INTRODUCTION

Concerns that pay-for-performance (P4P) and other value-based payment (VBP) programs financially disadvantage healthcare systems disproportionately caring for socially at-risk patients predate the 2010 Patient Protection and Affordable Care Act (ACA).1 These concerns have persisted as the ACA ushered in a new era of VBP.2, 3 VBP programs uniformly require risk adjustment to ensure performance and payments account for clinical complexity (primarily done at the beneficiary level). However, risk adjustment for social complexity (at the beneficiary or population level) has been far more controversial and is largely not utilized. Although unmet health-related social needs, driven by social determinants of health, have been clearly associated with both higher resource utilization and worse outcomes,4 concern that social risk adjustment may be perceived as lowering healthcare quality standards for socially vulnerable populations has limited its adoption.5

Yet, by not accounting for social factors, healthcare systems that care for vulnerable patients have been penalized by these models. The first two mandatory VBP programs launched by the Centers for Medicare and Medicaid Services (CMS) in the ACA era were the Hospital Value-Based Purchasing Program (HVBP) and the Hospital Readmission Reduction Program (HRRP) in 2012. Safety-net hospitals (frequently defined as those in the highest quartile of disproportionate share hospital patient percentage) were more frequently financially penalized under both programs.6,7,8,9 The Physician Value-Based Modifier (PVBM) began in 2014 with increasing magnitude of possible payment adjustments,10, 11 and safety-net systems again fared worse.12, 13 However, observational studies lacked the ability to assess whether lower performance was due to inadequate accounting of social complexity, lower quality of care, or both.

In 2014, the Improving Medicare Post-Acute Care Transformation Act (IMPACT) required the Office of the Assistant Secretary for Planning and Evaluation (ASPE) at the US Department of Health and Human Services to study the effect of social risk on quality and resource utilization in CMS’ VBP programs,14 leading to reports being published by multiple organizations between 2016-2020, including the National Quality Forum, National Academy of Medicine, and ASPE.15,16,17,18,19,20

As these evaluations were being conducted, the Medicare Access and CHIP Reauthorization Act (MACRA) was passed in 2015. This act created the Quality Payment Program (QPP) which further shifted Medicare from paying for volume to value.21 This has led to a new wave of payment programs based on quality and resource utilization, with increasing magnitudes of payment adjustments, thus increasing concerns of further disadvantage for safety-net systems and the patients they serve.22 In 2017 under the QPP, CMS launched the Merit-Based Incentive Payment System (MIPS), a P4P program which adjusts Medicare Fee-For-Service payments, and is the largest Medicare VBP program to date. MIPS is slated for complete enactment by 2023, at which time the maximum negative payment adjustment will be −9% and the maximum possible positive payment adjustment will be +37% of Medicare part B revenue.23 Since its inception, CMS has largely declined to account for social risk, stating plans to reconsider once ASPE’s final report was released in 2020.24 Initial studies published in 2020 demonstrated safety-net systems scored worse and received more penalties in MIPS,25, 26 similar to previous programs. In final rulemaking for 2021, CMS again deferred further decisions on accounting for social risk,27 despite the breadth of evidence suggesting its disproportionate impact on safety-net systems, increasing recommendations from policy reports and professional societies to account for social risk, and emerging guidance on how best to do so. In final rulemaking for 2022, CMS acknowledged the need to place more focus on equity but has not made long-term decisions on how to account for social risk in VBP programs.28

In this paper, we review the evidence to date evaluating the impact of mandatory Medicare VBP programs on safety-net systems, major policy reports regarding social risk in VBP, current policy options, and recommendations on behalf of the Society of General Internal Medicine (SGIM) to account for social risk in VBP programs.

METHODS

SGIM represents more than 3,000 of the world’s leading academic general internists, who are dedicated to improving access to care for all populations, eliminating healthcare disparities, and creating a just system of care where all people can achieve optimal health. This position statement was developed jointly by the SGIM Health Policy Committee and Health Equity Commission. We reviewed existing Medicare policy, the peer-reviewed literature, and relevant policy reports. The paper was reviewed by SGIM Council and approved on March 23, 2022.

Presently, dozens of Medicare VBP models exist, most of which are voluntary pilots within the CMS Innovation Center.29 We focused on mandatory CMS programs for which an evidence base exists in the peer-reviewed literature evaluating differences in the financial impact between safety-net and non-safety-net systems. We searched the MEDLINE database via PubMed using free text search terms for the names and acronyms of these programs: “hospital readmission reduction program,” “HRRP,” “hospital value-based purchasing,” “HVBP,” “hospital-acquired condition reduction program,” “HACRP,” “physician value-based modifier,” “PVBM,” and “merit-based incentive payment system” from 2012 (the year the first program was launched) through October 2021. We screened titles for relevance to safety-net systems.

