Association Between Patient Social Risk and Physician Performance Scores in the First Year of the Merit-based Incentive Payment System

Key Points

Question  Was there an association between patient social risk and physician performance in the first year of the Merit-based Incentive Payment System (MIPS), a major Medicare value-based payment program?

Findings  In this cross-sectional observational study of 284 544 physicians, physicians with the highest proportion of patients dually eligible for Medicare and Medicaid had significantly lower MIPS scores compared with physicians with the lowest proportion (mean, 64.7 vs 75.9; range, 0-100; higher scores reflect better performance).

Meaning  Physicians with the highest proportion of socially disadvantaged patients had significantly lower MIPS scores, although further research is needed to understand the reasons underlying the differences in MIPS scores by levels of patient social risk.

Importance  The US Merit-based Incentive Payment System (MIPS) is a major Medicare value-based payment program aimed at improving quality and reducing costs. Little is known about how physicians’ performance varies by social risk of their patients.

Objective  To determine the relationship between patient social risk and physicians’ scores in the first year of MIPS.

Design, Setting, and Participants  Cross-sectional study of physicians participating in MIPS in 2017.

Exposures  Physicians in the highest quintile of proportion of dually eligible patients served; physicians in the 3 middle quintiles; and physicians in the lowest quintile.

Main Outcomes and Measures  The primary outcome was the 2017 composite MIPS score (range, 0-100; higher scores indicate better performance). Payment rates were adjusted –4% to 4% based on scores.

Results  The final sample included 284 544 physicians (76.1% men, 60.1% with ≥20 years in practice, 11.9% in rural location, 26.8% hospital-based, and 24.6% in primary care). The mean composite MIPS score was 73.3. Physicians in the highest risk quintile cared for 52.0% of dually eligible patients; those in the 3 middle risk quintiles, 21.8%; and those in the lowest risk quintile, 6.6%. After adjusting for medical complexity, the mean MIPS score for physicians in the highest risk quintile (64.7) was lower relative to scores for physicians in the middle 3 (75.4) and lowest (75.9) risk quintiles (difference for highest vs middle 3, –10.7 [95% CI, –11.0 to –10.4]; highest vs lowest, –11.2 [95% CI, –11.6 to –10.8]; P < .001). This relationship was found across specialties except psychiatry. Compared with physicians in the lowest risk quintile, physicians in the highest risk quintile were more likely to work in rural areas (12.7% vs 6.4%; difference, 6.3 percentage points [95% CI, 6.0 to 6.7]; P < .001) but less likely to care for more than 1000 Medicare beneficiaries (9.4% vs 17.8%; difference, –8.3 percentage points [95% CI, –8.7 to –8.0]; P < .001) or to have more than 20 years in practice (56.7% vs 70.6%; difference, –13.9 percentage points [95% CI, –14.4 to –13.3]; P < .001). For physicians in the highest risk quintile, several characteristics were associated with higher MIPS scores, including practicing in a larger group (mean score, 82.4 for more than 50 physicians vs 46.1 for 1-5 physicians; difference, 36.2 [95% CI, 35.3 to 37.2]; P < .001) and reporting through an alternative payment model (mean score, 79.5 for alternative payment model vs 59.9 for reporting as individual; difference, 19.7 [95% CI, 18.9 to 20.4]; P < .001).

Conclusions and Relevance  In this cross-sectional analysis of physicians who participated in the first year of the Medicare MIPS program, physicians with the highest proportion of patients dually eligible for Medicare and Medicaid had significantly lower MIPS scores compared with other physicians. Further research is needed to understand the reasons underlying the differences in physician MIPS scores by levels of patient social risk.

Quiz Ref IDIn 2017, Medicare began making payments to physicians under the Merit-based Incentive Payment System (MIPS)—one of 2 major paths for physicians to engage in value-based care through the Medicare Access and CHIP Reauthorization Act. With MIPS, eligible physicians can receive rewards or penalties based on their performance across 4 domains: quality, costs, improvement activities, and promoting interoperability. More than 90% of eligible physicians participated in the program in 2017, and payments were adjusted up to 4% in 2019.¹ While several types of clinicians participate in MIPS, this study focuses on physicians.

