Abstract
This retrospective cohort study evaluated associations of race/ethnicity and gender with outcomes of diabetes complications severity, health care resource utilization (HRU), and costs among Medicare Advantage health plan members with type 2 diabetes (T2DM). Medical and pharmacy claims were evaluated for 333,576 members continuously enrolled from January 1, 2010, to December 31, 2011, aged 18–89 years, with ≥1 primary diagnosis medical claim, or ≥2 claims with a secondary diagnosis of T2DM (International Classification of Diseases, Ninth Revision, Clinical Modification code 250.x0 or 250.x2). Complications severity assessment by Diabetes Complications Severity Index ranged from 0 (no complications) to 5+. Mean (SD) all-cause medical, pharmacy, and total costs were reported alongside all-cause HRU by place of service (outpatient, inpatient, emergency room [ER]) and number of visits. Multivariate regression showed being Hispanic, black, or male was associated with higher prevalence of more severe complications. This racial/ethnic disparity was more pronounced among females, among whom odds of having more severe complications were higher for Hispanic and black as compared to white females [(Hispanic vs. white odds ratio [OR], 1.40; 95% confidence interval [CI], 1.32–1.48), and (black vs. white OR, 1.22; 95% CI, 1.19–1.25)]. Regardless of gender, blacks had more ER visits than whites. White females incurred the highest total health care costs (mean annual costs: $13,086; 95% CI, $12,935-$13,240, vs. Hispanic females: $10,732; 95% CI, $10,406–$11,067). These effects held regardless of other demographic and clinical attributes. These findings suggest racial/ethnic and gender differences exist in certain T2DM clinical and economic outcomes. (Population Health Management 2015;18:115–122)
Introduction
R
Studies suggest that black and Hispanic individuals are at an increased risk of developing T2DM as compared to non-Hispanic whites. In addition, they often have poorer outcomes, and tend to experience suboptimal quality of care. 1 –7 Moreover, their rates of T2DM incidence, complications, and mortality are up to 3 times higher than those of non-Hispanic whites. 2,3,5,11
In addition to racial/ethnic differences, gender differences may increase an individual's susceptibility for developing T2DM and its related complications. 12 –16 Women diagnosed with obesity and T2DM are more likely to experience emotional distress associated with T2DM, including depression and anxiety, and are more likely to have a poorer quality of life, use more health care services, and have a shorter life span as compared to men. 14 –18
An estimated $174 billion was spent on the treatment of diabetes in the United States in 2007. 19 These costs are expected to swell to $514 billion by 2025 as the US population ages. 19 T2DM-related health care costs can escalate because of T2DM-related complications 20 such as cardiovascular disease, 21,22 chronic kidney disease, 23,24 and visual impairment. 25 This study's aim was to understand whether racial/ethnic and gender differences were affecting clinical and economic outcomes in a large cohort of Medicare Advantage health plan members. The study evaluated the extent to which members' race/ethnicity and gender were associated with T2DM complication severity, health care resource use (HRU), and health care costs.
Methods
Study design
This retrospective cohort analysis evaluated the associations of race/ethnicity or gender with diagnosed T2DM complications, HRU, or costs. Medical and pharmacy claims, along with enrollment data for 333,576 Medicare Advantage members with T2DM from Humana (a large national health plan) were used. This was part of a larger study approved by an independent Institutional Review Board.
Study setting, time period, and selection of study cohort
The cohort's medical and pharmacy claims were assessed during the 24-month period of January 1, 2010, to December 31, 2011. Participants were Humana members enrolled in the Medicare Advantage with Prescription Drug coverage plan, continuously enrolled during the 24-month assessment period, aged between 18 and 89 years as of January 1, 2010, and diagnosed with T2DM, defined as having at least 1 medical claim with a primary diagnosis of T2DM (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 250.x0 or 250.x2) or at least 2 claims with a secondary diagnosis of T2DM. Members were excluded if they had a claim for type 1 diabetes (ICD-9-CM code 250.x1 or 250.x3), or a claim for gestational diabetes (ICD-9-CM code: 648.8) or pregnancy (630.xx-679.xx or v22.x-v24.x).
