Social Risk–Weighted Scoring to Improve Medicare Advantage Star Ratings

ABSTRACT

Objectives: Medicare Advantage (MA) Star Ratings influence bonus payments and health plan enrollment. Current approaches to accounting for enrollee social risk may not fully capture differences in beneficiary populations. This study assessed how MA contract performance rankings for statin receipt and adherence change under a social risk–weighted scoring approach.

Study Design: Cross-sectional analysis of 2019 MA encounter, prescription drug, and enrollment data for 165 MA contracts.

Methods: Contract-level statin receipt and adherence scores were calculated for disability and dual-eligibility subgroups and aggregated using empirically derived weights from a constrained optimization framework that equalized each subgroup's contribution to overall contract performance. Ordered logistic regression assessed associations between contract characteristics and weighted performance.

Results: Social risk–weighted scoring reclassified 24.2% of contracts for statin receipt and 20.6% for adherence relative to unweighted scoring. For statin receipt, 11.5% of contracts moved up, 12.7% moved down, and 75.8% were unchanged; for adherence, 7.3% moved up, 13.3% moved down, and 79.4% were unchanged. Under the weighted approach, higher disability prevalence remained associated with lower performance ratings even after weighting: Each 1–percentage point (PP) increase in disability prevalence corresponded to a 0.7–PP higher probability of receiving 3 or fewer stars for statin receipt and a 1.3–PP higher probability for adherence. Larger contracts also had lower predicted probabilities of receiving 4 stars, decreasing by 3.4 PPs for receipt and 3.8 PPs for adherence per 1-unit increase in log enrollment.

Conclusions: Social risk weighting rebalanced contract rankings but did not consistently improve ratings for high-risk plans. Remaining variation may reflect differences in quality or unmeasured factors.

Takeaway Points

Medicare Advantage Star Ratings influence billions in payments, yet they may penalize plans serving high-need populations.

• Social risk weighting reclassified 20% to 24% of contracts across the statin receipt and adherence measures.
• Performance rankings changed modestly, with approximately 1 in 5 contracts moving up or down a performance ranking.
• Social risk weighting did not uniformly raise ratings for contracts serving high–social risk populations. In several cases, contracts with lower proportions of dually eligible or disabled enrollees moved up in relative ranking.

Medicare Advantage (MA) performance ratings determine eligibility for billions in quality bonus payments annually, directly influencing the resources that plans can invest in supplemental benefits such as meals, transportation, and housing support for populations with greater health and social needs.^1-3 Yet concerns remain that the Star Ratings program insufficiently adjusts for social risk factors, systematically disadvantaging plans that serve a higher proportion of beneficiaries with disabilities, low incomes, and other barriers to care.^4,5 CMS modernization efforts, including Excellent Health Outcomes for All, reflect growing recognition that fair plan comparisons depend on accounting for social risk in performance measurement.^6,7 This shift moves beyond secondary, post hoc adjustments toward stronger, goal-aligned incentives within MA.⁸

Although some performance measures, such as hospital readmissions, are adjusted for clinical complexity, others, including statin receipt and adherence, are not routinely adjusted for enrollee characteristics, despite being influenced by both clinical and social risk.^9,10 In response to fairness concerns, CMS introduced the Categorical Adjustment Index (CAI) in 2017, which provides a bonus to contracts serving higher proportions of beneficiaries with dual eligibility or disability.⁷ Although the CAI represents progress, it addresses the incentive rather than the measurement, leaving underlying performance scores potentially biased by enrollee composition. Plan evaluations may not only distort bonus allocations but also limit beneficiaries’ access to high-quality plans in their region, especially if high–social risk plans are penalized for differences partly driven by enrollee characteristics rather than plan performance.¹¹

Recent policy proposals emphasize the importance of incorporating social risk directly into the construction of quality measures and contract ratings, rather than relying on downstream adjustments.^12-14 Direct integration of social risk into performance scoring could better align incentives, promote fairness, and encourage plans to invest in care improvement for high-need populations.¹⁴ However, the operational implementation of such methods, and their effects on performance-based groupings, remain underexplored.

In this study, we applied a social risk–weighted scoring method based on constrained optimization, derived empirical weights, and applied them to MA contracts’ statin receipt and adherence measures. This approach weights to subgroup performance so that outcomes among beneficiaries with and without social risk factors (disability or dual eligibility) contribute equally to the overall contract score. We then assessed how this method changed contract-level performance groupings and examined associations between contract characteristics and performance scores under social risk weighting.

