Comparison of Mortality Between Medicare Advantage and Traditional Medicare Beneficiaries With Kidney Failure

ABSTRACT

Objectives: To compare risk-adjusted 1-year mortality between Medicare Advantage (MA) and traditional Medicare (TM) enrollees with kidney failure who initiated dialysis.

Study Design: Longitudinal analysis of mortality and enrollment data for Medicare beneficiaries.

Methods: The study compared mortality between MA and TM enrollees with kidney failure who initiated dialysis in 2016, accounting for their enrollment switches between MA and TM during 12 months prior to dialysis initiation. Analyses were adjusted for risk scores and fixed effects for the month of dialysis initiation and county of residence.

Results: The difference in risk-adjusted 1-year mortality between MA stayers (Medicare beneficiaries who were continuously enrolled in MA prior to dialysis initiation) and TM stayers (those who were continuously enrolled in TM prior to initiating dialysis) was –0.1 percentage points (95% CI, –1.0 to 0.8); however, the difference increased to –1.0 percentage points (95% CI, –3.2 to 1.2) when comparing TM-to-MA switchers (those who switched from TM to MA before initiation) with TM stayers, a comparison more prone to favorable selection bias given our finding that TM-to-MA switchers were healthier than MA stayers.

Conclusions: Among Medicare beneficiaries with kidney failure who initiated dialysis, risk-adjusted 1-year mortality rate is not different between MA and TM stayers. If there is remaining favorable selection in MA due to unobserved health status, our finding provides a lower-bound estimate of the MA impact on mortality among beneficiaries with kidney failure.

Am J Manag Care. 2022;28(4):180-186. https://doi.org/10.37765/ajmc.2022.88861

Takeaway Points

The comparison of mortality between Medicare Advantage (MA) and traditional Medicare (TM) enrollees is empirically challenging because of favorable selection bias arising from MA plans attracting healthier individuals to enroll.

• Comparison of mortality between MA and TM “stayers,” which is less prone to favorable selection bias, indicated no difference in mortality among beneficiaries with kidney failure.
• Our finding provides a lower-bound estimate of the MA impact on mortality if there is remaining favorable selection in MA due to unobserved health status.
• It is important to account for enrollment switches between MA and TM when comparing health outcomes between MA and TM enrollees.

Over the past 15 years, enrollment in Medicare Advantage (MA) has quadrupled from 5.6 million in 2005 to 24.1 million in 2020, or 36% of all Medicare beneficiaries.¹ Unlike traditional Medicare (TM), MA plans—private managed care plans—receive capitated payments to bear the risk of paying for covered services for their enrollees.^2-4 In addition, MA plans receive bonus payments on the basis of a composite measure of quality, which includes kidney disease monitoring.⁵ The capitated and value-based payments may incentivize MA plans to improve quality of care while reducing costs, and therefore the health outcomes of MA enrollees could be better than those of TM enrollees.^6,7 In contrast, if the plans’ cost-containment efforts restrict access to effective care, the health outcomes of MA enrollees may be compromised. Previous studies on the effects of MA on health outcomes have shown mixed results, and few studies have focused on outcomes for individuals with complex health conditions such as kidney failure.^8-17

The comparison of health outcomes between MA and TM enrollees is empirically challenging because of favorable selection bias arising from MA plans attracting healthier and less-costly individuals to enroll.^18-23 In order to address the favorable selection by adjusting for underlying health status of MA and TM enrollees, many studies have attempted to use claims data that contain information on enrollees’ clinical diagnoses.^12,19-21 However, claims data for MA enrollees are often unavailable; most previous studies have compared MA enrollees who switched from TM (TM-to-MA switchers) with those who remained in TM (TM stayers) because the claims data for the TM-to-MA switchers are available for the period prior to the switch to MA.^19-21 It is important to note that the comparison between TM-to-MA switchers and TM stayers is more prone to favorable selection bias than the comparison between MA stayers and TM stayers if TM-to-MA switchers are healthier than MA stayers.