MIPS is Medicare’s latest attempt to link physician payments to the value of care and builds on other recent attempts at pay for performance, most notably the value-based payment modifier.² Prior studies have found mixed results on whether value-based purchasing improves quality or reduces costs,^3-5 but there is growing concern that these programs may exacerbate health disparities because they lack adequate adjustment for patient social risk.^6-9 Recognizing this concern, Medicare recently introduced several exemptions and bonuses, including a “complex patient bonus” to adjust for patient medical and social complexity, but it remains unclear how these changes will affect MIPS scores and payment.

Prior research has shown an association between social factors and physician group performance,^10-12 but to our knowledge, no studies have evaluated how patient social risk is related to performance in MIPS—a program that influences payment for hundreds of thousands of physicians and that will play a larger role in Medicare reimbursement over time. This study examines 3 questions. First, did physicians caring for more socially disadvantaged patients have lower scores in the first year of MIPS? Second, how did MIPS performance vary by medical specialty across the spectrum of patient social risk? Third, among physicians caring for the most socially disadvantaged patients, what characteristics were associated with better scores?

This study was deemed non–human subjects research by the institutional review board at Weill Cornell Medical College. Patient consent was not required because the study used administrative, deidentified data.

Using publicly available data in Medicare Physician Compare files, all clinicians who participated in MIPS in 2017 were identified and their MIPS composite scores, category scores, and relevant demographic information were extracted.¹³ National Provider Identifier (NPI) numbers were used to link to the 2017 Medicare Fee-For-Service Provider Utilization and Payment Data—Physician and Other Supplier Public Use File¹⁴ and Medicare Data on Provider Practice and Specialty (MD-PPAS) file¹⁵—to identify relevant clinician and practice characteristics.¹⁶ Physicians were included in the study population if they had a composite MIPS score in the Physician Compare database in 2017 that could be linked to clinician and practice characteristics relevant to the analysis. Nonphysician clinicians were excluded.

Quiz Ref IDThis study examined the relationship between physicians’ performance in MIPS and the proportion of socially disadvantaged patients they cared for in 2017. Dual eligibility was used as a proxy for social risk, as has been done in other studies.^7,17 The level of social risk of a physician’s patients was determined using variables from the Medicare Public Use File and was defined as the number of unique dually eligible patients divided by all unique Medicare patients seen by a physician. Beneficiaries were classified as dually eligible if they received full or partial Medicaid benefits in any month in a given calendar year. The Centers for Medicare & Medicaid Services (CMS) uses this variable for risk adjustment in the MIPS program.¹⁸

The Medicare Public Use File, which reflects 100% of Medicare enrollment data and fee-for-service Part B noninstitutional claims, was also used to obtain the mean Hierarchical Condition Category (HCC) score and total number of unique Medicare beneficiaries for each physician. The Physician Compare data were used to measure physician years in practice, sex, and mode of MIPS data submission (alternative payment model, group, individual, or other). The MD-PPAS file was used to construct a measure of physician group size (number of unique NPIs billing under the same primary tax identification number). Physicians’ zip code, as reported in the Physician Compare data, was used to create an indicator for practice in a rural area, as defined by the Federal Office of Rural Health Policy.¹⁹ Physicians were grouped into 6 broad specialty categories using data from the MD-PPAS file (primary care, medical subspecialty, surgical subspecialty, obstetrics/gynecology, psychiatry, and hospital-based specialty). For cases in which specialty data were not available from the MD-PPAS file, the primary specialty included in Physician Compare was used. An NPI was identified as belonging to a multispecialty group if the primary tax identification number included less than 80% of NPIs within a single broad specialty.