REVIEW OF THE LITERATURE

Table 1 summarizes the studies of payment equity in mandatory Medicare VBP programs. Figure 1 shows program enactment and maximum possible payment adjustments over time.

Table 1 Literature Investigating Differences in Payment Penalties Between Safety-Net and Non-safety-Net Systems in Mandatory Medicare Value-Based Programs
Figure 1
figure 1

Maximum payment adjustments of mandatory Medicare value-based programs enacted since the Affordable Care Act.

ACA Era Programs

After passage of the ACA, CMS launched four mandatory P4P programs linking quality measures to payment modifications.30 Three focused on payments to hospitals for acute care — HVBP and HRRP both launched in 2012, and the Hospital-Acquired Conditions Reduction Program (HACRP) in 2015. PVBM, launched in 2015, focused on physician payments including ambulatory care.

Under HVBP, 2% of Medicare payments are withheld from hospitals and redistributed based upon their performance on quality measures. Hospitals are scored on measures across multiple domains including clinical outcomes (e.g., mortality), patient safety (e.g., complications and healthcare-associated infections), patient experience, and efficiency/cost reduction.31 CMS adjusts payments to hospitals based on their total performance score in relation to all other hospitals and degree of improvement from their baseline. Studies have consistently demonstrated that safety-net hospitals fared worse in HVBP, with significantly more negative payment adjustments.7,8,9 Hospitals caring for higher proportions of black patients were also penalized more frequently.32

The HRRP deducts up to 3% of payments according to hospital performance in 30-day readmissions over a 3-year period based upon six hospital conditions/procedures: acute myocardial infarction, coronary artery bypass surgery, chronic obstructive pulmonary disease, heart failure, pneumonia and elective total hip or knee arthroplasty.33 Safety-net hospitals and hospitals caring for higher proportion of black patients were found to have a greater likelihood of receiving the highest penalty under HRRP.6, 9, 32 These disparities were in-part addressed by the 21st Century Cures Act of 2016 that mandated that HRRP use a stratified methodology to evaluate hospital performance relative to other hospitals within the same quintile of the proportion of dual eligible Medicare/Medicaid patients served (peer grouping).33 This went into effect in 2019 and has led to significantly reduced financial penalties for safety-net hospitals.34

The HACRP deducts 1% of Medicare payments from hospitals ranking in the worst performing quartile across six measures: one claims-based composite measure of patient safety and five chart-abstracted measures of healthcare-associated infections: central line–associated bloodstream infection, catheter-associated urinary tract infection, surgical site infection for abdominal hysterectomy and colon procedures, methicillin-resistant Staphylococcus aureus bacteremia, and Clostridium difficile infections.35 Safety-net hospitals and hospitals serving higher proportions of racial and ethnic minorities were more frequently penalized.32, 36, 37 Peer grouping has not been applied to this program as it has been for HRRP. However, modeling studies suggest a significant reduction in penalties to safety-net hospitals if this methodology were applied to HACRP.38

Under PVBM, physician payments were adjusted based on performance on measures across multiple quality (clinical care, person and caregiver experience, community/population health, patient safety, communication/care coordination and efficiency) and cost domains.39 One study demonstrated that systems serving higher proportions of dual eligible Medicare/Medicaid patients scored worse on quality measures and had higher financial penalties.12 Another study found the performance difference between low versus high social risk (highest quartile of dual eligible Medicare/Medicaid patients) systems was substantially reduced when further adjusted for measures of medical and social complexity, concluding that programs with inadequate social and clinical risk adjustment can lead to worsening disparities in payment.13 PVBM ended in 2019 and was replaced by MIPS.

