PREVENTING CHRONIC DISEASE

PUBLIC HEALTH RESEARCH, PRACTICE, AND POLICY

Volume 22, E50 AUGUST 2025

ORIGINAL RESEARCH

Medical Expenditure Differences Between Income

Levels Among US Adults With Diabetes
Yu Wang, PhD1; Hui Shao, PhD2; Elizabeth Bigman, PhD1; Christopher Holliday, PhD1; Ping Zhang, PhD1

Accessible Version: www.cdc.gov/pcd/issues/2025/25_0153.htm egorized by service type (inpatient, outpatient, prescription, home
health care services, emergency department, or other) and com-
Suggested citation for this article: Wang Y, Shao H, Bigman E,
pared across income groups based on the federal poverty level
Holliday C, Zhang P. Medical Expenditure Differences Between
(FPL): poor (<125% FPL), low (125% to <200% FPL), middle
Income Levels Among US Adults With Diabetes. Prev Chronic
(200% to <400% FPL), and high (≥400% FPL). One-way analysis
Dis 2025;22:250153. DOI: https://doi.org/10.5888/pcd22.250153.
of variance was used to test group differences, and a regression-
based decomposition identified factors driving expenditure dispar-
PEER REVIEWED ities. All expenditures were adjusted to 2021 US dollars.

Summary
Results
Mean total medical expenditures were significantly higher for the
What is already known on this topic?
poor-income group compared with the low-income, middle-
People from low-income families are disproportionately affected by dia-
betes and have higher medical expenditures due to greater health care
income, and high-income groups, though no significant differ-
needs. They also face significant barriers to accessing quality care. ences were observed among the latter 3 groups. Prescription drugs
What is added by this report? and home health care services in the poor-income group accoun-
We examined the relationship between income and medical spending. ted for most of this difference. Key factors associated with the
Adults from households with incomes below 125% of the federal higher expenditures in this group included elevated disability
poverty level had significantly higher medical expenditures than those rates, poorer physical health status, and dual Medicaid–Medicare
from higher income households. This difference was largely driven by
increased spending on prescriptions and home health care services. coverage.
High rates of disability and poor physical health among the low-income
group contributed to these elevated costs. Conclusion
What are the implications of our findings? Adults with diabetes from the poorest households incurred the
Efforts to improve the health of adults with diabetes from low-income highest medical expenditures, largely driven by poor physical
households may help lower overall health care expenditures.
health and higher rates of disability. Reducing disability and im-
proving health outcomes for this group may help lower their med-
ical expenses.
Abstract
Introduction
Introduction
Significant differences exist in the risk of diabetes and diabetes- Significant differences exist in the risk of diabetes and diabetes-
related complications by income level in the United States. We as- related complications by income level in the US. People with
sessed 1) to what extent medical expenditures in total and by lower incomes are disproportionately affected by diabetes (1);
health service type differ by income levels, and 2) how demo- adults with a family income below the federal poverty level (FPL)
graphic and socioeconomic factors and health status are associ- have the highest prevalence of diabetes. In 2019 through 2021, for
ated with these differences. example, 13.1% of people having a family income below the FPL
had diabetes compared with 5.1% of people with a family income
Methods of 500% or greater of the FPL (2). Low-income people with dia-
Data from the 2017 through 2021 Medical Expenditure Panel Sur- betes also face more challenges with diabetes management, with a
vey were analyzed to estimate annual per-person medical ex- higher rate of uncontrolled hemoglobin A1c, high blood pressure,
penditures for adults with diabetes. These expenditures were cat- and high lipid levels (3–5), as well as higher rates of diabetes-

The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health
and Human Services, the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions.

www.cdc.gov/pcd/issues/2025/25_0153.htm • Centers for Disease Control and Prevention 1

This publication is in the public domain and is therefore without copyright. All text from this work may be reprinted freely. Use of these materials should be properly cited.
PREVENTING CHRONIC DISEASE VOLUME 22, E50
PUBLIC HEALTH RESEARCH, PRACTICE, AND POLICY AUGUST 2025

