Association between mandatory bundled payments and changes in socioeconomic disparities for joint replacement outcomes

Abstract

Objective

To determine whether mandatory participation by hospitals in bundled payments for lower extremity joint replacement (LEJR) was associated with changes in outcome disparities for patients dually eligible for Medicare and Medicaid.

Data Sources and Study Setting

We used Medicare claims data for beneficiaries undergoing LEJR in the United States between 2011 and 2017.

Study Design

We conducted a retrospective observational study using a differences‐in‐differences method to compare changes in outcome disparities between dual‐eligible and non‐dual eligible beneficiaries after hospital participation in the Comprehensive Care for Joint Replacement (CJR) program. The primary outcome was LEJR complications. Secondary outcomes included 90‐day readmissions and mortality.

Data Extraction Methods

We identified hospitals in the US market areas eligible for CJR. We included beneficiaries in the intervention group who received joint replacement at hospitals in markets randomized to participate in CJR. The comparison group included patients who received joint replacement at hospitals in markets who were eligible for CJR but randomized to control.

Principal Findings

The study included 1,603,555 Medicare beneficiaries (mean age, 74.6 years, 64.3% women, 11.0% dual‐eligible). Among participant hospitals, complications decreased between baseline and intervention periods from 11.0% to 10.1% for dual‐eligible and 7.0% to 6.4% for non‐dual‐eligible beneficiaries. Among nonparticipant hospitals, complications decreased from 10.3% to 9.8% for dual‐eligible and 6.7% to 6.0% for non‐dual‐eligible beneficiaries. In adjusted analysis, CJR participation was associated with a reduced difference in complications between dual‐eligible and non‐dual‐eligible beneficiaries (−0.9 percentage points, 95% CI −1.6 to −0.1). The reduction in disparities was observed among hospitals without prior experience in a voluntary LEJR bundled payment model. There were no differential changes in 90‐day readmissions or mortality.

Conclusions

Mandatory participation in a bundled payment program was associated with reduced disparities in joint replacement complications for Medicare beneficiaries with low income. To our knowledge, this is the first evidence of reduced socioeconomic disparities in outcomes under value‐based payments.

What is known on this topic

• An urgent policy priority is to understand the impact of value‐based payment models on health equity and design future models to specifically improve equitable outcomes.

• Prior evaluations of the Comprehensive Care for Joint Replacement (CJR) program, a mandatory Medicare bundled payment model, found reduced spending and post‐acute care use with little change in quality outcomes.

• Some evidence indicates that access to surgery decreased for racial minorities among hospitals who participated in CJR. Changes in outcome disparities have received less attention, particularly for low‐income individuals.

What this study adds

• We found that mandatory participation in bundled payments was associated with reduced disparities in joint replacement complications for individuals with low income, as identified by dual eligibility for Medicare and Medicaid.

• To our knowledge, this is the first study to demonstrate improvements in disparities for a quality outcome under a value‐based payment model.

• Mandatory participation in payment models may have positive impacts for some dimensions of health equity. Understanding the effects of mandatory participation are salient given plans to launch a mandatory, equity‐oriented Medicare bundled payment program in 2026.

1. INTRODUCTION

Advancing health equity is a priority for value‐based payment reform. ¹ , ² Alternative payment models should seek to improve not only overall quality and cost‐efficiency but also equity in outcomes. ³ , ⁴ However, few models have been explicitly designed for this purpose, and in turn, few have demonstrated reductions in disparities. ⁵ , ⁶ Policymakers need insight into effective strategies as they implement new payment programs that seek to directly emphasize equity. ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹

Mandatory participation in payment models may offer specific benefits for improving health equity. ⁴ By requiring participation from healthcare organizations across geographic areas, mandatory programs may counteract patient selection observed in voluntary models and improve inclusion of historically marginalized communities. ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ Mandatory participation also enables rigorous evaluations that yield insight into the impact of value‐based reforms on disparities. ¹⁷ To that end, the Centers for Medicare and Medicaid Services (CMS) plans to implement an equity‐oriented bundled payment model for surgical procedures in 2026 that mandates participation among hospitals in specific areas. ¹⁸

