Trends in Medical Expenditures by Race/Ethnicity in Adults with Type 2 Diabetes 2002 – 2011

Abstract

Objective:

The objective of the study was to examine racial/ethnic differences in medical expenditures (prescription, office-based, in-patient, out-patient, emergency room, total) over time, overall and by type of expenditure, in a nationally representative sample of adults with diabetes.

Methods:

A weighted sample of 17,820,243 adults aged ≥ 18 with diabetes from the Medical Expenditure Panel Survey (MEPS) dataset from 2002 – 2011 were analyzed for this study. Multiple comparison testing and general linear modeling (GLM) were used to test for differences in expenditures by race. Total unadjusted expenditures by racial/ethnic category stratified by different insurance categories (privately insured, publicly insured, uninsured and overall) were also estimated.

Results:

Non-Hispanic Whites (NHW) had more than $4000 higher expenditures than Hispanics and Other races (p<0.0001). Prescription costs were about $410 less for Non-Hispanic Blacks (NHB) (p<0.0001), and more than $600 less for Hispanics (p<0.0001) and Others (p<0.0001) compared to NHW.

Conclusion:

Minority groups with type 2 diabetes were found to have significantly less total expenditures, with the exception of total expenditures for NHB compared to NHW. These findings indicate minorities with type 2 diabetes may be receiving less care than NHW, which has implications for the known disparities in health outcomes and complications in individuals with diabetes.

Introduction

Diabetes is a highly prevalent disease affecting millions of people (1). In the United States (US), it is the 7^th leading cause of death and leads to a loss of life expectancy of more than two years (2,3). It is predicted that the number of individuals living with diabetes in the US will increase to over 53 million by the year 2025 (4). Additionally, management of diabetes is very costly, totaling about $245 billion in direct and indirect costs in the US annually (2,5–7). About $176 billion were attributed to direct medical costs including outpatient, emergency, inpatient, prescription medication, and medical supply costs, while $69 billion is attributed to indirect costs such as disability, work loss, and premature death (1,2,5,6). Due to the increasing number of individuals living with diabetes, it is predicted that the cost to treat the disease will reach $514 billion by the year 2025 (4). After adjusting for population age and sex differences, average medical expenditures of those diagnosed diabetes were 2.3 times higher than people without diabetes, with 43% of their direct medical costs a result of inpatient stays (2,6,8,9). Over an individuals’ lifetime, diabetes increases health care expenditure by almost $9,000 and co-morbid conditions frequently associated with uncontrolled diabetes further exacerbate medical expenditures (8,10).

Disparities in the care of individuals diagnosed with diabetes is widely recognized in the US. Higher prescribing of insulin and lower prescribing of oral agents for African Americans with diabetes has been shown compared to NHW and Mexican-Americans (11). Hispanics with type 2 diabetes have been noted to have poorer access to care and poorer health status compared to NHW or NHB (12). For example, larger proportions of Hispanics and NHB reported not having had a hemoglobin A1c test in the prior 12 months, and Spanish-speaking Hispanic patients were significantly less likely than NHW patients to have had a physician visit (12–14).

Not surprisingly, the disparities in care and outcomes for patients with diabetes extend to disparities in costs of diabetes by race/ethnicity. Lifetime health care expenditures are higher for whites than NHB and Hispanics (8,9,15). NHBs and Hispanics with diabetes have been found to have significantly lower total healthcare costs, ambulatory care costs, and prescription costs than NHWs (16). Hispanics with diabetes were found to have fewer office-based visits and fewer prescriptions than non-Hispanics, therefore their expenditures were significantly less than for non-Hispanics. (17). Patients who have financial burdens are more likely to be non-adherent to medications because of the cost (18). In addition, individuals with higher income had higher physician office-based visits than those with lower income (9). Those with high out-of-pocket pay burdens, for example the under or uninsured who tend to be minority patients, are also less likely to seek adequate medical care to control their diabetes (19,20).