related complications and premature death (6,7) and barriers to was calculated by dividing the total family income with the offi-
quality care (8,9). cial FPL for that family size; this calculation is often used to de-
cide whether the income level of the person or family qualifies for
For people with diabetes, the relationship between income and certain federal benefits and programs (15). Four household in-
medical spending is complex. Low-income people with diabetes come levels were defined: poor income (<125% FPL), low in-
may have poorer health status (10), requiring more health services come (125% to <200% FPL), middle income (200% to <400%
leading to higher expenditures. Yet their barriers to accessing FPL), and high income (≥400% FPL). We used 125% of the FPL
quality care or new treatments could result in underuse of services as the threshold for poor income because it is an income that
and lower expenditures (8). Benefit coverage of insurance and so- would qualify a person or a family for many federal support pro-
cial programs for poor people could also affect their spending. In grams, including Medicaid in most states, Supplemental Nutrition
some countries, people with high incomes have high medical ex- Assistance Program, and legal aid. Low income was defined as
penditures (11), while in other countries people with the lowest in- 125% to less than 200% of the FPL to capture households that are
come have the highest medical expenditures (12,13). The pattern, still economically vulnerable and qualify for some federal support
magnitude, and factors associated with income-related differences programs, including childcare subsidies and the Children’s Health
in medical expenditures in people with diabetes have not been as- Insurance Program. At 400% of the FPL, income is where people
sessed in the US. would not receive federal support on health insurance premiums
In this study, we examined differences in medical expenditures by under the Affordable Care Act and, thus, is defined as the starting
income level among US adults with diabetes. We hypothesized point of high income.
that US adults with diabetes from low-income households may in- Variables
cur higher annual per-person medical expenditures than those in
The primary outcome was total annual medical expenditures per
higher-income households, potentially due to poorer physical
person, which is the sum of the direct payments for care provided
health status, but that the composition of their spending will differ.
during the year, including out-of-pocket payments and payments
Specifically, we assessed 1) to what extent medical expenditures
by private insurance, Medicaid, Medicare, or other sources. We
in total and by health service type differ by income levels and 2)
also examined annual per-person expenditure components by type
how demographic and socioeconomic factors and health status are
of health service: inpatient, outpatient, prescription, home health
associated with these differences.
care services, emergency department, or others (eg, glasses or con-
tact lenses, ambulance, disposable supplies, long-term use of
Methods equipment). Differences were defined as the difference in average
Data source and study population annual per-person expenditures between each pair of income cat-
egories by subtracting the average expenditure for adults from the
We used data from the 2017 through 2021 Medical Expenditure higher income group from the average expenditure for adults from
Panel Survey (MEPS), Household Component. MEPS is a nation- the lower income group. All expenditures were adjusted to 2021
ally representative household survey of the US civilian noninstitu- US dollars by using the Personal Consumption Expenditures in-
tionalized population that contains information about health condi- dex (16).
tions, health care service usage, and expenditures (14). The study
population included adults aged 18 years or older with self- We used a regression-based decomposition method to examine the
reported diabetes. We identified people with diabetes by the ques- association of differences in each pair of income categories where
tion, “Have you ever been told by a doctor or other health profes- their difference in medical expenditures was significant with
sional that you had diabetes?” Adults who answered yes to this demographic and socioeconomic factors and health status (17).
question were included in the analysis. We pooled 5 years of data The factors included in the regression models were age group
to achieve a sufficient sample size for the study. (18–44 y, 45–54 y, 55–64 y, ≥65 y), sex (female, male), race and
ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, oth-
Measures er [including American Indian or Alaska Native, Asian, Native
Hawaiian or Pacific Islander, or multiple races]), insurance (Medi-
Income groups caid–Medicare dual coverage, Medicare only or with private,
We defined 4 income categories based on FPL. FPL is a measure- Medicaid only, private insurance only, uninsured, other), educa-
ment that describes the minimum income a person or a family tion (less than a college degree, college degree or higher), disabil-
needs to pay for bare living essentials specific to family size and is ity status (yes, no), perceived physical health (good, poor), and
updated by the US government yearly (15). Percentage of FPL

The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services,
the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions.

2 Centers for Disease Control and Prevention • www.cdc.gov/pcd/issues/2025/25_0153.htm

PREVENTING CHRONIC DISEASE VOLUME 22, E50
PUBLIC HEALTH RESEARCH, PRACTICE, AND POLICY AUGUST 2025

perceived mental health (good, poor). Disability was defined by aimed to better describe how utilization, either individually or in
any self-reported daily living, functional, or activity limitations. combination, explained the observed difference.

Statistical analyses Results

Characteristics of people in the 4 income groups were summar-
The study included 14,227 adults with self-reported diabetes, ran-
ized as number and percentage with standard error (SE) for cat-
ging from 2,741 in 2017 to 3,125 in 2021. Demographic and so-
egorical variables and mean with SE for continuous variables. Dif-
cioeconomic characteristics and health status differed signific-
ferences in demographic and socioeconomic factors and health
antly across the 4 income groups. The adults from poor-income
status between income groups were compared using χ2 tests for
households, compared with those in other income groups, had the
categorical variables and t tests for continuous variables. We used
highest proportion of young people (aged 18–44 y); women; Black
the 1-way analysis of variance (ANOVA) method to test if ex-
and Hispanic people; people with dual Medicaid–Medicare cover-
penditure differences between adults from poor versus low, poor
age, Medicare only, or uninsured; people without a college degree;
versus middle, poor versus high, low versus middle, low versus
people reporting poor physical or mental health status; and people
high, and middle versus high household income groups were stat-
with a disability (Table 1).
istically different. A P value of .05 or less was used to define sig-
nificant differences. All variables were weighted according to the Differences in medical expenditures
MEPS guideline.
Mean total annual medical per-person expenditures were highest
For income groups with significant expenditure differences, we in adults from the poor-income households ($19,071), followed by
used the Blinder−Oaxaca decomposition method to examine how the low-income ($16,143), high-income ($15,961), and middle-
each demographic and socioeconomic factor and health status income ($14,930) households (Figure 1). The largest annual per-
variable was associated with the difference (17). Blinder–Oaxaca person expenditure difference, which was between adults from
decomposition is a regression-based method that predicts the mean poor-income households and those from the middle-income house-
expenditure of a group by using the mean value of each included holds, was $4,141 (Table 2). ANOVA showed the expenditures of
factor and its estimated regression coefficient. This method spe- adults from poor-income households significantly exceeded ex-
cifies how much of the difference in the mean expenditures penditures in the other 3 groups; other group differences were not
between 2 groups is explained by the difference in the mean val- significant.
ues of included factors. The remaining discrepancy cannot be ex-
plained by factors in the model (18). A factor can contribute posit-
ively or negatively to the difference in the mean value of the ex-
penditure between 2 income groups. Elevating factors are defined
as factors associated with higher difference; offsetting factors are
associated with lower difference. It is possible that the explained
difference could be larger than the actual difference, which indic-
ates that there should be an even bigger difference between the
groups based on the factors included in the model alone. This dis-
crepancy could be due to limitations in the model or unobserved
factors that neutralize the expected effect.