Comparisons of mortality rates for MA and TM are of particular interest for high-cost, high-need patients with severe chronic illnesses, given that MA plans’ quality-enhancing strategies (eg, chronic care management) may disproportionately benefit those with serious chronic conditions.²⁴ Further, plans’ cost-containment strategies, such as restricted networks, may have adverse consequences for high-need, high-cost patients. In this study, we compared 1-year mortality between MA and TM enrollees with kidney failure who initiated dialysis. We adjusted for differences in their health status and accounted for their enrollment switches in and out of MA plans prior to dialysis initiation. Specifically, we compared risk-adjusted 1-year mortality between MA stayers and TM stayers, adjusting for their underlying health status based on MA encounter data containing all claims of MA enrollees as well as TM claims data.

METHODS

Study Design

This longitudinal cohort study compared 1-year mortality between MA and TM enrollees with kidney failure initiating dialysis who lived in the same county. Those enrollees who received a kidney transplant after dialysis initiation were included. We particularly focused on the comparison between MA and TM enrollees who were continuously enrolled in MA or in TM during the 12 months prior to dialysis initiation (MA stayers and TM stayers). This comparison is preferable to the comparison between those who switch to MA prior to initiating dialysis (TM-to-MA switchers) and TM stayers because the latter comparison is more susceptible to favorable selection bias if TM-to-MA switchers are systematically healthier than MA stayers (eAppendix Figure 1 [eAppendix available at ajmc.com]). Note that in the 2 comparisons—(1) MA stayers vs TM stayers and (2) TM-to-MA switchers vs TM stayers—TM stayers are the common reference group, hence the main contrast is the difference between MA stayers and TM-to-MA switchers in favorable selection.

We also compared predialysis nephrology care and the use of an arteriovenous fistula or graft for dialysis among MA and TM enrollees. Given evidence that predialysis nephrology care is an important determinant of postdialysis outcomes, mortality following dialysis initiation should be attributed to the insurance program (MA or TM) in which the beneficiary was enrolled during the 12 months prior to starting dialysis. Further, the measures of predialysis nephrology care and fistula or graft placement in those starting dialysis occur in the year prior to initiation. Therefore, these measures should account for the enrollment status in MA or TM prior to dialysis initiation.

Data Sources and Study Population

The primary data sources were the CMS ESRD [end-stage renal disease] Medical Evidence Report (CMS form No. 2728) and the ESRD Death Notification Form (CMS form No. 2746), as well as MA encounter data, TM claims, and the Master Beneficiary Summary File (MBSF). The Medical Evidence Report provides detailed clinical information (eg, comorbid conditions, serum albumin and hemoglobin levels) for all patients with kidney failure. It was documented that comorbid conditions on this form were underreported^25,26; however, the percentage of patients who reported any comorbid condition other than their primary cause of kidney failure was similar between MA and TM enrollees (74.6% vs 73.0%), suggesting no differential underreporting between the 2 groups. The Death Notification Form includes information on date and cause of death reported by the treating nephrologist within 14 days of death for all patients who survive until the 91st day after dialysis initiation. MA encounter data contain claims (inpatient, outpatient, and carrier claims) for all MA enrollees. The MBSF provides monthly indicators of MA or TM enrollment status, as well as other demographic characteristics. We merged the CMS ESRD Medical Evidence Report with the MBSF to identify MA and TM status of beneficiaries with kidney failure who initiated dialysis in 2016 during 12 months prior to and at the time of their dialysis initiation. To calculate risk scores based on comorbidities during the year prior to dialysis initiation, we used the 2015 MA encounter data and TM claim files for MA and TM enrollees, respectively.

The study population included 49,160 Medicare beneficiaries (17,917 MA enrollees and 31,243 TM enrollees) 67 years and older who initiated dialysis in 2016 with continuous Medicare coverage from 1 year prior to dialysis initiation (eAppendix Figure 2).

Measures

The primary outcome was risk-adjusted 1-year mortality conditional on survival up to the 91st day following the initiation of dialysis. The US Renal Data System and previous studies have used this approach because death records for incident patients in the Death Notification Form are not reliably reported within the first 90 days following dialysis initiation.^27-30 The secondary outcomes were predialysis nephrology care and the use of an arteriovenous fistula or graft during the first dialysis treatment.

