Abstract
By March 2015, 30% of all Medicare beneficiaries were enrolled in Medicare Advantage (MA) plans. Research to date has not explored the impacts of MA market penetration on individual or population health outcomes. The primary objective of this study is to examine the relationships between MA market penetration and the beneficiary’s portfolio of cardiometabolic diagnoses. This study uses 2004 to 2008 Medical Expenditure Panel Survey (MEPS) Household Component data to construct an aggregate index that captures multiple diagnoses in one outcome measure (Chronic Disease Severity Index [CDSI]). The MEPS data for 8089 Medicare beneficiaries are merged with MA market penetration data from Centers for Medicare and Medicaid Services (CMS). Ordinary least squares regressions are run with SAS 9.3 to model the effects of MA market penetration on CDSI. The results suggest that each percentage increase in MA market penetration is associated with a greater than 2-point decline in CDSI (lower burden of cardiometabolic chronic disease). Spill-over effects may be driving improvements in the cardiometabolic health of beneficiary populations in counties with elevated levels of MA market penetration.
Keywords: cardiometabolic conditions, diabetes, cardiovascular disease, MEPS, Medicare Advantage, market penetration, spill-over effects
Introduction
Managed care is a common insurance form. In fact, 8 in 10 individuals covered by employer-sponsored private insurance, and one-third of individuals with Medicare, are enrolled in a managed care plan.1,2 Given the prominence of managed care as an insurance form, understanding the extent to which growth in managed care has influenced service utilization, health outcomes, and spending is essential to strong policy development.
This is what we know. Greater managed care market penetration is associated with better inpatient outcomes, including lower utilization of unnecessary inpatient procedures, a reduction in inpatient complications, and lower mortality posthospital discharge.3–5 Managed care is also linked to higher rates of prevention-oriented processes of care, including vaccinations and disease screenings among the general population.6–8
There is less agreement on the economic spill-over effects of Medicare Advantage (MA) market penetration.9–12 Some studies have found higher rates of Medicare managed care market penetration associated with reduced individual-level costs for the fee-for-service Medicare program.9 A more recent study found increases (or no savings) in total Medicare costs as Medicare managed care market penetration increases.10–12
Little evidence exists regarding the health effects of managed care market penetration. Studies suggest that strategies implemented by health care providers that contracted with managed care organizations (MCOs) spill over to patients who are not members of MCOs, particularly in terms of health service use and expenditures.3,6,7,13 If true, patients’ chronic conditions may be identified more quickly and controlled, leading to slower progression of the disease, even for those individuals not enrolled in managed care.6,9,10,13,14,15
The focus of this study is the association between MA market penetration and Medicare beneficiaries’ portfolio of cardiometabolic diseases. We hypothesized that greater county-level MA market penetration is associated with lower individual-level cardiometabolic disease complexity.
Methods
Data
In order to understand the associations between MA market penetration and the beneficiary’s portfolio of cardiometabolic diagnoses, we merged 2004 to 2008 MA market penetration data from the Centers for Medicare and Medicaid Services (CMS) with restricted data from the Medical Expenditure Panel Survey (MEPS) Longitudinal Files for MEPS panels 9 to 12. Beneficiary state and county of residence were used as the matching indicators.
CMS calculates MA market penetration as the percentage of Medicare beneficiaries in each county that has elected to enroll in MA plans. Our analysis used the market penetration data in 2 ways. First, we used the market penetration data in a continuous variable form, with a 1% (0.01) interval. Second, MA market penetration was divided into quartiles (Q1: 0%-3.00%, Q2: 3.01%-9.69%, Q3: 9.70%-28.19%, and Q4: 28.20%-54.77%).
The MEPS Household Component (MEPS-HC) is a rich data set with many variables relating to presence of disease, insurance coverage, and sociodemographic characteristics. The MEPS is a nationally representative survey of noninstitutionalized individuals residing in households sampled from the previous year’s National Health Interview Survey. Individuals surveyed in MEPS are grouped into panels, and each panel is surveyed 5 times over 2 years. A new panel begins each year, resulting in the overlapping panel design (one panel’s first year is concurrent with the previous panel’s second year).16
We narrowed our analysis to only those MEPS-HC respondents who were Medicare beneficiaries (mcrevy1=1 and mcrevy2=1 in panel 9 to 12 longitudinal files). Respondents aged 65 to 69 years were selected as the reference age-group. Having just aged on to Medicare, we hypothesized they would have lower chronic disease burdens than older beneficiaries. Beneficiaries younger than 65 typically have Medicare due to disability, making this a suboptimal reference group.