Therefore, the aim of this study was to examine racial/ethnic differences in medical expenditures, overall and by type of expenditure (prescription, office-based, in-patient, out-patient, emergency room, total), in a nationally representative sample of adults with diabetes. Past studies have looked at differences in Medicare or Medicaid populations, or by only examining one year of nationally representative data. We combined 10 years of data on medical expenditures collected by the Medical Expenditures Panel Survey (MEPS) to examine racial/ethnic differences and trends in medical expenditures, to better inform action on disparities in care for patients with diabetes.

Methods

Data Source and study subjects

We used pooled cross-sectional data analysis of Medical Expenditure Panel Survey (MEPS) among adults with diabetes for the years 2002-2011. A weighted population of 17,820,243 individuals among adults with diabetes were considered for analysis. The study used data from household component (HC), a survey of a nationally representative U.S. civilian non-institutionalized population. The survey is administered by the Agency for Healthcare Research and Quality (21). We used the full-year consolidated file from MEPS-HC, which contains information on participants’ use of medical care and their medical spending, as well as information on demographics, socioeconomics, and health conditions. Medical expenditures are defined as the payments that health care providers receive from all payers (including insurance providers, survey respondents, and other sources) (22).

Information on the HC is collected by self-report, and the Medical Provider Component (MPC) requests data on medical and financial characteristics from hospitals, physicians, home health care providers, and pharmacies in order to validate and supplement information received from the MEPS-HC respondents. (21).

We combined 10 years of MEPS data because they have a common variance structure necessary to ensure compatibility and comparatively of our variables within the complex sample design. We adjusted the analytic sampling weight variable by dividing it by the number of years being pooled. The analysis was adjusted for complex survey design consisting of clustering, stratification, and multistage and disproportionate sampling with oversampling of minorities (22). The 2002-2011 direct healthcare costs were adjusted to a common 2014 dollar using the consumers price index obtained from the Bureau of Labor Statistics (BLS) (23).

Measures

Dependent variables in this study are the total direct expenditures defined as the sum of direct payments for care provided during the year, including out-of-pocket payments and payments by private insurance, Medicaid, Medicare and other sources (22). The total medical healthcare expenditure is a sum of office-based medical provider expenditure, hospital outpatient expenditure, emergency room expenditure, inpatient hospital (including zero night stays) expenditure, prescription medicine expenditure, dental expenditure, home health care expenditure and other medical expenses (22). Analyses considered the total direct expenditures, and individual categories of cost including prescription, office-based, inpatient, outpatient, and ER expenditures.

The primary independent variable of interest was self-reported race/ethnicity. Race/ethnic groups were categorized as: Non-Hispanic White (NHW), Non-Hispanic Black (NHB), Hispanic, and Other.

Access to ‘Usual Source of Care’ was measured from a positive response to the question: Do you have a usual source of care provider? The variable ‘Medical Access’ was created from responses to three separate questions: 1) Do you have a usual source of care provider? 2) Were you unable to get necessary medical care? and 3) Were you delayed in getting necessary medical care? An answer of yes to the first question, and no to both of the second and third questions as coded as having medical access. ‘Prescription Access’ variable was created from responses to two questions: 1) Were you unable to get necessary prescription medication? and 2) Were you delayed in getting necessary prescription medication? An answer of no to both questions was coded as having prescription access.

All controlled covariates used for analysis were based on self-report included sex, age, marital status, education, insurance, living in metropolitan statistical area (MSA), region, and poverty income ratio. Covariates also included self-reported comorbidities of hypertension, cardio vascular disease (CVD), stroke, emphysema, joint pain, arthritis and asthma.

Analyses

The population characteristics of patients with diabetes were analyzed by race/ethnic categories, with differences tested for using Chi-square (X²) tests. We used standard pairwise comparison methods of Sidak, Scheffe and Bonferroni to compare the pooled mean total healthcare expenditure between racial/ethnic groups (24). We compared mean total expenditures between six groups (NHW vs NHB, NHW vs Hispanic, NHW vs Others, NHB vs Hispanic, NHB vs Others, Hispanic vs Others).