The Comprehensive Care for Joint Replacement (CJR) is the most prominent and well‐studied mandatory payment model to date. ¹⁹ , ²⁰ , ²¹ , ²² Through CJR, CMS required hospitals in certain geographic areas to accept bundled payments for lower extremity joint replacement surgery (hereafter, “joint replacement”). ²³ Understanding the implications of CJR for equity is highly salient to current debates over the future design of payment models for several reasons. First, individuals with low income, including those dually eligible for Medicare and Medicaid, have experienced long‐standing disparities in joint replacement outcomes, including higher rates of postoperative complications and readmissions. ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ Second, some evidence suggests that CJR may have inadvertently worsened disparities for certain populations. Prior studies have reported diminished access to joint replacement for Black individuals under CJR. ²⁹ , ³⁰ Other studies have found larger financial penalties for safety‐net hospitals. ³¹ , ³² , ³³ , ³⁴ Third, past evaluations have focused on the impact of CJR on disparities for outcomes related to healthcare utilization, such as use of post‐acute care. ³⁵ , ³⁶ , ³⁷ , ³⁸ However, no studies have explicitly focused on differential changes in quality outcomes for individuals with low‐income in the CJR program.

In this study, we examined whether disparities in outcomes changed among dual‐eligible patients at hospitals mandated to accept bundled payments through CJR compared to patients undergoing joint replacement at hospitals outside of the program. We hypothesized that CJR participation was associated with reduced disparities in joint replacement complications for dual‐eligible versus non‐dual‐eligible patients. Because previous experience with bundled payments is likely to influence hospitals' ability to redesign care delivery and address clinical or social factors that drive disparities, we also examined whether disparities changed in CJR based on hospitals' prior experience in joint replacement bundles through the voluntary Bundled Payments for Care Improvement (BPCI) model. ³⁹ , ⁴⁰

2. STUDY DATA AND METHODS

This study was a retrospective analysis of CJR as a randomized payment program. ²³ The University of Pennsylvania institutional review board approved this study with a waiver of informed consent. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines. ⁴¹

2.1. Data source and study population

We used Medicare enrollment files to identify hospitals in the 196 urban markets eligible for CJR as well as hospitals that participated in BPCI for joint replacement episodes. ⁴² Hospital characteristics were obtained from Medicare data and the American Hospital Association Annual Survey. ⁴³ We used 100% Medicare claims data to identify fee‐for‐service beneficiaries receiving joint replacement at hospitals in CJR‐eligible markets. The Medicare Master Beneficiary Summary File was used to obtain information about patient enrollment and demographics. Market characteristics were obtained from Medicare Provider Analysis and Review and Accountable Care Organization Research‐identifiable Files.

Consistent with CJR and BPCI rules, our sample (Supplement S1) consisted of Medicare fee‐for‐service beneficiaries with index admissions to short‐term acute hospitals designated under joint replacement‐related Medicare Severity‐Diagnosis Related Groups (MS‐DRGs 469 or 470). Beneficiaries were required to have at least 12 months of continuous enrollment in Medicare Parts A and B prior to the admission, as well as 90 days following hospital discharge. Patients were excluded if they expired during the index admission, left against medical advice, or had length‐of‐stay longer than 360 days. We also excluded patients with any Medicare Advantage claims for the index admission or within 12 months of hospital discharge, patients with admissions to hospitals not participating in the Inpatient Prospective Payment System (IPPS), and patients with end‐stage renal disease. For patients with multiple admissions, we included only the first.