The MEPS-HC includes Priority Conditions questions that ask respondents about a number of prevalent conditions. We used the Priority Condition variables for cardiometabolic conditions (ie, heart attack, coronary heart disease, angina, stroke/transient ischemic attack, diabetes, high blood pressure, and high cholesterol) and additional MEPS-HC variables for diabetes-related eye or kidney problems, as well as physical limitations.
In consultation with a team of internal medicine and family physicians (D. Barrett, MD, B. Godek, MD, and P. Latta, MD, 2011), we assigned a 1 to 10 clinical severity weight to each diagnosis (Table 1). These Chronic Disease Severity Index (CDSI) weights are additive, and beneficiaries with multiple comorbidities and additional diagnoses at year 2 of the MEPS survey have higher CDSI scores than in their first year. For example, an individual may enter the MEPS survey panel with high blood pressure, high cholesterol, and type 2 diabetes and receive 1, 2, and 5 points, respectively, for a total CDSI score of 8. If that individual were to have a stroke in year 2 of the panel, 7 more points would accrue, for a total CDSI score of 15 (Table 1).
Table 1.
Components of the CDSI Progression Scale.
| Diagnosis CDSI | Point Value |
|---|---|
| Myocardial infarction (heart attack) | 10 |
| Coronary heart disease | 8 |
| Angina | 8 |
| Stroke/TIA | 7 |
| Diabetes mellitus | 5 |
| “Walk-limit” physical limitationa | 3 |
| Diabetes-related eye or kidney problem | 3 |
| Multiple diagnoses of high-blood pressure | 2 |
| High cholesterol | 2 |
| High blood pressure (first time diagnosis) | 1 |
Abbreviations: CDSI, Chronic Disease Severity Index; TIA, transient ischemic attack or “mini-stroke.”
aAlthough not a specific diagnosis, physical limitation is included along with MEPS respondents’ other self-reported conditions and health problems.
The CDSI is particularly useful for analysis of survey data where diagnosis is based on self-report, rather than biometrics or health records. Used at a population level, the CDSI provides a rich perspective on the overall chronic disease portfolio. Descriptive statistics for the variables used in this study are presented in Table 2.
Table 2.
Descriptive Statistics.
| Variable | Percentage | Variable | Percentage | |
|---|---|---|---|---|
| Age-groups | MCO | |||
| <50 | 6.48 | Had MCO at end of first yeara | 14.14 | |
| 50-64 | 12.40 | Other Insurance | ||
| 65-69a | 24.14 | Had other insurancea | 51.27 | |
| 70-74 | 19.27 | Education | ||
| 75-79 | 17.07 | <High school | 25.86 | |
| 80-84 | 12.40 | High school | 35.48 | |
| ≥85 | 8.24 | Some college | 18.42 | |
| Gender | 4-year college | 11.40 | ||
| Femalea | 56.62 | >4-year collegea | 8.84 | |
| Race/ethnicity | Panel | |||
| Whitea | 84.77 | Panel 9 (2004-2005)a | 24.37 | |
| Hispanic | Panel 10 (2005-2006) | 24.37 | ||
| Not hispanica | 92.80 | Panel 11 (2006-2007) | 25.30 | |
| Income (% FPL) | Panel 12 (2007-2008) | 25.96 | ||
| <100% FPL | 13.54 | MA market penetration (MPen) | ||
| 100%-125% | 6.84 | Quartile 1: 0.00%-3.02% | 24.96 | |
| 126%-200% | 17.27 | Quartile 2: 3.02%-9.70% | 25.05 | |
| 201%-399% | 29.02 | Quartile 3: 9.70%-28.2% | 25.00 | |
| ≥400% FPLa | 33.33 | Quartile 4: >28.2%a | 25.00 | |
| Urban/rural | Mean | Range | ||
| Urbana | 79.28 | Continuous MA MPen | 15.6% | 0%-54.77% |
| CDSI scores | ||||
| Marital status | Year 1 | 8.73 pts. | 0-49 pts. | |
| Marrieda | 53.00 | Year 2 | 7.98 pts. | 0-49 pts. |
Abbreviations: CDSI, Chronic Disease Severity Index; FPL, federal poverty level; MA, Medicare Advantage; MCO, managed care organization; MPen, market penetration; pts: points.
aReference groups.
