Medicare Advantage and Postdischarge Quality: Evidence From Hospital Readmissions

December 8, 2020
Paul D. Jacobs, PhD

Jayasree Basu, PhD

The American Journal of Managed Care, December 2020, Volume 26, Issue 12

Relative readmission rates, a measure of care coordination, did not change over time for beneficiaries enrolled in Medicare Advantage and traditional Medicare.


Objectives: To compare relative readmission rates for beneficiaries enrolled in Medicare Advantage (MA) and traditional Medicare (TM) as suggestive evidence of changes in postdischarge care coordination and the quality of care delivered to Medicare beneficiaries.

Study Design: We used the Agency for Healthcare Research and Quality’s 2009 and 2014 Healthcare Cost and Utilization Project State Inpatient Databases for 4 states with reliable sources of payment identifiers, linking these data to local area characteristics. Our outcome was the probability of a hospital readmission within 30 days of an index admission. We computed readmission rates overall and by subgroups, including for patients with multiple chronic conditions, by patients’ state of residence, and by type of index admission.

Methods: We estimated linear probability models with hospital fixed effects including a wide array of patient-level characteristics relating to health status and sociodemographic characteristics. Standard errors were adjusted for clustering at the area level.

Results: Significantly lower all-cause readmission rates were found among MA enrollees relative to those in TM in both 2009 and 2014, suggesting an association between MA enrollment and higher quality of care. However, over the 2009-2014 period, MA enrollment was not associated with an increased reduction in readmission rates relative to TM.

Conclusions: Although our focus was on a single measure of performance, the claims that managed care plans are spearheading changes in the delivery system are not supported by our finding that relative readmission rates were stable over the 2009-2014 period.

Am J Manag Care. 2020;26(12):524-529.


Takeaway Points

Medicare Advantage (MA) plans have been increasingly focused on improving the delivery of care and coordinating postdischarge transitions through policy changes.

  • We compared relative readmission risks for beneficiaries enrolled in MA and traditional Medicare as evidence of changes in care coordination and practice outcomes.
  • Although delivery system reforms have been much lauded for years, changes in improved quality for postacute care do not appear to be differentially accruing to MA beneficiaries.
  • Additional research on specific managed care interventions to reduce readmissions is needed to strengthen the evidence base.


The Medicare Advantage (MA) program was designed to provide a range of insurance choices that can meet the specific health care needs of Medicare beneficiaries and help coordinate the delivery of care. As a result of both policy changes and the incentive to reduce costs in a capitated environment, MA plans have been increasingly focused on coordinating care. A variety of strategies to do so have been implemented over the past decade, including patient-centered care models, disease management programs, Special Needs Plans, and incentivizing primary care teams to better coordinate care delivery.1,2 Because of its size and focus on the elderly and disabled, the MA program has been lauded as a possible avenue for providing enrollees with better value and quality of care and potentially spurring larger reforms to the health care system.3

Readmission rates are one method for assessing the role that managed care can play in improving quality and value. Although readmissions are an imperfect proxy for quality, they can serve as a signal of the success that managed care can have at coordinating care through disease management, as well as postdischarge care and communication. Recognizing the role of readmissions in improving the value and quality of health care systems, since 2013 CMS has penalized acute care hospitals with relatively high readmission rates under the Hospital Readmissions Reduction Program (HRRP). Moreover, accountable care organizations are now being evaluated using all-cause readmissions.4 Although it may be impossible to target a specific readmission rate that could be considered appropriate or acceptable, MA plans have the potential to provide postacute care that is more coordinated than for those covered through traditional Medicare (TM). Thus, it is plausible that MA plans can improve so-called “care transitions,” which have been shown to reduce readmission rates.5

To evaluate the success of MA in improving readmission rates, we compared MA patients with their TM counterparts, assessing changes in the relative rate of readmissions over the 2009-2014 period. Although MA plans coordinate care in heterogeneous ways, we broaden the focus on aggregate outcomes to understand and compare how managed care may have influenced postdischarge coordination over a period of great interest for changes in the landscape for elderly beneficiaries. After controlling for differences in treatment patterns among hospitals and for differential health selection into MA and TM, we found that relative rates of readmission between MA and TM were not statistically different in 2009 and 2014. In 1 state—California—the change over time was modestly significant, indicating potential improvements in care delivery in a region with a history of higher managed care penetration. Earlier studies have compared readmission rates in MA and TM, but none to our knowledge has been able to directly address both concerns of selection bias and the possibility that treatment patterns may systematically differ among hospitals, which may have biased earlier efforts to identify the effect of MA on readmissions.