2.2. Hospital groups

In 2016, Medicare randomized 196 urban metropolitan areas around the country for CJR, selecting 75 eligible areas and requiring their hospitals to accept bundled payments for joint replacement. Hospitals in other areas continued to receive fee‐for‐service payment. Hospitals that were selected for CJR were defined as participant hospitals. A comparison group of nonparticipant hospitals comprised of those that were eligible, but not selected, for CJR. After randomization occurred, eight areas selected for CJR participation were later changed to nonparticipation. This study retained hospitals in those areas in the participation group, using an intention‐to‐treat analysis based on the initial randomization. This approach followed precedent from the formal CMS CJR program evaluation of CJR. ³⁸

For the purposes of secondary analysis, we divided participant hospitals into two subgroups based on prior experience with joint replacement bundles in BPCI. Between 2013 and 2017, hospitals voluntarily enrolled in BPCI on a time‐varying basis. The sequential implementation of BPCI followed directly by CJR offered a unique opportunity to conduct analyses for two subgroups of CJR hospitals: (1) those with prior experience in bundled payments after voluntarily adopting joint replacement bundles in BPCI prior to CJR; and (2) those without prior experience participating in joint replacement bundles prior to the CJR mandate.

2.3. Study periods

For all hospitals, the baseline period included January 2011–March 2016, the period prior to CJR. The intervention period spanned April 2016–December 2017. For the subgroup of BPCI participants, we defined hospital‐specific, time‐varying indicators for bundled payment participation between October 2013–March 2016 to account for the fact that hospitals could enroll in BPCI at different times prior to CJR. ⁴⁰ , ⁴⁴ , ⁴⁵ , ⁴⁶ Data through March 2018 were included for all hospitals to assess 90‐day post‐discharge outcomes.

2.4. Exposures

The exposure variable in our analysis was a three‐way interaction between patient dual‐eligibility status, hospital participation status, and a post‐CJR participation time indicator. Dual eligibility for Medicare and Medicaid is the most common indicator of socioeconomic status, often used by Medicare in policymaking and formal model evaluations. Dual‐eligibility was defined as having beneficiary premiums paid by a state for at least 1 month during the index admission year. Additional patient‐level covariates were chosen based on prior studies, including age, sex, race, ethnicity, clinical severity (Elixhauser comorbidities), disability status, and an indicator denoting joint replacement for hip fracture, and time‐varying market characteristics. ⁴⁰ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷

2.5. Outcome measures

The primary outcome was joint replacement complication rate, a composite measure for which higher rates indicate worse quality. ⁴⁸ Developed by the National Quality Forum, the measure is used by CMS in evaluations of CJR and contributes to determination of financial bonuses or penalties. ³⁸ The measure encompasses complications including acute myocardial infarction, pneumonia, and sepsis within 7 days of admission; surgical site bleeding, pulmonary embolism, and death within 30 days; and mechanical complications or wound infection within 90 days. We applied the measure to optimize sensitivity rather than specificity, by identifying surgical complications (bleeding and wound infection) when either a diagnosis or procedure code specified in the measure was present. This modification was consistent with our study objectives to identify changes in complications, as opposed to issuing financial penalties, particularly given that implementation of the complication measure has varied considerably in prior research. ¹⁹ , ²⁰ , ³⁷ Additional information on different versions of this measure, and how complication rates in this study compare to those reported in prior studies, can be found in Supplement S2. In secondary analyses, we tested sensitivity of the main findings to different measure definitions.

Secondary quality outcomes included 90‐day mortality (after the index hospitalization) and 90‐day unplanned hospital readmissions. We also examined 90‐day total episode spending, adjusted for inflation. We also examined utilization outcomes which we designated as exploratory, including discharge to post‐acute institutional care (PAC; includes skilled nursing facility (SNF), inpatient rehabilitation facility, and long‐term care hospital); for patients discharged to SNFs, SNF length‐of‐stay (LOS); receipt of home health (HH) services; and for patients who received HH services, HH days. We examined healthy days at home (HDAH), which counts the number of days that patients are alive and do not receive hospital, PAC, emergency department, or HH services for 90 days following discharge from the index hospitalization. ⁴⁹ These outcomes were included given their relevance to prior bundled payment analyses but deemed exploratory given our focus on disparities in quality rather than differences in utilization, as well as findings from assumption testing of the difference‐in‐differences methodology (see below). ⁴⁵

2.6. Data analysis

We used a difference‐in‐differences method to examine changes in disparities before and after implementing bundled payments. To do so, we used multivariable generalized linear models to estimate differential changes in outcomes for dual‐eligible versus non‐dual‐eligible patients receiving joint replacement at participant versus nonparticipant hospitals, before and after implementation of bundled payments (Supplement S3). ⁵⁰ This approach generated estimates of absolute changes in socioeconomic disparities at participating versus nonparticipating hospitals.