We then used generalized linear model (GLM) to estimate the adjusted direct medical expenditures by race/ethnic categories after controlling for confounding factors. The GLM has been widely employed in situations where the assumption of normality of the error term is not satisfied (25). The model addresses the positive skewness of expenditures (26) and allows users to calculate incremental effects and standard errors. The use of GLM has an advantage over log OLS since it relaxes the normality and homoscedasticity assumptions and avoids bias associated with retransforming to the raw scale (27). To control for confounding, socio-demographic factors comorbidities and time trend were included in the model.

The modified Park test was used as a diagnostic test to examine the model fit (27). The results of the modified Park test verified the use of a gamma distribution with a log link was the best–fitting GLM to get consistent estimation of coefficients and marginal effects of medical expenditure (28). The gamma model is used for data situations in which the response can take only values greater than or equal to zero (29). Multicollinearity was checked for predictors of the GLM taking in to account the complex survey design. Variance inflation factor (VIF) for all predictors used in the model indicated no multicollinearity problems.

As insurance status and access to care are both possible confounders to overall differences in expenditures, additional post hoc analyses were run to help understand and explain main differences in the original GLM models. In addition to considering total unadjusted expenditure by racial/ethnic categories overall, we estimated the total unadjusted expenditures by racial/ethnic category stratified by different insurance categories (privately insured, publicly insured, uninsured and overall). Finally, we ran adjusted Wald tests to determine if unadjusted mean total expenditures by race/ethnic group differed by access as defined by the access to care variables (usual source of care, medical access and prescription access). We used the Adjusted Wald test to account for skewedness of the cost data.

For all the analyses, we accounted for the complex sampling design of MEPS dataset in order to extrapolate the estimates to a U.S civilian non-institutionalized population. The data were analyzed with STATA 14 (30) and p value < 0.05 was considered statistically significant.

Results

Sample Demographics

Table 1 presents differences in demographic characteristics among adults with diabetes by racial/ethnic categories. Among adults with diabetes, of the participants who were male 68% were NHW, 13% NHB, 13% Hispanic, and 7% were Other. Of the participants who had college education or higher, or had private insurance similar proportions were found: 70 and 71% NHW, 14 and 13% NHB, 7 and 9% Hispanic, 9 and 7% Other. Lastly, of those with poor/negative vs. high income, 49 vs. 76% were NHW, 24 vs. 10% NHB, 20 vs. 7% Hispanic, and 7 vs. 7% Others.

Table 1.

Study sample characteristics

	NHW	NHB	Hispanic	Others	p-value
Sex					<0.0001***
Male	67.91	12.57	12.69	6.83
Female	60.87	18.12	14.23	6.78
Age					<0.0001***
18 – 44	52.62	18.78	20.83	7.77
45 - 64	62.61	16.11	13.89	7.40
≥ 65+	70.29	13.44	10.49	5.77
Marital Status					<0.0001***
Married	68.98	10.52	13.38	7.11
Unmarried	60.36	20.94	12.24	6.46
Never Married	48.35	27.14	18.42	6.09
Education					<0.0001***
<High School	48.90	17.87	27.55	5.68
High School	69.54	15.35	9.71	5.40
≥ College	70.48	13.70	7.19	8.62
Insurance					<0.0001***
Private	71.30	12.90	9.30	6.50
Public	55.53	20.11	17.23	7.13
Uninsured	44.76	16.13	31.22	7.89
MSA					<0.0001***
Non-MSA	79.64	9.72	5.83	4.81
MSA	60.43	16.85	15.42	7.31
Region					<0.0001***
Northeast	67.26	16.19	10.85	5.70
Midwest	78.55	12.43	4.86	4.15
South	61.73	21.53	11.93	4.81
West	52.25	5.94	27.50	14.30
Poverty Income Ratio					<0.0001***
Poor/Negative	48.91	24.00	20.22	6.87
Low Income	56.54	18.47	18.38	6.61
Middle Income	66.19	13.84	13.34	6.62
High Income	75.71	10.15	7.13	7.02
Hypertension					<0.0001***
Yes	64.06	17.07	12.36	6.51
CVD					<0.0001***
Yes	71.55	13.33	9.49	5.63
Stroke					0.0001***
Yes	67.75	17.31	9.83	5.12
Emphysema					<0.0001***
Yes	84.21	7.48	4.79	3.52
Joint Pain					<0.0001***
Yes	68.05	14.95	10.82	6.18
Arthritis					<0.0001***
Yes	69.07	15.26	10.23	5.44
Asthma					0.0045*
Yes	66.29	16.45	10.40	6.86

p-value + ≤0.05

^*≤0.01

^**≤0.001

^***≤0.0001;