Sensitivity analysis
We conducted 2 additional analyses. First, we applied the same
analytical approach to people without diabetes and compared these
results to the corresponding results for people with diabetes to ex-
plore the effect of diabetes on expenditure difference. Second, we Figure 1. Mean per person per year medical expenditures in US dollars,
conducted an analysis of utilization and per-unit expenditure for by income level among people with diabetes, Medical Expenditure Panel
medical services where the difference in expenditure of this ser- Survey, 2017–2021. Income categories were defined based on federal
poverty level (FPL), which incorporates both household income and size:
vice divided by the difference of the total medical expenditure poor income (<125% FPL), low income (125% to <200% FPL), middle
between income groups was greater than 20%. This analysis income (200% to <400% FPL), and high income (≥400% FPL).

The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services,
the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions.

www.cdc.gov/pcd/issues/2025/25_0153.htm • Centers for Disease Control and Prevention 3

PREVENTING CHRONIC DISEASE VOLUME 22, E50
PUBLIC HEALTH RESEARCH, PRACTICE, AND POLICY AUGUST 2025

Annual inpatient service expenditures per person ranged from

$3,522 (middle-income) to $4,022 (low-income), with no signific-
ant differences. Outpatient expenditures ranged from $3,991 (low-
income) to $4,897 (high-income), with expenditures for adults
from poor-income and low-income households significantly lower
than for high-income households. Prescription drug expenditures
ranged from $4,989 (middle-income) to $7,481 (poor-income),
with expenditures for adults from poor-income households signi-
ficantly higher than for the others; other differences were not sig-
nificant. Expenditures for home health care services ranged from
$471 (high-income) to $2,494 (poor-income), decreasing with de-
creasing income, with significant differences between all pairs ex-
cept middle-income and high-income. Emergency department ex-
penditures ranged from $328 (high-income) to $409 (poor-
income), significant only between these 2. Other medical ex-
penditures ranged from $748 (poor-income) to $1,141 (high- Figure 2. Decomposition results between income groups in total and by
income), with adults from high-income households significantly health service types, in US dollars, among adults with diabetes, Medical
Expenditure Panel Survey, 2017–2021. Each bar represents the total
higher than the others (Figure 1; Table 2). difference in expenditures between the respective income groups, with the
label underneath indicating the specific comparison (eg, Outpatient: poor
Decomposition Analysis vs high). A positive bar segment means that the factor is associated with
increase in the difference in medical expenditures between the income
Total medical expenditures groups. Conversely, a negative bar segment indicates that the factor is
associated with decrease in the difference in expenditures. The total bar
For total annual per-person medical expenditures, decomposition height (or depth) represents the net difference in medical costs,
analyses were performed on the 3 comparison groups whose dif- accounting for both positive and negative contributions. The unexplained
portion may include factors not accounted for by the selected variables,
ferences were significant in the ANOVA models (adults from representing residual disparities beyond what the model can explain.
poor-income households vs adults from each of the other income Income categories were defined based on federal poverty level (FPL),
level households) (Figure 2). Among adults from the poor-income which incorporates both household income and size: poor income (<125%
FPL), low income (125% to <200% FPL), middle income (200% to <400%
households relative to the low-income households, 99.7% of the FPL), and high income (≥400% FPL). Abbreviation: Dual coverage,
difference of the higher expenditure could be explained by the coverage with both Medicaid and Medicare.
factors included in the model. Significant elevating factors were a
higher percentage of Medicaid–Medicare dual coverage (43%), a Outpatient expenditures
higher disability rate (41%), and a higher proportion reporting Adults from the high-income households had higher annual per-
poor physical health (27%). For the poor-income versus middle- person outpatient care expenditures than either adults from the
income and high-income comparisons, the model predicted an poor-income or low-income households. However, in decomposi-
11% higher difference than was observed, indicating the existence tion analysis the model predicted higher outpatient expenditures
of offsetting factors that were not included our model. Dual cover- for adults from the poor-income and low-income households than
age, disability, and poor physical health were again the significant the high-income households. Thus, variables included in the re-
elevating factors. For the poor income versus high income com- gression analyses could not explain the higher outpatient medical
parison, the model predicted a 43% higher difference than was ob- expenditure in adults from the high-income households.
served, indicating the effect of offsetting factors beyond our mod-
el; a lower proportion of the White population, a higher propor- Prescription drug expenditures
tion of Black population, and a lower college education rate were Decomposition analysis was able to explain 63% of the higher ex-
significant factors, in addition to Medicaid–Medicare dual cover- penditures on prescription drugs for adults from the poor-income
age, disability, and poor physical health. versus the low-income households, 64% for adults from the poor-
income versus middle-income households, and 72% for adults
from the poor-income versus high-income households.
Medicaid–Medicare dual coverage, poor physical health status,
and disability were significant factors for all 3 comparisons. In ad-
dition, the higher proportion of the population aged 55 to 64 years
from poor-income households was associated with higher pre-

The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services,
the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions.