Empirical Model
Ordinary least squares multivariate linear regression was used to investigate the association between the CDSI variable and the MA market penetration. In addition to the main independent variable of interest, MA market penetration, the model also includes variables that have been shown to influence health, such as income, race, ethnicity, rurality, age, and sex. Individuals surveyed in MEPS were asked a series of questions relating to their health insurance.
Many of these questions ask about managed care. We created a composite managed care variable (MCO_y1) from the following MEPS-HC variables: mcdhmoy1 (covered by a Medicaid or Children’s Health Insurance Program HMO), prvhmoy1 (covered by a private HMO), or phmonpy1 (covered by an HMO, whether it pays nonplan doctors). If an individual responded yes to any of these 3 managed care questions, they were coded as having managed care type insurance. A Medicare managed care variable (mcrpho) was added in 2006 but did not exist in 2 of the 5 years used in this study (2004 and 2005).
The MEPS-HC also asks respondents about private insurance coverage. For those reporting coverage by private insurance at any time during the year in question (privaty1=1 or prvevyy1=1), we coded the private insurance dummy variable (otrins_y1) used in our models.
There are 4 sets of results that reflect the treatment of the market penetration variable and the time of measurement. Market penetration is treated as a continuous variable and then separately as a categorical variable. In addition, MA market penetration is measured at 2 different points in time, once at the end of year 1 and once at the end of year 2.
The data were analyzed using SAS 9.3, which allows for the analysis of data with complex survey sampling design. In our analyses, we used weights provided by MEPS to ensure that the data were representative of the US civilian, noninstitutionalized Medicare population at the time the data were collected.
Results
Descriptive Analysis
The population for this study included 8089 Medicare beneficiaries who participated in MEPS between 2004 and 2008 (panels 9-12). The data had roughly equal representation from each of the 4 panels. Women made up 56.6% of the sample, and 53% were married (Table 2). Nearly 85% were identified as white, and about 7% were Hispanic. Nearly 1 in 5 were younger than 65 years (18.8%), which is slightly more than the national average (17%), and 20.6% were at least 80 years of age. One-third resided in a household with income higher than 400% of the federal poverty level. Although more than 25% did not receive a high school education, almost 20% of the sample had a college education or greater. Nearly 80% of respondents lived in urban areas (as defined by the US Census Bureau). Slightly more than half had some insurance coverage besides Medicare. About 14% had coverage through MCOs. When MA market penetration is considered as a continuous variable, the county-level penetration rate ranged from 0% to 54.77%, with a mean of 15.59% (s = 0.51%). The CDSI scores at the end of year 1 ranged from 0 to 49, with a mean of 8.73. At the end of year 2, the range was 0 to 49, with a mean of 7.98.
Multivariate Regression Results
The results in Table 3 indicate a strong association between higher MA market penetration and lower burdens of cardiometabolic chronic disease (smaller CDSI scores in high-penetration counties).
Table 3.