The primary explanatory variable of interest was an indicator for enrollment in MA at the time of dialysis initiation without changing MA enrollment status during 12 months prior to the dialysis initiation. Covariates included beneficiaries’ age, sex, race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, or other), dual eligibility for Medicare and Medicaid, primary cause of kidney failure (diabetes, hypertension, or other), comorbid conditions, tobacco use, alcohol and drug dependence, hemoglobin and serum albumin levels, body mass index (BMI), risk scores, and proportion of individuals 65 years and older living in the beneficiary’s zip code with income below the federal poverty level and with educational attainment of a high school degree or higher. The risk scores were calculated following the CMS–Hierarchical Condition Categories ESRD algorithm.³¹ The zip code–level characteristics were obtained from the 2012-2016 American Community Survey.

Statistical Analysis

We used multivariable linear regression models adjusting for beneficiaries’ sociodemographic and clinical characteristics as well as risk scores to account for potential differences in underlying health status between MA and TM enrollees with kidney failure. All models included fixed effects for the month of dialysis initiation and county of residence. Heteroskedasticity-robust standard errors were clustered at the county level to allow for unrestricted serial correlation within a county. For observations with missing values on covariates (ie, BMI, hemoglobin and albumin levels), we imputed using the mean value of the nonmissing observations on the covariates. The primary variable of interest was an indicator for being an MA stayer at the time of dialysis initiation, with a reference group of TM stayers, which yielded the difference in outcomes between MA and TM stayers. We also compared outcomes between TM-to-MA switchers and TM stayers as done in most previous studies.

Sensitivity Analysis

In sensitivity analyses, we explored whether the composition of patients who initiated dialysis differed between MA and TM enrollees (eg, MA plans delayed dialysis initiation among low-risk patients) by comparing their estimated glomerular filtration rate,³² a measure of kidney function, at dialysis initiation. In addition, we conducted the main analysis by quartile of risk score to explore whether a mortality difference, if any, between MA and TM stayers varied across underlying health status. We also compared all MA enrollees (MA stayers and TM-to-MA switchers) and all TM enrollees (TM stayers and MA-to-TM switchers). Further, we applied a logit specification to the main regression model to account for the binary outcome—whether an enrollee died within a year since the 91st day following dialysis initiation.

RESULTS

Figure 1 shows MA and TM enrollment status of the study population during 12 months prior to and at the time of dialysis initiation, respectively. A total of 49,160 Medicare beneficiaries 67 years and older who initiated dialysis in 2016 included 16,445 (33.5%) MA stayers and 29,429 (59.9%) TM stayers at the time of dialysis initiation and 1472 (3.0%) TM-to-MA switchers and 1814 (3.7%) MA-to-TM switchers during 12 months prior to the dialysis initiation. The characteristics of enrollees in each group by MA-TM enrollment status are presented in Table 1. Compared with TM stayers, MA stayers were more likely to be Black (25.0% vs 18.6%) and dually eligible for Medicaid (25.4% vs 22.3%). Also, MA stayers tended to have fewer comorbid conditions and lower risk scores than TM stayers. Furthermore, TM-to-MA switchers were less likely to have cardiac-related comorbid conditions (congestive heart failure, atherosclerotic heart disease, and other cardiac disease) and had lower median risk scores than MA stayers, implying that TM-to-MA switchers were healthier than MA stayers based on these observable characteristics.

Difference in Mortality

The unadjusted 1-year mortality rate of MA stayers was lower than that of TM stayers, with a difference of 2.42 (95% CI, 1.62-3.23) percentage points (Table 2). The mortality gap was larger between TM-to-MA switchers and TM stayers, with a 5.23-percentage-point difference (95% CI, 3.18-7.28). The adjusted mortality difference between MA stayers and TM stayers was –0.11 (95% CI, –1.01 to 0.78) percentage points; however, the difference increased to –1.0 (95% CI, –3.22 to 1.21) percentage points when comparing TM-to-MA switchers with TM stayers. Of note, the top 10 leading causes of death were similar for all 4 groups (ie, MA stayers, TM-to-MA switchers, MA-to-TM switchers, and TM stayers) (eAppendix Table 1).