METHODS

This cross-sectional study used 2019 MA data to evaluate how contract-level performance groupings change under social risk–weighted scoring. Data sources included the 20% national sample of 2019 MA Encounter and Part D Prescription Drug Event files, the Medicare Beneficiary Summary File, and the MA/Part D Contract and Enrollment Data. We selected 2 quality measures from the CMS Star Ratings program: statin therapy for patients with cardiovascular disease (statin receipt) and statin adherence (≥ 80% proportion of days covered) (eAppendices 1 and 2 [eAppendices available at ajmc.com]). Measure specifications followed the Healthcare Effectiveness Data and Information Set Statin Therapy for Patients With Cardiovascular Disease guidelines. Unlike CMS, which calculates adherence across all eligible beneficiaries, we limited the adherence measure to those with cardiovascular disease to ensure consistency in denominators across both measures.

The analytic sample included MA beneficiaries 65 years and older who were continuously enrolled throughout 2019. Exclusions included individuals in hospice or long-term institutional care, with advanced illness, or with contraindications to statin therapy.

Social Risk Weighting Approach

We calculated contract-level social risk–weighted performance scores using a previously proposed constrained optimization framework.¹⁵ In contrast to standard regression-based risk adjustment, which removes social risk variation from performance estimates, this method incorporates social risk directly when calculating overall scores from subgroup performance. We used 2 policy-relevant social risk factors—dual eligibility for Medicaid and Medicare and disability status (defined by original reason for Medicare entitlement)—and calculated contract-level scores that applied empirically derived weights to subgroup performance for beneficiaries with and without each risk factor. Specifically, dually eligible and non–dually eligible enrollees contributed equally to the final score, as did disabled and nondisabled enrollees. Weights were derived using quadratic programming, a mathematical approach that minimizes the difference between weighted and unweighted performance scores while meeting policy-driven constraints (see eAppendix 3). These constraints included equal contributions from each social risk group, nonnegative weights, and normalization. This method does not change how plans perform within each subgroup but instead adjusts how those performances are combined, reflecting policy priorities (ie, improving fairness) and reducing the influence of population differences on contract comparisons.

Performance Group Assignment

We converted unweighted (ie, raw aggregate) and social risk–weighted scores into categorical performance groupings using a quantile-based method. To approximate CMS rating thresholds and support interpretability, we classified contracts into 3 ordinal tiers—3 or fewer stars, 4 stars, and 5 stars—using quintile cut points, with the top quintile corresponding to 5 stars, the second quintile to 4 stars, and the bottom 3 quintiles to 3 or fewer stars. These groupings reflect commonly used thresholds in CMS reporting and bonus eligibility. We excluded observations with extreme values, defined as those falling outside 3 times the IQR from the first or third quartile. Although CMS applies a clustering algorithm to assign official Star Ratings, we used quantile-based thresholds to approximate the empirical distribution of rating tiers and preserve methodological transparency.⁹

Statistical Analysis

Descriptive statistics summarized contract characteristics across unweighted and weighted performance groups for each statin measure. We calculated the proportion of contracts that shifted groups (up, down, or unchanged) under social risk weighting. To identify predictors of score variation, we estimated linear regression models for both unweighted and weighted scores, including covariates for log contract enrollment, mean risk score, out-of-pocket maximum, disability prevalence, and dual enrollment rate (eAppendix 3). To assess predictors of performance group assignment, we used generalized ordered logistic regression models with mean marginal effects to aid interpretation, relaxing the proportional odds assumption as needed.

As a sensitivity analysis, we compared 2 alternative aggregation approaches that combined the unweighted and social risk–weighted contract scores. The first assigned twice the weight (2:1) to subgroup-specific performance among disadvantaged beneficiaries (dually eligible and disabled), increasing the influence of outcomes for high-risk enrollees on the overall contract score. The second applied equal weights (1:1) to subgroup-specific and overall scores, reflecting an even balance between subgroup and overall performance. Unlike the social risk–weighted approach, these weighting ratios were fixed and not derived through optimization. Star quintiles were reassigned under each approach and compared with the unweighted baseline to quantify movement in contract classifications (eAppendiX Table 4). All analyses were conducted using Stata 18 (StataCorp LLC) and R 4.3.3 (R Foundation for Statistical Computing), with statistical significance defined as a P value less than .05 (2-sided). The study was approved by the Johns Hopkins Bloomberg School of Public Health Institutional Review Board.