All models included hospital, time, and MS‐DRG fixed effects and were adjusted for time‐varying market characteristics, including Medicare beneficiary population size, Medicare Advantage penetration, market concentration (as defined by Herfindahl–Hirschman Index), and ACO penetration. ⁵¹ The inclusion of hospital fixed effects allowed us to generate within‐hospital comparisons over time and mitigate potential confounding from selection along time‐invariant hospital and patient attributes that were not observed in our data. The model included an interaction between Elixhauser comorbidity index and time to allow for changing relationships between comorbidities and outcomes over time. ⁵² Given concern for changes in access based on beneficiary low‐income status, we also assessed for evidence of selection based on observable patient characteristics by plotting raw trends and using difference‐in‐difference models to assess changes in variables such as age and Elixhauser comorbidity index over time by hospital participation group and patient dual‐eligible status (Supplement S4). We also examined for evidence of selection based on market‐level changes in the number of procedures for dual‐eligible, non‐dual eligible, and all beneficiaries after the start of CJR (Supplement S4).

All models used robust standard errors clustered at the hospital level and applied market‐level sampling weights to account for strata used to randomize markets in CJR. ²⁰ , ²³ We plotted risk‐standardized complication rates over time for each study group by dual‐eligible patient status using the marginal standardization approach (Supplement S5). ⁵³ Statistical tests were 2‐tailed and considered significant at α equal to 0.05. Analyses were performed using Stata version 16.0 (StataCorp) and SAS version 9.4 (SAS Institute Inc).

We took steps to evaluate the validity of the difference‐in‐differences approach. First, we examined trends for outcomes in the baseline period for participant versus nonparticipant hospitals (Supplement S6). Second, we examined trends in outcomes for dual‐eligible versus non‐dual‐eligible patients at nonparticipant hospitals during the intervention period (Supplement S7). ⁵⁴

In secondary analyses, we examined whether the relationship between CJR participation and changes in disparities varied according to previous experience with bundled payments. First, we repeated the primary analysis among hospitals without prior bundled payment experience prior to CJR. Second, we repeated the primary analysis for the subgroup of hospitals with prior bundled payment experience, using event study methods to define an intervention period for each hospital as the interval between the time‐varying start of BPCI participation and December 2017. ⁵⁵

We conducted sensitivity analyses. First, we repeated analyses using eligibility for the Medicare Low‐Income Subsidy (LIS), rather than dual eligibility for Medicare and Medicaid, as an indicator of socioeconomic status. Second, we repeated the primary analysis using the original joint replacement complication rate measure. Finally, we examined differential changes for subtypes of complications, dividing complications into surgical (surgical site bleeding, wound/joint infection, mechanical complication, venous thromboembolism) and medical (acute myocardial infarction, pneumonia, sepsis) complications, as well as dividing complications into those occurring during the index hospitalization and those occurring during the post‐discharge period. ⁵⁶

3. STUDY RESULTS

3.1. Study population

Our sample consisted of 1,603,555 patients undergoing joint replacement at 1738 hospitals between 2011 and 2017, including 751,707 patients who received joint replacement at 845 participating hospitals and 86,842 (11.6%) of whom were dual‐eligible. A total of 851,848 patients received joint replacement at 893 nonparticipant hospitals, 89,657 (10.5%) of whom were dual‐eligible. During the baseline period, the mean age of individuals receiving joint replacement was 74.6 years (SD 8.5); 64.3% were women; 5.6% had Black race; and 3.5% had Hispanic ethnicity (Table 1). Patient characteristics for the post‐intervention period are shown in Supplement S8; hospital and market characteristics are included in Supplement S9.