Figure 2 plots 1-year mortality rates by risk score for MA and TM stayers. The unadjusted mortality rate of MA stayers was lower than that of TM stayers over almost the whole range of risk scores (panel A). However, once adjusted for patient sociodemographic and clinical characteristics with county fixed effects, the mortality rate of MA stayers was close to that of TM stayers whose risk scores were relatively low (< 1.0); and the mortality of MA stayers was even higher than that of TM stayers for those with higher risk scores (≥ 1.0) (panel B).

Differences in Predialysis Nephrology Care and Vascular Access for Dialysis

In adjusted analyses, there was no significant difference in predialysis nephrology care (0.32 percentage points; 95% CI, –0.89 to 1.54) or dialysis initiation with arteriovenous fistula or graft present (0.07 percentage points; 95% CI, –1.31 to 1.45) between MA and TM stayers, whereas the differences between TM-to-MA switchers and TM stayers were –3.57 (95% CI, –6.25 to –0.89) percentage points and –3.90 (95% CI, –6.58 to –1.22) percentage points, respectively.

Sensitivity Analysis

In adjusted analysis for the estimated glomerular filtration rate (Table 1), we found no significant difference between MA and TM stayers, implying a comparable patient composition between the 2 groups.

The analyses by quartile of risk score (eAppendix Table 2) showed no difference at the 5% significance level in mortality rates between MA and TM stayers across quartiles of risk scores. A similar result was observed for predialysis nephrology care: The difference was not significant except for the second quartile of risk scores (2.26 percentage points; 95% CI, 0.07-4.45). The use of arteriovenous fistula or graft exhibited no significant difference except for the first quartile of risk scores (4.42 percentage points; 95% CI, 1.78-7.05).

The comparison between all MA enrollees (TM-to-MA switchers and MA stayers) and all TM enrollees (MA-to-TM switchers and TM stayers) yielded consistent results with the comparison between MA and TM stayers: In adjusted analyses, there were no significant differences in mortality, predialysis nephrology care, and the use of arteriovenous fistula or graft (eAppendix Table 3). Results were also consistent in analyses that applied a logit model instead of a linear probability model (eAppendix Table 4).

DISCUSSION

After adjusting for sociodemographic characteristics and underlying health status based on risk scores and other clinical characteristics, we found no statistically significant difference in 1-year mortality between MA and TM stayers with kidney failure. Similarly, the comparison between TM-to-MA switchers and TM stayers also did not detect significant differences in adjusted mortality. The latter comparison (TM-to-MA switchers vs TM stayers) is more prone to favorable selection bias than the former comparison (MA stayers vs TM stayers) given our finding that TM-to-MA switchers are healthier than MA stayers. The stayers-to-stayers comparison yielded no difference in predialysis nephrology care and the use of arteriovenous fistula or graft, whereas the switchers-to-stayers comparison exhibited lower rates for these recommended services for MA enrollees compared with TM enrollees.

These findings highlight the importance of accounting for enrollment switches between MA and TM to address favorable selection bias when comparing outcomes between MA and TM enrollees. To our knowledge, our study is the first to compare mortality between MA and TM stayers, accounting for the MA-TM switching behavior among Medicare beneficiaries with kidney failure.

The lack of association of MA enrollment with mortality may be explained by similar rates of predialysis care and fistula use among MA and TM stayers. Both of these measures of dialysis preparation are associated with improved patient survival after the initiation of dialysis. If MA plans used fewer resources for the care and treatment for kidney failure than TM does, our finding of no difference in mortality between MA and TM stayers may suggest the relative efficiency of MA plans; however, additional research is needed to directly assess the efficiency advantages of MA plans for patients with kidney disease.

Our findings contrast with those of previous studies that reported lower mortality for MA enrollees than TM enrollees.^11,12,33 This discrepancy may be explained by at least 3 reasons. First, we accounted for enrollment switches between MA and TM prior to the index event (ie, dialysis initiation) by comparing MA and TM stayers. Second, we adjusted for underlying health status of MA enrollees using the MA encounter data containing all claims of MA enrollees; these data were unavailable until 2018, when they were first released. Third, we focused on a study population with a complex medical condition: Medicare beneficiaries with kidney failure who initiated dialysis.