Primary outcome measures
Primary outcome measures included diabetes complications severity, all-cause HRU, and health care costs. Complications severity was measured using the Diabetes Complications Severity Index (DCSI), 26 a claims-based measure of complications severity that predicts HRU and health care costs in managed care patient populations. 27 The DCSI scoring is based on the presence of ICD-9-CM codes for diabetic complications in the categories of cardiovascular, nephropathy, retinopathy, peripheral vascular, stroke, neuropathy, and metabolic complications. Each complications category that is present in the claims data is scored as a 1 or 2 (1=some abnormality, and 2=severe abnormality). For example, in the cardiovascular category, a diagnosis of atherosclerosis is scored as a 1 while myocardial infarction is scored as a 2. A DCSI summary score may range from 0 to 13. The unmodified DCSI is based on a summary score derived from diagnostic and laboratory data. A recent DCSI modification (used in this study) demonstrated that this DCSI version, omitting laboratory data, can be used to explain concurrent medical costs in a managed care setting. 27 The study cohort was divided into 6 subgroups based on DCSI scores: 0 (no complications) through 5+(score of 5 or more).
All-cause HRU and costs included all medical and pharmacy claims, irrespective of diagnosis or therapeutic class for the respective medications. All-cause HRU was measured for both study years. Outpatient and inpatient hospitalization use were defined as follows: If the place of service was a physician's office or outpatient facility, the use was considered to be outpatient; if it was an inpatient hospital facility, the use was considered to be inpatient hospitalization. When discharge and admittance dates were the same, the claim was considered a single hospitalization. Emergency room (ER) use was defined as all medical claims for which the place of service was an ER; however, ER claims with the same service date as a hospitalization or adjacent to a hospitalization were counted as a hospitalization only.
Total overall health care costs (reimbursements and co-pays) were defined as the sum of medical and pharmacy costs, and medical costs as the sum of costs associated with medical claims for outpatient, inpatient, or ER visits. Pharmacy costs were defined as the sum of costs associated with pharmacy claims. All medical and pharmacy claims were examined to assess total out-of-pocket medical and pharmacy spending. Costs were adjusted to 2011 dollars using the US Bureau of Labor Statistics' Consumer Price Index for Medical Care.
Study variables
Independent variables
Members' race/ethnicity was captured from Humana's Medicare member claims data as of January 1, 2010, and derived from Medicare's enrollment database, categorized as white, black, Hispanic, or Other. Medicare acquires race/ethnicity data primarily from the Social Security Administration, to which individuals self-report their race/ethnicity when applying for a social security card. 28 Members' gender (male or female) was captured from self-reported member enrollment data as of January 1, 2010.
Demographic covariates
The effects of the following demographic characteristics were controlled for when assessing the effects of race/ethnicity and gender on the study outcomes: Age was calculated from members' date of birth as of January 1, 2010. Geographic region was based on the US Census Bureau assignment of states to a geographic region (Northeast, Midwest, South, and West) as of January 1, 2010. Low-income subsidy (LIS) eligibility is the eligibility of Medicare beneficiaries with income below 150% of the poverty level and limited resources to qualify for additional premium and cost-share assistance for prescription drugs.
Clinical covariates
The effects of certain clinical characteristics also were controlled for in the study. Chronic disease burden was measured with the unweighted RxRisk-V Score, a prescription claims-based comorbidity index. The RxRisk-V is calculated for each member by mapping comorbid conditions to drug classes and individual drugs via Medi-Span generic product identifier codes. 29 Treatment by antidiabetic therapeutic classes was defined as the proportion of members with ≥1 pharmacy claim during the 24-month period for a medication within one of the following respective medication classes: biguanides, sulfonylureas, thiazolidinediones, amylin agonists, meglitinides, alpha-glucosidase inhibitors, glucagon-like peptide-1 receptor agonists, dipeptidyl peptidase-4 inhibitors, bile acid sequestrants, dopamine-2 agonists, insulin, and combination oral antidiabetics.
A hypoglycemic event was defined as having at least 1 outpatient claim with a diagnosis of hypoglycemia (ICD-9-CM codes: 251.0, 251.1, 251.2, and 250.8 [excluding claim with co-diagnoses of 259.8, 272.7, 681.xx, 682.xx, 686.9x, 707.1–707.9, 709.3, 730.0–730.2, and 731.8 in the same visit or hospitalization]), or at least 1 ER claim with a diagnosis of hypoglycemia in the primary position during the 24-month study period. Depression was defined as a member having at least 1 claim with an ICD-9-CM diagnosis of depression (296.2x, 296.3x, 300.4, and 311).