4 Centers for Disease Control and Prevention • www.cdc.gov/pcd/issues/2025/25_0153.htm

PREVENTING CHRONIC DISEASE VOLUME 22, E50
PUBLIC HEALTH RESEARCH, PRACTICE, AND POLICY AUGUST 2025

scription expenditure when compared with adults aged 55 to 64 play a more important role than the factors included in the model
years from low-income and middle-income households. Offset- regarding group differences.
ting factors included a lower proportion of White population
among adults from poor-income households compared with adults Sensitivity analyses
from high-income households, a lower Medicare-only insurance In US adults without diabetes, annual per-person medical ex-
population from poor-income households compared with adults penditures ranged from $5,769 (middle income) to $6,663 (high),
from the low-income and middle-income households, and a high- significantly lower than for those with diabetes in the same in-
er uninsured population from poor-income households compared come groups. Unlike the descending order in total medical ex-
with adults from the middle-income and high-income households. penditure (poor, low, high, middle) in the population with dia-
Home health care services expenditures betes, the order is high, poor, low, middle in the population
without diabetes. Total medical expenditure was significantly
Decomposition analysis explained 95% of the difference between
higher in adults from the poor-income ($553) and high-income
expenditures of adults from the poor-income and low-income
($894) households compared with middle-income group, with no
households and 77% between adults from the poor-income and the
significant differences for other pairs. The annual per-person ex-
middle-income households, and it overestimated the difference in
penditure gap was $894 between the income groups with the
expenditures between adults from the poor-income versus high-
highest and lowest expenditure among the population without dia-
income households by 2%. The top elevating factors to explain the
betes, versus $4,141 for the population with diabetes. The differ-
difference between adults from poor-income and both the low-
ence between adults from the middle-income and high-income
income and middle-income households included a higher Medi-
households was mainly from higher outpatient expenditures in
caid–Medicare dual coverage rate, a higher disability rate, a lower
adults from the high-income households, while the difference
Medicare-only rate, and a higher rate of perceived poor mental
between adults from middle-income and poor-income households
health. In contrast, a higher proportion of people aged 55 to 64
was mainly from higher inpatient expenditure in the poor-income
from the poor-income households was an offsetting factor. Higher
group in the population without diabetes. But in the population
Medicaid–Medicare dual coverage and disability rates were the 2
with diabetes, the difference in annual per person costs was from
elevating factors to explain the difference in home health care ser-
prescription and home health care services expenditure differ-
vices expenditure between adults from poor-income households
ences between the poor-income group and the other 3 income
and adults from high-income households.
groups.
For adults from low-income versus the middle-income and high-
Prescription and home health care services accounted for more
income households, 61% and 93%, respectively, of the differ-
than 20% of expenditure differences (prescription: poor vs low
ences were explained by the factors included in our model. The
67%, poor vs middle 60%, poor vs high 67%; home health: 35%,
factors significant in explaining the differences were similar
45%, 65%, respectively). Usage and per-unit expenditure analyses
between adults from the poor-income and high-income house-
compared prescription drugs and home health care services
holds, except that a higher rate of poor physical health was also a
between adults from the poor-income households and others. For
significant elevating factor.
prescription drugs, the unit expenditure per refill was significantly
Emergency department expenditures higher for adults from the high-income households compared with
Our model explained 74% of the difference in emergency depart- adults from the other 3 income level households; unit expenditure
ment expenditures between adults from the poor-income and high- per refill was higher in adults from the poor-income than the low-
income households. Higher disability rates and higher rates of re- income households. The total number of prescription refills in a
ported poor mental and physical health were significant elevating year decreased as income level increased, and the differences were
factors, while higher Medicaid–Medicare dual coverage rate was significant between all comparison pairs.
the only significant offsetting factor. The expenditure per home health care visit was highest in adults
Other medical expenditures from the high-income households, followed by low-income,
middle-income, and poor-income households. Expenditure per
Our model explained only a small portion of the difference in oth-
home health care visit was significantly lower for adults from
er medical expenditures among adults from the poor-income, low-
poor-income households compared with adults from low-income
income, and middle-income households versus the high-income
and high-income households. The total number of home health
households, indicating that factors not included in the model may
care services in a year decreased as income level increased, and
the differences were significant for all comparison pairs.

