetes diabetes Total with diabetes*
Total population* 301,736 17,486 5.8% 6,640 2.2% 24,126 8.0%
Non-Hispanic white 199,091 11,403 5.7% 4,520 2.3% 15,923 8.0%
Non-Hispanic black 37,002 2,775 7.5% 699 1.9% 3,474 9.4%
Non-Hispanic other 20,101 1,076 5.4% 317 1.6% 1,393 6.9%
Hispanic 45,541 2,231 4.9% 1,104 2.4% 3,335 7.3%
Male 148,744 8,543 5.7% 3,113 2.1% 11,656 7.8%
Female 152,992 8,943 5.8% 3,528 2.3% 12,471 8.2%
Age (years)
⬍18 73,878 157 0.2% 35 0.1% 192 0.3%
18–34 70,373 964 1.4% 669 1.0% 1,633 2.3%
35–44 43,356 1,686 3.9% 1,174 2.7% 2,860 6.6%
45–54 43,838 3,443 7.9% 1,327 3.0% 4,770 10.9%
55–59 18,235 2,307 12.7% 756 4.1% 3,063 16.8%
60–64 14,323 2,261 15.8% 775 5.4% 3,036 21.2%
65–69 10,690 1,879 17.6% 850 8.0% 2,729 25.5%
ⱖ70 27,042 4,788 17.7% 1,055 3.9% 5,843 21.6%
Private 169,886 7,057 4.2% 3,018 1.8% 10,075 5.9%
Government 91,794 8,997 9.8% 2,891 3.1% 11,888 13.0%
Uninsured 40,055 1,432 3.6% 731 1.8% 2,163 5.4%
Source: Combined information from the 2004 –2006 NHIS, 2004 NNHS, 1999 –2004 NHANES, and the U.S. Census Bureau population estimates for 2007.
*Numbers do not necessarily sum to totals because of rounding.

Underreporting of diabetes as a comor- Health resource use associated with annual per capita use for people
bidity. National data files for individual diabetes for three major health service de- with diabetesa,s,c
R H,a,s,c ⫽
health delivery settings are based on ad- livery settings (H)— hospital inpatient, annual per capita use for people
ministrative records and use admissions emergency departments, and other am- without diabetesa,s,c
or visits as the unit of analysis. Although bulatory visits—is estimated by combin-
diabetes is sometimes listed as a primary ing etiological fractions (␧) with total To help mitigate bias caused by correla-
or secondary condition in the diagnosis projected U.S. health service use (U) in tion between diabetes, the chronic medi-
codes, there is no way to link records 2007 as follows: cal conditions associated with diabetes,
across visits, admissions, and settings to and other factors, we use age-sex-setting–
create a profile that identifies all patients Attributed health resource use H specific etiological fractions for each med-
with diabetes.
Because of the above limitations we ⫽ 冘冘 冘 medical
age sex condition
ε H,a,s,c ⫻ U H,a,s,c
ical condition. The primary data source
for calculating etiological fractions is the
estimate health resource use attributed to Ingenix MCURE database. This database
diabetes using one of two approaches for contains a complete set of medical claims
each cost component. For some cost com- The etiological fraction for each age- for over 16.3 million people in 2006.
ponents we employ an attributed risk group (a), sex (s), and medical condition These claims primarily come from large
methodology using population etiological (c) is calculated using the diagnosed dia- employers, but the database also contains
fractions similar to the approach used in betes prevalence rate (P) and the rate ratio some Medicare claims. Despite the over-
previous ADA studies and in the cost of (R): representation of the privately insured
illness literature (16). For other cost com- population, the MCURE database proves
ponents (home health, pharmaceuticals, P a,s ⫻ 共R H,a,s,c ⫺ 1兲 to be an extremely powerful tool because
ε H,a,s,c ⫽
medical supplies, and podiatric services) P a,s ⫻ 共R H,a,s,c ⫺ 1兲 ⫹ 1 it allows us to link patient records during
we use a simple comparison of annual per the year and across health delivery set-
capita health resource use for people with The rate ratio for hospital inpatient tings. Also, due to its sheer size, it allows
and without diabetes controlling for age days, emergency visits, and other ambu- us to generate age-sex-setting–specific
and sex. Both approaches are equivalent latory visits (physician office visits com- relative rate ratios for each medical con-
under a reasonable set of assumptions, bined with hospital outpatient visits) dition that are more stable than rates es-
but the second approach cannot be used represents how annual per capita health timated using the MEPS.
with some of the data sources analyzed service use for the population with dia- Unlike the MEPS, the MCURE data-
because of underreporting of diabetes as a betes compares with the population base does not contain information on un-
comorbidity. without diabetes: derlying health status and unhealthy


American Diabetes Association

Figure 1—Rate of health resource use by men with diabetes relative to men without diabetes for visits/days where heart failure is the primary

lifestyle that could contribute to both di- being a contributing factor to both health logical fractions) are biased for the other
abetes and other health problems. For the problems. To remedy the attributed rate medical conditions modeled.
10 medical conditions that are the largest overestimation for these three categories, The prevalence of diabetes and co-
contributors to the overall cost of diabe- we scale the MCURE rate ratios using the morbidities, and the rate at which people
tes, we estimated a series of multivariate regression results from the MEPS analysis with diabetes use more comorbidity-
Poisson regressions, using data from the by applying the following scalar: related health care services, varies by de-
MEPS, to determine the extent to which mographic and by health care delivery
controlling only for age and sex might log(RRFull) setting. The rate ratios decline with age for
bias the rate ratios. We estimated two re- ScalarMEPS ⫽ (4) most medical conditions and are gener-
gressions: a naı̈ve and a full model. The ally higher for hospital inpatient days as
dependent variable for each regression compared to ambulatory visits. For illus-
is annual health resource use. Separate RRFinal ⫽ (RRMCURE)ScalarMEPS tration, we use male patients with heart
regressions were run using hospital inpa- failure as their primary diagnosis for the
tient days, emergency visits, and ambula- For general medical conditions, health visit or admission (Fig. 1). Men with dia-
tory visits (physician office and hospital care utilization is scaled down by 7% for betes use significantly more health re-
outpatient visits combined) as the depen- emergency visits, hospital outpatient vis- sources than men of similar age who do
dent variable. The naı̈ve model contains its, and physician office visits. Inpatient not have diabetes. For example, among
only a flag variable (1 ⫽ yes, 0 ⫽ no) for days are scaled down by 48%. Visits men age 60 – 64, those with diabetes have
diabetes and variables to control for age where hypertension is the primary diag- eight times the number of hospital inpa-
and sex. The full model contains the vari- nosis were scaled down by 33, 35, and tient days, seven times the number of
ables in the naı̈ve model, and also controls 34% for emergency visits, hospital outpa- emergency visits, and six times the num-
for education level, income, marital sta- tient visits, and physician office visits, re- ber of physician office and outpatient vis-
tus, medical insurance status, and race spectively. Visits where renal failure is the its for heart failure compared to their
and ethnicity. The rate ratio coefficients primary diagnosis were scaled down by 8, peers without diabetes.
for the diabetes flag variable in the naı̈ve 52, and 47% for emergency visits, hospi- We use an attributable risk approach
and full models are then compared. We tal outpatient visits, and physician office to estimate the portion of nursing facility
found statistically significant overesti- visits. use attributed to diabetes. Two analyses
mates in the rate ratio when using the na- We did not use the rate ratios from were conducted using the 2004 NNHS.
ı̈ve model for three condition categories: the MEPS analysis directly because the First, we compared the prevalence of di-
general medical conditions that are not MCURE data have a much larger sample abetes among nursing facility residents
identified as chronic conditions associ- size and the estimates are more stable with the prevalence of diabetes among the
ated with diabetes, hypertension, and re- than estimates generated from the MEPS. overall population in the same age-sex
nal failure (a major complication of both Analysis of the MEPS found no consistent group. Second, we analyzed the primary
diabetes and hypertension). Diabetes and evidence that rate ratios from the naı̈ve diagnosis codes associated with the nurs-
hypertension often coexist, with obesity model (which are used to calculate etio- ing facility admission, attributing to dia-


Economic costs of diabetes in the U.S. in 2007


29% 30%
Percent of Category Expenditures











General medical

Chronic Complications of Diabetes

Figure 2—Percent of category expenditures associated with diabetes. See APPENDIX for diagnosis codes for each category of complications.