CDSI Ordinary Least Squares Multivariate Regression Model.a
| Variable | End of Year 1 | End of Year 2 | Variable | End of Year 1 | End of Year 2 |
|---|---|---|---|---|---|
| bCounty MA market penetration rate (MA Pen. Rate) | Demographics | ||||
| Continuous | −2.12e | −2.43e | Not married | 0.26e | 0.25f |
| cAlternate model Using MA Pen. rate quartilesd | Male | 1.66e | 1.60e | ||
| Quartile 1 | 0.81e | 1.17e | Nonwhite | 0.00 | 0.11 |
| Quartile 2 | 0.62e | 0.57e | Hispanic | 0.05 | 0.65e |
| Quartile 3 | 0.38 | 0.37e | Rural/non-MSA | −0.42e | −0.44e |
| MEPS panelsd | Educationd | ||||
| Panel 12 | 2.48e | 1.99e | <High school | 2.07e | 1.44e |
| Panel 11 | 1.19e | 0.19e | High school | 0.72e | 0.25e |
| Panel 10 | 1.12e | 0.06 | Some college | 1.43e | 1.23e |
| 4-year college | 0.11 | 0.19 | |||
| Insurance | |||||
| No supplement to Medicare | 0.59e | 0.38e | Age-groupsd | ||
| No managed care coverage | −0.92e | −0.59e | <50 | −3.20e | −3.52e |
| 50-64 | 2.02e | 1.64e | |||
| Income (%FPL)d | 70-74 | 1.31e | 1.43e | ||
| <100% | 0.61e | 1.03e | 75-79 | 2.50e | 2.64e |
| 100%-125% | 1.13e | 1.04e | 80-84 | 2.68e | 2.91e |
| 126%-200% | 0.89e | 1.21e | 85+ | 1.92e | 2.52e |
| 201%-399% | 0.81e | 0.78e | |||
| Intercept | 3.49e | 5.22e | |||
Abbreviations: CDSI, Chronic Disease Severity Index; FPL, federal poverty level; MA, Medicare Advantage; MEPS, Medical Expenditure Panel Survey; MPen, market penetration; MSA, metropolitan statistical area.
aPositive coefficients are interpreted as the increases in CDSI versus the reference group.
bResults shown for regression model using MA market penetration continuous variable.
cMA market penetration quartile coefficients added to this table for comparison. Other covariates did not differ, qualitatively, from the model using MA market penetration as a continuous variable.
dReference groups for multilevel categorical variables are high MA market penetration rate (fourth quartile), panel 9, incomes ≥400% FPL, ≥4-year college, and ages 65 to 69 years.
e p-value <0.01.
f p-value <0.05.
For example, as indicated in Table 3, when market penetration is a continuous variable, a 1-percentage-point increase in MA market penetration was associated with a 2.12-point decline in CDSI at the end of year 1 and 2.43 point decline at the end of year 2 (P = .0012 and P = .0017, respectively). We reran the analysis using dummy variables for each quartile of MA market penetration, assessing significance using an α of 0.10. Compared to the fourth quartile (those living in the 25% of counties with the highest MA market penetration), quartiles 1 to 3 had higher CDSI scores, and all differences were statistically significant (P = .001-.049).
Variables that had a statistically significant association with higher chronic disease scores include low income, male, single, lower education, lack of supplemental insurance, and increased age. Rural respondents and individuals without managed care type coverage had lower CDSI scores (lower burdens of chronic cardiometabolic disease). For rural residents, this may indicate a lower chronic disease burden or may reflect lower access to health care services, resulting underdiagnosis of chronic conditions. Beneficiaries without managed care had CDSI scores approximately 0.4 points lower than those reporting they had managed care. This result may be indicative of healthy individuals self-selecting into traditional Medicare and choosing either not to buy supplemental coverage or to purchase unmanaged Medigap policies.17
Discussion
This research is the first since the Medicare Modernization Act (MMA) to explore the relationships between the market penetration of MA plans and the prevalence of cardiometabolic chronic diseases among Medicare beneficiaries. A CDSI scale was constructed to represent the beneficiary’s overall chronic disease portfolio for survey or claims-based data.
The results from this analysis suggest that greater county-level MA market penetration is associated with lower individual-level cardiometabolic disease complexity. The results and methods used do not allow for a causal conclusion. That said, it may be the case that the MA program, through disease management programs, prevention and wellness initiatives, or other beneficiary outreach, is producing the desired health outcomes in the counties in which it is most highly penetrated. If true, then it warrants policymakers to consider MA an important public health program.