TABLE 1.

Patient characteristics by hospital participation in bundled payments for lower extremity joint replacement and dual‐eligibility status, baseline period.

Patient characteristics	Participant hospitals		Nonparticipant hospitals
	Dual eligible	Non‐dual eligible	Dual eligible	Non‐dual eligible
Total beneficiaries, no.	66,848	495,647	68,969	566,864
Age at admission, mean (SD) years	70.2 (13.0)	75.6 (8.0)	68.6 (13.0)	75.3 (7.8)
Sex, no. (%)
Female	50,755 (75.9)	314,620 (63.5)	52,469 (76.1)	356,891 (63.0)
Male	16,093 (24.1)	181,027 (36.5)	16,500 (23.9)	209,973 (37.0)
Race and ethnicity, no. (%)
White	38,492 (57.6)	451,529 (91.2)	45,672 (66.2)	521,456 (92.0)
Black	11,402 (17.1)	22,109 (4.5)	11,807 (17.1)	22,468 (4.0)
Hispanic	11,110 (16.7)	11,555 (2.3)	7917 (11.5)	11,633 (2.1)
Other a	5844 (8.7)	10,454 (2.1)	3573 (5.2)	11,307 (2.0)
Disabled, no. (%)	21,120 (31.6)	21,343 (4.3)	25,338 (36.7)	24,343 (4.3)
Total number of elixhauser comorbidities, mean (SD)	5.8 (3.1)	4.2 (2.6)	5.5 (3.0)	4.0 (2.6)
Most common elixhauser comorbidities, no. (%)
Hypertension	58,593 (87.7)	417,537 (84.2)	59,209 (85.8)	478,532 (84.4)
Hypothyroid	19,652 (29.4)	149,548 (30.2)	19,610 (28.4)	162,652 (28.7)
Diabetes mellitus	29,500 (44.1)	143,590 (29.0)	28,049 (40.7)	157,007 (27.7)
Chronic lung disease	28,052 (42.0)	128,087 (25.8)	29,506 (42.8)	145,859 (25.7)
Obesity	20,934 (31.3)	107,802 (21.7)	23,067 (33.5)	127,325 (22.5)
Prior hospitalization in past 1 year, no. (%)	18,250 (27.3)	84,558 (17.1)	18,477 (26.8)	94,939 (16.8)
Prior skilled nursing facility or inpatient rehabilitation use in preceding 1 year, no. (%)	7652 (11.5)	28,950 (5.9)	7198 (10.4)	31,434 (5.6)
Procedure performed for hip fracture, no (%)	16,968 (25.6)	77,649 (15.7)	15,907 (23.1)	82,436 (14.5)

^a Includes Asian, North American Native, and Unknown race/ethnicity.

3.2. Primary analysis

Among participant hospitals, the joint replacement complication rate decreased between the baseline and intervention periods from 11.0% to 10.1% for dual‐eligible patients and 7.0% to 6.4% for non‐dual‐eligible patients (Supplement S10). Among nonparticipant hospitals, the joint replacement complication rate decreased from 10.3% to 9.8% for dual‐eligible patients and from 6.7% to 6.0% for non‐dual‐eligible patients.

In adjusted analyses, participation in bundled payments was associated with a differentially lower rate of joint replacement complications for dual‐eligible compared to non‐dual‐eligible individuals (−0.9 percentage points; 95% CI −1.6 to −0.1). Plots of risk‐standardized trends illustrated the differential reduction in disparities between dual and non‐dual eligible beneficiaries for hospitals following the start of bundled payment participation (Figure 1). Dual‐eligibility status was not associated with differential changes in 90‐day readmissions or mortality (Figure 2) but was associated with differentially lower total episode spending ($1230 lower for dual vs. non‐dual eligible individuals; 95% CI −$1648 to −$812).

FIGURE 1.

Risk‐standardized plot of joint replacement complication rates for dual‐eligible (solid) and non‐dual‐eligible individuals (dashed), for participating (green) and nonparticipating hospitals (gray), before and after mandatory participation in bundled payments (red vertical line).