We note that there may remain favorable selection bias even in the stayers-to-stayers comparison. First, MA stayers may be healthier than TM stayers on unobservable dimensions not measured by risk scores and other clinical characteristics. Second, MA plans may have incentives to code enrollees’ diagnoses more intensely (“upcoding”) to increase risk scores, which in turn yields higher risk-adjusted capitated payments that they receive.^34-36 To the extent that MA stayers are healthier than TM stayers in ways that are unobserved in our data or if MA enrollees have more intensely coded comorbidities, our findings provide a lower-bound estimate of the impact of MA on mortality.

Effective January 1, 2021, the 21st Century Cures Act for the first time allowed persons with kidney failure to enroll in or switch to MA after their dialysis initiation, with an expectation that MA plans may improve quality of care for these patients. Our results suggest that an evaluation of the impacts of the Cures Act for persons with kidney failure needs to account for possible enrollment switches between MA and TM before and after dialysis initiation. Further, our findings do not provide evidence that outcomes and access to predialysis nephrology care will improve following the expansion of MA plans to the kidney failure population.

Limitations

Our study has several limitations. First, our data include Medicare beneficiaries with kidney failure who initiated dialysis only in 2016 due to limited availability of the MA encounter data necessary for the risk adjustment for MA stayers during 12 months prior to the dialysis initiation. The MA encounter data were available only for 2015 claims when we conducted the analysis. Second, mortality rates were measured conditional on survival until the 91st day following dialysis initiation because death records in the Death Notification Form are complete beginning with the 91st day.^29,30 We may not capture a difference, if any exists, in mortality within 90 days from the dialysis initiation. Third, our data lacked individual-level information on income and educational attainment, although we included the zip code–level measures based on enrollees’ residence. Fourth, we could not adjust for a possible difference in coding intensity between MA and TM enrollees. Fifth, we were unable to adjust fully for unobserved underlying health status of MA and TM enrollees.

CONCLUSIONS

Among Medicare beneficiaries with kidney failure who initiated dialysis, we did not find evidence of a difference in risk-adjusted 1-year mortality between MA and TM stayers, a comparison less prone to favorable selection bias than a comparison between TM-to-MA switchers and TM stayers. If there is remaining favorable selection in MA due to unobserved health status, our finding provides a lower-bound estimate of the MA impact on mortality among beneficiaries with kidney failure. Our findings suggest that after accounting for enrollment switches between MA and TM, MA plans do not differentially improve survival in the high-cost, high-need population of patients with kidney failure who initiate dialysis.

Author Affiliations: Department of Health Services, Policy, and Practice, Brown University (DK, YL, SS, MR-H, RT, KHN, ANT), Providence, RI; Providence VA Medical Center (SS, RT, ANT), Providence, RI; Department of Medicine, University of Washington School of Medicine (RM), Seattle, WA.

Source of Funding: National Institute of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health (R01DK113298-02 and R01DK129388).

Author Disclosures: Dr Rivera-Hernandez’s work is supported by the National Institutes of Health (grant K01AG057822-03). The remaining 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 (DK, SS, RM, MR-H, RT, KHN, ANT); acquisition of data (ANT); analysis and interpretation of data (DK, YL, SS, RM, MR-H, ANT); drafting of the manuscript (DK, SS, RM, MR-H, RT, KHN); critical revision of the manuscript for important intellectual content (DK, SS, RM, MR-H, RT, KHN, ANT); statistical analysis (DK, YL); and obtaining funding (ANT); administrative, technical, or logistic support (YL).

Address Correspondence to: Daeho Kim, PhD, Department of Health Services, Policy, and Practice, Brown University, 121 S Main St, Providence, RI 02903. Email: Daeho_Kim@brown.edu.