MSA – Metropolitan Statistical Area; CVD – Cardiovascular Disease

Unadjusted Mean Expenditures

The unadjusted mean total expenditure differences for race/ethnic categories among adults with diabetes is presented in Table 2 and Figure 1. The unadjusted mean total healthcare expenditures for NHW was $12,972 (95% CI $12,418, $13,527), NHB was $12,400 (95% CI $11,517, $13,282), Hispanic was $8,705 (95% CI $8,089, $9,321), and Other race/ethnic group was $8,613 (95 % CI $7,653, $9,573). The unadjusted mean total healthcare expenditures for NHW increased from $12,486 in 2002/03 to $14,155 in 2004/05 and then decreased to $12,498 in 2010/2011. NHB experienced an increase in mean total expenditure from $11,578 in 2002/2003 to $12,475 in 2004/2005, to $13,501 in 2006/2007; and then a decrease to $12,435 in 2010/2011. Hispanics followed a similar trend with mean total expenditure of $7978 in 2002/2003 increasing to $8544 in 2004/2005, to $9604 in 2006/2007, and decreased to $8568 in 2010/2011. Mean total expenditure varied across years for the ‘Other’ race/ethnicity group. In 2002/2003 mean total expenditure was $9157, $8290 in 2004/2005, $9222 in 2006/2007, and $7025 in 2010/2011.

Table 2.

Mean total expenditure by race (95% Confidence Interval)

	2002/03	2004/05	2006/07	2008/09	2010/11	Pooled
NHW	$12,486.02 (11596.1 – 13375.93	$14155.40 (12680.63 – 15630.16)	$12914.79 (11837.12 – 13992.46)	$12913.07 (11856.49 – 13969.66)	$12498.13 (11498.54 – 13497.72)	$12972.42 (12417.78 – 13527.06)
NHB	$11,578.04 (9862.86 – 13293.22)	$12475.35 (10561.06 – 14389.65)	$13501.02 (11152.71 – 15849.32)	$11912.49 (10641.64 – 13183.34)	$12434.55 (10714.37 – 14154.73)	$12399.56 (11517.18 – 13281.94)
Hispanic	$7978.28 (6900.12 – 9056.45)	$8544.06 (7225.36 – 9862.76)	$9604.04 (8195.52 – 11012.57)	$8627.64 (7468.75 – 9786.52)	$8567.60 (7205.01 – 9930.19)	$8705.38 (8089.47 – 9321.29)
Other	$9157.03 (7557.37 – 10756.68)	$8289.55 (6461.42 – 10117.68)	$9222.23 (6704.42 – 11740.04)	$9576.19 (6962.99 – 12189.38)	$7025.36 (5526.91 – 8523.81)	$8612.72 (7652.93 – 9572.50)

Figure 1:

Total unadjusted expenditures by race/ethnicity for all individuals, private insurance, public insurance, uninsured

Adjusted Mean Expenditures

Table 3 presents the adjusted incremental expenditure difference by race/ethnic categories, with NHW as the reference group. All analyses were adjusted for socio-demographic factors, comorbidity and time trends. The adjusted total expenditures for Hispanic were lower by $1,985 (95% CI: −$2,716, −$1,254) and Other race/ethnic group were lower by $2,590 (95% CI: −$3,304, −$1,877) compared with NHW. Compared with NHW, the adjusted prescription expenditure for NHB were lower by $410 (95% CI: −$591, −$228), Hispanic were lower by $646 (95% CI: −$836, −$455) and Other race/ethnic group were lower by $657 (95% CI: −$878, −$435). Compared with NHW, the adjusted office-based expenditure for Other race/ethnic group were lower by $425. Compared with NHW, inpatient expenditure were lower by $593 for Hispanic and were lower by $824 for other race/ethnic group. The adjusted outpatient and ER expenditure for Hispanics were $372 (p<0.0001) and $51 (p=0.02) less than those for NHW respectively. The adjusted outpatient and ER expenditure for Others were $428 (p<0.0001) and $55 (p=0.05) less than those for NHW respectively.