betes 100% of the admissions where comparing this number to total health re- groups). Setting-specific etiological
diabetes is the primary diagnosis. Both source use, Fig. 2 summarizes the propor- fractions provide more precise estimates
approaches give similar results, so we use tion of medical expenditures for each for attributing health resource use be-
the first approach as it is more consistent condition attributed to diabetes. The eti- cause people with diabetes might have
with the approach used for other settings. ological fractions, however, vary substan- both increased incidence of use (e.g.,
Estimates of disease-related health re- tially by demographic and setting. more hospital admissions) and increased
source use for home health services, am- Differences between these overall eti- intensity of use (e.g., longer average
bulance services, podiatric services, ological fractions and those used in the length of hospital stay). Compared to the
pharmaceuticals, diabetes supplies, and 2003 study are due primarily to three fac- 2003 study, use of setting-specific etio-
other medical supplies were based on the tors: logical fractions increases the estimate of
MEPS. For these categories we calculated One, the etiological fractions used in hospital inpatient costs attributed to dia-
the average age-sex–specific annual ex- the 2003 study were estimated using odd betes and lowers the estimate of ambula-
penditures for people with diabetes com- ratios from logistic regression rather than tory costs attributed to diabetes.
pared to people without diabetes. relative rate ratios. While odds ratios are Three, for hypertension, renal failure,
Multiplying estimates of per capita health good proxies for the relative rate ratios for and general medical conditions, we adjust
resource use attributed to diabetes by the comorbidities with low prevalence, for the relative rate ratios to take into account
number of people with diabetes produced comorbidities with high prevalence (e.g., that controlling only for age and sex will
national estimates of health resource use general medical conditions) the etiologi- overestimate the impact of diabetes on
attributed to diabetes. Although we use cal fractions will be overestimated. prevalence of these three categories of
the NIS to estimate inpatient care attrib- Two, etiological fractions used in this medical conditions. Both hypertension
uted to diabetes, the NIS contains only study are based on a data source with and diabetes are correlated with obesity,
hospital costs (and not the cost for profes- much larger sample sizes, so we were able with obesity directly increasing the risk of
sional services by physicians who bill sep- to compute age-sex-setting–specific rela- hypertension independently of the in-
arate from the hospital). Therefore, we tive rate ratios for each comorbidity. For creased risk via diabetes. Likewise, the
use the MEPS to calculate use of physician hospital inpatient, emergency, and other clinical relationship between diabetes and
services per inpatient day by medical (office and hospital outpatient) ambula- renal failure, a comorbidity of both hyper-
condition. tory care settings, we calculate etiological tension and diabetes, might be overesti-
Summarizing the attributed health re- fractions for each of the eight age-groups mated if only controlling for age and sex.
source use across age, sex, health delivery modeled (whereas the 2003 study calcu- For general medical conditions, the
setting, and detailed comorbidities, and lated fractions for three larger age- hypothesis is that diabetes increases use


American Diabetes Association



Inpatient Days


Neurological Peripheral Cardiovascular Renal Metabolic Other General medical
vascular conditions

Chronic Complication of Diabetes

ALOS for admissions where diabetes is a secondary diagnosis

Predicted ALOS if diabetes were not a complicating factor

Figure 3—Average hospital length of stay (ALOS) when diabetes is a secondary diagnosis.

of health care services because diabetes longer length of stay (in addition to the tient visits, as well as for visits to physician
has a compounding influence. Analysis of impact of diabetes on increased risk of offices (Table 2). In addition, we use the
the MCURE database finds that people admission). For example, patients admit- MEPS to calculate the average cost per
with diabetes are slightly more likely to be ted for general medical conditions that are prescription filled (Table 3). We analyzed
admitted to the hospital for general med- not identified comorbidities of diabetes the NHIS to determine the percent of peo-
ical conditions, controlling for age and had an average length of stay of 13 days ple with diabetes (by age and sex) who use
sex, but they are also more likely to have a when diabetes was listed as a secondary insulin and oral medications. We com-
longer length of stay. Analysis of the 2005 complication. The multivariate regression bined this information with MEPS data on
NIS finds that hospital admissions where results suggest that if diabetes were not a the annual cost for insulin and oral med-
diabetes is listed as a secondary diagnosis complication, the length of stay would ications for each person who uses these
have longer average length of stay than have been 4 days shorter, indicating that items. Other cost components calculated
would occur if diabetes were not a com- diabetes increased the length of stay by on an annual per capita basis include
plicating factor. Using multivariate ordi- nearly 50%. home health services, ambulance ser-
nary least-squares regression (results not Estimates of health resource use at- vices, podiatric services, diabetes sup-
shown) with discharge-level data, we es- tributed to diabetes are combined with plies, and all other medical supplies. For
timate the relationship between length of estimates of the average medical cost per these categories we calculated the average
stay and having diabetes listed as a sec- event to compute total medical costs at- age-sex–specific annual cost per person
ondary diagnosis code, controlling for the tributed to diabetes. Cost data come pri- with and without diabetes. The excess in
average length of stay associated with the marily from the MEPS and the NIS, with cost for people with diabetes is consid-
primary diagnosis, patient age and sex, some cost estimates obtained from pub- ered to be attributed to diabetes.
and insurance status. We use the regres- lished sources. For many cost compo- We use the NIS, rather than the
sion findings to predict length of stay for nents, we use medical cost-per-event MEPS, to estimate hospital costs associ-
the admissions where diabetes is a sec- estimates specific to the medical condi- ated with inpatient services received.
ondary diagnosis, both with and in the tion modeled. Hospital-specific cost-to-charge ratios are
absence of diabetes as a complicating fac- Using the 2003–2005 combined available for the 2004 NIS and were used
tor (Fig. 3). The results, while not used to MEPS files, we calculate the average cost to calculate average cost per hospital day.
calculate the cost estimates presented, il- per visit for each category of medical con- The NIS is a much larger sample than the
lustrate that diabetes contributes to ditions for hospital emergency and outpa- MEPS, which is particularly important for


Economic costs of diabetes in the U.S. in 2007

Table 2—Average cost ($) per medical event, by complication, 2007

Chronic complications
Peripheral General
Medical event Diabetes Neurological vascular Cardiovascular Renal Metabolic Ophthalmic Other medical
Hospital inpatient day*† 1,853 2,363 2,226 3,225 1,872 2,156 2,408 1,718 2,188
Outpatient visit† 331 487 714 577 592 295 753 403 745
Emergency visit† 696 798 811 680 576 594 1,613 636 701
Physician office visit† 132 160 199 151 169 118 182 162 162
Source: *2004 NIS; †2003–2005 MEPS. Estimates adjusted to 2007 dollars.

estimating the average cost for medical years of lost work due to chronic disabil- not in the labor force as individuals in the
conditions that can have large variance in ity and premature mortality. The produc- labor force tend to have higher education
hospital cost per day. In addition, the NIS tivity loss for people employed in the levels (and by assumption greater pro-
contains five-digit diagnosis codes neces- workforce is calculated by combining ductivity, on average). We use 75% of the
sary for identifying many comorbidities of age-sex–specific estimates of workforce average earnings for people in the labor
diabetes, whereas the public use files for participation rates (estimated from the force as a proxy for the value of produc-
the MEPS contain only three-digit diag- 2006 NHIS), average earnings (from the tivity for people under age 65 not in the
nosis codes. However, costs captured by Bureau of Labor Statistics), and the size of labor force. We conduct sensitivity anal-
the NIS do not include physician costs so the population with diabetes. ysis to determine how this 75% assump-
we use the MEPS to calculate the physi- The cost of lost productivity is diffi- tion affects the overall cost estimates.
cian services cost per inpatient day by cult to quantify, especially for those not in
medical condition. Combining physician the labor force. Some studies have valued Absenteeism and presenteeism
costs from the MEPS and facility costs the time of those not in the labor force People with diabetes have higher rates of
from the NIS yields the cost-per- using minimum wage (i.e., to reflect the health-related absenteeism than their
inpatient-day estimates in Table 2. Over- cost to hire help in the home), but this peers without diabetes (21–23). A synthe-
the-counter medications are excluded likely underestimates the value of pro- sis of the published literature found that
from the analysis as there is no informa- ductivity for nonretired individuals who people with diabetes have health-related
tion to determine whether people with di- choose not to be in the workforce. On the absenteeism rate that was, on average,
abetes have higher rates of use than other hand, using the average earnings of 0.8% higher than people without diabetes
people without diabetes. those in the labor force will likely overes- (equivalent to 1.9 workdays per worker
timate the value of productivity for those with diabetes per year). Estimates vary
Productivity foregone
The indirect costs associated with many Table 3—Average cost per medical event, 2007
diseases are substantial, and for popula-
tions such as working-age adults, the lost Medical event Unit cost ($)
productivity costs to society can exceed
disease-related medical costs (17–20). Prescription (excluding insulin and oral agents)* 72
The indirect costs associated with diabetes Oral agents (per user per year)* 697
include health-related days absent from Insulin (per insulin user per year)* 751
work, reduced job performance due to Diabetic supplies (per person with diabetes per year)*, a 102
health problems, reduced labor force par- Home health visits (cost per person per day of use)*, a 204
ticipation and reduced earnings capacity Hospice care day† 147
from permanent disabilities, and lost pro- Nursing facility day‡ (excluding food and rent) b 131 (120)
ductivity from premature mortality. Other medical supplies (excess cost per person with diabetes per year)*, a
Lost productivity among adults with Glasses/contacts 5.57
diabetes in the labor force or who are pre- Ambulance services 5.92
vented from working by disability di- Orthopedic items 5.52
rectly affects the nation’s economic Hearing devices 9.02
output as measured by the gross domestic Prosthesis 4.09
product. Productivity loss also occurs Bathroom aids 1.52
among the population not in the labor Medical equipment 7.09
force and includes the impact of health Disposable supplies 10.85
problems on the ability to provide ser- Alterations/modifications 7.10
vices in the home and volunteer work. Other 0.14
To estimate the value of lost produc- Source: *2003–2005 MEPS; †Hospice Association of America 2006. ‡Average of cost for semi-private room
tivity, we calculate the number of full- and private room per the Genworth Financial 2007 Cost of Care Survey (
comweb/consumer/pdfs/long_term_care/Cost_Of_Care_Survey.pdf). Note: aCost estimate varies by age and
time equivalent workdays lost due to sex. bAn estimated 38% of the cost per day in a nursing home is for food and rent. For long-term residents,
increased absenteeism and reduced per- we exclude from the cost estimates expenses for room and board that would still have been incurred if the
formance at work (presenteeism), and the person were living at home. Estimates are adjusted to 2007 dollars.