The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services,
the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions.

www.cdc.gov/pcd/issues/2025/25_0153.htm • Centers for Disease Control and Prevention 5

PREVENTING CHRONIC DISEASE VOLUME 22, E50
PUBLIC HEALTH RESEARCH, PRACTICE, AND POLICY AUGUST 2025

To better understand the relationship between income and health, fit: improving health outcomes and preserving peoples’ ability to
we conducted an additional analysis comparing comorbidity pro- maintain stable employment and income, thus reducing health care
files across income levels. Most chronic conditions — such as needs.
high blood pressure, arthritis, high cholesterol, asthma, coronary
heart disease, myocardial infarction, and stroke — were more pre- Dual Medicaid–Medicare coverage status was significantly associ-
valent among adults from lower-income households, with signific- ated with the income-related medical expenditure difference in our
ant differences observed between adults from the poor-income or study, highlighting the role of the dual-coverage policy in support-
low-income households and adults from the high-income house- ing people with greater needs. Dually covered beneficiaries have
holds. Overall, these findings are consistent with our main results, low incomes and are either elderly or have long-term disabilities,
which indicate poorer physical health among adults from the leading to higher medical care needs and a larger share of ex-
lowest-income households. penditure from both programs (26,27). Among the
Medicaid–Medicare dually covered population with diabetes,
Discussion higher rates of diabetes-related complications and comorbidities
were observed compared with people with other insurance types.
We examined differences in medical expenditures among US Dual coverage helped address barriers to health care access (28)
adults with diabetes by income level and found that adults from and was shown to be effective in meeting the complex health
the lowest income (poor) households faced significantly higher ex- needs of eligible people.
penditures — $2,928 to $4,141 or 15% to 22% more than adults
Diabetes influenced both the magnitude and pattern of income-
from higher income households. This discrepancy primarily stems
related differences in medical expenditure. People without dia-
from spending on prescription drugs and home health care ser-
betes had 34% to 42% lower annual per-person medical expendit-
vices. Higher disability rates, poor physical health status, and
ures than those with diabetes (Table 2). Furthermore, the income-
Medicaid–Medicare dual coverage were the most important
related expenditure gap was much greater among people with dia-
factors explaining the difference between adults from the lowest
betes. The difference in medical expenditures between income
income households and other income level households. Our study
groups was $905 (13%) for those without diabetes compared with
is the first to describe expenditure differences by income level and
$4,141 (22%) for those with diabetes (Table 2).
to explore the associated factors in US adults with diabetes. Our
cost estimates by income group could serve as a measure for plan- Our findings suggest that efforts to improve health status and pre-
ning the budget needed for social programs targeting low-income vent disability in the lowest income group could yield financial
people. By identifying cost sources and factors behind expendit- benefits due to their high medical expenditures. For instance, stud-
ure gaps, our findings may help inform efforts to ease the finan- ies evaluating the health and economic impact of diabetes preven-
cial burden of diabetes across income levels. tion programs in Medicaid populations showed that such pro-
grams were cost-effective, could result in cost savings over a 25-
Our results align with studies focused on the general US popula-
year horizon, and could improve health equity (29). Additionally,
tion, showing that people from families with lower incomes use
as diabetes is a strong risk factor for disability, preventing dia-
more health care services and face higher medical costs (19,20),
betes at a population level would also be effective in reducing dis-
primarily due to poorer health status and higher rates of disability
ability rates, further lowering medical expenditures in the lowest
(19–21). Poor health and disabilities increase care needs but can
income group. Among people with diagnosed diabetes, manage-
also limit work, reducing income and worsening financial strain.
ment education programs aimed at reducing complications have
In our study, the rate of disability in adults from the poor-income
been particularly cost-effective and even cost-saving, especially
households was nearly twice as high as for adults from the high-
for low-income patients (30,31).
income households, which explains a large proportion of higher
medical expenditures among adults from the poorest income Limitations
households (22). Previous studies have also found that disability
status is strongly associated with elevated medical expenditures in This study has several limitations. First, MEPS is limited to the ci-
people with diabetes (23). Despite the availability of various gov- vilian noninstitutionalized population, so expenditures for people
ernment benefit programs, many people still face long-term in- in long-term care for disability-related reasons were not included,
come loss due to disability (24). Diabetes increases the risk of potentially underestimating medical expenditure differences.
physical disability by 50% to 90% (25). Preventing disability Second, results from our decomposition analyses should be inter-
among people with diabetes could therefore offer a twofold bene- preted with caution, as the associations between factors and ex-
penditures are not causal. Medical expenditures may be affected

The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services,
the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions.