FIGURE 2.

Adjusted differences for quality, utilization, and spending outcomes between dual‐eligible and non‐dual eligible individuals receiving care at participant hospitals, as compared to nonparticipant hospitals.

For exploratory utilization outcomes, there were differential changes observed in discharge to post‐acute institutional care (−1.6 percentage points; 95% CI −3.1 to −0.02), SNF LOS (−1.1 days; 95% CI −1.6 to −0.6). Dual‐eligible individuals had differentially greater healthy days at home compared to non‐dual‐eligible individuals (1.0 days, 95% CI 0.3 to 1.6). There were no differential changes observed for receipt of HH (−0.1 percentage points; 95% CI −1.9 to 1.7) or HH days (0.1 days; 95% CI −0.2 to 0.4). Estimates for all outcomes are provided in Supplement S11.

3.3. Secondary analyses

Among hospitals without prior experience in bundled payments, CJR participation was associated with a differentially lower rate of joint replacement complications for dual‐eligible compared to non‐dual‐eligible individuals (−0.8 percentage points; 95% CI −1.4 to −0.2). For these hospitals, CJR participation was also associated with differentially lower total episode spending ($1254 lower for dual vs. non‐dual eligible individuals, 95% CI −$1573 to −$934).

Among participant hospitals with experience in bundled payments, there were no significant differential changes in the rate of joint replacement complications following adoption of bundled payments (−0.8 percentage points; 95% CI −1.9 to 0.3). However, there were significantly decreased differences in total episode spending by −$1284 (95% CI −1867 to −700). Figure 3 compares findings for the two subgroups of CJR participant hospitals.

FIGURE 3.

Adjusted differences for quality, utilization, and spending outcomes between dual‐eligible and non‐dual eligible individuals receiving care at participant hospitals, stratified by no prior experience with bundled payments (blue) and prior experience with bundled payments (yellow), compared to nonparticipant hospitals.

Participation in bundled payments was associated with a differentially lower joint replacement complication rates for LIS‐eligible Medicare patients compared to other patients (−0.6 percentage points; 95% CI −1.4 to −0.1). There were also differentially lower 90‐day readmissions for LIS‐eligible patients (−0.7 percentage points; 95% CI −1.3 to −0.1). There were no other substantial differences in this analysis from the main study findings (Supplement S12). Analyses using the original joint replacement complication rate measure revealed qualitatively similar findings compared to primary analyses, although differential reductions in complications were not statistically significant by a small margin (Supplement S13).

Finally, participation in bundled payments was associated with differentially lower surgical complications for dual‐eligible patients (−0.8 percentage points vs. non‐dual eligible patients; 95% CI −1.3 to −0.3) but not medical complications (−0.1 percentage points vs. non‐dual eligible patients; 95% CI −0.4 to 0.2). There were no significant differential reductions in complications for either the index admission (−0.3 percentage points; 95% CI −0.9 to 0.2) or post‐discharge period (−0.5 percentage points; 95% CI −1.0 to 0.03, Supplement S13).

4. DISCUSSION

In this study of Medicare's CJR program, mandatory hospital participation was associated with a differential reduction in joint replacement complications for dual‐eligible versus other patients. We also observed differentially lower healthcare spending but no differential changes in mortality or hospital readmissions. Further, observable measures of selection such as characteristics of patients receiving LEJR or surgical volumes did not exhibit changes. To our knowledge, this is the first study to demonstrate disparity reductions for a quality outcome among individuals with low income under a contemporary value‐based payment model. These findings are notable given mixed evidence to date on the impact of mandatory bundled payments for other dimensions of health equity. ³⁰ , ³⁶ , ³⁷ , ³⁸ These findings are immediately relevant to the newly announced Transforming Episode Accountability Model (TEAM), a new Medicare bundled payment model that will be mandatory and include LEJR starting in 2026. ¹⁸