American Diabetes Association

Age 18-34 Age 35-44 identify these cases using the 2004 –2006
Age 45-54
Age <18 1,918 5,564
NHIS, and use multivariate logistic re-
488 1% 2% gression (results not shown) to compare
0% Age 55-59 rates of disability by diagnosed diabetes
13,907 status controlling for other factors hy-
pothesized to be correlated with the like-
lihood of receiving disability payments.
Age 60-64 The rate of people unable to work because
17,634 of diabetes-related disability increases
6% with age.
The cost per person not working per
Age 65-69
day is calculated using average daily earn-
20,656 ings for those working. SSI payments are
7% not included in the cost estimate, as this is
considered a transfer payment (i.e., a cost
to one person is a benefit to another

Diabetes has become one of the leading
Age 70+ causes of death in the U.S. Death certifi-
206,791 cates underreport deaths attributed to di-
73% abetes because diabetes is often a
secondary cause of death. We combine
Figure 4—Age distribution of deaths associated with diabetes. CDC data for 2004 on cause of death, eti-
ological fractions derived from several
across the four studies reviewed— multaneously control for other factors sources, and estimates of total diabetes
ranging from 0.5% to 1.2% (23). Analysis that might be correlated with diabetes, cases in 2004 by age and sex to estimate
of the 2006 NHIS finds that self-reported and these same studies find that hyper- mortality rates for diabetes-attributed
workdays absent is higher for people with tension is associated with an average 6.9% deaths from renal disease, cerebrovascu-
diabetes, with part of this increase associ- decline in productivity. After adjusting lar disease, and cardiovascular disease (as
ated with higher rates of hypertension the diabetes presenteeism estimate for the well as directly from diabetes). We ex-
among people with diabetes. After adjust- portion of hypertension not attributed to trapolate these rates to the 2007 popula-
ing for the effect of hypertension not at- diabetes, we calculate a productivity loss tion using the growth in number of
tributed to diabetes, we calculate that the associated with diabetes of 6.6% (or 14 diabetes cases between 2004 and 2007.
average lost number of workdays per em- days per worker with diabetes per year). People with diabetes are at increased risk
ployed person per year is 1.8, ranging The effect of lost productivity on the job for many other diseases, and heart disease
from a low of 0.9 days for the population due to chronic illness, such as diabetes, and stroke account for an estimated 65%
aged 18 –34 years to a high of 2.5 days for could be reflected in the wage of the af- of deaths in people with diabetes (32).
the population ages 45–54 years. flicted worker rather than a cost to the Using relative risk information from a
Reduced productivity at work gener- employer. Our calculation of lost produc- meta analysis of the cardiovascular dis-
ally is measured using surveys, asking tivity reflects the cost to society without ease literature, we estimate that diabetes is
workers, for example, if during the past differentiating between the cost to the responsible for ⬃16% of deaths where
week they have experienced a loss of con- afflicted worker and the cost to the cardiovascular diseases (excluding cere-
centration, had to repeat a job, worked employer. brovascular diseases) is the primary cause
more slowly than usual, been fatigued at of death (32). Using the etiological frac-
work, or were unable to do their work. Disability tions for emergency department use as a
These surveys often ask for a self- Diabetes increases the likelihood that proxy for the mortality etiological frac-
assessment of the degree to which work chronic disability prevents working, or in tions, we estimate that 38%of deaths list-
productivity suffered on these days. Find- some cases limits employment opportu- ing cerebrovascular disease as the primary
ings from multiple studies suggest that nities reducing earnings (27–30). Be- cause and 57% of deaths listing renal fail-
people with diabetes have higher rates of tween 60 and 70% of people with ure as the primary cause can be attributed
reduced work productivity than do their diabetes have a form of neuropathy, such to diabetes. The majority of diabetes at-
peers without diabetes, although such as sensory impairment or pain in the foot tributed deaths occur among the elderly,
studies lack the rigor of a clinically con- or hands, and about 82,000 lower-limb with 73% of deaths occurring among peo-
trolled trial (23–26). amputations are performed each year on ple age 70 and older, and 7% of deaths
A synthesis of the literature found people with diabetes (31). The cost esti- occurring among people age 65– 69 (Fig.
that annual health-related at work pro- mates for diabetes-related disability are 4).
ductivity loss associated with diabetes is based on estimated disability cases that To estimate the national productivity
9.2%, although the rate from the four are sufficient to result in social security loss of early mortality we compute the net
studies reviewed ranged from 1.9% to supplemental insurance (SSI) pay- present value of future productivity
21.8% (23). These estimates do not si- ments—a conservative assumption. We (PVFP) for men and women by age. The


Economic costs of diabetes in the U.S. in 2007



Present value of future productivity






<18 18-34 35-44 45-54 55-59 60-64 65-69 70+
Age at death
Male Female

Figure 5—Net present value of future lost earnings from premature death.

approach combines Bureau of Labor Sta- The PVFP declines with age and is higher ployment from chronic disability, and
tistics (BLS) information on average an- for men than for women, reflecting differ- premature mortality. In addition to these
nual earnings (E) and labor force ences in annual earnings, labor force par- quantified costs, diabetes imposes high
participation rates (L) by age (A) and sex ticipation rates, and expected remaining intangible costs on society in terms of re-
(S), and CDC information on expected years of life. The PVFP is sensitive to the duced quality of life, and the pain and
mortality rates (M) for men and women in chosen discount rate, with each percent- suffering of people with diabetes, their
the U.S (33). We use a real discount rate age point increase in the discount rate families, and their friends.
(d) of 3%, a rate often used in public causing the U.S. total PVFP productivity ADA’s previous estimate of the cost of
health studies and assume that productiv- loss from diabetes-attributed mortality to diabetes ($132 billion in 2002, consisting
ity growth (p) will increase real earnings drop by $1.5–$2 billion (or 6 –7% of total of $92 billion in medical costs and $40
by 1% per year (34 –36). Also, we assume mortality costs). billion in lost productivity), if adjusted for
that the value of productivity generated inflation would be equivalent to $153 bil-
for the population age 18 – 65 who are not RESULTS lion in 2007 dollars. The $21 billion in-
employed is 75% of the value of em- crease between the 2002 estimate (in
ployed people of the same age and sex. Summary 2007 dollars) and new 2007 estimate of
Labor force participation rates and aver- The total national cost of diabetes in the $174 billion reflects three factors: 1)
age annual earnings are low for the pop- U. S. is estimated at $174 billion in 2007. growth in diabetes prevalence, 2) medical
ulation age 75 and older, so we make the Diabetes is responsible for an estimated costs rising faster than general inflation,
simplifying assumption that PVFP for this $116 billion in medical costs, as well as and 3) improvements made in the meth-
population is zero. NPV is calculated us- $58 billion in reduced productivity from ods and data sources to estimate the cost
ing the following formula, with estimates work-related absenteeism, reduced pro- of diabetes. Coincidentally, the $116 bil-
by age and sex provided in Fig. 5: ductivity at work and at home, unem- lion in excess medical expenditures is al-
most the same number obtained by
inflating the 2002 estimate ($92 billion)
by growth in diagnosed diabetes preva-

冢 冣
E a,S ⫻ 共0.75 ⫹ 0.25 ⫻ L a,S兲 ⫻ 共1 ⫹ p兲
⫻ 兿 共1 ⫺ M a,S兲 lence and general medical inflation as
兺 measured by BLS.
共1 ⫹ d兲
共a⫺A兲 The indirect costs of diabetes are
higher than the 2002 estimates, even after
adjusting for growth in diabetes preva-
a lence and general inflation, in large part

冢 冣
E a,S ⫻ L a,S ⫻ 共1 ⫹ p兲
⫻ 兿 共1 ⫺ M a,S兲
because of the change in data and meth-
ods to model lost productivity due to

⫹ absenteeism and presenteeism. The cost
a⫽65 共1 ⫹ d兲 of unemployment from permanent dis-


American Diabetes Association

Table 4—Health resource use (in millions of units) in the U.S., by diabetes attribution and cost component, 2007