6 Centers for Disease Control and Prevention • www.cdc.gov/pcd/issues/2025/25_0153.htm

PREVENTING CHRONIC DISEASE VOLUME 22, E50
PUBLIC HEALTH RESEARCH, PRACTICE, AND POLICY AUGUST 2025

by other factors not included in our model, such as health insur- References
ance policies, purchaser options, national and state policies, and
drug patents. Third, self-reported health status may not accurately 1. Robbins JM, Vaccarino V, Zhang H, Kasl SV. Socioeconomic
reflect actual health care needs. The large portion of the expendit- status and diagnosed diabetes incidence. Diabetes Res Clin
ure difference explained by Medicaid–Medicare dual coverage Pract. 2005;68(3):230–236. doi:10.1016/j.diabres.2004.09.007
does not suggest that eliminating these benefits would reduce dis- 2. Centers for Disease Control and Prevention. National Diabetes
parities; rather, dual coverage provides the financial means neces- Statistics Report. 2024. Accessed January 15, 2025. https://
sary to meet health care needs for the poor. Lastly, while we at- www.cdc.gov/diabetes/php/data-research/?CDC_AAref_Val=
tempted to use specific comorbidities and complications as prox- https://www.cdc.gov/diabetes/data/statistics-report/index.html
ies, sample size limitations and missing data prevented further 3. Houle J, Lauzier-Jobin F, Beaulieu MD, Meunier S, Coulombe
analysis. S, Côté J, et al. Socioeconomic status and glycemic control in
adult patients with type 2 diabetes: a mediation analysis. BMJ
Conclusion Open Diabetes Res Care . 2016;4(1):e000184. doi:10.1136/
Significant differences in medical expenditures per person per year bmjdrc-2015-000184
exist by income level among US adults with diabetes, with the 4. Bird Y, Lemstra M, Rogers M, Moraros J. The relationship
poorest income group incurring 15% to 22% higher expenditures between socioeconomic status/income and prevalence of
than higher income groups. Higher spending on prescription drugs diabetes and associated conditions: a cross-sectional
and home health care services were the 2 main drivers. Higher dis- population-based study in Saskatchewan, Canada. Int J Equity
ability rates, poor reported physical health status, and Health. 2015;14(1):93. doi:10.1186/s12939-015-0237-0
Medicaid–Medicare dual coverage largely explain these expendit- 5. C o l l i e r A , G h o s h S , H a i r M , W a u g h N . I m p a c t o f
ure differences. Programs aimed at improving health status and socioeconomic status and gender on glycaemic control,
preventing disability in low-income populations with diabetes may cardiovascular risk factors and diabetes complications in type 1
help reduce medical expenditure disparities between income and 2 diabetes: a population based analysis from a Scottish
groups in the US. region. Diabetes Metab. 2015;41(2):145–151. doi:10.1016/j.
diabet.2014.09.004
Acknowledgments 6. Osborn CY, de Groot M, Wagner JA. Racial and ethnic
disparities in diabetes complications in the northeastern United
Conflict of interest disclosure: None reported. Funding: The au- States: the role of socioeconomic status. J Natl Med Assoc.
thors received no financial support for the research, authorship, or 2013;105(1):51–58. doi:10.1016/S0027-9684(15)30085-7
publication of this article. 7. Saydah S, Lochner K. Socioeconomic status and risk of
diabetes-related mortality in the U.S. Public Health Rep. 2010;
The findings and conclusions in this report are those of the au- 125(3):377–388. doi:10.1177/003335491012500306
thors and do not necessarily represent the official position of the 8. Lazar M, Davenport L. Barriers to health care access for low
Centers for Disease Control and Prevention. income families: a review of literature. J Community Health
No copyrighted material, surveys, instruments, or tools were used Nurs. 2018;35(1):28–37. doi:10.1080/07370016.2018.1404832
in the research described in this article. 9. Gunja MZ, Collins SR. Who are the remaining uninsured, and
why do they lack coverage. Commonwealth Fund; 2019.
Author Information 10. Tatulashvili S, Fagherazzi G, Dow C, Cohen R, Fosse S, Bihan
H. Socioeconomic inequalities and type 2 diabetes
Corresponding Author: Yu Wang, PhD, Division of Diabetes complications: a systematic review. Diabetes Metab . 2020;
Translation, National Center for Chronic Disease Prevention and 46(2):89–99. doi:10.1016/j.diabet.2019.11.001
Health Promotion, Centers for Disease Control and Prevention, 11. Zhu X, Cai Q, Wang J, Liu Y. Determinants of medical and
4770 Buford Hwy, Atlanta, GA 30341 ([email protected]). health care expenditure growth for urban residents in China: a
systematic review article. Iran J Public Health. 2014;43(12):
Author Affiliations: 1Division of Diabetes Translation, Centers for 1597–1604.
Disease Control and Prevention, Atlanta, Georgia. 2 Hubert 12. Pashchenko S, Porapakkarm P. Medical spending in the US:
Department of Global Health, Rollins School of Public Health, facts from the Medical Expenditure Panel Survey data set. Fisc
Emory University, Atlanta, Georgia. Stud. 2016;37(3-4):689–716. doi:10.1111/j.1475-5890.2016.
12100

The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services,
the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions.

www.cdc.gov/pcd/issues/2025/25_0153.htm • Centers for Disease Control and Prevention 7