RESULTS

The final analytic sample included 610,028 MA beneficiaries eligible for statin receipt and 485,618 beneficiaries eligible for statin adherence, across 165 contracts.

Contract Characteristics by Performance Group

MA contract characteristics varied across social risk–weighted performance groups for both statin receipt and adherence (Table 1). For statin receipt, contracts with higher Star Ratings tended to have lower mean (SD) monthly enrollment: 24,185.8 (36,028.3) enrollees in contracts with 3 or fewer stars, 18,300.1 (13,571.3) in 4-star contracts, and 17,030.6 (15,907.5) in 5-star contracts. The mean disability prevalence decreased from 26.7% (13.1%) in contracts with 3 or fewer stars to 24.4% (12.7%) among 5-star contracts, whereas dual enrollment increased slightly from 15.2% (10.9%) to 17.2% (13.8%). For statin adherence, mean enrollment dropped from 23,430.1 (35,493.8) in contracts with 3 or fewer stars to 20,314.2 (14,461.4) in 5-star contracts. Disability prevalence was highest among contracts with 3 or fewer stars (28.5 [13.2]) and lowest among 5-star contracts (17.2% [7.4%]), with 4-star contracts falling in between (25.2% [8.6%]). Social risk–weighted performance scores increased monotonically with star level: from 0.73 (0.03) to 0.83 (0.03) for statin receipt and from 0.56 (0.03) to 0.67 (0.04) for adherence.

Changes in Performance Group With Social Risk Weighting

Social risk–weighted scoring reclassified 24.2% of contracts for statin receipt and 20.6% for statin adherence (Table 2 and Figure 1). For statin receipt, 19 contracts (11.5%) moved up, 21 (12.7%) moved down, and 125 (75.8%) were unchanged. For statin adherence, 12 contracts (7.2%) moved up, 22 (13.3%) moved down, and 131 (79.4%) remained the same. Contracts that moved up for statin receipt had lower mean disability (23.6%) and dual enrollment (13.0%) than those that remained unchanged (26.6% and 16.2%, respectively). For statin adherence, contracts that moved down had slightly lower disability prevalence (24.7%) than those that remained unchanged (26.3%).

Predictors of Performance Scores

Figure 2 shows associations between social risk factors and both unweighted and weighted performance scores (eAppendices 4 and 5). For statin receipt, disability prevalence was not significantly associated with either the unweighted score (β = –0.00025; 95% CI, –0.00081 to 0.00031) or the weighted score (β = –0.00047; 95% CI, –0.00113 to 0.00020). For statin adherence, disability was inversely associated with performance in both the unweighted score (β = –0.00106; 95% CI, –0.00174 to –0.00039; P = .002) and the weighted score (β = –0.00129; 95% CI, –0.00205 to –0.00052). Dual enrollment was not significantly associated with scores in either model.

Predictors of Change in Performance Group Assignment

Ordered logistic regression models identified predictors of performance group classification based on social risk–weighted scores (Table 3). A 1-unit increase in log enrollment was associated with a 9.7-percentage point (PP) decrease in the probability of receiving 5 stars for statin receipt (95% CI, –17.4 to –2.1; P = .013) and a 6.2-PP decrease for adherence (95% CI, –12.1 to –0.3; P = .039). Higher disability prevalence was also negatively associated with top performance for both measures. Each 1-PP increase in disability prevalence was associated with a 0.5-PP decrease in the likelihood of receiving 5 stars for statin receipt (95% CI, –1.0 to –0.07; P = .047) and a 2.3-PP decrease for adherence (95% CI, –2.9 to –1.6; P < .001). Disability was also positively associated with low performance: A 1-PP increase was linked to a 0.7-PP increase in the probability of receiving 3 stars or fewer for statin receipt and a 1.3-PP increase for adherence.

Sensitivity Analysis

Under the fixed 2:1 weighting approach, in which subgroup-specific performance among dually eligible and disabled beneficiaries received twice the weight of overall scores, 17% to 19% of contracts moved up, and a similar proportion moved down, whereas approximately two-thirds remained unchanged. The equal-weighted (1:1) approach produced smaller shifts, with 13% to 16% of contracts changing groups. The mean percentages of disabled and dually eligible enrollees were similar to the main analysis, and enrollee-weighted results showed that fewer than one-third of total beneficiaries were affected under either alternative scheme (eAppendix 7).