The primary outcome was the final composite MIPS score (range, 0-100). Higher scores reflect better performance; actual 2017 scores ranged from 0 to 100. Secondary outcomes included the 3 MIPS category scores included in the program in 2017: quality (range, 0-100), advancing care information (now called “promoting interoperability”; range, 0-100), and improvement activities (range, 0-40). The cost domain was not scored by the Centers for Medicare & Medicaid Services (CMS) in 2017. Final MIPS scores reflect a weighted sum of the category scores and other program bonuses.²⁰ The MIPS program is budget neutral, such that more “low performers” leads to larger payment for “high performers” and vice versa. A separate $500 million is reserved for “exceptional performance.” Based on 2017 performance, payments were adjusted in 2019. A composite score difference between 0 and 2.99 resulted in a 4% reduction in payment rates. Each 10-point increase in the composite score between 3.0 and 69.99 resulted in a 0.03% increase in payment. In 2017, the “exceptional performance” threshold was set at 70, and each 10-point increase above 70 resulted in a 0.53% increase in 2019 payment.

To examine the relationship between MIPS scores and social risk, the mean composite MIPS scores across ventiles of the proportion of physicians’ patients who were dually eligible for Medicare and Medicaid was graphed. (Ventiles comprised 20 groups of physicians ordered by the proportion of their patients dually eligible for Medicare and Medicaid, with each group containing 1/20th of the study population.) This analysis was repeated after adjusting for patients’ clinical risk (HCC score) and, separately, after adjusting for clinical risk along with a broad set of physician and practice characteristics. The extent to which physician and practice characteristics, as well as MIPS scores, differed across physicians with a low, medium, and high proportion of dually eligible patients was then analyzed using mean-comparison (t) tests. Physicians in the second, third, and fourth quintiles were collapsed into a single “middle” category. The relationship between MIPS scores and social risk was also examined by specialty and by MIPS reporting mode. Less than 10% of observations were excluded for missing covariate data. The characteristics of excluded relative to included observations are presented in eTable 1 in the Supplement. For all of these analyses, MIPS scores were reported after adjusting for mean HCC risk score using ordinary least-squares regression to control for mean HCC score by specialty.

For physicians in the highest quintile, ordinary least-squares multivariable regression with robust standard errors was used to examine the physician and practice characteristics associated with higher MIPS scores. This regression also controlled for mean HCC risk score. Sensitivity analyses included the addition of state fixed effects and the use of the quality domain score as the outcome. Using state fixed effects, we estimated differences in MIPS scores across physicians within the same state. A model with state fixed effects includes a binary indicator for each state. Because of the potential for type I error due to multiple comparisons, findings for analyses of secondary end points should be interpreted as exploratory.

The study was conducted using Stata statistical software version 14.1 (StataCorp). Two-sided P < .05 was considered statistically significant.

The final population included 284 544 physicians. Of the initial analytic sample of 413 233 observations in the 2017 Physician Compare file, 37 063 observations were excluded because of duplicate NPI numbers; 60 205 were excluded because they were not physicians; and 31 411 were excluded because of missing covariate data (eFigure 1 in the Supplement). The mean composite MIPS score across all physicians was 73.3 (of a possible 100). Among physicians in the highest risk quintile, 52.0% of patients were dually eligible for Medicare and Medicaid. For physicians in the middle 3 quintiles group and lowest risk quintile, 21.8% and 6.6%, respectively, of patients were dually eligible (Table 1).

Quiz Ref IDAfter adjustment for patient medical complexity, physicians with a higher proportion of dually eligible patients had lower MIPS scores, a relationship driven by physicians in the highest quintile (top 4 ventiles) of dually eligible patients served (Figure 1). The mean MIPS score for physicians in the highest risk quintile (64.7) was significantly lower than for physicians in the middle 3 quintiles (75.4) and lowest (75.9) risk quintile (difference for highest vs middle, –10.7 [95% CI, –11.0 to –10.4]; highest vs lowest, –11.2 [95% CI, –11.6 to –10.8]; P < .001) (Table 2). Results were similar before and after adjustment for medical complexity. Adjusting for physician and practice characteristics attenuated this relationship modestly, but a significant difference remained (eFigure 2 in the Supplement).