This analysis builds upon existing literature in several ways. Prior research has emphasized spending or utilization outcomes, but not direct measures of clinical quality. ³⁵ , ³⁶ , ³⁷ For instance, one study examining changes across racial and ethnic groups reported differential reductions in the rates of discharge to post‐acute institutional care for Black patients as compared to White patients, although this may or may not represent improvement in quality. ³⁷ With respect to individuals with low income, prior evaluations of CJR have found overall reductions in readmissions across both dual‐eligible and non‐dual eligible groups but no change in disparities. ³⁵ , ³⁷ , ³⁸ Our study did observe differentially lower discharges to institutional post‐acute care, in contrast to one prior report, which may be due to differences in study design such as inclusion of hospitals participating in the earlier BPCI program. ³⁵

Our study, however, focused on a quality measure that hospitals can monitor and improve in response to payment reforms. Joint replacement complications represent an unambiguously negative outcome for patients. ⁴⁸ At a high level, there are two potential explanations for the modest reduction in joint replacement complication disparities, which require further examination through more granular approaches. ⁵⁷ Reduced disparities may reflect steps taken by hospitals to implement targeted process changes specifically for dual‐eligible patients with recognition that these beneficiaries have incurred greater cost and have worse outcomes. Alternatively, it is possible that broader changes to improve care delivery have disproportionate benefits for dual‐eligible individuals.

Several factors could influence the ability of hospitals to address disparities under value‐based arrangements like bundled payments. Our findings that disparity reductions were only observed among hospitals without prior bundled payment experience was unexpected. This finding may reflect differences in organizational capacity, given that hospitals with prior bundled payment experience differed from those without experience along key characteristics such as volume, teaching status, and ownership status. ¹⁵ However, hospitals with prior experience also tended to be larger, nonprofit, and teaching facilities, characteristics which are generally associated with greater capacity to improve care delivery. ¹⁵ A more plausible explanation is that hospitals with experience in joint replacement bundle payments under BPCI had more favorable patient populations (e.g., those without challenging social circumstances) with fewer opportunities to reduce disparities. Importantly, the design of this study, which used a baseline period that predated the start of the BPCI program, precluded the possibility that improvements occurring after BPCI participation yet before CJR participation were the reasons for these findings. Regardless, these findings suggest that experience with bundled payments is not necessarily a prerequisite to improving disparities for individuals with low income.

Taken together, our findings and past evidence suggest that policymakers should consider multiple dimensions of payment model design with respect to health equity. One dimension of equity is organizational participation. ⁴ , ⁷ , ⁵⁸ To date, nearly all models have been voluntary, allowing healthcare organizations to determine whether to enroll and assume financial accountability for the quality and costs of care. This approach has led to uneven participation across communities, precluding historically marginalized populations from accessing the potential benefits of innovative payment models. ⁵⁹ Mandatory models may guard against inequitable participation by including all healthcare delivery organizations in a given area, prompting care improvements among a broad range of healthcare organizations that serve varied populations. ⁴ , ²² For example, hospitals serving communities with higher proportions of racial minorities were more likely to participate in CJR, and hospitals selected for CJR were also more likely to be safety‐net institutions. ¹⁴ , ¹⁵ However, there are also concerns that mandatory models could exacerbate disparities by compelling participation from organizations with insufficient resources or expertise to meet the needs of more complex and socially vulnerable patients. While this study offers reassurance that patients with low‐income do not have worsened outcomes, its findings must be interpreted alongside evidence on other outcomes and other patient populations. For example, prior studies have demonstrated that safety‐net hospitals incurred disproportionate financial penalties under CJR than other hospitals. ³¹ , ³² , ³³ , ³⁴

Another crucial dimension is access to care. ⁶⁰ Prior studies suggest that participation in CJR was associated with reduced access to joint replacement surgery for historically marginalized groups. One study reported fewer procedures for total knee replacement, but not total hip replacement, for Black compared to White patients as well as dual‐eligible compared to non‐dual eligible individuals. ³⁰ Another study confirmed that Black patients underwent relatively fewer hip and knee replacements following hospital participation in CJR. ²⁹ Qualitative evaluation suggests that some CJR hospitals changed guidance for surgeons to prompt consideration of social risk factors when deciding to perform the procedure. ³⁸