DISCUSSION

Incorporating social risk–weighted scoring meaningfully reshaped contract rankings for 2 MA Star Rating program measures but did not uniformly benefit plans serving high–social risk populations.⁹ These measures reflect both provider action and patient adherence, making them sensitive to social barriers that affect performance across Star Ratings quality domains.¹⁰ The constrained optimization approach offers policy makers greater control over the incentives encoded in Star Ratings, with methodological advantages over traditional regression-based risk adjustment, stratification, or CMS’ CAI. Unlike regression adjustment, which conditions on patient mix and removes social risk variation, the CAI applies post hoc adjustments that happen to disproportionately benefit contracts with the highest shares of dually eligible and disabled beneficiaries. As an adjustment to the overall score, the impact of subgroup-specific performance among dually eligible and disabled beneficiaries is necessarily limited when applying the CAI. Most contracts’ scores will still be dominated by their performance among non–dually eligible and nondisabled beneficiaries even after applying the CAI. Social risk weighting allows policy makers to change this calculus by specifying directly how much weight to give to these at-risk groups.^7,16 Furthermore, adjustments due to the CAI are due solely to the composition of contract beneficiaries, rather than performance among at-risk groups.

The constrained optimization method allows structured control over the weighting scheme and flexibility to align weights with evolving policy priorities.¹⁵ Although conventional case-mix adjustment could stratify and reaggregate by population prevalence, this approach would continue to favor larger, lower-risk groups such as nondisabled, non–dually eligible enrollees.^17,18 In contrast, social risk–weighted scoring allows policy makers to explicitly define optimal weights aligned with social risk factors, embedding fairness into the scoring logic and promoting improvement across diverse enrollee groups.

We found that social risk–weighted scoring reclassified 11.5% of contracts upward and 12.7% downward for statin receipt, and 7.2% upward and 13.3% downward for adherence, indicating that weighting modestly affects plan classification.¹⁹ Disability and dual enrollment rates were inversely associated with both unweighted and social risk–weighted scores. Notably, social risk weighting did not uniformly increase the ranking of contracts serving high-risk populations. In some cases, contracts with lower proportions of dually eligible or disabled enrollees moved upward. This pattern may reflect true differences in care quality or resources across plans, but it could also indicate the need for further adjustment for case mix (including additional measures of patient health status and complexity), which is possible in social risk weighting and should be included.²⁰

The effects of social risk weighting were more pronounced for statin adherence than for statin receipt, consistent with prior research showing adherence’s greater sensitivity to social barriers such as cost and transportation.¹⁵ Although score differences were modest in absolute terms, the reclassification of more than 20% of contracts across both measures suggests material impacts. Nevertheless, unmeasured social risk not captured by indicators for dual eligibility and disability surely continues to affect observed performance. Setting realistic expectations for such reforms will require acknowledging these enduring challenges. Sensitivity analyses using alternative weighting based on social risk yielded similar patterns of contract reclassification. The choice of weighting approach depends on the underlying conceptual framework. The 2:1 and 1:1 schemes represent intuitive but arbitrary alternatives that assign greater or equal weight to performance among high-risk subgroups. In contrast, the constrained optimization approach supports equal influence of subgroup performance on overall scores based on empirical distributions, whereas fixed 2:1 or 1:1 ratios do not account for subgroup size, variance, or correlation with overall performance.

We also observed distinct patterns in predictors of Star Rating reassignment. Disability prevalence remained a consistent predictor of lower ratings, suggesting that clinical and functional limitations may continue to influence measured performance even when social risk weighting is applied (eAppendiX Table 9).²¹ In contrast, dual eligibility was not significantly associated with performance group, suggesting that disability status may independently capture additional dimensions of vulnerability not reflected in low-income status alone. The method produced Star distributions comparable to CMS benchmarks yet added flexibility for aligning scoring with policy goals.²² As a transparent, scalable alternative to case-mix adjustment, social risk–weighted scoring offers a pathway to enhance fairness and encode policy priorities into plan evaluation. Rather than rewarding plans for enrollee mix alone, this method creates incentives for improvements in care delivery for at-risk groups. Policy makers could implement similar approaches through dashboards, pilot programs, or targeted payment models.⁸ Future studies should model the fiscal and enrollment implications of social risk–weighted scoring to evaluate how alternative scoring methods might redistribute bonus payments or affect plan competitiveness.