Alternatively, MA plans may be engaging in risk selection, seeking greater market penetration in areas of lower chronic disease prevalence, and attempting to attract disproportionately healthier beneficiaries. Previous research has examined the behavior of MCOs and demonstrated their risk selection activities designed to enroll lower risk individuals.18,19 However, those studies predated the MMA, which risk-adjusted CMS payments to MA plans for enrolling beneficiaries with chronic diseases (including the cardiometabolic diagnoses considered in this analysis).
As MA enrollment continues to climb, our findings support the call for more extensive research on the mechanisms that are driving lower cardiometabolic disease prevalence in higher penetration MA markets. This study was limited by the consolidation of MA plan types into 1 MA variable and by the use of a new outcome measure (CDSI). Future work includes testing the CDSI measure using multiple data sets and conducting a longitudinal study that controls for self-selection. Future work will also test different measures of chronic disease severity and will control for different MA plan types.
Author Biographies
Steven W. Howard, Ph.D. earned his PhD at Oregon State University (Corvallis, Oregon USA), and MBA at the University of Oregon (Eugene, Oregon USA). He currently serves as Assistant Professor on the faculty at Saint Louis University in the Department of Health Management and Policy, and in the Center for Outcomes Research (St. Louis, Missouri USA). Prior to academia, Dr. Howard worked in the Medicaid and Medicare managed care industry.
Stephanie Lazarus Bernell, PhD. received a B.A. and M.A. in Economics from the American University in Washington, DC, USA. She subsequently received a PhD in Health Economics from the Johns Hopkins University’s Bloomberg School of Public of Health in Baltimore, Maryland USA. Dr. Bernell currently serves as Associate Professor and Program Coordinator of the Health Management and Policy programs at Oregon State University’s School of Public Health and Human Sciences in Corvallis, Oregon USA.
Faizan M. Casim, MPH, received his B.S. in Biology at University of Mary Washington in Fredericksburg, Virginia, USA. He then received his MPH at Saint Louis University School of Public Health at Saint Louis, Missouri, USA. Mr. Casim worked as a research consultant prior to joining USDA Food and Nutrition Service, where he now works as a Program Analyst.
Jennifer Wilmott, MPH MSW, received a B.S. in Social Work at Saint Louis University in St. Louis, Missouri, USA. She then received a MSW and MPH degree (Biostatistics and Epidemiology) at Saint Louis University. Since 2011, Ms. Wilmott has worked for the Department of Pediatrics at Saint Louis University, first as a research coordinator, and now as a biostatistician and data manager.
Lindsey Pearson, MHA, received a B.S. in Management at Missouri State University and a Master of Health Administration degree at Saint Louis University in St. Louis, MIssouri, USA. Lindsey is currently an Administrative Fellow at Saint Luke’s Health System in Kansas City, Missouri, USA.
Caitlin M. Byler, MHA, received a dual BA/BS in International Business at the University of Missouri in Columbia, Missouri, USA. She completed graduate school at Saint Louis University in St. Louis, Missouri USA where she earned a Master of Health Administration. Caitlin completed an administrative fellowship program at The University of Texas MD Anderson Cancer Center in Houston, Texas USA where she now serves as a Program Manager for MD Anderson Cancer Network.
Zidong Zhang, M.P.H., M.S., received a B.S. in Optometry at Donghua University in Shanghai, China. He subsequently received his M.S. in Vision Science at the State University of New York College of Optometry in New York, New York, USA. He did ocular infection research at University of Mississippi Medical Center where he consolidated his research interest in epidemiology and health service research. Then, he earned an MPH in epidemiology and worked with Dr. Steven Howard on health service research at Saint Louis University in St. Louis, Missouri USA. Mr. Zhang currently serves as the Epidemiologist at Jefferson County Health Department, Hillsboro, Missouri, USA.
Footnotes
Declaration of Conflicting Interests: The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding: The author(s) received no financial support for the research, authorship, and/or publication of this article.
References
- 1. Jacobson G, Damico A, Neuman T, Gold M. KFF report: Medicare Advantage 2015 Data Spotlight: Overview of Plan Changes. Kaiser Family Foundation; 2014. Website http://kff.org/medicare/issue-brief/medicare-advantage-2015-data-spotlight-overview-of-plan-changes/. Published December 10, 2014. Accessed 21 September, 2015.