Population with diabetes

Incurred by people
Incurred by
Attributed to diabetes with diabetes
% of % of without
Health resource Units U.S. total Units U.S. total diabetes U.S. total*
Institutional care
Hospital inpatient days 24.3 13 40.7 22 145.7 186.4
Nursing/residential facility days 56.1 10 140.6 25 433.3 573.9
Outpatient care
Physician’s office visits 64.7 7 137.6 14 822.1 959.7
Emergency visits 5.6 5 11.5 10 104.4 115.9
Hospital outpatient and freestanding ambulatory 5.7 6 10.9 12 81.7 92.7
surgical center visits
Home health visits 27.4 13.8 46.0 23.2 152.7 198.8
Hospice care days 0.2 ⬍1 9.6 11 81.5 91.1
Retail prescriptions 175.3 8 359.6 17 1,762.0 2,121.6
*Numbers do not necessarily sum to totals because of rounding.

ability and premature mortality are con- earlier estimate, reflecting a change in shows the proportion of U.S. health re-
sistent with the 2002 estimates, adjusting method for calculating attributed days source use attributed to diabetes. Nursing
for growth in diabetes prevalence and (i.e., use of a population etiological frac- facility days, home health visits, and hos-
inflation. tion that is specific to nursing homes pice care days are not calculated by com-
rather than separate etiological fractions plication. These tables reveal several
Health resource use attributed to for each medical condition that were de- trends:
diabetes rived from the MEPS).
Combining estimates of per capita use of Table 5 shows estimated health re-
● A large portion of health resource use
health care services, diabetes prevalence source use attributed to diabetes, by type
rates, etiological fractions, and the 2007 of service and aggregated into three broad attributed to diabetes is for general
population in the U.S. produces national age categories. The population age 65 and medical conditions that are not chronic
estimates of total health resource use, older uses a large portion of services, re- complications of diabetes. As discussed
health resource use by people with diabe- flecting the burden diabetes places on the earlier, we find that diabetes contrib-
tes, and health resource use attributed to Medicare program. utes to longer hospital length of stay
diabetes (Table 4). For example, of the Tables 6 – 8 provide a breakdown of regardless of the reason for admission
projected 186 million hospital inpatient health resource use attributed to diabetes (and controlling for other factors that
days in the U.S. in 2007, an estimated by medical condition. Table 6 shows total affect hospital length of stay). Com-
40.7 million days (22%) are incurred by use attributed to diabetes; Table 7 shows pared to their peers without diabetes,
people with diabetes and 24.3 million the share of attributed use associated with people with diabetes have much higher
(13%) are attributed to diabetes. While 1 each medical condition; and Table 8 rates of physician office visits and emer-
in 4 nursing facility days is incurred by a
person with diabetes, 1 in 10 days is at-
Table 5—Health resource use attributed to diabetes in the U.S., by age and type of service,
tributed to diabetes. About half of all phy-
2007 (in thousands)
sician office visits, emergency visits,
hospital outpatient visits, and outpatient
prescriptions (excluding oral agents and Age (years)
insulin) incurred by people with diabetes Health resource ⬍ 45 45–64 ⱖ 65 Total*
are attributed to their diabetes.
Differences between these estimates Institutional care
and estimates from the ADA 2003 study Hospital inpatient days 2,115 7,586 14,562 24,262
reflect, in part, the use of setting-specific Nursing/residential facility days 1,269 11,103 43,687 56,059
etiological fractions and a new data source Outpatient care
to estimate the relative rate ratios used to Office-based physician visits 7,353 26,552 30,808 64,713
calculate the etiological fractions. Attrib- Emergency visits 1,499 1,984 2,084 5,567
uted hospital inpatient days, and physi- Hospital outpatient and freestanding 1,307 2,535 1,888 5,730
cian office and emergency visits all ambulatory surgical center visits
increased from the earlier study, with Home health visits 0 8,939 18,449 27,388
hospital outpatient visits remaining rela- Hospice care days 4 22 165 192
tively unchanged. Attributed nursing fa- Retail prescriptions 15,181 71,295 88,841 175,317
cility days are substantially lower than our *Numbers do not necessarily sum to totals because of rounding.


Economic costs of diabetes in the U.S. in 2007

Table 6—Health resource use attributed to diabetes in the U.S., by medical condition (in thousands)

Chronic complications
Neuro- Peripheral Cardio- medical
Health resource Diabetes logical vascular vascular Renal Metabolic Ophthalmic Other conditions Total*
Hospital inpatient days 828 1,318 1,221 6,446 1,755 82 15 1,871 10,726 24,262
Office-based physician visits 21,931 2,386 1,405 6,635 1,905 403 4,934 1,443 23,672 64,713
Emergency visits 336 173 53 592 229 22 7 280 3,874 5,567
Hospital outpatient and 2,544 154 189 549 147 27 173 173 1,773 5,730
freestanding ambulatory
surgical center visits
*Numbers do not necessarily sum to totals because of rounding.

gency visits for general medical condi- associated with admissions where the that would have occurred in the absence
tions. primary diagnosis code is not an iden- of diabetes.
● A substantial amount of attributed tified chronic complaint associated Table 9 summarizes national expen-
health resource use is for chronic com- with diabetes. For men age 18 – 44 ditures for the cost components analyzed,
plications of diabetes. In particular, car- years, those with diabetes have admis- accounting for over one trillion dollars in
diovascular disease, neurological sion rates that are about twice that of projected expenditures for 2007. Ap-
symptoms, and renal complications are men without diabetes (Fig. 6). This ra- proximately $205 billion in expenditures
associated with high resource use at- tio declines with age, such that men age is incurred by people with diabetes, re-
tributed to diabetes. 70 and older with diabetes have only flecting $1 of every $5 health care dollars.
● Diabetes is listed as the primary diag- 5% more admissions than men in this Costs attributed to diabetes total $116 bil-
nosis code for a large portion of physi- age-group without diabetes. For the 70 lion, or 57% of total medical costs in-
cian office visits and hospital outpatient and older age-group, men with diabetes curred by people with diabetes. For the
visits attributed to diabetes. Diabetes is still have 70% more inpatient days than cost components analyzed, over $1 of ev-
listed as the primary diagnosis code for men without diabetes for admissions ery $10 health care dollars is attributed to
a small proportion of inpatient days with a primary diagnosis code listed as diabetes.
and emergency visits attributed to dia- a general medical condition. The rate National health-related expenditures
betes. Because inpatient days for gen- ratios are similar for women, with dia- are projected to exceed $2 trillion in
eral medical conditions constitutes a betes associated with an increase in ad- 2007, but only half of these expenditures
sizable (44%) portion of total inpatient missions of between 8 and 54% and are included in our analysis (37,38).
days attributed to diabetes, additional total inpatient days associated with an These cost estimates omit national expen-
attention was paid to this cost compo- increase of between 63 and 111% ditures (and any portion of such expendi-
nent. In Fig. 3 we presented findings across different age-groups. tures that might be attributable to
from an analysis of the 2005 NIS that diabetes) for administering government
shows that hospital stays for general Health care expenditures attributed health and private insurance programs,
medical conditions are longer when di- to diabetes investment in research and infrastructure,
abetes is listed as a complication (con- Health care expenditures attributed to di- over-the-counter medications, disease
trolling for other factors that affect abetes reflect the additional expenditures management and other programs tar-
length of stay). Analysis of the MCURE the nation incurs because of diabetes. geted to people with diabetes, and office
database shows that people with diabe- This equates to the total health care ex- visits to nonphysician providers other
tes have higher rates of hospital admis- penditures for people with diabetes mi- than podiatrists (e.g., dentists, optome-
sions and more inpatient days nus the projected level of expenditures trists, etc.). Expenditures for health re-

Table 7—Share of total attributed health resource use, by medical condition

Chronic complications
Neuro- Peripheral Cardio- medical
Health resource Diabetes logical vascular vascular Renal Metabolic Ophthalmic Other conditions Total*
Hospital inpatient days 3 5 5 27 7 0 0 8 44 100
Office-based physician visits 34 4 2 10 3 1 8 2 37 100
Emergency visits 6 3 1 11 4 0 0 5 70 100
Hospital outpatient and freestanding 44 3 3 10 3 0 3 3 31 100
ambulatory surgical center visits
*Numbers do not necessarily sum to totals because of rounding.