Data analysis
Regression analyses were used to assess the associations of race/ethnicity or gender with severity of T2DM complications, HRU, and costs. The cohort's demographic and clinical characteristics were assessed with univariate descriptive statistics. Medication use was evaluated using mean number of antidiabetic drug classes.
All-cause HRU was reported descriptively by place of service. Overall HRU among members was reported by count and percentage (ie, as a dichotomous status variable indicating the proportion of total members utilizing the specific service type). Mean (standard deviation [SD]) number of health care visits was calculated annually.
All-cause health care costs were reported descriptively by cost component. Mean (SD) all-cause medical, pharmacy, and total health care costs (medical+pharmacy) are reported. Chi-square analyses, analysis of variance pairwise comparisons, and t tests were used to assess significant differences based on demographic and clinical characteristics.
Regression models were used to evaluate the association of race/ethnicity and gender with DCSI score, HRU, and costs, controlling for demographic and clinical characteristics. HRU measures included in the models were number of outpatient visits, inpatient stays, and ER visits; the cost measures included total health care, medical, and pharmacy costs. The HRU and cost measures were averaged over a 2-year time period (2010–2011). Regression models were adjusted for age, region, LIS status, RxRisk-V score, diagnosis of depression, diagnosis of hypoglycemia, and number of unique antidiabetic drug classes.
Linear regression models were not used because HRU and cost data were nonnegative with skewed distributions (ie, a small number of high utilizers). Ordinary least squares linear models assuming normal distribution may yield imprecise estimates of means. Therefore, generalized linear models (GLMs) were employed, allowing for non-normal distribution, nonconstant variance, and different types of dependent variables. For HRU, GLM with negative binomial distribution and log link was used, and for health care costs measures, GLM with gamma distribution and log link function was used. All analyses were performed using SAS Enterprise Guide 5.1 statistical software (SAS Institute Inc., Cary, NC), and statistical significance was determined at the 0.05 level.
Additional analyses (not shown) indicated the vast majority of the cohort (84.9%) was aged 65 years and older. The above-mentioned analyses were conducted on the age 65+cohort exclusively, and the results were very comparable to those of the entire cohort. As a result, the data for the entire cohort, aged 18–89 years, were included in the data reported for this study.
Results
Table 1 shows the number of members who met the study inclusion criteria (333,576 members), and depicts the cohort's demographic and clinical characteristics, stratified by race/ethnicity. The sample comprised more than 45% males, and was predominantly white, followed by black, Other, and Hispanic. Members in the Other category were younger compared to the others, and there were fewer males in the black group. There was greater black and Hispanic representation in the Southern United States compared to other regions under examination. More blacks and Hispanics qualified for LIS. Depression rates were high overall, but markedly lower among black and Other members, and highest among Hispanic members. The rates of hypoglycemic events were higher among blacks and Hispanics. Women had higher rates of LIS eligibility (8.2% vs. 5%; P<001), much higher rates of diagnosed depression (24.7% vs. 14.2%; P<001), and lower rates of hypoglycemia compared to men (6.7% vs. 7.5%, P<001; data not shown).
Tables 2 and 3 contain the descriptive results of DCSI groupings and all-cause HRU and costs by race/ethnicity and gender for the combined calendar years of 2010 and 2011. Most members had either no DCSI complications or the highest severity (DCSI=5+). The Other group had the highest percentage of members with no complications, whereas the Hispanic group had the lowest rate (Table 2). Hispanic members had the highest representation in the more severe DCSI, followed by blacks. With regard to HRU, the Other group had fewer outpatient, inpatient, and ER visits, whereas Hispanic and black members had more ER visits compared to whites. Total health care and medical costs were the highest for whites. As reported in Table 3, a higher percentage of female members had no complications and a lower percentage had a DCSI score of 5+, whereas the reverse was the case for males (P<.001 for both), and females showed higher HRU across all 3 domains, and higher pharmacy costs.
