Association of Payment Model Changes With the Rate of Total Joint Arthroplasty in Patients Undergoing Kidney Replacement Therapy

Abstract

Background

To encourage high-quality, reduced-cost care for total joint arthroplasty (TJA), the Centers of Medicare & Medicaid Services mandated a pay-for-performance model, the Comprehensive Care for Joint Replacement (CJR), as part of the Patient Protection and Affordable Care Act (PPACA). The CJR incentivizes cost containment, and it was anticipated that its implementation would reduce access to TJA for high-cost populations. Patients with end-stage kidney disease (ESKD) undergoing kidney replacement therapy (dialysis and kidney transplant) are costly compared with healthier patients, but it was unknown whether this population lost access to hip and knee replacement because of CJR implementation. This population allows study of whether TJA is accessible for medically complex patients whose risk of surgical complications has been mitigated, as kidney transplantation improves outcomes compared with dialysis, allowing evaluation as to whether access improved when patients crossed over from dialysis to transplantation. Because all patients with ESKD are included in a mandated national registry, we can quantify whether access changed for patients who underwent dialysis and transplantation.

Questions/purposes

(1) How did the rate of TJA change amid the shift to bundled payments for patients with ESKD receiving dialysis? (2) How did the rate of TJA change amid the shift to bundled payments for patients with ESKD after kidney transplant?

Methods

This was an observational cohort study from 2008 to 2018 using the United States Renal Data System, a mandatory national registry that allows for the opportunity to study all individuals with ESKD. During the study period, we identified 1,324,614 adults undergoing routine dialysis and 187,212 adult kidney transplant recipients; after exclusion for non-Medicare primary insurance (n = 785,224 for dialysis and 78,011 for transplant), patients who were 100 years or older (n = 79 and 0, respectively), those who resided outside of 50 US states and Puerto Rico (n = 781 and 87, respectively), missing dialysis status for the dialysis cohort (n = 8658), and multiorgan transplant recipients for the transplant cohort (n = 2442), our study population was 40% (529,872) of patients who underwent routine dialysis and 57% (106,672) of adult kidney transplant recipients, respectively. TJA was ascertained using Medicare Severity Diagnosis Related Groups and ICD-9 and ICD-10 codes. We divided the study period by PPACA (January 1, 2014, to March 31, 2016) and CJR (April 1, 2016, to December 31, 2018) implementation and compared the incidence of TJA by era using mixed-effects Poisson regression adjusting for calendar time and clinical and demographic variables.

Results

After adjustment for linear temporal trend and patient case mix, there was no evidence of association between policy implementation and the incidence of TJA. In the dialysis cohort, the adjusted incidence rate ratio (IRR) for TJA was 1.06 (95% confidence interval [CI] 0.98 to 1.14; p = 0.2) comparing PPACA with the previous period and 1.02 (95% CI 0.96 to 1.08; p = 0.6) comparing CJR with the previous periods. Similarly, in the transplant cohort, the adjusted IRR for TJA was 0.82 (95% CI 0.67 to 1.02; p = 0.07) comparing PPACA with the previous period and 1.10 (95% CI 0.94 to 1.28; p = 0.9) comparing CJR with the previous periods.

Conclusion

There was no loss in access to TJA for medically complex patients receiving kidney replacement therapy. The increase in TJA incidence for patients after kidney transplant and decrease for patients receiving dialysis suggest that surgeons continued to provide care for higher risk patients whose risk of morbidity or mortality with total joint replacement has been maximally improved after transplantation.

Level of Evidence

Level III, prognostic study.

Introduction

Because total joint arthroplasty (TJA) is a common procedure with relatively high costs [5], policymakers have attempted to devise ways to lower or at least constrain costs. To provide high-quality care at reduced cost, the Centers for Medicare & Medicaid Services began to transition away from the traditional fee-for-service model to a pay-for-performance model under the Patient Protection and Affordable Care Act (PPACA) in 2010 [4]. This culminated in the Comprehensive Care for Joint Replacement (CJR) model for TJA in 2016, which mandated that (1) hospitals were responsible for the costs associated with the procedure and 90 days after discharge and (2) penalties or rewards would be imposed on hospitals if the episode of care resulted in above- or below-average reimbursement, respectively.

These policy changes raise concern for inequitable reimbursement for patients with numerous or severe medical comorbidities, including those undergoing kidney replacement therapy (dialysis and kidney transplant), eventually imposing a barrier to TJA in such subgroups [7, 10, 12]. Reasons for this have been multifactorial, including candidate complexity and lack of care coordination throughout an entire episode, which has led to increased complication and readmission rates [3]. With nearly 75% of patients with end-stage kidney disease (ESKD) being older than 65 years [9, 11, 13], there is demand for TJA in these patients.