Physicians in the highest risk quintile had different characteristics compared with physicians in the middle 3 quintiles and lowest risk quintile (Table 1). Quiz Ref IDFor example, compared with physicians in the lowest risk quintile, physicians in the highest risk quintile were more likely to work in small practices of 1 to 5 clinicians (40.0% vs 37.8%; difference, 2.3 percentage points [95% CI, 1.7 to 2.8]; P < .001) and in rural locations (12.7% vs 6.4%; difference, 6.3 percentage points [95% CI, 6.0 to 6.7]; P < .001). They were less likely to care for more than 1000 Medicare beneficiaries (9.4% vs 17.8%; difference, –8.3 percentage points [95% CI, –8.7 to –8.0]; P < .001) or to have more than 20 years in practice (56.7% vs 70.6%; difference, –13.9 percentage points [95% CI, –14.4 to –13.3]; P < .001). Compared with physicians in the lowest risk quintile, physicians in the highest risk quintile were more likely to practice psychiatry (6.6% vs 0.2%; difference, 6.4 percentage points [95% CI, 6.2 to 6.6]; P < .001) and in hospital-based specialties (31.0% vs 11.3%; difference, 19.7 percentage points [95% CI, 19.2 to 20.2]; P < .001) but were less likely to practice in medical specialties (22.0% vs 26.4%; difference, –4.4 percentage points [95% CI, –4.9 to –3.9]; P < .001) or surgical specialties (11.4% vs 29.8%; difference, –18.4 percentage points [95% CI, –18.9 to –17.9]; P < .001). In addition, they were less likely to report to Medicare through alternative payment models (18.7% vs 19.7%; difference, –1.0 percentage points [95% CI, –1.4 to –0.5]; P < .001) and more likely to report as a group (50.0% vs 39.7%; difference, 10.3 percentage points [95% CI, 9.7 to 10.9]; P < .001).

There were significant differences by physician specialty in MIPS performance, both overall and across quintiles of social risk (Table 2). Obstetrician-gynecologists and psychiatrists had lower MIPS scores compared with other physicians. MIPS scores across specialty groups except psychiatry were significantly lower for physicians in the highest risk quintile compared with those in the middle 3 group and lowest risk quintile. For example, among physicians practicing in a medical subspecialty, the mean MIPS score for physicians in the highest risk quintile was 68.7 compared with 79.3 for physicians in the lowest risk quintile (difference, –10.6 [95% CI, –11.4 to 9.7]; P < .001). For obstetrician-gynecologists, the decrease in MIPS scores was even greater (mean score, 40.5 for highest risk vs 68.7 for lowest risk; difference, –28.2 [95% CI, –33.4 to –23.0]; P < .001).

The mode through which physicians reported measures to Medicare was also associated with differences in MIPS scores across the spectrum of patient social risk (eTable 2 in the Supplement). Physicians in the highest risk quintile who reported through an alternative payment model had scores that did not significantly differ from physicians in the middle and lowest risk quintiles, but those who reported as a group (mean score, 64.8 for highest risk vs 75.5 for lowest risk; difference, –10.7 [95% CI, –11.4 to –10.0]; P < .001) or as individuals (mean score, 49.2 for highest risk vs 66.9 for lowest risk; difference, –17.8 [95% CI, –18.5 to –17.0]; P < .001) had lower scores.