MACRA Era Programs

The passage of MACRA in 2015 led to the creation of the QPP with two tracks — MIPS, and the Advanced Alternative Payment Models. Under MIPS, providers (including physicians and other clinicians) are scored in four domains: quality, cost, improvement activities, and promoting interoperability. Quality measures are selected by the provider or healthcare system and must be submitted to CMS, although this has been criticized for being burdensome and complex.19 CMS’ own data demonstrate providers in rural and small practices fare worse under MIPS.40 Recent studies have demonstrated that providers in safety-net systems and those serving the highest proportion of dual eligible Medicare/Medicaid patients performed worse and had more frequent financial penalties in MIPS, just as they did under the legacy PVBM program upon which MIPS was built.25, 26, 41 Providers in systems serving the highest proportion of racial and ethnic minority patients also fared worse.42

REVIEW OF MAJOR POLICY REPORTS

Table 2 summarizes major policy reports regarding measurement and accounting for social risk factors in Medicare’s VBP programs. The recommendations are not uniform, but some themes are clear. They include the need for better measurement of social risk data, the need to create equity-focused outcome measures, and the need to change incentive payment structures to support safety-net systems.

Table 2 Summary of Major Findings and Recommendations of Policy Reports Focused on Medicare Value-Based Payment Programs and Social Risk Factor Adjustment

In ASPE’s 2016 Report to Congress, two major findings were noted. First, patients with social risk factors had worse outcomes on many quality measures regardless of provider characteristics, and that dual Medicare/Medicaid enrollment status was the most powerful predictor of poor outcomes. Second, providers that disproportionately cared for patients with social risk factors had worse performance on quality measures, and these providers had higher penalties under all existing VBP programs.15

In their 2017 report, the National Quality Forum evaluated hundreds of quality measures for consideration of social risk adjustment and tested adjustment for a subset of measures with a variety of social risk variables (e.g., race, ethnicity, dual eligible Medicare/Medicaid status, income, education). Adjustment for patient-level social risk variables largely did not impact predictions on quality and resource utilization. However, they note substantial limitations in obtaining meaningful social risk data at the patient-level.16 Their subsequent Roadmap for Promoting Health Equity and Eliminating Health Disparities recommended increased measurements of social risk factors, implementation of evidence-based methods to reduce disparities, creation of health equity measures, and changes to financial payments to support safety-net systems.17

In their 2017 report, the National Academy of Medicine noted that VBP programs do not account for the role of social risk factors in healthcare outcomes, which may exacerbate healthcare disparities by disincentivizing providers from managing high social-risk populations.18 They provided detailed guidance on social risk adjustment methodology, including adjustment for dual Medicare/Medicaid eligibility, income, education, neighborhood deprivation (composite indices based on census data), self-reported race and ethnicity, marital status, and homelessness. Social risk factors could be addressed by stratified public reporting, adjustment in performance measure scores, direct adjustments of payments, and restructuring payment incentive design.

In 2018, the Medicare Payment Advisory Commission further underscored the point that social risk factors should be accounted for in payment adjustments in VBP programs. The Commission specifically recommended comparing quality and outcomes within peer groups of healthcare systems which serve a larger number of low-income or high-risk patients as a way to avoid financially penalizing these systems.19

ASPE’s 2020 final report reconfirmed that dual Medicare/Medicaid eligibility status was the most powerful existing patient-level predictor of social risk. The report further stated that social risk information is not routinely or systematically collected and is not standardized, limiting the development of processes and models for incorporating individual-level social risk in payment models. The committee called for standardizing, measuring, and reporting social risk factors, continuing to develop strategies for addressing social risk factors, and encouraging clinicians to build community links to services to address social needs. They recommended that resource use and patient experience measures should adjust for social risk factors in VBP programs, but quality measures should not.20

GAPS IN THE EVIDENCE AND POSSIBLE FUTURE DIRECTIONS

While there are methodologic differences and limitations in the studies conducted to date, they have consistently demonstrated that safety-net systems face higher financial penalties in Medicare’s VBP programs. These observational studies cannot identify precisely why safety-net systems fare worse, but there is little doubt that they are penalized for disproportionately caring for populations at high social risk. Yet, how best to account for social risk in payment models remains unclear, due to limited data collection and evaluation to date. Dual Medicare/Medicaid eligibility is the only current variable used in CMS’ mandatory programs but is an imperfect measure of social risk — Medicaid eligibility differs by state, and healthcare systems in non-expansion states are at an inherent disadvantage. Moreover, eligibility for Medicaid is an insensitive measure of poverty.43

Nonetheless, models using additional social risk data do exist. For example, the Massachusetts Medicaid program tied P4P incentives to hospitals that reduced racial and ethnic disparities across various quality measures44. More recently, Massachusetts incorporated housing status and neighborhood stress into risk adjustment methodology for prospective payments,45 while the Department of Veteran’s Affairs accounts for homelessness in its global budget mode.46