A core goal of managed care organizations is to provide value for enrollees because of their ability to provide more closely coordinated care than individuals would otherwise receive in an unmanaged health care plan. However, given the complexity of the inpatient setting and the variety of causes that can lead to readmissions, it is not clear to what extent managed care can improve readmission risks. In the Medicare setting, limited but conflicting evidence exists regarding how MA may contribute to a patient’s readmission risks after an initial hospitalization compared with beneficiaries in a TM setting. Most studies have found that unadjusted readmission rates were lower for MA than TM patients. However, contradictory evidence exists after adjusting for risk, which is necessary because enrollees are enrolled by default into TM and most have a choice of several MA plans. Most earlier studies found that readmission rates for MA patients were lower than those for TM patients.6,7 On the other hand, other studies using different methods and data sources have found that MA patients were significantly more likely than TM patients to be readmitted.8,9

Previous research efforts have generally focused on readmission outcomes for a single year, including work on racial and ethnic disparities in readmission rates,10 readmission comparisons for single diseases,7,11,12 and an early assessment of readmission rates shortly after the introduction of the HRRP.13 Other studies focusing on readmission outcomes over time in the wake of the HRRP did not compare outcomes for MA and TM beneficiaries.14,15

Our study differs from earlier work by comparing MA and TM readmission rates over time while using robust methods to adjust for selection bias between MA and TM and simultaneously controlling for hospital-specific patterns of selection and treatment. Heterogeneous effects at the hospital level could occur either through selection or via treatment quality and decision-making. Specifically, MA plans might steer enrollees to particular hospitals (selection), and hospitals can vary widely in their discharge protocols, as well as incentives and arrangements for utilizing postacute care (treatment), both of which could have contaminated previous results but, to our knowledge, have not yet been addressed.


Our main sources of data are the Agency for Healthcare Research and Quality 2009 and 2014 Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases. HCUP data are discharge-level records, and the unit of analysis is each inpatient hospital admission (index admission), which could have been for any cause between January and November in each calendar year. Our outcome was the patient’s risk of readmission, defined as the probability of a hospital readmission within 30 days of discharge from an index admission. (A single patient can be counted multiple times during the course of the January through November observation period, so a hospital stay may be both a readmission for a prior stay and the index admission for a subsequent readmission.) We included index admissions for beneficiaries 65 years and older. Planned readmissions and trauma-related readmissions are excluded, based on methods used in previous research.16

We looked at the performance of MA plans in 4 states (New York, California, Florida, and Tennessee), which had a range of MA penetration rates (from 32% to 38%). Together these 4 states accounted for 32% of all Medicare managed care enrollees in 2014.17 We limited the sample to these states because they have robust and validated data on patient identifiers to calculate readmissions and indicators for MA and TM for both 2009 and 2014.

To identify MA vs TM enrollment, we used patient-level indicators for the expected source of payment as recorded in the intramural versions of the discharge-level HCUP data. MA plans included all managed care plan types (eg, health maintenance organizations [HMOs], preferred provider organizations, private fee-for-service [FFS]); we were not able to distinguish among them.

We computed readmission rates overall and by subgroups, including for patients with multiple chronic conditions (MCC), by patients’ state of residence, and by type of index admission. Relative probabilities of readmission in MA and TM were compared over time. We explored effects on patients with MCCs because of the attention this population has long garnered and the possibility that stronger effects could be found for a population with the most to gain from better postacute coordination.18 Because of the high number of chronic conditions among the elderly, we defined MCC as 6 or more conditions identified in an index admission, which narrowed the sample for these analyses to about 64% of the size of our elderly study population. We computed readmission rates by index admission types, including emergency, urgent, and elective admissions. Our hypothesis was that if managed care organizations can effectively reduce readmission rates, we would expect larger improvements for elective admissions, for which plans can anticipate and plan postacute care more effectively than for complex, unforeseen cases.

We computed multivariate linear probability models (LPMs) with hospital fixed effects, which control for the effect of hospital-specific treatment patterns and potential steering by MA plans to specific hospitals. All models controlled for individual and regional characteristics that may be associated with the probability of choosing MA vs TM and overall treatment patterns. Estimates shown are predicted probabilities of readmission while adjusting for model covariates. Standard errors were adjusted for clustering at the primary care service area (PCSA) level, and reported differences were significant using a P value of .05.