REFERENCES

1. Freed M, Damico A, Neuman T. A dozen facts about Medicare Advantage in 2020. Kaiser Family Foundation. January 13, 2021. Accessed February 1, 2021. https://www.kff.org/medicare/issue-brief/a-dozen-facts-about-
medicare-advantage-in-2020/

2. Medicare Managed Care Manual: chapter 8 – payments to Medicare Advantage organizations. CMS. September 19, 2014. Accessed August 1, 2019. https://www.cms.gov/Regulations-and-Guidance/Guidance/Manuals/Downloads/mc86c08.pdf

3. Medicare Payment Advisory Commission. Report to the Congress: Medicare and the Health Care Delivery System. Medicare Payment Advisory Commission; 2019.

4. Medicare Advantage program payment system. Medicare Payment Advisory Commission. October 2020. Accessed January 30, 2021. https://www.medpac.gov/wp-content/uploads/import_data/scrape_files/docs/default-source/payment-basics/medpac_payment_basics_20_ma_final_sec.pdf

5. Medicare 2020 Part C & D star ratings technical notes. CMS. October 2019. Accessed October 23, 2020. https://www.cms.gov/Medicare/Prescription-Drug-Coverage/PrescriptionDrugCovGenIn/Downloads/Star-Ratings-Technical-Notes-Oct-10-2019.pdf

6. McGuire TG, Newhouse JP, Sinaiko AD. An economic history of Medicare part C. Milbank Q. 2011;89(2):289-332. doi:10.1111/j.1468-0009.2011.00629.x

7. Neuman P, Jacobson GA. Medicare Advantage checkup. N Engl J Med. 2018;379(22):2163-2172. doi:10.1056/NEJMhpr1804089

8. Miller RH, Luft HS. HMO plan performance update: an analysis of the literature, 1997-2001. Health Aff (Millwood). 2002;21(4):63-86. doi:10.1377/hlthaff.21.4.63

9. Dowd B, Maciejewski ML, O’Connor H, Riley G, Geng Y. Health plan enrollment and mortality in the Medicare program. Health Econ. 2011;20(6):645-659. doi:10.1002/hec.1623

10. Gowrisankaran G, Town R, Barrette E. Managed care, drug benefits and mortality: an analysis of the elderly. B E J Econom Anal Policy. 2011;11(2):1-32. doi:10.2202/1935-1682.2719

11. Afendulis CC, Chernew ME, Kessler DP. The effect of Medicare Advantage on hospital admissions and mortality. Am J Health Econ. 2017;3(2):254-279. doi:10.1162/AJHE_a_00074

12. Beveridge RA, Mendes SM, Caplan A, et al. Mortality differences between traditional Medicare and Medicare Advantage: a risk-adjusted assessment using claims data. Inquiry. 2017;54:46958017709103. doi:10.1177/0046958017709103

13. Abaluck J, Bravo MMC, Hull P, Starc A. Mortality effects and choice across private health insurance plans. National Bureau of Economic Research working paper No. 27578. July 2020. Accessed November 4, 2020. https://www.nber.org/papers/w27578

14. Landon BE, Zaslavsky AM, Saunders R, Pawlson LG, Newhouse JP, Ayanian JZ. A comparison of relative resource use and quality in Medicare Advantage health plans versus traditional Medicare. Am J Manag Care. 2015;21(8):559-566.

15. Huckfeldt PJ, Escarce JJ, Rabideau 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

16. Kumar A, Rahman M, Trivedi AN, Resnik L, Gozalo P, Mor V. Comparing post-acute rehabilitation use, length of stay, and outcomes experienced by Medicare fee-for-service and Medicare Advantage beneficiaries with hip fracture in the United States: a secondary analysis of administrative data. PLoS Med. 2018;15(6):e1002592. doi:10.1371/journal.pmed.1002592

17. 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

18. Morrisey MA, Kilgore ML, Becker DJ, Smith W, Delzell E. Favorable selection, risk adjustment, and the Medicare Advantage program. Health Serv Res. 2013;48(3):1039-1056. doi:10.1111/1475-6773.12006

19. Brown J, Duggan M, Kuziemko I, Woolston W. How does risk selection respond to risk adjustment? new evidence from the Medicare Advantage program. Am Econ Rev. 2014;104(10):3335-3364. doi:10.1257/aer.104.10.3335