Ultimately, mandatory participation in payment models can have complex consequences that preclude simple categorization of being “good” or “bad” for health equity. This study lends support to such efforts by highlighting the potential for payment models, even those not explicitly designed to address inequity, to reduce disparities in outcomes for historically marginalized populations. These findings also underscore the reality that achieving a more focused and coordinated approach to designing equity‐oriented programs will require a more sophisticated policymaking approach, lest models create a complex milieu of effects on equity. Such insights are highly policy relevant to the newly announced Transforming Episode Accountability Model, a mandatory bundled payment model proposed by CMS to address health disparities as well as improve outcomes. ¹⁸

5. LIMITATIONS

This study had limitations. First, as with all observational analyses, findings were subject to residual confounding. However, we mitigated these concerns by using a difference‐in‐differences design that incorporated hospital fixed effects to account for unobserved heterogeneity as well as accounting for multiple observed patient, hospital, and market characteristics. We also conducted additional analyses to assess the robustness of our analytic approach. Second, differential reductions in complications by dual‐eligibility status could have been related to patient selection rather than quality improvements. However, we did not find evidence of substantial changes in the case‐mix or volumes for joint replacement across our study period (Supplement S4). Third, hospitals that participated in BPCI differed from CJR‐eligible hospitals that did not participate in BPCI. ¹⁵ However, differences were stable during the pre‐intervention period, and we also conducted subgroup analyses to separately examine these groups of hospitals. Fourth, we examined disparities in several measures, but due to limitations in claims data, other aspects of quality such as patient‐reported outcomes and experience should be assessed in future research. Finally, episode‐based models represent an important type of payment model, but future work should assess the impact of other types of programs, such as population‐based models, on health equity.

6. CONCLUSION

Mandatory participation in joint replacement bundles was associated with reduced disparities in complications among individuals with low income. The reduction in disparities was observed among hospitals without prior bundled payment experience. Among the different ways that payment reform may impact equity, policymakers should consider evidence for reduced disparities in outcomes when designing and implementing future value‐based payment models, including participation mechanism.

FUNDING INFORMATION

This work was funded by the National Institute on Minority Health and Health Disparities (R01‐MD013859) and Agency for Health Research and Quality (R01HS027595). Dr. Kilaru received support from the National Heart, Lung, and Blood Institute (K23HL171859). Sponsors did not have any role in the research.

CONFLICT OF INTEREST STATEMENT

Dr. Liao reported receiving personal fees from the Washington Health Alliance, Comagine Health, and Marcus Evans. Dr. Navathe reports grants from Hawaii Medical Service Association, grants from Commonwealth Fund, grants from Robert Wood Johnson Foundation, grants from Donaghue Foundation, grants from Pennsylvania Department of Health, grants from the Veterans Affairs Administration, grants from Ochsner Health System, grants from United Healthcare, grants from Blue Cross Blue Shield of NC, grants from Blue Shield of CA, grants from Humana Inc., personal fees from Navvis Healthcare, personal fees and equity from Navahealth, personal fees from YNHHSC/CORE, personal fees from Maine Health Accountable Care Organization, personal fees from Singapore Ministry of Health, personal fees from Elsevier Press, personal fees from Medicare Payment Advisory Commission, personal fees from Cleveland Clinic, personal fees from Analysis Group, personal fees from VBID Health, personal fees from Advocate Physician Partners, personal fees from the Federal Trade Commission, personal fees from Catholic Health Services Long Island, and equity from Clarify Health, and non‐compensated board membership for Integrated Services, Inc. outside the submitted work in the past 3 years.

Supporting information

Data S1. Supporting information.

Kilaru AS, Liao JM, Wang E, et al. Association between mandatory bundled payments and changes in socioeconomic disparities for joint replacement outcomes. Health Serv Res. 2024;59(5):e14369. doi: 10.1111/1475-6773.14369