To control for self-selection into managed care plans, we included the following individual patient characteristics: number of chronic conditions, age group, severity of illness, admission type, risks of mortality, race/ethnicity, gender, and patient comorbidities. We used the All Patient Refined Diagnosis Related Groups (APR-DRGs) severity score in the HCUP database to control for patient severity and risk of mortality; these were summarized into 4 categories. Severity of illness and risk of mortality were assigned by applying a base APR-DRG and then adjusting for the severity of secondary diagnoses and interactions with age, principal diagnosis, and procedures.19 To control for variations in readmission risks across APR-DRGs, our model includes APR-DRG weights, calculated as standardized values of overall readmission rates in 4 states by APR-DRG categories.

We defined regional controls using data from the Health Resources and Services Administration that define markets as PCSAs (rather than larger, more heterogeneous, counties). PCSAs approximate geographic markets for primary care services based on Medicare FFS patient flows to primary care physicians’ offices.20 PCSA-level variables included primary care physician density, population density, and rural-urban location.

eAppendix Table 1 (eAppendix available at provides summary statistics for all covariates by MA and TM status in each year.

Sensitivity Analyses

We tested the robustness of our findings to different models, including LPMs without hospital fixed effects, propensity score matching (PSM) with inverse probability weights, and PSM with augmented inverse probability weights. PSM models were used to understand the sensitivity of our estimates to a more conservative approach for matching comorbidities to adjust for selection bias in MA or TM. eAppendix Table 2 shows that the results for these alternative models did not differ meaningfully from the results in the text. For all PSM models, we confirmed the models were correctly specified for each covariate by checking the balance in both the differences in the means and the ratio of variances between the treated and control groups (eAppendix Tables 3 and 4).

We also tested the sensitivity of our findings to (1) including number of procedures and total charges for each admission, (2) including mortality rates at the hospital level, (3) including length of stay for the index admission, and (4) excluding patients who died in the hospital. No meaningful changes in our findings were observed.


Figure 1 shows readmission rates in 2009 and 2014 after adjusting for differences in beneficiary health status and differences arising from hospital treatment patterns and compositional effects. In 2009, readmission rates for MA enrollees were lower (14.4%) compared with those for beneficiaries with TM (15.1%). Likewise, in 2014, readmission rates for MA enrollees (13.4%) were lower than for those in TM (14.1%). The difference in MA and TM readmission rates in 2009 (–0.6%) was not statistically different from the difference in 2014 (–0.8%) (P = .30). (eAppendix Table 2 shows that the relative unadjusted readmission rate changed over time, but this comparison was statistically insignificant after controlling for health status and other covariates. eAppendix Tables 5 and 6 provide the full regression results for the LPMs for each year.)

Figure 2 shows that, among those with 6 or more chronic conditions in 2009, MA enrollees had lower rates of readmission (16.9%) than enrollees in TM (17.5%). Similarly, in 2014, enrollees with MCC in MA had lower readmission rates (15.2%) than those in TM (16.1%). However, the differences in the MA and TM readmission rates (–0.7% in 2009 and –0.9% in 2014) did not statistically differ over time (P = .27).

Figure 3 shows differences in the MA and TM readmission rates in each year and by state. The difference between the MA and TM readmission rates did not change over the 2009-2014 period in Florida, New York, or Tennessee. Differences between these years in each state were not statistically or meaningfully different. However, the difference in MA and TM readmission rates in 2014 (–1.6%) was significantly different from the difference in 2009 (–1.1%) in California (P = .04).

Figure 4 displays differences in MA and TM readmission rates by category of index discharge in each year of our study. None of the differences in discharge types over time suggested changes in the relative rate of MA and TM readmissions. Although all 3 of the relative rates of readmission were not statistically different in 2014 than in 2009, the 2014 relative rates were modestly lower than the 2009 rates across all 3 categories.


Our results are consistent with those of the majority of earlier studies in finding lower rates of readmission for MA enrollees than TM enrollees after controlling for health risk. We confirmed that lower rates of readmission in MA were present in both 2009 and 2014 and were generally consistent over time for the subgroups we analyzed, including beneficiaries with MCC, in 3 of the 4 states in our sample, and for different admission types. Our estimates extend the prior literature by using a methodology that accounts for both differential health selection into MA and for treatment and selection patterns that can differ across hospital settings. After controlling for a wide array of measures of health risk, including chronic conditions and comorbidities, and removing hospital-specific variability, our results provide further support for the association between MA enrollment and lower risk of hospital readmission, which is one measure by which quality of care and value to enrollees can be assessed.

