Association of Clinician Minority Patient Caseload With Performance in the 2019 Medicare Merit-based Incentive Payment System

Abstract

This study assesses the association between US clinicians’ caseload of minority patients and their 2019 Medicare Merit-based Incentive Payment System performance score.

Disparities in health care delivered to individuals in the US who belong to racial and ethnic minority groups may be exacerbated by systemic inequities in value-based payment systems.¹ Prior research on the Medicare Merit-based Incentive Payment System (MIPS) found that clinicians who served high caseloads of low-income patients received lower performance scores in 2019.^2,3

It is unknown how clinicians who serve high caseloads of minority patients perform on MIPS. We assessed the association between clinicians’ caseload of minority patients and their 2019 MIPS performance score.

Methods

This study was exempted by the Saint Louis University institutional review board, with informed consent waived.

We obtained MIPS performance scores and caseload data for clinicians who participated in the program in 2019 (details appear in the eMethods in the Supplement).^4,5 We classified clinicians into low (bottom quintile), middle (middle 3 quintiles), and high (top quintile) caseloads of minority race/ethnicity patients; minority was defined as Black, Hispanic, Asian/Pacific Islander, Native American, or other as classified by Medicare.⁵

We used multivariable regression to assess the association between clinicians’ minority patient caseload and MIPS performance scores (range, 0-100; higher scores represent better performance) and receipt of payment penalties (score <3) and bonuses (score ≥70), adjusting for clinician, practice, patient, and area characteristics, including patient poverty as identified by Medicaid enrollment (Table 1). In separate models, we stratified regression results by low (bottom quintile), middle (middle 3 quintiles), and high (top quintile) poverty caseload to identify racial disparities potentially hidden by poverty status.

Table 1. Outpatient Clinicians Participating in the 2019 Merit-based Incentive Payment System by Minority Patient Caseload Classification.

	Minority patient caseloada
	Low (0%-8.2%)	Middle (8.3%-33.6%)	High (33.7%-100%)
Clinician and practice characteristics
No. of clinicians	98 264	294 760	98 256
Sex, No. (%)
Female	31 309 (31.9)	107 097 (36.3)	37 724 (38.4)
Male	66 955 (68.1)	187 663 (63.7)	60 532 (61.6)
Time since medical school graduation, mean (SD), y	21.5 (11.9)	20.1 (12.1)	20.1 (12.5)
Specialty, No. (%)
Primary careb	18 162 (18.5)	61 294 (20.8)	28 932 (29.5)
Advanced primary carec	14 991 (15.3)	39 796 (13.5)	10 923 (11.1)
Otherd	65 111 (66.3)	193 670 (65.7)	58 401 (59.4)
Affiliation, No. (%)
Major health systeme	39 736 (40.4)	135 255 (45.9)	45 878 (46.7)
Major teaching hospitalf	18 307 (18.6)	92 198 (31.3)	43 017 (43.8)
Safety-net facilityg	4458 (4.5)	26 720 (9.1)	21 710 (22.1)
Medicare patient caseload characteristics
No. of Medicare beneficiaries, mean (SD), %h	654 (596.2)	500 (568.0)	394 (487.6)
Minority beneficiaries, mean (SD), %i	5.6 (0.02)	18.2 (0.07)	49.9 (0.15)
Low-income beneficiaries (dually eligible for Medicaid), mean (SD), %	18.3 (0.13)	23.5 (0.15)	42.7 (0.19)
Age, mean (SD), y	73.1 (3.4)	71.8 (4.3)	69.6 (5.1)
Sex, mean (SD), %
Female	56.9 (0.10)	57.0 (0.11)	56.5 (0.11)
Male	43.1 (0.10)	43.0 (0.11)	43.5 (0.11)
CMS Hierarchical Condition Categories risk score, mean (SD)	1.44 (0.46)	1.75 (0.67)	2.31 (1.09)
Local practice area characteristicsj
Patients with Medicaid or uninsured, mean (SD), %	24.2 (0.14)	24.8 (0.15)	29.1 (0.18)
Setting, No. (%)
Rural	27 223 (27.7)	29 697 (10.1)	7320 (7.5)
Urban	78 504 (79.9)	279 146 (94.7)	93 853 (95.5)
Census region, No. (%)
Northeast	24 596 (25.0)	65 088 (22.1)	16 984 (17.3)
Midwest	35 552 (36.2)	64 646 (21.9)	13 136 (13.4)
South	27 510 (28.0)	116 577 (39.6)	41 841 (42.6)
West	12 367 (12.6)	54 211 (18.4)	27 657 (28.2)
Merit-based Incentive Payment System performance and reimbursement
Final performance score, mean (SD)	72.9 (34.4)	72.2 (34.1)	67.3 (36.2)
Payment penalty, No. (%)k	6318 (6.4)	19 862 (6.7)	8413 (8.6)
Payment bonus, No. (%)l	67 791 (69.0)	199 624 (67.7)	61 305 (62.4)