Tables 4 and 5 contain the results of the regression models of the associations of gender and race/ethnicity with DCSI and health care costs, controlling for other clinical and demographic variables. Among males, the odds of having a higher DCSI score were higher for members who were Hispanic compared to white, and higher for those who were black compared to white. Among females, the odds of having a higher DCSI score were higher for Hispanics (OR, 1.396; 95% CI, 1.32-1.48), and higher for blacks as compared to whites. The findings in Table 4 indicate that regardless of gender, Hispanic and black members were more likely to have higher DCSI scores than white members, and the observed racial/ethnic differences were more pronounced among females than among males. The analyses also indicate that a diagnosis of depression was associated with an 84% increase in the odds of having more severe complications (OR, 1.8350; P<.001), and hypoglycemic events were associated with a 187% increase in those same odds (OR, 2.869; P<.001; data not shown).
The results in Table 5 suggest that regardless of gender, whites incurred the highest total costs of all groups. Among all males, being Hispanic or black was associated with a 10.4% and 7.5% increase in total costs, respectively; a similar pattern held for medical and pharmacy costs. A diagnosis of depression was associated with a 42.1% increase in total costs, and a diagnosis of hypoglycemia was associated with a 27.2% increase (data not shown).
With regard to HRU, men were less likely to use outpatient, inpatient, and ER services than women (data not shown). Hispanic and black males were 13% and 11% more likely to use inpatient services, and 22% and 9% more likely to use ER services when compared to all other combined gender-race/ethnic member categories. A diagnosis of depression was associated with an 18% increase in outpatient services, a 75% increase in inpatient services, and an 83% increase in ER use. A similar pattern of increased HRU was observed for hypoglycemic events.
Discussion
Previous research has demonstrated that racial/ethnic and gender differences can impact the likelihood of patients developing T2DM-related complications, 30 –32 and can affect the manner in which patients with T2DM utilize and access health care resources. 33 –35 The findings of the present study support those reports, and suggest that these differences in clinical and economic outcomes can occur regardless of a patients' age, comorbidities, or whether or not he or she is LIS eligible.
The findings of higher prevalence of severe T2DM complications among Hispanics and blacks, and among black and Hispanic females, are consistent with research demonstrating that these patients may be at greater risk for more severe complications because of challenges with medication adherence and managing blood glucose, blood pressure, and low-density lipoprotein cholesterol. 1,36 These self-management goals are difficult for most patients to achieve, and existing studies suggest that these challenges may be caused by psychosocial and environmental factors including low health literacy, concerns about medication side effects, inadequate social support, inadequate patient-clinician communication, and suboptimal access to health care. 36 The exact reasons for these challenges are not entirely clear and warrant further study.
Consistent with this study's findings on gender and race/ethnicity, Finkelstein et al found that Medicare recipients diagnosed with T2DM and major depression were more likely to be female and to be Hispanic compared to other racial/ethnic groups, particularly blacks. 37 The exact reasons for these patterns are unknown and warrant more research, given the greater health care costs and utilization associated with diagnoses of depression and T2DM. Also of concern is this study's finding that a diagnosis of depression was linked to a marked increase in the odds of having a higher DCSI score.
The findings concerning resource use support previous work suggesting that black and Hispanic patients are more likely to utilize ER and inpatient services as compared to whites. 38,39 Studies of T2DM and other chronic diseases suggest that black patients may overuse ER resources because of inadequate access to primary care physicians, poor medication adherence, and having a poor understanding of their disease. 33,38 Enhanced access to primary care may increase the availability of preventive care, which in turn may reduce ER visits and inpatient hospital stays. 33,38,39
This study also found that whites incurred the highest total health care and medical costs, and in contrast, being black or Hispanic was associated with lower all-cause health care costs. Although few studies have explored this issue, a cross-sectional study of insured patients with T2DM by Lee and colleagues indicated that white patients may incur higher overall costs than black or Hispanic patients, and that the differences may be related to racial/ethnic differences in patterns of health care use. 38 The authors also surmised that the different utilization patterns could have been the result of differences in health care coverage, easier access by whites to more costly health care resources, and variations in supplier-induced demand.