Limitations

This study has several limitations. First, the analysis focused on 2 measures—statin receipt and adherence—as a use case to examine the application of social risk weighting. Although these measures are consequential within the CMS Star Ratings program (statin receipt is weighted once and statin adherence is weighted 3 times in the overall summary score), findings may not generalize to the broader set of measures used in CMS ratings across multiple clinical and operational domains. Although we focused on 2 measures, they serve as a proof of concept. Contracts may perform better on some measures and worse on others, so the full policy impact of weighting can be assessed only across the full set of measures included in Star Ratings. Second, the adherence measure in this study was restricted to beneficiaries with cardiovascular disease, whereas the CMS measure includes all eligible beneficiaries (eAppendix 1). However, individuals with cardiovascular disease constitute the majority of statin users and represent a clinically meaningful subgroup. This consistency in denominator definition across both measures enhances comparability between receipt and adherence outcomes in our analysis. Third, the categorization of plans into Star Rating groups was based on a quantile approach rather than CMS’ clustering algorithm. Although this method ensured monotonic groupings and produced thresholds similar to those reported by CMS (as shown in the eAppendix), it may not fully replicate CMS’ star grouping methodology. However, it preserves transparency and reproducibility. Fourth, social risk weights were derived using a constrained optimization framework designed to align contract performance scores with equal subgroup representation while minimizing distortion. These are example policy priorities, and other relative weightings of social risk factor groups or alternative stratification approaches may yield different results. For instance, prioritizing racial and ethnic subgroups, area deprivation, or other indicators of social need could lead to different scoring and reclassification patterns. Finally, this was a cross-sectional analysis and did not assess the potential effects of social risk weighting on plan behavior, beneficiary outcomes, or the redistribution of potential bonus payments over time. Future longitudinal studies could examine how plans respond to alternative scoring incentives, particularly with regard to investment in care for high-need populations.

CONCLUSION

Applying social risk–weighted scoring meaningfully reclassified MA contract rankings for statin receipt and adherence. By integrating social risk directly into performance aggregation, this approach provides policy makers with a flexible, transparent framework to better align incentives with the challenges faced by health plans serving high-need populations.²³ Although social risk weighting may not eliminate structural barriers, it offers a scalable method to enhance the fairness and validity of quality measurement without altering individual measure specifications. As CMS continues efforts to modernize Star Ratings, constrained optimization–based weighting could help redesign plan evaluations to address the influence of social risk in performance measurement.

Author Affiliations: Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health (AA, RSH), Baltimore, MD; RAND Corporation (DA), Santa Monica, CA.

Source of Funding: Hopkins Business of Health Initiative and Johns Hopkins Center for Health Disparities Solutions.

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 (AA, DA); acquisition of data (AA); analysis and interpretation of data (AA, DA, RSH); drafting of the manuscript (AA, RSH); critical revision of the manuscript for important intellectual content (AA, DA, RSH); statistical analysis (AA, RSH); provision of patients or study materials (AA); obtaining funding (AA); administrative, technical, or logistic support (AA); and supervision (AA).

Address Correspondence to: Andrew Anderson, PhD, Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, 1812 Ashland Ave, Baltimore, MD 21205. Email: aander86@jh.edu.

REFERENCES

1. Newhouse JP. Risk adjustment: where are we now? Inquiry. 1998;35(2):122-131.

2. Roberts ET, Burke R, Haddad K. Medicare Advantage supplemental benefits: origins, evolution, and issues for policy making. Health Affairs Forefront. September 19, 2024. Accessed April 28, 2025. https://www.healthaffairs.org/content/forefront/medicare-advantage-supplemental-benefits-origins-evolution-and-issues-policy-making

3. Biniek JF, Damico A, Neuman T. Spending on Medicare Advantage quality bonus payments will reach
at least $12.8 billion in 2023. KFF. August 9, 2023. Accessed November 9, 2023. https://www.kff.org/medicare/issue-brief/spending-on-medicare-advantage-quality-bonus-payments-will-reach-at-least-12-8-billion-in-2023/

4. DuGoff EH, Boyd C, Anderson G. Complex patients and quality of care in Medicare Advantage. J Am Geriatr Soc. 2020;68(2):395-402. doi:10.1111/jgs.16236

5. Wright B. Structural inequities in Medicare Advantage—a growing cause for concern. JAMA Netw Open. 2024;7(7):e2424096. doi:10.1001/jamanetworkopen.2024.24096