20. 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

21. Newhouse JP, Price M, Huang J, McWilliams JM, Hsu J. Steps to reduce favorable risk selection in Medicare Advantage largely succeeded, boding well for health insurance exchanges. Health Aff (Millwood). 2012;31(12):2618-2628. doi:10.1377/hlthaff.2012.0345

22. Rahman M, Keohane L, Trivedi AN, Mor V. High-cost patients had substantial rates of leaving Medicare Advantage and joining traditional Medicare. Health Aff (Millwood). 2015;34(10):1675-1681. doi:10.1377/hlthaff.2015.0272

23. Meyers DJ, Mor V, Rahman M. Medicare Advantage enrollees more likely to enter lower-quality nursing homes compared to fee-for-service enrollees. Health Aff (Millwood). 2018;37(1):78-85. doi:10.1377/hlthaff.2017.0714

24. Improving care for Medicare beneficiaries with chronic conditions. Medicare Payment Advisory Commission. May 14, 2015. Accessed May 21, 2019. https://www.medpac.gov/wp-content/uploads/import_data/scrape_files/docs/default-source/congressional-testimony/testimony-improving-care-for-beneficiaries-with-chronic-conditions-senate-finance-.pdf

25. Longenecker JC, Coresh J, Klag MJ, et al. Validation of comorbid conditions on the end-stage renal disease medical evidence report: the CHOICE study. Choices for Healthy Outcomes in Caring for ESRD. J Am Soc Nephrol. 2000;11(3):520-529. doi:10.1681/ASN.V113520

26. Krishnan M, Weinhandl ED, Jackson S, Gilbertson DT, Lacson E Jr. Comorbidity ascertainment from the ESRD Medical Evidence Report and Medicare claims around dialysis initiation: a comparison using US Renal Data System data. Am J Kidney Dis. 2015;66(5):802-812. doi:10.1053/j.ajkd.2015.04.015

27. United States Renal Data System. 2017 USRDS Annual Data Report: Epidemiology of Kidney Disease in the United States. National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases; 2017.

28. Researcher’s guide to the USRDS database. US Renal Data System. 2017. Accessed May 21, 2019. https://www.usrds.org/media/2188/2017_usrds_researcher_guide.pdf

29. Swaminathan S, Sommers BD, Thorsness R, Mehrotra R, Lee Y, Trivedi AN. Association of Medicaid expansion with 1-year mortality among patients with end-stage renal disease. JAMA. 2018;320(21):2242-2250. doi:10.1001/jama.2018.16504

30. Foley RN, Chen SC, Solid CA, Gilbertson DT, Collins AJ. Early mortality in patients starting dialysis appears to go unregistered. Kidney Int. 2014;86(2):392-398. doi:10.1038/ki.2014.15

31. Report to Congress: risk adjustment in Medicare Advantage. CMS. December 2018. Accessed March 2, 2020. https://www.cms.gov/Medicare/Health-Plans/MedicareAdvtgSpecRateStats/Downloads/RTC-Dec2018.pdf

32. Levey AS, Coresh J, Greene T, et al; Chronic Kidney Disease Epidemiology Collaboration. Expressing the Modification of Diet in Renal Disease Study equation for estimating glomerular filtration rate with standardized serum creatinine values. Clin Chem. 2007;53(4):766-772. doi:10.1373/clinchem.2006.077180

33. Powers BW, Yan J, Zhu J, et al. The beneficial effects of Medicare Advantage special needs plans for patients with end-stage renal disease. Health Aff (Millwood). 2020;39(9):1486-1494. doi:10.1377/hlthaff.2019.01793

34. Geruso M, Layton T. Upcoding: evidence from Medicare on squishy risk adjustment. J Polit Econ. 2020;128(3):984-1026. doi:10.1086/704756

35. Kronick R, Welch WP. Measuring coding intensity in the Medicare Advantage program. Medicare Medicaid Res Rev. 2014;4(2). doi:10.5600/mmrr2014-004-02-a06

36. Kronick R. Projected coding intensity in Medicare Advantage could increase Medicare spending by $200 billion over ten years. Health Aff (Millwood). 2017;36(2):320-327. doi:10.1377/hlthaff.2016.0768