Among physicians in the highest risk quintile, several characteristics were associated with higher MIPS scores (Figure 2). Quiz Ref IDFor example, better MIPS performance was associated with practicing in a larger group (mean score, 82.4 for >50 physicians vs 46.1 for 1-5 physicians; difference, 36.2 [95% CI, 35.3 to 37.2]; P < .001) and caring for more Medicare beneficiaries (mean score, 71.8 for >1000 beneficiaries vs 62.0 for 11-200 beneficiaries; difference, 9.8 [95% CI, 8.7 to 10.8]; P < .001). Even among physicians in the highest risk quintile, those with a relatively higher proportion of dually eligible patients had lower MIPS scores. For example, physicians in the highest risk quintile who cared for the highest proportion of dually eligible patients had mean MIPS scores 5.8 points lower (95% CI, –6.6 to –5.0; P < .001) than those caring for the smallest proportion (mean score, 60.5 for highest proportion vs 66.3 for smallest proportion). Results were not meaningfully different in sensitivity analyses using state fixed effects or using the quality domain score as the outcome (eTables 3 and 4 in the Supplement).

In 2017, CMS calculated MIPS scores across 3 domains: quality, advancing care information, and improvement activities. (The cost domain was scored starting in 2018.) Scores did not differ meaningfully between physicians in the middle 3 quintiles group and lowest risk quintile in any score domain (Table 3). However, physicians in the highest risk quintile had significantly lower scores in all 3 domains. For example, compared with physicians in the lowest risk quintile, those in the highest risk quintile had lower quality scores (64.0 vs 75.4; difference, –11.4 [95% CI, –11.8 to –10.9]; P < .001), lower advancing care information scores (63.4 vs 73.6; difference, –10.2 [95% CI, –10.7 to –9.7]; P < .001), and lower improvement activities scores (29.4 vs 33.3; difference, –3.9 [95% CI, –4.1 to –3.7]; P < .001).

In this cross-sectional analysis of physicians participating in the first year of the Medicare MIPS program, physicians in the highest quintile of proportion of dually eligible patients served had composite scores more than 11 points lower compared with other physicians. Even among physicians in the highest risk quintile, those who cared for the most dually eligible patients had MIPS scores 6 points lower than those who cared for the fewest dually eligible patients. The relationship between patient social risk and MIPS performance held across specialties, with the exception of psychiatry. Lower composite MIPS scores for physicians in the highest risk quintile was not driven by any one score domain but rather reflected lower scores in each of the quality, improvement activities, and advancing care information domains.

For physicians caring for the most socially disadvantaged patients, several factors were associated with higher MIPS scores, some of which are modifiable. For example, physicians in the highest risk quintile who practiced in larger groups, practiced in multispecialty practices, or submitted information through alternative payment models had higher MIPS scores, possibly reflecting the greater infrastructure and resources these practices have to collect, analyze, and report measures to CMS.^21,22 Specialties that care for relatively few Medicare patients, such as psychiatry and obstetrics/gynecology, may choose not to invest significant attention or resources in Medicare quality reporting, resulting in relatively low MIPS scores compared with other specialties.

To our knowledge, this is the first study to examine the relationship between MIPS scores and patient social risk. These results are consistent with prior research in other value-based programs, suggesting that clinicians and health care organizations serving poorer patients tend to have lower performance scores.²³ For example, studies have found that greater patient social risk was associated with lower scores for physicians in the value-based modifier program and that adjustment for social factors substantially narrowed these differences.^7,17 Similarly, social factors such as poverty and housing instability seem to be associated with differences in hospital readmission rates but are not currently incorporated into Medicare’s Hospital Readmissions Reduction Program.^24,25 Many value-based payment programs may thus penalize clinicians for social factors outside their control and inadvertently transfer resources from those caring for less affluent patients to those caring for more affluent patients—the so-called reverse Robin Hood effect.²⁶

In the first year of MIPS, the potential financial penalties and rewards were relatively small (4% of payment rates).^27,28 However, because the program is budget-neutral and because there were more high than low performers in 2017, actual adjustments ranged from a 4.0% penalty to a 1.9% bonus.²⁷ For example, physicians who had a composite MIPS score of 60 received a 0.16% rate increase for services delivered, while those who had a score of 80 received a 0.82% increase.²⁸ This gap may grow substantially in the coming years, as the maximum MIPS penalties and bonuses are set to increase, such that physicians could receive payment adjustments of up to 9% in 2022 based on performance in 2020.²⁷ If there is an imbalance between low and high performers, payment adjustments could be higher than 9%, and small differences in scores could lead to large differences in penalties and rewards.²⁹