6. CMS Framework for Healthy Communities. CMS. Accessed April 28, 2025. https://www.cms.gov/priorities/health-equity/minority-health/equity-programs/framework

7. Tolpadi A, Sorbero ME, DeYoreo M, Elliott MN, Damberg CL. Understanding the social risk factor adjustment’s effect on Star Ratings. Am J Manag Care. 2023;29(12):e372-e377. doi:10.37765/ajmc.2023.89471

8. Meyers DJ, Trivedi AN, Ryan AM. Flaws in the Medicare Advantage Star Ratings. JAMA Health Forum. 2025;6(1):e244802. doi:10.1001/jamahealthforum.2024.4802

9. 2024 Medicare Advantage and Part D Star Ratings. CMS. October 13, 2023. Accessed April 28, 2025.
https://www.cms.gov/newsroom/fact-sheets/2024-medicare-advantage-and-part-d-star-ratings

10. Adeoye-Olatunde OA, Hastings TJ, Blakely ML, Boyd L, Aina AB, Sherbeny F. Social determinants of health and medication adherence in older adults with prevalent chronic conditions in the United States: an analysis of the National Health and Nutrition Examination Survey (NHANES) 2009-2018. Pharmacy (Basel). 2025;13(1):20. doi:10.3390/pharmacy13010020

11. Gupta A, Silver D, Meyers DJ, Glied S, Pagán JA. Medicare Advantage plan Star Ratings and county social vulnerability. JAMA Netw Open. 2024;7(7):e2424089. doi:10.1001/jamanetworkopen.2024.24089

12. Johnston KJ, Bynum JPW, Joynt Maddox KE. The need to incorporate additional patient information into risk adjustment for Medicare beneficiaries. JAMA. 2020;323(10):925-926. doi:10.1001/jama.2019.22370

13. Chopra A, Boehler A, Bacher G. Risk adjustment: it’s time for reform. Health Affairs Forefront. January 9, 2023. Accessed April 28, 2025. https://www.healthaffairs.org/content/forefront/risk-adjustment-s-time-reform

14. McWilliams JM. Risk adjustment reform: navigating ideas and tradeoffs (part 1). Health Affairs Forefront. March 26, 2025. Accessed April 28, 2025. https://www.healthaffairs.org/content/forefront/risk-adjustment-reform-navigating-ideas-and-tradeoffs-part-1

15. Agniel D, Cabreros I, Damberg CL, Elliott MN, Rogers R. A formal framework for incorporating equity into health care quality measurement. Health Aff (Millwood). 2023;42(10):1383-1391. doi:10.1377/hlthaff.2022.01483

16. Jaffery JB, Safran DG. Addressing social risk factors in value-based payment: adjusting payment not performance to optimize outcomes and fairness. Health Affairs Forefront. April 19, 2021. Accessed April 28, 2025. https://www.healthaffairs.org/content/forefront/addressing-social-risk-factors-value-based-payment-
adjusting-payment-not-performance

17. Romano PS. Should health plan quality measures be adjusted for case mix? Med Care. 2000;38(10):977-980. doi:10.1097/00005650-200010000-00001

18. Salem-Schatz S, Moore G, Rucker M, Pearson SD. The case for case-mix adjustment in practice profiling: when good apples look bad. JAMA. 1994;272(11):871-874. doi:10.1001/jama.1994.03520110051028

19. Zaslavsky AM, Hochheimer JN, Schneider EC, et al. Impact of sociodemographic case mix on the HEDIS measures of health plan quality. Med Care. 2000;38(10):981-982. doi:10.1097/00005650-200010000-00002

20. Shwartz M, Restuccia JD, Rosen AK. Composite measures of health care provider performance: a description of approaches. Milbank Q. 2015;93(4):788-825. doi:10.1111/1468-0009.12165

21. Sorbero ME, Paddock SM, Damberg CL, et al. Adjusting Medicare Advantage star ratings for socioeconomic status and disability. Am J Manag Care. 2018;24(9):e285-e291.

22. Xu P, Burgess JF Jr, Cabral H, Soria-Saucedo R, Kazis LE. Relationships between Medicare Advantage contract characteristics and quality-of-care ratings. Ann Intern Med. 2015;162(5):353-358. doi:10.7326/M14-0332

23. Reid RO, Deb P, Howell BL, Shrank WH. Association between Medicare Advantage plan star ratings and enrollment. JAMA. 2013;309(3):267-274. doi:10.1001/jama.2012.173925