Health Outcomes of Socially Vulnerable Populations in Different Medicare Payment Arrangements

ABSTRACT

Objectives: Two-sided risk Medicare Advantage (MA) arrangements (at-risk MA) outperform Medicare fee-for-service (FFS) arrangements in quality and efficiency. The influence of social determinants of health (SDOH) remains unclear. Using the Social Vulnerability Index (SVI) as a proxy for SDOH, we compared outcomes for beneficiaries with varying levels of social vulnerability across at-risk MA, FFS MA, and FFS Medicare (traditional Medicare [TM]) arrangements.

Study Design: Retrospective cohort study.

Methods: We analyzed TM claims from 2019 and MA encounter data for 1,703,060 beneficiaries attributed to 16 physician groups in at-risk MA. Two models were estimated: (1) adjusting for SVI and (2) modeling SVI as interactions. We used multivariable logistic regression to analyze outcomes.

Results: At-risk MA outperformed FFS in all 8 outcomes regardless of SVI. We found that at-risk MA outperformed FFS TM in 5 of 8 outcomes for beneficiaries with higher social vulnerabilities. Compared with TM, at-risk MA beneficiaries had 13.5% to 16.6% fewer acute admissions, 2.0% to 3.6% fewer emergency department (ED) visits, 10.9% to 14.4% fewer inpatient admissions through the ED, less than 1% fewer office visits, and 23.4% to 30.0% fewer instances of high-risk medication use, with the greatest differences observed in beneficiaries living in high SVI areas. At-risk MA also exhibited better quality metrics than FFS MA, but to a lesser extent than FFS TM.

Conclusions: This observational study found that beneficiaries in at-risk MA received higher-quality, more efficient care—especially in groups with elevated social vulnerability—suggesting that these populations benefit from value-based care (VBC) arrangements. Additional research is required to identify the specific VBC components driving quality and efficiency differences.

Am J Manag Care. 2026;32(Spec. No. 6):SP258-SP266. https://doi.org/10.37765/ajmc.2026.89964

More than 50% of Medicare recipients choose Medicare Advantage (MA) over traditional Medicare (TM), although coverage and care delivery vary widely across plans.¹ MA plans typically offer enhanced coordination, caps on out-of-pocket spending, and additional benefits such as prescription drug, dental, and vision coverage that TM plans do not; however, MA plans may require use of specified networks and require prior authorization.^2,3 MA and TM plans differ in provider incentives, particularly in the degree of financial risk, if any, assumed by providers for achieving positive health outcomes.⁴ As health care policy increasingly embraces value-based care (VBC), evidence shows that 2-sided risk MA arrangements consistently yield superior health outcomes compared with fee-for-service (FFS) arrangements, including FFS MA and FFS TM.^5-7 In 2-sided risk MA arrangements, providers share in savings when patients meet or exceed health benchmarks but face financial penalties if care quality declines or costs exceed benchmarks (ie, upside and downside risk). In upside-only MA arrangements, providers may receive lower financial savings but do not face financial penalties for missing quality or cost benchmarks. Risk arrangements encourage implementation of robust care coordination, preventive care, and earlier health interventions for patients. CMS supports 2-sided risk through programs such as the Medicare Shared Savings Program and Accountable Care Organization (ACO) Realizing Equity, Access, and Community Health (REACH) in TM and through various risk-sharing models in MA.

Clinicians, researchers, and policy makers increasingly recognize that social determinants of health (SDOH)—including income, housing, education, and neighborhood characteristics—strongly impact health outcomes.⁸ Despite the growth of MA and strong support for VBC, limited evidence exists on how beneficiaries’ social vulnerability interacts with these care models. Understanding VBC performance for socially vulnerable communities is vital, as disadvantaged populations tend to choose MA over TM coverage, likely due to MA’s additional benefits and caps on out-of-pocket spending compared with FFS TM.^9,10 Data on the relationship between risk-sharing arrangements and health outcomes in these populations remain sparse. Barkett et al found that individuals from racial minority categories enrolled in MA, compared with FFS TM, received higher rates of preventive screening and primary care services and had lower rates of potentially avoidable hospital admissions.¹¹ In contrast, Jacobson et al reported fewer differences in care performance between MA and TM for underserved communities.¹²