Furthermore, the exceptional performance threshold, above which physicians receive additional bonuses, will increase to 85 in 2020.^28,30 In 2017, it was given for a score of 70 or above, meaning that the median physician in our highest risk quintile did not receive it, while the median physician in the middle group and lowest risk quintile did.²⁸ Similarly, the minimum score needed to avoid a penalty was 3 in 2017 but increased incrementally each year to 45 in 2020 (and will increase to 60 in 2021).³¹

While the Medicare Payment Advisory Commission has recommended eliminating MIPS in its current form, Congress has not provided any indication it intends to do so.³² Regardless of what drives differences in MIPS scores across the spectrum of patient social risk, CMS could consider levers to ensure it compares physicians with similar patients. CMS could, for example, compare physicians only within the same category of patient social risk.^23,33 While CMS has recently introduced policies aimed at supporting practices serving high proportions of medically and socially complex patients, these may not be effective in reducing differences in MIPS scores and payment associated with patient social risk. For example, CMS has introduced a “complex patient bonus” of up to 5 points based on the average HCC score and proportion of dually eligible patients under a clinician’s care. However, CMS’ own analysis suggests this is unlikely to substantially reduce differences in MIPS scores.³⁴

This study has several limitations. First, dual eligibility was used as a proxy for social risk. While many other studies have also done so, dual eligibility may not fully capture patients with social disadvantage, and dual-eligibility criteria vary from state to state.^35-37 Second, because this study was cross-sectional, it could not determine whether a causal link exists between patient social risk and MIPS performance, although efforts were made to control for confounders, including medical risk and various practice characteristics. Third, this study focused on the first year of MIPS, the only year for which data are available. Fourth, in the absence of patient-level data on outcomes, it was not possible to examine whether differences in scores reflected worse care quality or more limited ability to effectively collect and report various MIPS measures.

In this cross-sectional analysis of physicians who participated in the first year of the Medicare MIPS program, physicians with the highest proportion of patients dually eligible for Medicare and Medicaid had significantly lower MIPS scores compared with other physicians. Further research is needed to understand the reasons underlying the differences in physician MIPS scores by levels of patient social risk.

Corresponding Author: Dhruv Khullar, MD, MPP, Department of Population Health Sciences, Department of Medicine, Weill Cornell Medical College, 402 E 67th St, New York, NY 10065 (Khd9010@med.cornell.edu).

Accepted for Publication: July 6, 2020.

Author Contributions: Drs Schpero and Bond had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Khullar, Schpero, Bond, Casalino.

Acquisition, analysis, or interpretation of data: Khullar, Schpero, Bond, Qian.

Drafting of the manuscript: Khullar, Schpero, Bond.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Schpero, Bond, Qian.

Obtained funding: Casalino.

Administrative, technical, or material support: Bond, Qian.

Supervision: Khullar, Casalino.

Conflict of Interest Disclosures: Dr Khullar reported receiving grants from the American Medical Association and from the Patient-Centered Outcomes Research Institute (PCORI), outside this work. Dr Bond reported receiving grants from Arnold Ventures. Dr Casalino reported receiving personal fees from the Medicare Payment Advisory Commission and receiving an honorarium for participating until December 2019 in quarterly meetings as a member of the American Medical Association Professional Satisfaction and Practice Sustainability Advisory Committee. Dr Schpero reported receiving grants from PCORI, outside this work. No other disclosures were reported.

Funding/Support: This work was supported by the Physicians Foundation Center for the Study of Physician Practice and Leadership at Weill Cornell Medicine.

Role of the Funder/Sponsor: The Physicians Foundation Center had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We thank Rebecca Onie, JD, and Rocco Perla, EdD, of The Health Initiative for conversations motivating this work and for their thoughtful comments on a prior draft of this article. They did not receive compensation for their involvement.