Table 3.

Generalized Linear Model (GLM) – Incremental effects of healthcare expenditures by racial/ethnic categories among US adults with diabetes (reported as dollars in 2014)

	NHB	Hispanic	Others
Prescription
Incremental Cost	−$410.16***	−$646.00***	−$657.20***
95% CI	−$591.44 - −$228.88	−$836.63 - −$455.37	−$878.71 - −$435.70
P-value	<0.0001	<0.0001	<0.0001
Office-Based
Incremental Cost	$13.89	−$200.62	−$425.00***
95% CI	−$235.76 – 263.53	−$484.09 - $82.86	−$662.90 - −$187.09
P-value	0.91	0.17	<0.0001
In-Patient
Incremental Cost	$240.51	−$593.42+	−$824.15*
95% CI	−$210.63 - $691.65	−$1078.25 - −$108.59	−$1338.21 - −$310.10
P-value	0.30	0.016	0.002
Out-patient
Incremental Cost	−$151.33	−$372.47***	−$428.31***
95% CI	−$308.05 - $5.39	−$526.64 - −$218.30	−$558.11 - −$298.51
P-value	0.058	<0.0001	<0.0001
ER Visit
Incremental Cost	$15.98	−$50.60+	−$54.87+
95% CI	−$39.95 - $71.91	−$93.86 - −$7.35	−$108.76 - −$0.97
P-value	0.58	0.022	0.046
Total
Incremental Cost	−$419.15	−$1985.93***	−$2590.95***
95% CI	−$1121.91 – 283.62	−$2716.97 - −$1254.89	−$3304.71 - −$1877.20
P-value	0.24	<0.0001	<0.0001

Access to Care

Adults in the US who reported having access to a usual source of care had a mean total expenditure of $13208 (95% CI $12630, $13785), compared to those without a usual source of care had mean total expenditure of $8652 ($6924, $10381) (p<0.0001). However, NHW without access to medical care were the only group who had significantly higher ER expenditures, $439 (95% CI $340, $538); compared to those who had access, $286 (95% CI $257, $316) (p=0.003). NHW without access to prescriptions had significantly higher total expenditures, $15600 (95% CI $13710, $17489), compared to those who had access, $12695 (95% CI $12130, $13260) (p=0.003). NHW without access to prescriptions also had significantly higher prescription expenditures, $14417 (95% CI $4080, $4754), compared to those who had access, $3829 (95% CI $3692, $3965) (p=0.001). NHW without access to prescriptions also had significantly higher ER expenditures, $503 (95% CI $372, $634), compared to those who had access, $286 (95% CI $256, $316) (p=0.001). NHB without access to prescriptions had significantly higher prescription expenditures, $4522 (95% CI $3891, $5152), compared to those who had access, $3340 (95% CI $3095, $3585) (p=0.0003). Those from the ‘Other’ group without access to prescriptions had significantly higher total expenditures, $14722 (95% CI $9448, $19996), compared to those who had access, $8049 (95% CI $7177, $8922) (p=0.013). Those from the ‘Other’ group without access to prescriptions had significantly higher prescription expenditures, $4160 (95% CI $3017, $5303), compared to those who had access, $2735 (95% CI $2435, $3035) (p=0.015).