We examined whether health outcomes varied across Medicare risk-sharing arrangements among beneficiaries living in areas with either high or low social vulnerabilities. To isolate the effects of MA plan design and provider incentives, we focused on non–dual-eligible beneficiaries, excluding the confounding influence of Medicaid coverage. This approach enables clearer assessment of how at-risk MA, FFS MA, and FFS TM arrangements perform across varying levels of social vulnerability. To our knowledge, no studies have specifically examined how SDOH may influence outcomes under at-risk MA vs FFS MA or FFS TM arrangements.

METHODS

Study Design

To evaluate whether health outcomes vary by Medicare payment arrangement across different levels of social vulnerability, we conducted an observational, retrospective cohort study of non–dual-eligible beneficiaries enrolled in at-risk MA, FFS MA, and FFS TM (hereafter, TM) in 2019. An external institutional review board (Solutions IRB) approved the study. Because the analysis used deidentified, retrospective data, it qualified as non–human participants research and was exempt from further review. The study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology reporting guideline.

Study Data

We used a subset of data from earlier published research that stemmed from 2 sources: (1) the CMS Virtual Research Data Center (VRDC), which tracks services provided under TM and MA, and (2) a nonpublic data set from 16 physician groups participating in at-risk MA payment arrangements.^7,13 All 16 participating physician groups were in 2-sided risk MA arrangements, and we labeled this cohort as at-risk MA. Detailed methods are provided in eAppendix Methods 1 (eAppendix available at ajmc.com). All beneficiaries in the study received care from physicians assigned to the 16 physician groups. The study period covered January 1 to December 31, 2019, prior to the COVID-19 pandemic and the launch of the ACO REACH Model.¹⁴

We used the Social Vulnerability Index (SVI), developed by the CDC, which is widely used in health research as a proxy for SDOH.^15,16 SVI scores range from 0.0 to 1.0, with higher values indicating greater vulnerability, and are updated biennially using census data and 5 years of American Community Survey data.¹⁷ Using SVI data scored at the census tract level—subdivisions of counties—and data from the US Department of Housing and Urban Development and US Postal Service, we crosswalked data to 5-digit zip codes assigning each beneficiary a composite SVI score for analysis.¹⁸ Using the crosswalk, we aligned 2020 SVI data to beneficiaries’ zip codes available in the CMS VRDC. CMS revised the domains in 2020 to the current themes; therefore, we used the 2020 SVI data for 2019 claims data, rather than SVI 2018 data, to reflect the most current SDOH domains. We used composite SVI scores, which comprise 16 variables grouped into 4 themes: socioeconomic status, household characteristics, race/ethnicity, and house type/transportation.

Sample and Cohorts

We limited the sample to beneficiaries living in the US who were non–dually eligible, were aligned with 1 of the 16 participating physician groups, and had MA or TM with both Part A and Part B coverage for 12 months of continuous enrollment (eAppendix Methods 1, eAppendix Figure 1, and eAppendix Table 1). Analyses of pharmacy outcome measures were restricted to MA and TM beneficiaries with Part D benefits. Comorbidities were statistically adjusted using Hierarchical Condition Categories (HCCs). We used HCC version 28 to diminish effects of potential MA coding intensity.¹⁹ Diagnoses from chart reviews were not included to mitigate potential coding pattern differences between the cohorts. Beneficiaries who switched between TM or MA (or vice versa) or with no HCCs were excluded. Beneficiaries in each cohort were categorized as living in an area with either low-to-medium (SVI of 0.00-0.50) or medium-to-high (0.51-1.00) social vulnerability.

Outcomes

Outcomes were calculated at the beneficiary level and included a combination of established health resource utilization measures and Agency for Healthcare Research and Quality Prevention Quality Indicators (PQIs) (eAppendix Table 2). We examined 8 outcomes: acute inpatient (IP) admissions, 30-day readmissions, emergency department (ED) visits, IP admissions through the ED, PQI 91–acute preventable admissions (acute composite), PQI 92–chronic preventable admissions (chronic composite), office visits, and reduction of high-risk medication use.