But, importantly, we also found that the lower relative rate of readmission in MA was not statistically different in 2014 compared with 2009. Since 2009, increasing attention has been paid to how to improve the quality of care for the Medicare population and, specifically, how MA plans might better coordinate care using disease management programs, postdischarge planning, and Special Needs Plans, which may benefit those with MCC.18 With greater capacity to focus on designing integrated delivery systems to meet the needs of high-risk, high-need individuals, it would be reasonable to expect differential improvement in readmission rates among MA enrollees compared with those in TM. An alternative explanation of our findings is that managed care organizations have been less involved in delivery system changes than hospital administrators and thus reductions in readmissions are accruing across Medicare beneficiaries without respect to payer status.

In 3 of the 4 states in our study, we were unable to confirm that any such significant changes had taken place over the 2009 to 2014 time period. We found that the relative readmission rate in MA vs TM for California was lower in 2014 compared with 2009. Of note, California has consistently had higher local HMO penetration in its MA market compared with other states (97% enrolled in HMOs in California in 2014 compared with 70%-73% in the 3 other states in our sample). It is possible that the higher percentage of enrollees in HMO plans in California may have enabled improved quality of inpatient or postacute delivery through greater coordination and management of care. However, our methodological design did not allow us to determine the specific causes of the difference in California.

We found no evidence that readmission rates are differentially improving in MA for some types of index admissions but not others. The null finding for any improvement in MA relative to TM over time seems intuitive for the complex and heterogeneous treatment patterns associated with emergency and urgent care admissions. One area where managed care plans could theoretically have more ability to coordinate and control readmissions would be elective surgeries. However, we found similar relative rates of readmission in MA and TM within any particular year and no change over time for any of the subcategories of readmissions.


By focusing on readmissions in the HCUP data, we did not specifically measure how MA plans may be coordinating postdischarge care. Although we have controlled for a broad array of comorbidities and health status measures, unobserved health selection between MA and TM may still influence our results. For instance, one characteristic we do not observe consistently in the HCUP data is dual eligibility for Medicare and Medicaid. Because dual-eligible patients are sicker and poorer and more likely to be in TM, this could have resulted in higher observed rates of readmission among TM patients than among MA patients. Ultimately, we find that controls for severity of illness, including an extensive list of patient comorbidities, did not substantially affect our estimated probabilities. Thus, further refinements in this vein would likely not have changed our results and, moreover, would have only narrowed the statistically insignificant differences we found over time between MA and TM readmission rates.

Compositional changes in the MA and TM populations over the 2009-2014 period may also be influencing our results. Beneficiaries may change their status over time by switching from one sector to the other, perhaps in response to a health event. These effects are likely mitigated by the introduction of lock-in periods for MA enrollees that began in 2006.21 Finally, because we studied only 4 states, our conclusions may not be generalizable to the nation as a whole.


We consistently found lower readmission rates for MA beneficiaries than for those in TM, which suggests that managed care might be successful at providing higher quality in the postacute care setting. However, over the period we observed, we did not find that the differential between MA and TM enrollees was increasing, although a broad range of attempts were made to systematically change how MA plans deliver care to beneficiaries. Future work in this area should investigate more recent data on how readmissions and other quality metrics differ in MA and TM and more systematically assess specific managed care interventions to improve the delivery system. Although delivery system reforms have been much lauded for years, changes in improved readmission rates do not appear to be differentially accruing to MA beneficiaries.

Author Affiliations: Agency for Healthcare Research and Quality (PDJ, JB), Rockville, MD.

Source of Funding: None.

Author Disclosures: The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (PDJ, JB); acquisition of data (JB); analysis and interpretation of data (PDJ, JB); drafting of the manuscript (PDJ, JB); critical revision of the manuscript for important intellectual content (PDJ, JB); and statistical analysis (PDJ, JB).

Address Correspondence to: Paul D. Jacobs, PhD, Agency for Healthcare Research and Quality, 5600 Fishers Ln, Mailstop 07W41A, Rockville, MD 20850. Email:


1. Johnson G, Lyon ZM, Frakt A. Provider-offered Medicare Advantage plans: recent growth and care quality. Health Aff (Millwood). 2017;36(3):539-547. doi:10.1377/hlthaff.2016.0722

2. Thomas KS, Durfey SNM, Gadbois EA, et al. Perspectives of Medicare Advantage plan representatives on addressing social determinants of health in response to the CHRONIC Care Act. JAMA Netw Open. 2019;2(7):e196923. doi:10.1001/jamanetworkopen.2019.6923

3. Broussard BD, Shrank WH. Medicare Advantage and the future of value-based care. Health Affairs. July 3, 2019. Accessed February 26, 2020.