PREVENTING CHRONIC DISEASE VOLUME 22, E50
PUBLIC HEALTH RESEARCH, PRACTICE, AND POLICY AUGUST 2025

13. Cunningham PJ, Green TL, Braun RT. Income disparities in 24. Meyer BD, Mok WK. Disability, earnings, income and
the prevalence, severity, and costs of co-occurring chronic and consumption. J Public Econ. 2019;171:51–69. doi:10.1016/j.
behavioral health conditions. Med Care. 2018;56(2):139–145. jpubeco.2018.06.011
doi:10.1097/MLR.0000000000000864 25. Pilla SJ, Rooney MR, McCoy RG.Disability and diabetes in
14. Cohen JW, Cohen SB, Banthin JS. The Medical Expenditure adults. In: Lawrence JM, Casagrande SS, Herman WH, Wexler
Panel Survey: a national information resource to support DJ, Cefalu WT, eds. Diabetes in America. National Institute of
healthcare cost research and inform policy and practice. Med Diabetes and Digestive and Kidney Diseases; 2023.
Care . 2009;47(7 suppl 1):S44–S50. doi:10.1097/MLR. 26. Hackbarth G, Berenson R, Miller M. Report to the Congress:
0b013e3181a23e3a Medicare and the health care delivery system . Medicare
15. Agency for Healthcare Research and Quality. MEPS HC 243 Payment Advisory Comission; 2013.
2022 full year consolidated data file. 2024. Accessed October 27. Figueroa JF, Lyon Z, Zhou X, Grabowski DC, Jha AK.
18, 2024. https://meps.ahrq.gov/data_stats/download_data/ Persistence and drivers of high-cost status among dual-eligible
pufs/h243/h243doc.shtml Medicare and Medicaid beneficiaries: an observational study.
16. Agency for Healthcare Research and Quality. Using Ann Intern Med . 2018;169(8):528–534. doi:10.7326/M18-
appropriate price indices for analyses of health care 0085
expenditures or income across multiple years. Accessed 28. Stuart B, Yin X, Davidoff A, Simoni-Wastila L, Zuckerman I,
February 21, 2024. https://meps.ahrq.gov/about_meps/Price_ Shoemaker JS, et al. Impact of Part D low-income subsidies on
Index.shtml medication patterns for Medicare beneficiaries with diabetes.
17. Rahimi E, Hashemi Nazari SS. A detailed explanation and Med Care . 2012;50(11):913–919. doi:10.1097/MLR.
graphical representation of the Blinder–Oaxaca decomposition 0b013e31826c85f9
method with its application in health inequalities. Emerg 29. Laxy M, Zhang P, Ng BP, Shao H, Ali MK, Albright A, et al.
Themes Epidemiol. 2021;18(1):12. doi:10.1186/s12982-021- Implementing lifestyle change interventions to prevent type 2
00100-9 diabetes in US Medicaid programs: cost effectiveness, and
18. Jann B. The Blinder–Oaxaca decomposition for linear cost, health, and health equity impact. Appl Health Econ
regression models. Stata J. 2008;8(4):453–479. doi:10.1177/ Health Policy. 2020;18(5):713–726. doi:10.1007/s40258-020-
1536867X0800800401 00565-w
19. Lemstra M, Mackenbach J, Neudorf C, Nannapaneni U. High 30. Prezio EA, Pagán JA, Shuval K, Culica D. The Community
health care utilization and costs associated with lower socio- Diabetes Education (CoDE) program: cost-effectiveness and
economic status: results from a linked dataset. Can J Public health outcomes. Am J Prev Med. 2014;47(6):771–779. doi:10.
Health. 2009;100(3):180–183. doi:10.1007/BF03405536 1016/j.amepre.2014.08.016
20. Fitzpatrick T, Rosella LC, Calzavara A, Petch J, Pinto AD, 31. Gilmer TP, Roze S, Valentine WJ, Emy-Albrecht K, Ray JA,
Manson H, et al. Looking beyond income and education: Cobden D, et al. Cost-effectiveness of diabetes case
socioeconomic status gradients among future high-cost users of management for low-income populations. Health Serv Res .
health care. Am J Prev Med . 2015;49(2):161–171. doi:10. 2007;42(5):1943–1959. doi:10.1111/j.1475-6773.2007.00701.
1016/j.amepre.2015.02.018 x
21. Booth GL, Bishara P, Lipscombe LL, Shah BR, Feig DS,
Bhattacharyya O, et al. Universal drug coverage and
socioeconomic disparities in major diabetes outcomes.
Diabetes Care. 2012;35(11):2257–2264. doi:10.2337/dc12-
0364
22. Ng BP, Shrestha SS, Lanza A, Smith B, Zhang P. Medical
expenditures associated with diabetes among adult Medicaid
enrollees in eight states. Prev Chronic Dis . 2018;15:E116.
doi:10.5888/pcd15.180148
23. Williams JS, Egede LE. Differences in medical expenditures
for men and women with diabetes in the Medical Expenditure
Panel Survey, 2008–2016. Womens Health Rep (New
Rochelle). 2020;1(1):345–353. doi:10.1089/whr.2020.0050

The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services,
the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions.