Statistical and Sensitivity Analyses

Analyses were conducted between January and May 2025. We used multivariable logistic regression to estimate the probability of an outcome occurring, all of which were modeled as binary indicators rather than counts. We adjusted models to account for differences in sex, age, health status (using a standardized risk HCC version 28 score), health maintenance organization status, physician group, and social vulnerability (SVI). We excluded race and ethnicity directly in our models because one of the social vulnerability measures already incorporates those factors. We ran 2 versions of each regression model: one that included social vulnerability as a covariate for standard adjustment (ie, model 1), and another regression that modeled SVI as an interaction term (ie, model 2). This allowed us to examine whether the effect of different Medicare risk arrangements on health outcomes varied between the 2 SVI groups. Details for the statistical methods and model are presented in eAppendix Methods 2.

For all analyses, we calculated and reported the marginal risk differences between the different Medicare arrangements, and results for outcomes were scaled per 1000, as some events were rare. All analyses were performed using SAS 9.4 (SAS Institute Inc). Results were statistically significant if the P value was less than .05. To account for multiple comparisons, we used a Bonferroni correction to set a stricter P-value threshold (P ≤ .00313) to account for the 8 outcomes (each with 2 SVI subgroup comparisons) to minimize false positives.

RESULTS

This study included 16 physician groups (9072 primary care physicians) and 1,703,060 beneficiaries: 52.3% in TM, 23.5% in FFS MA, and 24.2% in at-risk MA. The mean age across all groups was approximately 74 years, and women comprised more than half the sample in each group. The mean composite SVI scores were 0.51, 0.60, and 0.64 for TM, FFS MA, and at-risk MA, respectively. Nearly 68% of beneficiaries in the at-risk MA group lived in higher-SVI regions compared with approximately 63% in the FFS MA group and 50% in the TM group. The Table (part A and part B) summarizes additional cohort characteristics. Unadjusted results are presented in eAppendix Table 3. Figure 1 presents a bar graph showing how the composite SVIs were distributed across each Medicare arrangement (see eAppendix Figure 2 for specific SVI domains).

Using our first model, which adjusted for social factors using SVI as a standard variable, we found significant differences in all 8 health outcomes between beneficiaries in at-risk MA and those in FFS plans (FFS MA or TM), with the most pronounced differences between at-risk MA and TM (average effects in Figure 2 and eAppendix Table 4). When we modeled SVI as an interaction term to explore differences between beneficiaries in the 3 payment arrangement groups, we examined more closely whether social factors influenced the performance of risk arrangements. We found that 5 of the 8 outcomes were statistically different across the 3 cohorts: acute IP admissions, ED visits, IP admission through ED, office visits, and high-risk medication use (subgroups SVI in Figure 2 and eAppendix Table 5). There were no significant differences in 30-day readmissions, PQI 91–acute composite conditions, and PQI 92–chronic composite conditions.

Next, we examined how outcomes compared between the different risk arrangements at lower and higher social vulnerability levels. Between at-risk MA and TM, at-risk MA beneficiaries exhibited significantly fewer acute IP admissions, corresponding to a 13.5% and 16.6% reduction in hospitalizations in the lower and higher SVI groups, respectively. ED visits were also lower among at-risk MA beneficiaries, indicating 2.0% and 3.6% reductions in ED visits in the lower and higher SVI groups, respectively. We observed a 10.9% and 14.1% reduction of IP admissions through the ED for the lower and higher SVI groups, respectively. Office visits were only modestly lower in the at-risk MA group, with less than 1% reduction in both lower and higher SVI groups. Use of high-risk medications was substantially lower among at-risk MA beneficiaries compared with TM, with 23.4% and 30.0% reductions in the lower and higher SVI groups, respectively (Figure 2 and eAppendix Table 5).