American Diabetes Association

Table 8—Proportion of total health resource use attributed to diabetes in the U.S. by medical condition

Chronic complications
Peripheral Cardio- medical
Health resource Diabetes Neurological vascular vascular Renal Metabolic Ophthalmic Other conditions Total
Hospital inpatient days 100 26 36 34 35 51 27 44 7 13
Office-based physician visits 100 26 18 11 15 4 19 13 3 7
Emergency visits 100 22 12 19 9 41 9 7 4 5
Hospital outpatient and 100 27 24 10 11 4 18 14 2 6
freestanding ambulatory
surgical center visits

sources unrelated to diabetes (e.g., care in Over half (56%) of all health care ex- and above at $3,808, $5,094, and $9,713,
residential mental retardation facilities) penditures attributed to diabetes are for respectively.
are likewise excluded from the analysis. health resources used by the population Table 12 summarizes diabetes-
Almost half of all health care expen- age 65 years and older, much of which is attributed health care expenditures for
ditures attributed to diabetes come from borne by the Medicare program (Table those cost components modeled by med-
higher rates of hospital admission and 10). The population age 45– 64 incurs ical condition, with inpatient days being
longer average lengths of stay per admis- 35% of diabetes-attributed costs, with the the largest component of attributed costs.
sion. Of the projected $430 billion in na- remaining 9% incurred by the population General medical conditions comprise
tional expenditures for hospital inpatient under age 45. The annual attributed 76% of the projected $430 billion in na-
care (including both facility and profes- health care cost per person with diabetes tional expenditures for hospital inpatient
sional services costs), approximately $97 increases with age, primarily as a result of care. With 7% of inpatient days for this
billion (or 23%) is incurred by people increased use of hospital inpatient and condition group attributed to diabetes,
who have diabetes and $58.3 billion nursing facility resources, physician office this category constitutes the single largest
(14%) is directly attributed to their diabe- visits, medications (other than insulin contributor to the attributed medical cost
tes. Retail prescriptions (excluding insu- and oral agents), and home care (Table of diabetes. The second largest category of
lin and oral agents) are another major 11). Dividing total attributed health care diabetes costs is inpatient days associated
expense category, with 17% of prescrip- expenditures by the number of people with cardiovascular disease. Cardiovascu-
tion costs incurred by people with diabe- with diabetes, we estimate the average an- lar disease, as the primary diagnosis for
tes and 8% of costs ($12.7 billion) nual excess expenditures for the popula- admissions, generates $66 billion per year
attributed to their diabetes. tion under age 45, age 45– 64, and age 65 in inpatient expenditures in the U.S. (or

(with diabetes versus without diabetes)

Rate Ratio

<18 18-34 35-44 45-54 55-59 60-64 65-69 70+
Age Group

Hospital admissions Hospital days

Figure 6—Rate ratios for inpatient care where the primary diagnosis code is a general medical condition: males. A ratio of 1.0 means that per capita
hospital admissions (or days) for people with diagnosed diabetes is the same as the per capita admissions (or days) for people without diabetes.


Economic costs of diabetes in the U.S. in 2007

Table 9—Health care expenditures in the U.S., by diabetes status and type of service, 2007 (in millions of dollars and % of U.S. total)

Population with diabetes

Total incurred by people
Attributed to diabetes with diabetes
% of % of Population
Cost component Dollars U.S. total Dollars U.S. total without diabetes U.S. total*
Institutional care
Hospital inpatient 58,344 14 96,974 23 332,902 429,875
Nursing/residential facility 7,486 10 18,525 25 56,692 75,217
Outpatient care 21,739
Physician’s office 9,897 6 8,065 14 132,984 154,723
Emergency department 3,870 5 7,879 10 73,381 81,446
Ambulance services 103 5 370 18 1,726 2,096
Hospital outpatient 2,985 4 6,770 10 60,054 66,824
Home health 5,586 14 9,391 23 31,149 40,546
Hospice 28 ⬍1 1,411 11 10,622 12,033
Podiatry 273 19 408 28 1,028 1,437
Outpatient medications and supplies
Insulin 3,733 100 3,733 100 NA 3,733
Diabetic supplies 1,783 100 1,783 100 NA 1,783
Oral agents 8,586 100 8,586 100 NA 8,586
Retail prescriptions 12,692 8 26,035 17 127,562 153,597
Other equipment and supplies 890 5 2,714 14 16,901 19,615
Total* 116,257 11 205,092 20 834,379 1,051,505
*Numbers do not necessarily sum to totals because of rounding.

15% of total U.S. medical expenditures Indirect costs attributed to diabetes per worker per year, and lower average
for inpatient care). Together, the general The national cost of lost productivity as- daily earnings.
medical conditions and cardiovascular sociated with diabetes in 2007 is esti- More substantial are the estimated
disease categories are responsible for 91% mated at $58.2 billion. This includes costs associated with presenteeism. Be-
of U.S. expenditures for inpatient care diabetes-attributed absenteeism from cause of high labor force participation
and 76% of hospital inpatient costs attrib- work, reduced productivity while at work, rates and high daily earnings, the male
uted to diabetes. reduced productivity for those not in the population age 35– 44 has the highest an-
The population with diabetes is older labor force, inability to work because of nual per capita cost of presenteeism
and sicker than the population without disability, and premature mortality. ($2,883 per person with diabetes). At the
diabetes. Table 13 shows that annual The number of workdays absent be- national level presenteeism loss is equiv-
medical expenditures for the cost compo- cause of diabetes in 2007 is estimated at alent to 120 million workdays lost, with
nents analyzed are over four times higher 15 million, at a national cost of $2.6 bil- an estimated national cost of $20 billion
($11,744 vs. $2,935). A more important lion (Table 14). These statistics are calcu- in 2007.
comparison is the health care expendi- lated by combining age-sex–specific Reduced productivity for those not in
tures for people with diabetes compared estimates of the number of people with the labor force is calculated by extrapolat-
to what expenditures for this same popu- diabetes, labor force participation rates, ing the age-sex work absenteeism rates to
lation (i.e., age-sex adjusted) would be in annual workdays lost from absenteeism the population not in the labor force and
the absence of diabetes. People with dia- per worker, and average daily earnings. valuing each lost productivity day using
betes have health care expenditures that The population with the highest per cap- 75% of daily earnings of their peers in the
are 2.3 times higher ($11,744 vs. $5,095) ita productivity loss from absenteeism is labor force. The estimated productivity
than expenditures for this same popula- males age 45–54. With 1.7 million people loss is equivalent to six million days, with
tion would be in the absence of diabetes, with diabetes, a labor force participation a national cost of $800 million.
suggesting that diabetes is responsible for rate of 84%, average days absent of 2.5 At any given time during 2007, ⬃16
$6,649 in excess expenditures per year days per worker per year, and average million people were unemployed and re-
per person with diabetes. daily earnings of $240, the absenteeism ceiving SSI payments due to disability.
The 2.3 multiple is similar to the 2.4 cost per person with diabetes per year is Over one million of these people had di-
multiple in the 2003 study. The ratio of $493 (Table 15). Per capita cost is signif- abetes, with over 445,000 cases of unem-
costs with to without diabetes, adjusting icantly lower for the youngest and oldest ployment attributed to diabetes, equating
for demographic mix, ranges from a low to 107 million lost workdays at a national
age-groups—populations with lower
of 1.4 for ambulance services to a high of cost of $7.9 billion.
rates of employment, lower days absent
3.0 for podiatry services. An estimated 284,000 deaths in 2007


American Diabetes Association

Table 10—Health care expenditures attributed to diabetes in the U.S., by age group and type this study. The 2003 study used the
of service, 2007 (in millions of dollars) overall comorbidity etiological fraction
as a proxy for the mortality etiological
Age (years) fraction. For this study we use the emer-
gency room etiological fraction as a
Cost component ⬍45 45–64 ⱖ65 Total* proxy for the mortality etiological frac-
Institutional care tion. For cardiovascular disease we cal-
Hospital inpatient 4,551 18,447 35,346 58,344 culated the etiological fraction using
Nursing/residential facility 166 1,527 5,793 7,486 mortality risk data from the literature
Outpatient care (32).
Physician’s office 1,113 3,982 4,802 9,897
Emergency department 1,047 1,384 1,438 3,870
Ambulance services 4 79 21 103 CONCLUSIONS — Diabetes cost the
Hospital outpatient 645 1,312 1,028 2,985 nation an estimated $174 billion in 2007.
Home health 0 1,823 3,763 5,586 This burden is spread across society, al-
Hospice 1 3 24 28 though the burden falls disproportion-
Podiatry 14 58 202 273 ately on the people with diabetes and
Outpatient medications and supplies their families. On the surface it appears
Insulin 788 1,564 1,381 3,733 that the burden falls primarily on insurers
Diabetic supplies 217 859 707 1,783 who pay a significant portion of medical
Oral agents 967 4,163 3,456 8,586 costs, employers who experience produc-
Retail prescriptions 1,099 5,161 6,432 12,692 tivity loss, and the people with diabetes
Other equipment and supplies 77 448 365 890 and their families who incur higher out-
Total* 10,689 40,810 64,758 116,257 of-pocket medical costs and reduced
*Numbers do not necessarily sum to totals because of rounding. earnings potential or employment oppor-
tunities. Ultimately, though, the burden is
passed along to all of society in the form of
are attributed to diabetes (Table 16). This loss for people who die prematurely in higher insurance premiums and taxes, re-
includes 77,000 deaths where diabetes is 2007 counted in 2007. duced earnings, and reduced standard of
listed as the primary cause of death, Estimates of diabetes deaths attrib- living. Furthermore, with just under 1 in
123,000 deaths where cardiovascular dis- uted to renal disease and cerebrovascu- 10 people having diabetes, this disease di-
ease is listed as the primary cause of death lar disease are substantially higher than rectly or indirectly touches almost every-
(with 16.5% of national deaths from estimates in the 2003 study, reflecting one in society.
cardiovascular disease attributed to the larger etiological fraction used in Because the risk of developing
diabetes), 59,000 deaths where cerebro-
vascular disease is the listed primary
cause of death, and 25,000 deaths where Table 11—Annual per capita health care expenditures attributed to diabetes by age group
renal disease is listed as the primary cause 2007
of deaths. Taking into account the age and
sex distribution of these deaths, we esti- Age (years)
mate that the value of lost productivity
Cost component ⬍45 45–64 ⱖ65 Total*
from premature mortality is $26.9 billion.
Using net present value of future produc- Institutional care
tivity as a proxy for productivity loss is a Hospital inpatient 1,621 2,303 5,302 3,337
departure from the method used to calcu- Nursing/residential facility 59 191 869 428
late the value of productivity loss associ- Outpatient care
ated with the other components of Physician’s office 397 497 720 566
indirect costs. Absenteeism, presentee- Emergency department 373 173 216 221
ism, and unemployment from disability Ambulance services 1 10 3 6
are all calculated to have occurred during Hospital outpatient 230 164 154 171
2007. The productivity loss in 2007 asso- Home health 0 228 564 319
ciated with premature mortality should Hospice 0 0 4 2
be calculated based on the expected pro- Podiatry 5 7 30 16
ductivity of people who would have been Outpatient medications and supplies
alive in 2007 but who died prematurely in Insulin 281 195 207 214
earlier years (2006, 2005, etc.) because of Diabetic supplies 77 107 106 102
diabetes. However, such an approach Oral agents 344 520 518 491
would be difficult to implement. The ap- Retail prescriptions 392 644 965 726
proach used does provide practical con- Other equipment and supplies 27 56 55 51
sistency and there is no “double counting” Total* 3,808 5,094 9,713 6,649
with the discounted future productivity *Numbers do not necessarily sum to totals because of rounding.