Methods to detail social risk exist, but issues remain. Z-codes are sets of International Classification of Disease codes that were introduced in 2015 to document social risk. However, use of these codes remains low.47 Increased collection and documentation of individual-level socioeconomic data is also not without concern. This can be a sensitive topic and, if collected without care, may distress patients already affected by bias, barriers, and structural racism. An alternative approach to using individual social risk factors is the use of census data to evaluate the social vulnerability of the communities where individuals live, which would not require further data collection from patients. Examples of such area-level, neighborhood indices include the Neighborhood Stress Score,45 Area Deprivation Index,48 and Social Vulnerability index.49 Numerous studies have evaluated the potential use of these indices in Medicare payments,44, 50, 51 demonstrating their utility in predicting outcomes and improving equity in payments to safety-net systems for specific programs such as HRRP and HVBP.52,53,54 Such indices are publicly available and only require patient address for use.

Given the current lack of granular social risk data, adjusting the financial incentive structure of payment models is another potential means to address the penalization of safety-net systems. Despite its limitations, peer grouping based on proportion of dual eligible Medicare/Medicaid beneficiaries reduced the penalizations for safety-net systems in HRRP after the policy was implemented in 2019.34 Other policies such as the MIPS complex patient bonus also exist, which again uses proportion of dual eligible Medicare/Medicaid patients as the measure for overall social complexity of the population served. However, this bonus has thus far been of insufficient magnitude to provide meaningful balance to penalties incurred by safety-net systems.25 If properly designed, such an approach could also succeed in improving equity in MIPS payment adjustments. However, identifying the degree of payment adjustment is challenging, and this issue would be eliminated by peer grouping.

RECOMMENDATIONS FROM SGIM

The evidence that safety-net systems fare worse in Medicare’s VBP programs is clear. By not acknowledging or accounting for social risk factors that are beyond the control of those providing care, current payment models contribute to disparities through the “Reverse Robin Hood” effect — financially penalizing already under-resourced safety-net systems that serve vulnerable populations. They also create a disincentive to provide care to these populations. Policy makers must change the structural inequities in these regressive payment programs and must do so without further delay. SGIM strongly urges simultaneous action to immediately change current policies to financially support safety-net systems and the populations they serve, while also taking longer term approaches to determine best practices for addressing social risk and unmet health-related social needs, with the overall goal of improving equity in VBP programs. ASPE’s 2020 final report, as well as other policy institutes,55, 56 has made similar recommendations. However, they have not yet been implemented and these recommendations do not go far enough in addressing social risk and equity in VBP.

SGIM recommends the following changes:

  1. 1.

    Make equity an explicit goal for Medicare VBP

The pursuit of equitable health outcomes regardless of race, ethnicity, social or medical vulnerability should be a core goal of VBP. This must be articulated clearly in the motivations for transforming American healthcare from volume to value-based. CMS has recently made this explicit in their goals,57 but not in their program rules. Their policies must match their goals.

  1. 2.

    Make immediate changes to Medicare VBP programs to level the playing field

While Medicare’s current P4P programs are widely seen as an on-ramp for advanced payment models, CMS has not given clear indication of when these programs will end. Structural inequities within these programs must be proactively ended now. While multiple options exist to improve equity within these programs, peer grouping based on dual eligible status provides the best option to immediately mitigate the current inappropriate financial penalization of safety-net systems, and CMS should implement it across its mandatory programs. While dual eligible status remains an imprecise measure of social risk at the patient level, it can serve as a proxy for an overall population of patients served and allow for a fairer comparison of healthcare system performance within strata of social complexity, until a time when optimal methods for accounting for social risk are determined.

  1. 3.

    Move towards area-level indices of neighborhood vulnerability

Area-level indices using census data to evaluate the social vulnerability of communities are immediately available, do not require additional individual patient data to be captured, and have been used in various studies of Medicare payment models.44, 47,48,49,50,51 Such indices should be employed to improve social risk adjustment, and programs such as the Maryland Primary Care Program and ACO Reach model have begun to move in this direction.58, 59 Neighborhood indices can be used to risk-adjust individual quality measures at the patient level, or to directly adjust payments based on the overall population served by a healthcare system. Such indices can be used until individual social risk data can be collected in a standardized and sensitive manner, and optimal methods to use such data are determined. There may be ongoing benefits to using neighborhood indices even after individual data become available, as relying on collecting individual data would likely disadvantage vulnerable patients who may be less likely to access care and report social needs.60

  1. 4.