Abbreviation: CMS, Centers for Medicare & Medicaid Services.

^aLow vs middle vs high for patient minority caseload is the bottom vs middle 3 vs top quintiles of clinicians for percentage of minority Medicare patients treated in 2017. All differences in proportions (χ² test) or means (analysis of variance or F test) across low vs middle vs high patient minority caseload were statistically significant at P < .001.

^bIncludes geriatric medicine, internal medicine, family medicine, general practice, obstetrics/gynecology, or pediatric medicine.

^cIncludes nurse practitioners and physician assistants. Some also may have been specialty trained.

^dIncludes all other clinicians.

^eIdentified by the Agency for Healthcare Research and Quality for clinicians’ group practices listed in Physician Compare.

^fDefined as a general acute care hospital with a resident-bed ratio of 0.25 or greater in the 2017 CMS Impact File. Identified by hospital affiliation in Physician Compare.

^gIdentified by hospital affiliation in Physician Compare linked to Disproportionate Share Hospital percentage in the top quartile in the 2017 CMS Impact File and at least 1 practice location in a Census block group in the top quartile nationally for uninsured and Medicaid residents as a percentage of total residents.

^hData for beneficiaries seen as patients in 2017 were winsorized.

ⁱAll Medicare patients not self-identified as White (including Black, Hispanic, Asian, Pacific Islander, Alaska Native, and other) as a percentage of total patients.

^jCalculated across all practice areas for the subset of clinicians with multiple practice locations. Rural vs urban location and Census region for clinicians identified by geocoding their practice addresses listed in Physician Compare and then linking this to rural-urban commuting area codes and state of residence.

^kFinal performance scores of less than 3 received payment penalties in 2019.

^lFinal performance scores of 70 or greater received payment bonuses in 2019.

The threshold for statistical significance was P < .05 with 2-sided tests. The analyses were performed using Stata version 16 (StataCorp).

Results

Of 752 400 clinicians with MIPS performance scores, there were 210 321 mostly low-volume clinicians with missing data on minority caseload excluded and 50 799 with missing data for other variables, leaving 491 280 in the study. There were 98 256 clinicians with high-minority caseloads (mean, 49.9% minority patients) vs 294 760 with middle (mean, 18.2%) vs 98 264 with low (mean, 5.6%) (Table 1). Clinicians with high- vs middle- vs low-minority caseloads were more likely to be primary care physicians (29.5% vs 20.8% vs 18.5%, respectively), affiliated with a major teaching hospital (43.8% vs 31.3% vs 18.6%), and serving more low-income patients (42.7% vs 23.5% vs 18.3%) (P < .001 for each). Patients treated by clinicians with high-minority caseloads also were more medically complex.

The mean unadjusted MIPS score for clinicians with high- vs middle- vs low-minority caseloads was 67.3 vs 72.2 vs 72.9 (P < .001). After adjustment, high-minority caseload was associated with a 1.0-point (95% CI, 0.6-1.4) (1.4% of clinicians) lower MIPS score compared with a low-minority caseload in the overall sample (Table 2). Among clinicians serving high-poverty caseloads, middle- and high-minority caseloads were associated with 3.9-point (95% CI, 2.9-4.9) and 4.2-point (95% CI, 3.1-5.3) (equal to 6.1% and 6.6%) lower MIPS scores.

Table 2. Association of Clinician Patient Minority Caseload Classification Stratified by Poverty Caseload With 2019 Merit-based Incentive Payment System Performance and Reimbursement, Adjusted Results.