- 2. Claxton G, Rae M, Panchal N, et al. Employer Health Benefits Survey: The Kaiser Family Foundation, Health Research and Educational Trust; 2014. Website http://kff.org/report-section/ehbs-2014-summary-of-findings/. Published September 10, 2014. Accessed 21 September, 2015.
- 3. Sari N. Do competition and managed care improve quality? Health Econ. 2002;11(7):571–584. [DOI] [PubMed] [Google Scholar]
- 4. Escarce JJ, Jain AK, Rogowski J. Hospital competition, managed care, and mortality after hospitalization for medical conditions: evidence from three states. Med Care Res Rev. 2006;63(6 suppl):112s–140s. [DOI] [PubMed] [Google Scholar]
- 5. Rogowski J, Jain AK, Escarce JJ. Hospital competition, managed care, and mortality after hospitalization for medical conditions in California. Health Serv Res. 2007;42(2):682–705. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Baker LC, Phillips KA, Haas JS, Liang SY, Sonneborn D. The effect of area HMO market share on cancer screening. Health Serv Res. 2004;39(6 pt 1):1751–1772. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Gillies RR, Chenok KE, Shortell SM, Pawlson G, Wimbush JJ. The impact of health plan delivery system organization on clinical quality and patient satisfaction. Health Serv Res. 2006;41(4 pt 1):1181–1199. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Scanlon DP, Swaminathan S, Chernew M, Lee W. Market and plan characteristics related to HMO quality and improvement. Med Care Res Rev. 2006;63(6 suppl):56s–89s. [DOI] [PubMed] [Google Scholar]
- 9. Chernew M, Decicca P, Town R. Managed care and medical expenditures of Medicare beneficiaries. J Health Econ. 2008;27(6):1451–1461. [DOI] [PubMed] [Google Scholar]
- 10. Nicholas LH. Does Managed Care Reduce Federal Spending? Evidence from Medicare. Ann Arbor, MI: Population Studies Center, University of Michigan; 2009. [Google Scholar]
- 11. Bundorf MK, Schulman KA, Stafford JA, Gaskin D, Jollis JG, Escarce JJ. Impact of managed care on the treatment, costs, and outcomes of fee-for-service Medicare patients with acute myocardial infarction. Health Serv Res. 2004;39(1):131–152. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Nicholas LH. Medicare Advantage? Managed Care and Medicare quality, Cost and Enrollment. New York, NY: Columbia University; 2009. [Google Scholar]
- 13. Meara E, Landrum MB, Ayanian JZ, McNeil BJ, Guadagnoli E. The effect of managed care market share on appropriate use of coronary angiography among traditional Medicare beneficiaries. Inquiry. 2004;41(2):144–158. [DOI] [PubMed] [Google Scholar]
- 14. Berenson RA. Medicare disadvantaged and the search for the elusive ‘level playing field’. Health Aff (Millwood). 2004;Suppl Web Exclusives:W4–572–585. [DOI] [PubMed] [Google Scholar]
- 15. Baker LC. The effect of HMOs on fee-for-service health care expenditures: evidence from Medicare. J Health Econ. 1997;16(4):453–481. [DOI] [PubMed] [Google Scholar]
- 16. Agency for Healthcare Research and Quality. MEPS-HC Sample Design and Collection Process. Agency for Healthcare Research and Quality, Rockville, Md; 2013. Website http://www.meps.ahrq.gov/survey_comp/hc_data_collection.jsp. Published June 9, 2013. Accessed 21 September, 2015. [Google Scholar]
- 17. Cao Z, McGuire TG. Service-level selection by HMOs in Medicare. J Health Econ. 2003;22(6):915–931. [DOI] [PubMed] [Google Scholar]
- 18. Mehrotra A, McNeil BJ, Landon BE. Congestive heart failure disease management in Medicare-managed care. Am Heart J. 2007;154(6):1153–1159. [DOI] [PubMed] [Google Scholar]
- 19. Atherly A, Hebert PL, Maciejewski ML. An analysis of disenrollment from Medicare managed care plans by Medicare beneficiaries with diabetes. Med Care. 2005;43(5):500–506. [DOI] [PubMed] [Google Scholar]