    Implement a standard approach to collecting social risk data

A standard approach to assessing social risk is critical. However, the collection of socioeconomic data at the patient level in a manner which is both sensitive and robust requires time, resources, and care. Promoting collection of this data without a well-established framework and trained staff can have negative consequences including inaccurate data collection and patient mistrust (particularly among individuals historically traumatized by the healthcare system). Social and demographic data must be discussed with trusted healthcare team members in the context of community resources and partnerships for the sake of addressing unmet health-related social needs, not primarily for financial and administrative purposes. Models such as the Accountable Health Communities and Maryland Total Cost of Care have begun to move in this direction.61,62,63

  1. 5.

    Develop a research agenda to identify and evaluate appropriate methods to adjust for social risk in payment programs

The lack of patient-level social risk data has hampered efforts to create and evaluate methodologies to equitably finance value-based healthcare. At present it is unclear whether the use of social risk data to adjust quality measures scores or to directly adjust payments via peer grouping, bonus point systems or alternative methods would best address the funding inequities in current VBP programs. Furthermore, as mainstream Medicare VBP programs move away from P4P, accounting for social risk to determine appropriate prospective payments will be critical. A research agenda, with support and funding from the Agency for Healthcare Research and Quality, must be developed to (1) assess existing methods aiming to account for social risk in VBP and (2) develop and validate new methods to expand risk adjustment methods, which link individual with area-level risk.

  1. 6.

    Develop health equity quality measures

While the above recommendations can serve to level the playing field such that safety-net systems are not inappropriately financially penalized, they do not create incentives to improve care for at-risk patients. To achieve this, health equity measures must be developed, validated, and tested in real-world settings. Stratifying traditional quality measures by race and ethnicity has been proposed to identify and incentivize closure of care gaps for traditionally marginalized groups. However, such an approach alone would not identify or address the social factors which have led to poor health outcomes. Collecting social data should first and foremost be for the purpose of screening the unmet social needs of patients (e.g., housing insecurity, food insecurity) so that they can be addressed. As such, SGIM supports the creation of health equity measures that specifically evaluate the care being provided to address these needs. Drawing from examples such as the Accountable Health Communities model,64 the Johns Hopkins community health partnership,65 and National Quality Forum roadmap,17 SGIM believes measures must be developed to evaluate aspects of care delivery across the entire spectrum of healthcare, such as the following: screening for unmet health-related social needs and referral/receipt of social services; meaningful completion of Community Health Needs Assessment (required for hospital non-profit status) and subsequent actions taken; level of community partnerships, including collaboration with county and state health departments; demonstration of language-concordant care delivery; level of investment in social services; and implementation of evidence-based approaches to reducing disparities.

  1. 7.

    Create policies which integrate healthcare and social services

The creation of health equity measures can incentivize focus on social needs and community partnerships. However current referral mechanisms can be cumbersome, slow, and challenging to track. Policies are needed to promote coordination of health and social services, including support for better data-tracking systems. Better integration has been demonstrated to improve health outcomes and resource utilization.66

  1. 8.

    Support policies that address upstream social determinants of health

Changes to VBP programs can incentivize healthcare systems to improve care for vulnerable populations. However, such interventions still remain downstream of the factors which have driven inequities in the first place.67 States and countries with a higher ratio of spending on social services to healthcare services have better health outcomes.68, 69 While payment reform to allow healthcare dollars to be redistributed to address unmet health-related social needs has been piloted in Medicaid waivers,70 the Accountable Health Communities model and some private sector models,71 over-medicalizing social issues may have unintended consequences such as linking housing or others supports to insurance coverage/eligibility. Ultimately, adverse social determinants of health are a consequence of long-standing laws, policies, cultures, and institutions derived from our nation’s history of racism and exclusion.72 As such, direct anti-poverty policies are likely to have far more impact on improving health and equity than adjustments to Medicare’s VBP programs.

CONCLUSION

Current mandatory Medicare VBP programs lead to disparate financial penalties of safety-net systems, creating a Reverse Robin Hood effect where funds are redirected from already under-resourced safety-net systems to well-resourced ones serving less vulnerable populations. Steps can be taken at many levels to improve equity, support safety-net systems, and reduce health disparities. Immediate steps to account for social risk include peer grouping according to proportion of dual eligible Medicare/Medicaid patients served. Short-term steps include using area-level indices of neighborhood deprivation to account for social risk and developing standardized approaches to collecting individual-level socioeconomic data in a robust but sensitive way. Longer term steps include implementing a research agenda to evaluate best practices for accounting for social risk, developing validated health equity specific measures of care, and promoting policy changes to allow better integration with and funding of social services.