Economic costs of diabetes in the U.S. in 2007

Table 12—Health care expenditures attributed to diabetes in the U.S., by medical condition for select settings, 2007 (in millions of dollars)

Chronic complications
Neuro- Peripheral Cardio- medical
Setting Diabetes logical vascular vascular Renal Metabolic Ophthalmic Other conditions Total*
Hospital inpatient 1,535 3,115 2,719 20,790 3,285 176 36 3,215 23,473 58,344
Physician’s office 2,899 382 279 1,004 323 48 899 233 3,830 9,897
Emergency department 234 138 43 403 132 13 11 178 2,717 3,870
Hospital outpatient 842 75 135 317 87 8 130 70 1,321 2,985
*Numbers do not necessarily sum to totals because of rounding.

chronic diseases increases with age, the stantial burden on the individual and on and administration costs (e.g., to admin-
aging population is expected to drive a the health care system (41). ister the Medicare and Medicaid pro-
substantial increase in the number of grams, to process insurance claims).
people with diabetes even if other risk Administration costs for government
factors remain unchanged. The U.S. The cost estimates presented might health programs and private insurers are
be conservative for several reasons:
Census Bureau projects a rise in the per- ● Omitted from this analysis due to data
⬃$150 billion per year. Expenditures for
cent of the population age ⱖ65 years, investment in medical research and
limitations is whether diabetes is associ-
from 12.4% in 2003 to 19.7% by 2030 ated with increased use of over-the- health infrastructure total over $130 bil-
(39). A recent study estimates that prev- counter medications and higher rates of lion per year. If a portion of these costs
alence of diagnosed diabetes will more visits to optometrist and dentist offices. were attributed to diabetes, the national
than double between 2005 and 2050, Diabetes increases the risk of periodontal cost of diabetes would be billions of dol-
from 5.6 to 12% (40). A recent nation- disease, so one would expect dental costs lars higher than our estimate suggests.
ally representative, longitudinal study to be higher for people with diabetes. ● The cost estimates presented are based
found that elderly people newly diag- Also omitted from the cost estimates are on the assumption that people with un-
nosed as having diabetes experience expenditures for prevention programs diagnosed diabetes incur no excess
much higher rates of complications in targeted to people with diabetes (e.g., dis- medical and indirect costs attributed to
the years after diagnosis than do their ease management programs), research diabetes. Additional research is needed
peers without diabetes, implying a sub- activities (e.g., to develop new drugs), to quantify whether people with undi-

Table 13—Annual per capita health care expenditures by diabetes status, 2007

Unadjusted Age-sex adjusted

Ratio with Ratio with
With Without to without Without to without Attributed
Cost component diabetes diabetes diabetes diabetes diabetes to diabetes
Institutional care
Hospital inpatient 5,546 1,171 4.7 2,209 2.5 3,337
Nursing/residential facility 1,059 199 5.3 631 1.7 428
Outpatient care
Physician’s office 1,243 468 2.7 677 1.8 566
Emergency department 461 258 1.8 240 1.9 221
Ambulance services 21 6 3.5 15 1.4 6
Hospital outpatient and freestanding ambulatory surgical center 387 211 1.8 216 1.8 171
Home care 551 110 5.0 232 2.4 319
Hospice care NA NA NA NA NA 2
Podiatry 23 4 6.5 8 3.0 16
Outpatient medications and supplies
Insulin 214 NA NA NA NA 214
Diabetic supplies 102 NA NA NA NA 102
Oral agents 491 NA NA NA NA 491
Retail prescriptions 1,489 449 3.3 763 2.0 726
Other equipment and supplies 155 59 2.6 104 1.5 51
Total* $11,744 $2,935 4.0 $5,095 2.3 $6,649
*Numbers do not precisely sum to totals because of rounding.


American Diabetes Association

Table 14—Indirect costs attributed to diabetes, 2007

Total cost
attributable to Proportion
diabetes of indirect costs
Cost component Productivity loss ($ billions) (%)
Work days absent 15 million days 2.6 4
Reduced performance at work 120 million days 20.0 34
Reduced productivity days for those not in labor force 6 million days 0.8 1
Permanent disability 445,000 people, 107 million days 7.9 14
Mortality 284,000 deaths 26.9 46
Total 58.2 100

agnosed diabetes incur disease-related elderly parent is not included in the One challenge for this study was to con-
costs in excess of their peers without cost estimates. The value of informal trol for correlation between diabetes and
diabetes. That people with diabetes al- care giving is excluded from our cost use of health resources for reasons not di-
ready have complications at the time of estimate. rectly attributed to diabetes. Health behav-
diagnosis suggests that diabetes-related ● Our estimate of lost productivity attrib- ior that affects both the presence of diabetes
costs are present among the undiag- uted to chronic disability from diabe- and the presence of other comorbidities,
nosed. The population with undiag- tes is likely conservative due to three unless controlled for, could result in an
nosed diabetes might represent a factors: One, using SSI payments to overestimate of the link between diabetes
population less inclined to seek medi- identify cases of disability likely under- and use of health resources. Controlling for
cal attention, contributing to a delay in estimates disability cases because the age and sex helps to control for this corre-
diagnosis. If people with undiagnosed criteria for SSI eligibility includes re- lation. In addition, for the top-10 cost driv-
diabetes have disease-related medical quirements for documentation of dis- ers we conducted additional analysis with
costs that are 10, 20, or 30% as high as ability from a health professional and the MEPS, and based on the results, we re-
people with diagnosed diabetes, then apply income limits. Two, our esti- duced the etiological fractions for hyperten-
the national cost of diabetes could be mates omit the value of productivity sion, renal complications, and general
$7, $13 or $20 billion higher, respec- loss that results in reduced earnings po- medical conditions. This potential limita-
tively, than our estimates suggest. tential but does not prevent working. tion also applies to the estimates of indirect
● The lost productivity estimates are for Three, we do not include productivity costs attributed to diabetes.
those individuals with diagnosed dia- loss associated with early retirement, Other study limitations discussed
betes and exclude lost productivity and a longitudinal study using the previously include small sample size for
associated with diabetes of family Health and Retirement Survey found some data sources used, the use of a data
members. For example, the produc- that people with diabetes tend to retire source (MCURE) that overrepresents the
tivity loss associated with adults who ⬃1.2 years earlier than their peers commercially insured population, and
take time off from work to care for an without diabetes (27). the need to use different approaches to

Table 15—Annual productivity loss per person with diabetes by cause, 2007

Reduced productivity
for those not Unemployment Premature
Sex Age Absenteeism Presenteeism in labor force from disability mortality
Male ⬍18 — — — — 4,306
18–34 22 1,458 17 936 3,366
35–44 99 2,883 10 728 4,476
45–54 493 2,688 72 652 4,468
55–59 360 2,196 109 315 3,081
60–64 181 1,517 131 806 1,822
65–69 86 721 — 274 569
70⫹ 45 378 — 146 284
Female ⬍18 — — — — 2,070
18–34 50 844 21 743 1,095
35–44 47 1,378 14 1,216 1,456
45–54 240 1,310 70 449 1,388
55–59 179 1,093 90 491 994
60–64 72 604 80 492 642
65–69 27 228 — 448 140
70⫹ 17 140 — 74 116