Between at-risk MA and FFS MA, at-risk MA beneficiaries exhibited 3.5% fewer acute IP admissions in both the lower and higher SVI groups. For the outcome of IP admissions through the ED, both the lower and higher SVI groups exhibited significant differences, with at-risk MA groups exhibiting 4.5% and 6.1% reductions, respectively. Differences in office visits were negligible, with the higher SVI group showing a less than 1% reduction for at-risk MA. Use of high-risk medication was significantly different only in the higher SVI group, where we noted a 6.0% reduction for at-risk MA (Figure 3 and eAppendix Table 5). Results for comparisons between the 2 FFS cohorts (TM and FFS MA) are provided in eAppendix Figure 3 and eAppendix Table 5.

Sensitivity Analysis

We conducted sensitivity analyses using 2 models—average effects and SVI interactions—on dual-eligible beneficiaries (eAppendix Tables 6 and 7). Results for comparisons between the 3 cohorts (model 1) remained statistically significant, mirroring the results of the primary analyses with the strongest differences observed in at-risk MA vs TM cohorts. The sensitivity analysis for the dual-eligible cohort aligned with the primary analysis for average effects, but significant differences were observed only for reductions in ED and high-risk medication use in the interaction analyses trending toward at-risk MA exhibiting more efficient care.

DISCUSSION

In this analysis, we examined 2019 health outcomes for non–dually eligible beneficiaries in at-risk MA, FFS MA, and TM arrangements. The sample was drawn from previously published multiyear cohorts that found that at-risk MA arrangements generally outperformed TM and FFS MA arrangements in quality and efficiency metrics; however, those did not assess whether these performance differences varied by beneficiaries’ socioeconomic and demographic characteristics. To address this gap, we evaluated the differential impact of risk-sharing arrangements on beneficiaries living in regions with lower vs higher social vulnerability, using the SVI as a proxy tool to measure SDOH. We found that overall, and notably for beneficiaries living in areas with higher social vulnerability, at-risk risk MA exhibited higher-quality care and efficiency compared with either FFS MA or TM, suggesting that a fully accountable VBC model (ie, at-risk MA) exhibited more-efficient and higher-quality performance than FFS models in this observational sample.

The interaction models assessed whether social vulnerability modified the observed relationship between at-risk MA arrangements and positive health outcomes compared with the FFS arrangements. For outcomes with significant interactions, the data revealed that beneficiaries enrolled in at-risk MA plans with higher social vulnerability scores exhibited better care quality and utilization than those in either FFS MA or TM. Reductions in acute IP admissions, ED visits, inpatient admission through the ED, and usage of high-risk medications were most pronounced in the at-risk MA beneficiaries who lived in more socially vulnerable regions. Although office visits were reduced in the at-risk MA beneficiaries, these differences were less than 1%. The remaining outcomes (30-day readmission, PQI 91, and PQI 92) exhibited no significant differences between SVI groups, suggesting comparable care quality. In this study, FFS MA generally outperformed TM, but neither FFS arrangement outperformed at-risk MA, reinforcing the potential of 2-sided risk VBC models to improve care quality. The sensitivity analysis for the dual-eligible cohort held for broader average effects comparison, but only 2 of the outcomes (reductions in ED visits and in high-risk medication use) exhibited significant differences between SVI groups.

Unlike FFS models that incentivize volume, at-risk MA aligns payment with outcomes, which encourages plans and providers to focus on preventive care, care coordination, and efficient resource use. The observations found in this study suggest that at-risk MA arrangements may be more effective at delivering high-quality care in socially vulnerable communities, potentially due to stronger provider incentives for care coordination, preventive services, and population health management. Future research should explore alternative VBC models (ie, ACO REACH) and the specific mechanisms within VBC models—such as care coordination strategies, provider incentives, and community partnerships—that may contribute to higher quality and efficiency outcomes in high-vulnerability settings.

Limitations

This study has several limitations. The data on snowbird status are not available because our data provide only a single beneficiary zip code per year, so we cannot track month-to-month changes. We may not have fully accounted for observed and unobserved differences between cohorts or individual physician practices. Our sample skewed toward regions with higher SVI scores, limiting generalizability and introducing potential selection bias; prior research described a similar trend of MA enrollees being associated with higher social vulnerabilities compared with TM enrollees, suggesting that this study follows a previously noted trend.¹⁰ Geographic coverage was restricted to participating physician groups. When 5-digit zip code data were crosswalked to SVI census tracts, there was a possibility that some zip codes were misaligned, and SVI data may misclassify individual beneficiary vulnerability status, increasing the risk for ecological fallacy. Using the composite SVI score limits insight into specific SVI domains. As our measurement year was limited to 2019, the study is also subject to temporal bias. As this was a claims-based study, there is potential of missing data; however, our methods reflect standard approaches to mitigate this limitation. Lastly, our results reflect observed associations and do not establish causal relationships between Medicare arrangements and beneficiary outcomes, including those stratified by SVI levels.