The cost-related findings of the present study may be related to the above-mentioned underuse of primary care resources on the part of minority patients. 33 Some researchers posit that in order to increase minority patients' use of preventive care services, programs such as patient education, nurse-led disease management programs, and matching of primary care patients with clinicians who are proficient in their language and culture are warranted. 33 –36 For example, a study by Davidson et al demonstrated that a nurse-directed diabetes disease management program resulted in a lower incidence of ER and urgent care visits as compared to usual care in a minority population receiving care at a county health clinic. 34
Health care professionals may need to address these T2DM management challenges with Hispanic and black patients as early in treatment as possible by encouraging them to attend follow-up outpatient visits and to adhere to their medication and lifestyle regimens. This attention may increase the likelihood that these patients' progress would be followed more closely so as to prevent or delay the onset of complications, which can be costly. From a health plan perspective, educational programs to improve patients' health literacy and to address cultural barriers to effective clinician-patient communication may be warranted.
Limitations
This study has limitations to consider when interpreting its findings. First, the identification of race/ethnicity using the Medicare enrollment database may lend itself to ambiguity, and may be insufficient for coding of populations other than black, white, and to some degree, Hispanic individuals.
28,40
The subsample of Hispanic patients was relatively small (2.3% of the study population) and was concentrated in the Southern United States. Thus, the results may have limited generalizability to other Hispanic Medicare populations. Nevertheless, the present study's findings appear to be consistent with those of other studies of health disparities that used diverse and adequately sized samples.
37
–39
Further, the study team did not comprehensively assess the outcomes of the Other racial/ethnic category, which likely comprised several subgroups including Asians and American Indians, among whom disparities in T2DM-related outcomes have been identified previously.
41,42
The likely small sizes of individual Other subgroups would have posed challenges to drawing reliable conclusions about health disparities. Some researchers posit that Medicare claims coding can be enhanced with the Spanish Surname list or geocoding.
40,43
Second, the study sample was not limited to patients who were newly diagnosed with T2DM. Because of the study's cross-sectional design, there was no predetermined baseline period prior to the 24-month study period. Therefore the sample included both incident and prevalent cases of T2DM. Readers should be cautious when generalizing the results of this study to other T2DM populations. Third, the DCSI had not been validated on a Medicare population prior to the present study. Although the DCSI had been validated on younger commercially insured populations, the index seems to be appropriate for use with Medicare populations as it has shown comparable robustness in predicting health care utilization and costs in the present study. A fourth limitation is that the study's use of administrative claims exposes it to threats to validity, including errors in claims coding, missing data, lack of data on indirect cost and cash purchas
Conclusions
The study's findings suggest a need to better understand and address health disparities among patients with T2DM. Consistent with previous studies, differences in race/ethnicity and gender were found in T2DM complication severity, use of ER services, and health care costs. These differences were evident in spite of other demographic and clinical attributes such as age, comorbidities, and low-income eligibility status. Overall, the total health care costs for Hispanics and blacks were lower than those of whites, but a segment of black and Hispanic patients, males in particular, incurred greater costs as well as higher use of ER services. Health plan officials should evaluate the potential for investments in interventions to reduce these racial/ethnic and gender differences in T2DM clinical and economic outcomes, which may ultimately reduce costs.
Footnotes
Author Disclosure Statement
Drs. Hazel-Fernandez, Li, Nero, Moretz, Slabaugh, Meah, Costantino, Patel, Ms. Baltz, and Mr. Bouchard declared the following potential conflicts of interest with regard to the research, authorship, and/or publication of this article: This study was sponsored by Novo Nordisk, Inc. Drs. Hazel-Fernandez, Moretz, Slabaugh, Costantino, and Patel are employees of Comprehensive Health Insights, a Humana company, which received funding from Novo Nordisk to conduct this study. Drs. Li and Nero, former employees of Comprehensive Health Insights, are now employed by Cigna and Magellan Health Services, respectively, and declare no conflicts of interest. Dr. Meah is a Humana employee who declares no conflicts of interest. Ms. Baltz was an employee of Humana at the time that this manuscript was prepared and declares no conflict of interest. Mr. Bouchard is an employee and stockholder of Novo Nordisk Inc., which provided funding for this study.
Prior Presentation
An abstract of these findings was accepted for presentation at the International Society of Pharmacoeconomics and Outcomes Research 19th Annual Meeting, June 3, 2014, in Montreal, Quebec, Canada.