8 Centers for Disease Control and Prevention • www.cdc.gov/pcd/issues/2025/25_0153.htm

PREVENTING CHRONIC DISEASE VOLUME 22, E50
PUBLIC HEALTH RESEARCH, PRACTICE, AND POLICY AUGUST 2025

Tables
Table 1. Characteristics of US Adult Population With Diabetes, by Income Level, 2017–2021 Medical Expenditure Panel Surveya

Poor income Low income Middle income High income

Characteristic (<125% FPL) (125% to <200% FPL) (200% to <400% FPL) (≥400% FPL) P value
Sample size 4,009 2,386 3,913 3,919
Weighted sample 28,508,850 21,108,144 39,644,109 48,866,689
Mean age, y 61.2 (0.4) 63.4 (0.5) 61.7 (0.4) 61.9 (0.4) <.001
Age, %
18–44 15.0 (1.0) 14.3 (1.3) 13.7 (1.1) 8.4 (0.8) <.001
45–54 12.8 (1.0) 10.3 (1.1) 15.5 (1.1) 15.5 (1.3)
55–64 27.6 (1.2) 20.3 (1.5) 22.3 (1.2) 28.8 (1.4)
≥65 44.6 (1.5) 55.1 (2.2) 48.4 (1.6) 47.3 (1.7)
Sex, %
Female 59.4 (1.4) 53.1 (1.8) 47.6 (1.2) 40.9 (1.4) <.001
Male 40.5 (1.4) 46.9 (1.8) 52.4 (1.2) 59.1 (1.4)
Race and ethnicity, %
Non-Hispanic White 45.6 (2.1) 53.8 (2.0) 60.2 (1.6) 68.9 (1.5) <.001
Non-Hispanic Black 22.0 (1.6) 17.3 (1.6) 13.8 (1.0) 11.0 (1.0)
Hispanic 23.0 (1.9) 20.7 (1.7) 17.5 (1.3) 10.2 (0.9)
Otherb 9.4 (1.0) 8.2 (0.8) 8.5 (0.8) 9.9 (0.9)
Insurance, %
Medicaid–Medicare dual 30.6 (1.3) 17.0 (1.2) 6.7 (0.6) 3.1 (0.6) <.001
coverage
Medicare only or with private 28.7 (1.1) 46.3 (1.7) 45.2 (1.3) 43.8 (1.4)
Medicaid only 27.4 (1.4) 14.5 (1.2) 7.8 (0.6) 2.4 (0.3)
Private only 6.3 (0.7) 15.2 (1.1) 35.7 (1.3) 48.0 (1.4)
Uninsured 5.8 (0.6) 5.6 (0.7) 3.4 (0.4) 1.3 (0.3)
c
Other insurance 1.2 (0.3) 1.4 (0.4) 1.2 (0.2) 1.4 (0.3)
Education, %
College degree or higher 8.7 (0.8) 12.4 (1.3) 19.6 (1.1) 38.7 (1.4) <.001
Self-reported health status, %
Poor physical health 49.6 (1.1) 37.2 (1.4) 30.2 (1.1) 23.2 (0.9) <.001
Poor mental health 26.8 (1.0) 17.8 (1.1) 12.1 (0.7) 8.1 (0.6) <.001
Disability 64.6 (1.4) 54.4 (1.4) 42.3 (0.4) 35.8 (1.2) <.001

Abbreviation: FPL, federal poverty level.

a
Unless otherwise noted, the data are weighted means (standard errors).
b
American Indian or Alaska Native, Asian, Native Hawaiian or Pacific Islander, or multiple races.
c
Other public insurance and other hospital or physician coverage.

The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services,
the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions.

www.cdc.gov/pcd/issues/2025/25_0153.htm • Centers for Disease Control and Prevention 9

PREVENTING CHRONIC DISEASE VOLUME 22, E50
PUBLIC HEALTH RESEARCH, PRACTICE, AND POLICY AUGUST 2025

Table 2. Total Mean Per Person Per Year Medical Expenditure (US Dollars) and Comparisons by Health Service Type and Income Level Among the US
Adult Population With Diabetes, 2017–2021 Medical Expenditure Panel Surveya

Home health Emergency

Difference Total Inpatient Outpatient Prescription care services department Other
Poor income 19,071 (767) 3,921 (343) 4,017 (264) 7,481 (379) 2,494 (298) 409 (24) 748 (44)
Poor vs Low 2,928b (926) −100 (544) 102 (362) 1,964b (383) 1,019b (363) 62 (48) −118 (83)
b b b
Poor vs Middle 4,141 (872) 400 (395) −546 (382) 2,492 (375) 1,878 (299) 19 (47) −101 (63)
Poor vs High 3,110b (909) 204 (482) −875b (339) 2,069b (412) 2,022b (301) 82b (39) −393b (66)
Low income 16,143 (743) 4,022 (438) 3,991 (78) 5,517 (276) 1,475 (216) 348 (42) 866 (77)
Low vs Middle 1,213 (848) 500 (502) −648 (343) 528 (293) 860b (212) −43 (56) 16 (88)
b b
Low vs High 181 (994) 304 (535) −977 (340) 105 (320) 1,004 (234) 20 (48) −275b (94)
Middle income 14,930 (543) 3,522 (239) 4,533 (267) 4,989 (177) 616 (68) 391 (38) 850 (45)
Middle vs High −1,031 (693) −196 (382) −328 (314) −423 (238) 144 (234) 63 (46) −291b (58)
High income 15,961 (515) 3,718 (316) 4,897 (216) 5,412 (193) 471 (100) 328 (29) 1,141 (45)
a
The presented numbers are mean (standard error).
b
Significantly different at P < .05.

The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services,
the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions.

10 Centers for Disease Control and Prevention • www.cdc.gov/pcd/issues/2025/25_0153.htm