CONCLUSIONS

This study highlights the potential of at-risk MA arrangements to deliver higher-quality and more-efficient care in socially vulnerable, non–dually eligible beneficiaries who are likely to face additional care barriers due to lack of access to dual-eligible benefits (eg, Medicaid). By emphasizing care coordination and aligning financial incentives with outcomes, at-risk MA models demonstrate strong performance across recognized indicators of quality and efficiency. These results support the continued expansion and refinement of VBC strategies to better address the complex needs of populations with elevated social risks. /

Acknowledgments

The authors thank Dr Chris Hane for sharing his expertise in SVI data. They also thank Amy Okaya for her critical review of the manuscript.

Author Affiliations: Optum Center for Research and Innovation (KCo, KCa, OA, MSJ, BN), Eden Prairie, MN; America’s Physician Groups (JP, SD), Washington, DC; CareJourney by Arcadia (CF), Arlington, VA.

Source of Funding: None.

Author Disclosures: Dr Cohen is an employee of Optum Health, which participates in both Medicare Advantage and traditional Medicare; he owns stock in UnitedHealth Group, is a member of the America’s Physician Groups board, and attends the American Medical Group Association and America’s Physician Groups national conferences. Drs Catlett and Omeli are employees of Optum Health and own stock in UnitedHealth Group. Ms Podulka is employed by America’s Physician Groups. Ms Jarvis is employed by Optum Health. Ms Dentzer is the president and CEO of America’s Physician Groups and an ex officio member of the organization’s board of directors. 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 (KCo, KCa, JP, OA, MSJ, SD); acquisition of data (KCo, OA, CF); analysis and interpretation of data (KCo, KCa, JP, OA, CF, BN, SD); drafting of the manuscript (KCo, KCa, JP, OA, MSJ, BN, SD); critical revision of the manuscript for important intellectual content (KCo, KCa, JP, OA, BN, SD); statistical analysis (OA, CF); administrative, technical, or logistic support (KCa, MSJ, SD); and supervision (KCo, CF).

Send Correspondence to: Kenneth Cohen, MD, Optum Center for Research and Innovation, 11000 Optum Circle, Eden Prairie, MN 33554. Email: ken.cohen@optum.com.