Economic costs of diabetes in the U.S. in 2007

Table 16—Mortality costs attributed to diabetes, 2007

Deaths attributed to diabetes

Value of lost
Total U.S. Deaths % of total productivity
Primary cause of death deaths (thousands) (thousands) U.S. deaths (millions of dollars)
Diabetes 77 77 100.0 9,520
Renal disease 43 25 57.4 2,116
Cerebrovascular disease 155 59 37.6 3,849
Cardiovascular disease 739 123 16.5 11,417
Grand total NA* 284 NA* 26,902
*Grand total comprises mortality for reasons other than diabetes, renal disease, cerebrovascular disease, and cardiovascular disease.

model different cost components because ties, and improved care for people with di- cardiovascular complications and events
of data limitations. One sensitivity analy- abetes to reduce the need for costly in diabetes mellitus. Drugs 67:997–1026,
sis conducted was to compare relative rate complications. Improved understanding of 2007
ratios for the population age 65– 69 and the economic cost of diabetes and the major 10. Fox CS, Larson MG, Leip EP, Meigs JB,
age 70 and older calculated from two determinants of costs helps to inform and Wilson PW, Levy D: Glycemic status and
development of kidney disease: the
sources—MCURE and the Medicare 5% motivate decisions that can reduce the na- Framingham Heart Study. Diabetes Care
Sample. The relative rate ratios for each tional burden of this disease. 28:2436 –2440, 2005
health care delivery setting (hospital inpa- 11. Geerlings SE, Meiland R, Hoepelman IM:
tient, emergency, and ambulatory) and [Urinary tract infections in women with
comorbidity group were similar for the References diabetes mellitus]. Ned Tijdschr Geneeskd
two data sources, and if we had used the 1. Economic consequences of diabetes mel-
145:1832–1836, 2001
litus in the U.S. in 1997. American Diabe-
rate ratios calculated from the Medicare tes Association. Diabetes Care 21:296 –
12. Gordois A, Scuffham P, Shearer A,
5% Sample, our estimate of total national Oglesby A, Tobian JA: The health care
309, 1998
medical costs attributed to diabetes costs of diabetic peripheral neuropathy in
2. Ettaro L, Songer TJ, Zhang P, Engelgau
would be ⬃1% higher. MM: Cost-of-illness studies in diabetes
the US. Diabetes Care 26:1790 –1795,
The medical cost analysis is largely 2003
mellitus. Pharmacoeconomics 22:149 –
13. Janghorbani M, Hu FB, Willett WC, Li TY,
based on claims data, and claims data tend 164, 2004
3. Hogan P, Dall T, Nikolov P: Economic Manson JE, Logroscino G, Rexrode KM:
to be less accurate than clinical reports in Prospective study of type 1 and type 2
identifying patients with specific condi- costs of diabetes in the U.S. in 2002. Dia-
betes Care 26:917–932, 2003 diabetes and risk of stroke subtypes: the
tions. The direction of bias for the cost of Nurses’ Health Study. Diabetes Care 30:
diabetes is unknown, as use of claims data 4. Ray N, Willis S, Thamer M: Direct and In-
direct Costs of Diabetes in the United States 1730 –1735, 2007
might overdiagnose for some conditions in 1992. Alexandria, VA, American Diabe- 14. Trautner C, Icks A, Haastert B, Plum F,
and underdiagnose for other conditions. tes Association, 1993 Berger M: Incidence of blindness in rela-
All models are simplifications of reality 5. Centers for Disease Control and Preven- tion to diabetes: a population-based
and their conclusions are affected by both tion: Diabetes: disabling disease to double study. Diabetes Care 20:1147–1153,
structural boundaries and the uncertainties by 2050 [article online], 2007. Available 1997
of the data with which they are calibrated. from 15. Brown DA, Unrein C: A review of cardio-
Despite limitations, the estimates presented lications/aag/ddt.htm. Accessed 23 Octo- vascular comorbidities of diabetes. Am J
ber 2007 Managed Care 13:3–10, 2007
here show a consistent picture that diabetes
6. National Health Interview Survey (NHIS). 16. Benichou J: A review of adjusted estima-
places an enormous burden on society—in tors of attributable risk. Stat Methods Med
both economic terms and reduced quality 2006. Atlanta, GA, National Center for
Health Statistics, Centers for Disease Con- Res 10:195–216, 2001
of life. Although this study improves on the trol and Prevention, 2006 17. Begley CE, Famulari M, Annegers JF, Lair-
methods and uses new data sources com- 7. Cowie CC, Rust KF, Byrd-Holt DD, Eber- son DR, Reynolds TF, Coan S: The cost of
pared to the 2003 study, the overall cost of hardt MS, Flegal KM, Engelgau MM, Say- epilepsy in the United States: an estimate
diabetes estimates are consistent with ear- dah SH, Williams DE, Geiss LS, Gregg from population-based clinical and sur-
lier estimates after adjusting for increasing EW: Prevalence of diabetes and impaired vey data. Epilepsia 41:342–351, 2000
prevalence of diabetes and price increases. fasting glucose in adults in the U.S. pop- 18. Honeycutt A, Dunlap L, Chen H, Homsi
Some of the estimates by cost component, ulation: National Health And Nutrition G, Grosse S, Schendel D: Economic costs
medical condition, and age differ from the Examination Survey 1999 –2002. Diabe- associated with mental retardation, cere-
2003 study. tes Care 29:1263–1268, 2006 bral palsy, hearing loss, and vision im-
8. Okura Y, Urban LH, Mahoney DW, Ja- pairment? United States, 2003. MMWR
The present findings demonstrate that
cobsen SJ, Rodeheffer RJ: Agreement be- Morb Mort Weekly Rep 53:57–59, 2004
the burden of diabetes and its complications tween self-report questionnaires and 19. Rice DP, Miller LS: Health economics and
on the individual and on the health care medical record data was substantial for cost implications of anxiety and other
system is significant. Much of this cost is diabetes, hypertension, myocardial in- mental disorders in the United States. Br J
preventable through improved diet and ex- farction and stroke but not for heart fail- Psychiatry Suppl 34:4 –9, 1998
ercise, prevention initiatives to reduce the ure. J Clin Epidemiol 57:1096 –1103, 2004 20. Rice DP: Economic costs of substance
prevalence of diabetes and its comorbidi- 9. Coccheri S: Approaches to prevention of abuse, 1995. Proc Assoc Am Physicians


American Diabetes Association

Table A1—Summary of data sources

Data source and description Used to estimate/analysis Strengths and limitations

National Health Interview Survey (NHIS): Diabetes diagnosed prevalence by age, sex, ⫹Large sample size
combined 2004–2006 surveys to and race/ethnicity ⫹Contains employment-related information
increase sample size. Prevalence of insulin and oral agents use ⫺Diabetes status self-reported of whether ever
Impact of diabetes on employment/hours been told by physician you have diabetes
worked ⫺Excludes institutionalized population where
Activity limitation and restriction diabetes is overrepresented

National Health and Nutrition Verify diagnosed prevalence ⫹Contain both self-reported and lab
Examination Survey (NHANES): Estimate undiagnosed prevalence test–identified diabetic persons
combined the three biannual surveys ⫺Excludes institutionalized population where
(1999–2000, 2001–2002, and 2003– diabetes is overrepresented
2004) to increase sample size.

Medical Expenditure Panel Survey Average cost per physician office, ⫹Rich source of health resource use and cost
(MEPS): combined 2003–2005 surveys outpatient, and emergency visit and information
to increase sample size. outpatient prescription ⫺Relatively small sample size per year
Average annual expenditures for podiatry, ⫺Contains only three-digit diagnosis codes;
home health, insulin, oral agents, many chronic conditions of diabetes require
diabetes-related supplies use and cost, four-digit codes to identify
other medical equipment and supplies ⫺Excludes institutionalized population where
diabetes is overrepresented

National Ambulatory Medical Care National number of physician office visits ⫹Larger sample size than MEPS
Survey (NAMCS): combined 2003– by medical condition(using primary ⫹Contains five-digit diagnosis codes to identify
2005 surveys to increase sample size. diagnosis code) chronic conditions of diabetes
Average number of scripts written per visit ⫺Visits are the unit of observations, with
incomplete information on patients (including
whether they have diabetes)

National Hospital Ambulatory Medical National number of hospital outpatient ⫹Same as for NAMCS
Care Survey (NHAMCS): combined and emergency visits by medical ⫺Same as for NAMCS
2003–2005 surveys to increase sample condition (using primary diagnosis
size. code)
Average number of scripts written per visit

Nationwide Inpatient Sample (NIS): used National number of hospital inpatient days ⫹Same as for NAMCS
2004–2005 surveys. for diabetes and comorbidities of ⫺Same as for NAMCS
diabetes (using primary diagnosis)
Cost per inpatient day calculated using
hospital-specific cost-to-charge ratios

National Home and Hospice Care Survey Hospice care use (also validate home ⫹Same as for NAMCS
(NHHCS): used 2000 data. health) ⫺Same as for NAMCS

National Nursing Home Survey (NNHS): Nursing facility use ⫹Same as for NAMCS
used 2004 data. ⫺Same as for NAMCS