Discussion

This is the first study to our knowledge to examine racial/ethnic differences in multiple categories of medical expenditures in a nationally representative sample of adult patients with type 2 diabetes over time. Results are consistent with previous studies looking at certain populations, or during one year, and show that disparities in healthcare expenditures by race/ethnicity continue to be a problem for the US population. We found that Hispanics and Others had significantly lower expenditures in total expenditures and most expenditure categories compared to NHW; while NHB showed no differences in expenditures with the exception of prescription costs. Hispanics had the largest proportion of uninsured individuals compared to NHW, NHB, and Other race/ethnicity; and Hispanics and Others had the smallest proportion of individuals who had private or public insurance. We also found that not having access to prescriptions resulted in increased prescription expenditures in NHW, NHB, and Others. Based on these results, some of the most important areas to target to decrease disparities in expenditures may be through improving insurance access for Hispanics and Other races, and increasing access to prescription coverage.

Health expenditures offer information on the economic burden of disease and can also provide insight on missed opportunities for engaging the healthcare system if lower expenditure is due to lack of appropriate care (31,32). Overall, NHW had higher medical expenditures than NHB, Hispanics, and Others suggesting that NHW are receiving more care for their disease than minority populations. NHB had significantly lower prescription costs than NHW, indicating prescription coverage may differ between the two groups. Not having access to prescriptions increases prescription costs, which may result in less prescription purchase and thus lower prescription expenditures (16,20). This can be seen in our post-hoc analyses that showed individuals with delays or difficulty in obtaining prescriptions had higher prescription costs overall. These findings are consistent with the results from a study by Hu et al that examined differences in medical expenditures in patients with diabetes (9). Hu et al also found that NHB reported lower mean prescription drug expenditures than NHW (9). Lee et al also found that NHBs and Hispanics had lower health care costs and prescription costs than NHWs (16). Possible explanations for the differences included: 1) differences in health care use, or 2) differences in costs due to differences in health insurance coverage; private vs public. Individuals receiving Medicaid showed lower documented costs than individuals with private insurance (16). Those with Medicare had lower costs than those with private insurance, and it is expected that cost savings from Medicare could lead to increased medication adherence and in turn lead to better diabetes control. Medication coverage should be taken into account when prescribing treatment regimens because of insurance components such as the Medicare doughnut hole that can negatively impact prescription access for low-income patients (33,34). In addition, policy changes to address underinsurance or lack of prescription coverage plans should be developed to enhance the reach of newly developed interventions that will address this component of disparities in care.

Hispanics had significantly lower expenditures in all categories except office-based costs. Based on these results, Hispanics may have access to medical providers for an office-based visit, however they may have insufficient access for other care expenditures. This supports prior suggestions that Hispanics are likely to be underinsured rather than uninsured (35,36). Strategies for increasing access to care should take cultural and social factors into account and should increase the affordability of essential medications such as insulin and oral agents for those who need them (1). Based on these results, interventions that address access to care, and financial barriers should be developed and targeted to reach Hispanics with type 2 diabetes.

The strengths of this study are that a large nationally representative sample was used to collect the data that has been reported, and statistical methods were used that addressed the positive skewness of expenditure data and reduced the likelihood of introducing bias into the data. Study limitations include being unable to determine causality due to the cross-sectional nature of the study. Secondly, the study is at risk for potential bias due to the reliance on memory recall for patients completing the survey. Thirdly, different households were sampled overtime instead of following the same individuals annually, therefore the results of the study may not be used for longitudinal interpretation.

Conclusion

The implications of our study findings suggest that there is a differential impact of medical expenditures, which has continued overtime. Our results showed that access to care is an issue for Hispanics, where prescription coverage specifically is a more critical issue for NHBs. Not having access to prescriptions resulted in an increase in total expenditure, prescription expenditure, and ER expenditure in most race/ethnic groups. Differences in medical expenditures may indicate minority patients do not have the same access to care and prescriptions needed to effectively treat type 2 diabetes. Clinicians and researchers will need to use a multifaceted approach to treat the disease that involves in-office methods combined with medications to effectively manage individuals who may not have a usual source of care, who may only come in once a year, or who may not be able to obtain necessary prescriptions. Possible policy changes include an increase in prescription coverage for patients with type 2 diabetes, reduction in cost and greater availability of generic medications, and increase in insurance coverage for Hispanics and other minority populations.