REFERENCES

1. Neuman T, Freed M, Biniek JF. 10 reasons why Medicare Advantage enrollment is growing and why it matters. KFF. January 30, 2024. Accessed February 6, 2024. https://www.kff.org/medicare/issue-brief/10-reasons-why-medicare-advantage-enrollment-is-growing-and-why-it-matters/
2. Freed M, Biniek JF, Damico A, Neuman T. Medicare Advantage in 2024: enrollment update and key trends. KFF. August 8, 2024. Accessed September 25, 2025. https://www.kff.org/medicare/medicare-advantage-in-2024-enrollment-update-and-key-trends/
3. Medicare Advantage. In: Medicare Payment Advisory Commission. July 2025 Data Book: Health Care Spending and the Medicare Program. Medicare Payment Advisory Commission; 2025: 129-143. Accessed September 30, 2025. https://www.medpac.gov/document/july-2025-data-book-section-9-medicare-advantage/
4. Jacobson G, Cicchiello A, Sutton JP, Shah A. Medicare Advantage vs. traditional Medicare: how do beneficiaries’ characteristics and experiences differ? The Commonwealth Fund. October 14, 2021. Accessed September 30, 2025. https://www.commonwealthfund.org/publications/issue-briefs/2021/oct/medicare-advantage-vs-traditional-medicare-beneficiaries-differ
5. Gondi S, Li Y, Drzayich Antol D, Boudreau E, Shrank WH, Powers BW. Analysis of value-based payment and acute care use among Medicare Advantage beneficiaries. JAMA Netw Open. 2022;5(3):e222916. doi:10.1001/jamanetworkopen.2022.2916.
6. Cohen K, Ameli O, Chaisson CE, et al. Comparison of care quality metrics in 2-sided risk Medicare Advantage vs fee-for-service Medicare programs. JAMA Netw Open. 2022;5(12):e2246064. doi:10.1001/jamanetworkopen.2022.46064
7. Cohen KR, Vabson B, Podulka J, et al. Medicare risk arrangement and use and outcomes among physician groups. JAMA Netw Open. 2025;8(1):e2456074. doi:10.1001/jamanetworkopen.2024.56074
8. Social determinants of health. Healthy People 2030. Accessed September 30, 2025. https://odphp.health.gov/healthypeople/priority-areas/social-determinants-health
9. Bilder S, Brot-Goldberg Z, Jones B, et al. Harvard-Inovalon Medicare study: who enrolls in Medicare Advantage vs. Medicare fee-for-service. Inovalon. June 2023. Accessed September 30, 2025. https://www.inovalon.com/resource/harvard-inovalon-medicare-study-who-enrolls-in-medicare-advantage-vs-medicare-fee-for-service/
10. ATI Advisory. Comparing Medicare Advantage and FFS Medicare across race and ethnicity. Better Medicare Alliance. July 2023. Accessed September 25, 2025. https://bettermedicarealliance.org/wp-content/uploads/2023/08/ATI_BMA-Chartbook-Comparing-MA-and-FFS-Across-Race-and-Ethnicity_FIN.pdf
11. Barkett J, Tabak R, Riley C. Black, Hispanic, and Asian American/Pacific Islander MA beneficiaries receive more primary care and less potentially avoidable care than similar beneficiaries in traditional Medicare. Berkeley Research Group. April 2025. Accessed September 30, 2025. https://media.thinkbrg.com/wp-content/uploads/2025/04/24142844/HTS_BMA-White-Paper-Race-Ethnicity-in-MA_2025.pdf
12. Jacobson G, Cicchiello A, Sutton JP, Shah A. Medicare Advantage vs. traditional Medicare: how do beneficiaries’ characteristics and experiences differ? The Commonwealth Fund. October 14, 2021. Accessed September 30, 2025. https://www.commonwealthfund.org/publications/issue-briefs/2021/oct/medicare-advantage-vs-traditional-medicare-beneficiaries-differ
13. Vabson B, Cohen K, Ameli O, et al. Potential spillover effects on traditional Medicare when physicians bear Medicare Advantage risk. Am J Manag Care. 2025;31(8):390-396. doi:10.37765/ajmc.2025.89686
14. Accountable Care Organization (ACO) Realizing Equity, Access, and Community Health (REACH) model. CMS. February 24, 2022. Accessed September 30, 2025. https://www.cms.gov/newsroom/fact-sheets/accountable-care-organization-aco-realizing-equity-access-and-community-health-reach-model
15. Flanagan BE, Gregory EW, Hallisey EJ, Heitgerd JL, Lewis B. A social vulnerability index for disaster management. J Homel Secur Emerg Manag. 2011;8(1):0000102202154773551792. doi:10.2202/1547-7355.1792
16. Mah JC, Penwarden JL, Pott H, Theou O, Andrew MK. Social vulnerability indices: a scoping review. BMC Public Health. 2023;23(1):1253. doi:10.1186/s12889-023-16097-6
17. Social Vulnerability Index (SVI). CDC. Updated July 22, 2024. Accessed October 3, 2025. https://www.atsdr.cdc.gov/place-health/php/svi/index.html
18. HUD USPS zip code crosswalk files. US Department of Housing and Urban Development, Office of Policy Development and Research. Accessed October 3, 2025. https://www.huduser.gov/portal/datasets/usps_crosswalk.html
19. Albanese A, Aramanda A, Brooks J, Klomp C. An updated analysis of coding pattern differences in Medicare Advantage. Health Aff Sch. 2026;4(1):qxag010. doi:10.1093/haschl/qxag010