Ingenix MCURE database Calculate age-sex specific relative rate ⫹Large sample size
ratios for each medical condition for ⫺All medical records can be linked for the year
hospital inpatient days, emergency to identify people with diabetes based on
visits, and ambulatory visits (physician whether they have any diabetes diagnosis code
office and hospital outpatient during the year
combined) ⫺Lacks detailed data on health behavior found in


Economic costs of diabetes in the U.S. in 2007

Table A2 —Chronic complications of diabetes 111:119 –125, 1999

21. Burton WNM, Conti DJP, Chen CYP,
Schultz ABM, Edington DWP: The role of
Chronic complications of diabetes ICD-9 codes health risk factors and disease on worker
Neurological symptoms productivity. J Occup Environ Med 41:
Myasthetic syndromes in diseases classified 358.1 863– 877, 1999
22. Goetzel RZP, Hawkins KP, Ozminkowski
elsewhere (amyotrophy)
RJP, Wang SP: The health and productiv-
Other specified idiopathic peripheral neuropathy 356.8 ity cost burden of the “Top 10” physical
Mononeuritis of upper and lower limbs 354, 355 and mental health conditions affecting six
Arthropathy associated w/neurological disorders 713.5 large U.S. employers in 1999. J Occup En-
(Charcot’s arthropathy) viron Med 45:5–14, 2003
Peripheral autonomic neuropathy 337.1 23. Goetzel RZP, Long SRM, Ozminkowski
Polyneuropathy in diabetes 357.2 RJP, Hawkins KP, Wang SP, Lynch WP:
Neuralgia, neuritis, and radiculitis, unspecified 729.2 Health, Absence, disability, and presen-
Diabetes with neurological complications 250.6 teeism cost estimates of certain physical
Occlusion of cerebral arteries 434 and mental health conditions affecting
u.s. employers. J Occup Environ Med 46:
Hemorrhagic stroke 430–432
398 – 412, 2004
Late effects of cerebrovascular disease 438 24. Boles MP, Pelletier BM, Lynch WP: The
Occlusion of stenosis of pre-cerebral arteries 433 Relationship between health risks and
Other and ill-defined cerebrovascular disease 437 work productivity. J Occup Environ Med
Acute, but ill-defined, cerebrovascular disease 436 46:737–745, 2004
TIAs 435 25. Goetzel RZP, Ozminkowski RJP, Baase
Peripheral vascular disease CMM, Billotti GMM: Estimating the re-
Atherosclerosis 440 turn-on-investment from changes in em-
Embolism and thrombosis, structure of artery 444, 447.1 ployee health risks on the Dow Chemical
Other peripheral vascular disease 443 Company’s health care costs. J Occup En-
viron Med 47:759 –768, 2005
Other disorders of circulatory system 459
26. Stewart WFP, Ricci JAS, Chee ES, Hirsch
Phlebitis and thrombophlebitis, portal vein 451,452
AGM, Brandenburg NAP: Lost productive
throbosis and thrombolism and venous time and costs due to diabetes and dia-
thrombolism betic neuropathic pain in the US work-
Other venous embolism and thrombolism 453 force. J Occup Environ Med 49:672– 679,
Varicose veins of lower extremities 454 2007
Gangrene and amputations 785.4, 885–887, 895–897 27. Vijan S, Hayward RA, Langa KM: The im-
Chronic ulcer of skin 707 pact of diabetes on workforce participa-
Cardiovascular disease tion: results from a national household
Aortic and other aneurysms 441, 442 sample. Health Serv Res 39:1653–1669,
Hypotension 458 2004
28. Ng YC, Jacobs P, Johnson JA: Productivity
Angina 413
losses associated with diabetes in the US.
Conduction disorders and cardiac dysrhythmias 426–427 Diabetes Care 24:257–261, 2001
ASCVD 429.2 29. Tunceli KP, Bradley CJP, Nerenz DP, Wil-
Cardiomegaly 429.3 liams LKM, Pladevall MM, Lafata JEP: The
Cardiomyopathy 425 impact of diabetes on employment and
Other acute and subacute forms of ischemic heart 411 work productivity. Diabetes Care 28:
disease 2662–2667, 2005
Heart failure 428 30. Von KM, Katon W, Lin EH, Simon G,
Diabetes w/peripheral circulatory disorders 250.7 Ciechanowski P, Ludman E, Oliver M,
Myocardial degeneration 429.1 Rutter C, Young B: Work disability among
individuals with diabetes. Diabetes Care
Myocardial infarction 410, 412
28:1326 –1332, 2005
Other chronic ischemic heart disease 414 31. Centers for Disease Control and Preven-
Hypertension 401–405 tion: National diabetes fact sheet: General
Renal Complications information and national estimates on di-
Infections of kidney 590 abetes in the United States [article online],
Other disorders of bladder 596 2005. Available at
Cystitis 595 diabetes/pubs/factsheet.htm. Accessed 23
Renal sclerosis, unspecified 587 October 2007
Glomerulonephritis, nephrotic syndrome, 580–583 32. Kanaya AM, Grady D, Barrett-Connor E:
nephritis, and nephropathy Explaining the sex difference in coronary
heart disease mortality among patients
Proteinuria 791.0
with type 2 diabetes mellitus: a meta-anal-
Renal failure and its sequelae 584, 586, 588 ysis. Arch Intern Med 162:1737–1745,
Other disorders of kidney and ureter 593 2002
Urinary tract infection 599.0 33. United States Life Tables, 2003. National
Continued on following page Vital Statistics Reports, vol. 54 no. 14.


American Diabetes Association

Table A2 —Continued Hyattsville, Maryland, National Center

for Health Statistics, 2007 [article on-
line]. Available from http://www.cdc.
Chronic complications of diabetes ICD-9 codes
Diabetes and renal complications 250.4 pdf. Accessed 4 November 2007
Chronic renal failure (ESRD) 585 34. Max W, Rice DP, Sung HY, Michel M: Val-
Endocrine/metabolic complications uing Human Life: Estimating the Present
Dwarfism-obesity syndrome 259.4 Value of Lifetime Earnings, 2000. San Fran-
Glycogenosis and galactosemia 271.0, 271.1 cisco, CA, Institute for Health & Aging,
Disorders of iron metabolism 275.0 2002
Hypercholesterolemia 272.0 35. Morrow RH, Bryant JH: Health policy ap-
proaches to measuring and valuing hu-
Hyperchylomicronemia 272.3
man life: conceptual and ethical issues.
Hyperkalemia 276.7
Am J Public Health 85:1356 –1360, 1995
Hypertriglyceridemia 272.1
36. West RR, McNabb R, Thompson AG,
Macroglobulinemia 273.3 Sheldon TA, Grimley EJ: Estimating im-
Lancereaux’s disease 261 plied rates of discount in healthcare deci-
Lipidoses 272.7 sion-making 3. Health Technol Assess 7:1–
Other specified endocrine disorders 259.8 60, 2003
Other and unspecified hyperlipidemia 272.4 37. Poisal JA, Truffer C, Smith S, Sisko A,
Mixed hyperlipidemia 272.2 Cowan C, Keehan S, Dickensheets B:
Renal glycosuria 271.4 Health spending projections through
Ophthalmic complications 2016: modest changes obscure part D’s
Other retinal disorders 362 impact. Health Affairs 26:242–253, 2007
Vascular disorders of the iris and ciliary body 364.0, 364.4 38. Smith C, Cowan C, Heffler S, Catlin A: Na-
Disorders of the optic nerve and visual pathways 377 tional health spending in 2004: recent slow-
Diabetes with ophthalmic complications 250.5 down led by prescription drug spending.
Cataract 366 Health Affairs 25:186 –196, 2006
Glaucoma 365 39. U.S. Interim Projections by Age, Sex, Race,
Visual disturbance, low vision, blindness 368–369 and Hispanic origin. U.S. Census Bureau,
Other complications Population Division, Population Projec-
Bacteremia, bacterial infection, Coxsackie virus 079.2, 790.7 tions Branch [Article online]. Available
Candidiasis of skin and nails 112.3 from
nvsr54/nvsr54_14.pdf. Accessed 15 Sep-
Chronic osteomyelitis of the foot 730.17
tember 2007
Other and unspecified noninfectious 558.9
40. Narayan KMV, Boyle JP, Geiss LS, Saad-
gastroenteritis and colitis
dine JB, Thompson TJ: Impact of recent
Impotence of organic origin 607.84 increase in incidence on future diabetes
Infective otitis externa 380.1 burden. Diabetes Care 29:2114 –2116,
Degenerative skin disorders 709.3 2006
Candidiasis of vulva and vagina 112.1 41. Bethel MA, Sloan FA, Belsky D, Feinglos
Cellulitis 681, 682 MN: Longitudinal incidence and preva-
Diabetes with other specified manifestations 250.8 lence of adverse outcomes of diabetes
Diabetes with unspecified complication 250.9 mellitus in elderly patients. Arch Intern
Other bone involvement in disease classified 731.8 Med 167:921–927, 2007