	All clinicians by minority caseload		Serving low patient poverty caseloadsa		Serving middle patient poverty caseloadsb		Serving high patient poverty caseloadsc
	Middle vs Low	High vs Low	Middle vs low	High vs low	Middle vs low	High vs low	Middle vs low	High vs low
Total No. of clinicians	491 280		98 336		294 689		98 255
Final performance score, mean change score (95% CI)d,e	0 (−0.3 to 0.2)	−1.0 (−1.4 to −0.6)	1.4 (0.9 to 1.9)	−0.6 (−2.2 to 0.9)	−1.3 (−1.6 to −1.0)	−2.5 (−3.0 to −2.0)	−3.9 (−4.9 to −2.9)	−4.2 (−5.3 to −3.1)
Payment penalty, odds ratio (95% CI)d,f	1.06 (1.03 to 1.10)	1.06 (1.01 to 1.12)	0.94 (0.88 to 1.00)	0.98 (0.81 to 1.19)	1.16 (1.10 to 1.22)	1.14 (1.06 to 1.22)	1.44 (1.30 to 1.59)	1.33 (1.20 to 1.48)
Payment bonus, odds ratio (95% CI)d,f	0.96 (0.95 to 0.98)	0.92 (0.90 to 0.95)	1.07 (1.04 to 1.11)	0.99 (0.89 to 1.10)	0.91 (0.89 to 0.93)	0.86 (0.83 to 0.89)	0.78 (0.73 to 0.84)	0.77 (0.72 to 0.82)
Dependent variable base rates
Final performance score, mean (SD)	71.4 (34.7)		72.8 (34.4)		73.4 (33.2)		63.6 (37.9)
Payment penalty, %	7.0		6.5		5.8		11.2
Payment bonus, %	66.9		69.9		68.9		57.9

^aClinicians in the bottom quintile for proportion of Medicare beneficiaries dually enrolled in Medicaid (0%-11.8%). The distribution was 36.2% for low vs 61.6% for middle vs 2.3% for high patient minority caseload.

^bClinicians in the middle 3 quintiles for proportion of Medicare beneficiaries dually enrolled in Medicaid (11.9%-38.3%). The distribution was 18.9% for low vs 66.7% for middle vs 14.4% for high patient minority caseload.

^cClinicians in the top quintile for proportion of Medicare beneficiaries dually enrolled in Medicaid (38.4%-100%). The distribution was 7.1% for low vs 38.4% for middle vs 54.5% for high patient minority caseload.

^dAdjusted for individual clinician, Medicare patient caseload, and local practice area characteristics (patient poverty adjustment by quintile of clinician caseload of Medicare beneficiaries dually enrolled in Medicaid except in models already stratified by top vs middle vs bottom quintile of poverty). State fixed effects were used in place of Census region to control for the states in which clinicians practiced, which differ in state eligibility rules for Medicaid that affect the patient poverty caseload classification.

^eMultivariable ordinary least-squares regression models.

^fMultivariable logistic regression models.

The unadjusted percentage of clinicians receiving a payment penalty with a high- vs middle- vs low-minority caseload was 8.6% vs 6.7% vs 6.4%. After adjustment, middle- and high-minority caseloads were associated with higher odds of a payment penalty among all clinicians (odds ratio [OR], 1.06 [95% CI, 1.03-1.10] and OR, 1.06 [95% CI, 1.01-1.12], respectively) and higher odds of a payment penalty among clinicians with high-poverty caseloads (OR, 1.44 [95% CI, 1.30-1.59] and OR, 1.33 [95% CI, 1.20-1.48]). Clinicians with middle- and high-minority caseloads had lower odds of a payment bonus, particularly among those with high-poverty caseloads.

Discussion

Clinicians with high-minority caseloads performed worse in the 2019 MIPS than their peers, even after adjustment for poverty. This association was stronger among clinicians with high-poverty caseloads. As penalties in the MIPS increase to 9% of reimbursement in the coming years,⁶ the program has the potential to widen inequities in care by taking resources from already resource-constrained practices and the patients they serve. There is also a risk that the MIPS will create disincentives for providing care to minority patients, which could widen existing inequities in access to outpatient services.

This study had limitations. Low-volume clinicians were excluded; the findings may not generalize to them. This was an observational study and there was likely residual unmeasured confounding. More research is needed to identify causal mechanisms underlying the findings and to track whether the MIPS is associated with widening inequities in the future.

Section Editor: Jody W. Zylke, MD, Deputy Editor.

Supplement.

eMethods