4. RTI International. Accountable care organization 2016 program quality measure narrative specifications. CMS. January 13, 2016. Accessed February 26, 2020.

5. Naylor MD, Aiken LH, Kurtzman ET, Olds DM, Hirschman KB. The care span: the importance of transitional care in achieving health reform. Health Aff (Millwood). 2011;30(4):746-754. doi:10.1377/hlthaff.2011.0041

6. Lemieux J, Sennett C, Wang R, Mulligan T, Bumbaugh J. Hospital readmission rates in Medicare Advantage plans. Am J Manag Care. 2012;18(2):96-104.

7. Huckfeldt PJ, Escarce JJ, Radieau B, Karaca-Mandic P, Sood N. Less intense postacute care, better outcomes for enrollees in Medicare Advantage than those in fee-for-service. Health Aff (Millwood). 2017;36(1):91-100. doi:10.1377/hlthaff.2016.1027

8. Friedman B, Jiang HJ, Steiner CA, Bott J. Likelihood of hospital readmission after first discharge: Medicare Advantage vs. fee-for-service patients. Inquiry. 2012;49(3):202-213. doi:10.5034/inquiryjrnl_49.03.01

9. Panagiotou OA, Kumar A, Gutman R, et al. Hospital readmission rates in Medicare Advantage and traditional Medicare: a retrospective population-based analysis. Ann Intern Med. 2019;171(2):99-106. doi:10.7326/M18-1795

10. Li Y, Cen X, Cai X, Thirukumaran CP, Zhou J, Glance LG. Medicare Advantage associated with more racial disparity than traditional Medicare for hospital readmissions. Health Aff (Millwood). 2017;36(7):1328-1335. doi:10.1377/hlthaff.2016.1344

11. Smith MA, Frytak JR, Liou JI, Finch MD. Rehospitalization and survival for stroke patients in managed care and traditional Medicare plans. Med Care. 2005;43(9):902-910. doi:10.1097/01.mlr.0000173597.97232.a0

12. Wong ES, Hebert PL, Maciejewski ML, et al. Does favorable selection among Medicare Advantage enrollees affect measurement of hospital readmission rates? Med Care Res Rev. 2014;71(4):367-383. doi:10.1177/1077558714533823

13. Henke RM, Karaca Z, Gibson TB, et al. Medicare Advantage and traditional Medicare hospitalization intensity and readmissions. Med Care Res Rev. 2018;75(4):434-453. doi:10.1177/1077558717692103

14. Carey K, Meng-Yun L. Readmissions to New York hospitals fell for three target conditions from 2008 to 2012, consistent with Medicare goals. Health Aff (Millwood). 2015;34(6):978-985. doi:10.1377/hlthaff.2014.1408

15. Demiralp B, He F, Koenig L. Further evidence on the system-wide effects of the Hospital Readmissions Reduction Program. Health Serv Res. 2018;53(3):1478-1497. doi:10.1111/1475-6773.12701

16. Friedman B, Basu J. The rate and cost of hospital readmissions for preventable conditions. Med Care Res Rev. 2004;61(2):225-240. doi:10.1177/1077558704263799

17. Gold M, Jacobson G, Damico A, Neuman T. Medicare Advantage 2014 spotlight: enrollment market update. Kaiser Family Foundation. May 1, 2014. Accessed February 26, 2020.

18. Parekh AK, Goodman RA, Gordon C, Koh HK; HHS Interagency Workgroup on Multiple Chronic Conditions. Managing multiple chronic conditions: a strategic framework for improving health outcomes and quality of life. Public Health Rep. 2011;126(4):460-471. doi:10.1177/003335491112600403

19. Averill RF, Goldfield N, Steinbeck B, et al. Development of the All Patient Refined DRGs (APR-DRGs). 3M Health Information Systems. 2000. Accessed February 26, 2020.

20. Goodman DC, Mick SS, Bott D, et al. Primary care service areas: a new tool for the evaluation of primary care services. Health Serv Res. 2003;38(1, pt 1):287-309. doi:10.1111/1475-6773.00116

21. Newhouse JP, Price M, McWilliams JM, Hsu J, McGuire TG. How much favorable selection is left in Medicare Advantage? Am J Health Econ. 2015;1(1):1-26. doi:10.1162/AJHE_a_00001