Patients requiring kidney replacement therapy are a unique model for studying access in medically complex patients. Bundled payment models frequently focus on medical optimization before surgery. Based on current studies, patients after kidney transplant have improved outcomes compared with patients receiving dialysis [2, 6, 16, 17]. Therefore, this population—unlike populations who have undergone other solid organ transplants or most medically complex conditions—has a clear before and after medical optimization state [2, 6, 8, 14, 16, 17].

Therefore we asked: (1) How did the rate of TJA change amid the shift to bundled payments for patients with ESKD receiving dialysis? (2) How did the rate of TJA change amid the shift to bundled payments for patients with ESKD after kidney transplant?

Patients and Methods

Data Source

This study used data from the United States Renal Data System (USRDS), a national data registry that collects, analyzes, and distributes information on ESKD in the United States, including treatment and outcomes. Established and overseen by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), the USRDS is a compilation of various data sources, primarily from the mandatory reports submitted to the Centers for Medicare & Medicaid Services and the Organ Procurement and Transplantation Network for all eligible patients in the US.

Study Population

This study consisted of two cohorts of adult (age 18 years or older) patients who either initiated dialysis or underwent kidney-only transplantation, both of whom had Medicare Part A, B, or Other as their primary form of insurance. We identified both cohorts from January 1, 2008, to December 31, 2018 (Fig. 1).

Fig. 1.

Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) flow diagram of the study population.

We categorized the study period into three periods according to the policy changes (pre-PPACA = January 1, 2008, to December 31, 2013; PPACA = January 1, 2014, to March 31, 2016; and mandatory CJR = April 1, 2016, to December 31, 2018). The starting date of the PPACA period reflects when the law was fully implemented, despite being enacted on March 23, 2010 [18]. Time period was treated as a time-varying exposure; in other words, participants who were in the risk set on the calendar date of transition were censored and transitioned to the subsequent time period.

We identified 529,872 unique individuals with ESKD who had ESKD during our study period and used Medicare as primary insurance, of whom 56% (296,695) initiated dialysis in the pre-PPACA period, 20% (108,288) initiated dialysis in the PPACA period, and 24% (124,889) initiated dialysis in the CJR period (Table 1).

Table 1.

Clinical characteristics of the dialysis cohort

Characteristic	Pre-PPACA (2008-2014) (n = 296,695)	PPACA (2014-2016) (n = 108,288)	CJR (2016-2018) (n = 124,889)	p value
Age at first ESKD service in years	70 ± 13	70 ± 12	70 ± 12	< 0.001
Women	45 (132,256)	43 (46,793)	43 (53,563)	< 0.001
Racea				< 0.001
American Indian or Alaska Native	1 (2898)	1 (1026)	1 (1281)
Asian	3 (9098)	3 (3614)	4 (4511)
Black	24 (70,924)	23 (24,772)	22 (26,856)
Native Hawaiian or Pacific Islander	1 (1758)	1 (786)	1 (958)
Other or Multiracial	< 1 (818)	< 1 (255)	< 1 (294)
White	71 (211,178)	72 (77,826)	73 (90,967)
Hispanic/Latino ethnicitya	10 (29,986)	10 (10,568)	10 (12,093)	< 0.001
Primary cause of ESKD				< 0.001
Diabetes	45 (133,865)	47 (50,344)	48 (59,751)
Hypertension	32 (92,976)	32 (34,041)	32 (39,066)
Glomerulonephritis	6 (16,641)	6 (6095)	5 (6545)
Cystic kidney disease	1 (3830)	1 (1514)	2 (2004)
Other urologic	2 (4625)	2 (1612)	1 (1840)
Other cause	11 (31,931)	10 (10,850)	11 (13,374)
Unknown cause	4 (10,601)	2 (2245)	1 (798)
Comorbidities
Cancer	9 (27,588)	9 (9543)	9 (11,380)	< 0.001
Smoking	6 (16,090)	5 (5740)	6 (7350)	< 0.001
Chronic obstructive pulmonary disease	12 (36,228)	12 (12,903)	12 (14,926)	< 0.001
Congestive heart failure	38 (109,365)	34 (36,392)	33 (41,147)	< 0.001
Peripheral vascular disease	16 (46,632)	12 (13,201)	11 (13,845)	< 0.001
BMI in kg/m2	28 (24-33)	28 (24-33)	28 (24-33)	< 0.001
Medicaid dual eligibility	39 (116,553)	39 (42,357)	38 (47,214)	< 0.001

Data presented as mean ± SD, % (n), or median (IQR). Percentages were calculated after removing observations with missing values.

^aThe social constructs of race and ethnicity as collected by CMS 2728 (2005-2018 versions) and reported to USRDS were considered as “perceived race” or “perceived ethnicity,” as these questions on the form were completed by the attending physician, head nurse, or social worker involved in the patient’s treatment of kidney disease.

We identified 106,672 Medicare-primary transplant recipients who underwent TJA, of whom 52% (55,433) were transplanted in the pre-PPACA period, 20% (21,690) were transplanted in the PPACA period, and 28% (29,549) were transplanted in the CJR period (Table 2).

Table 2.

Clinical characteristics of the transplant cohort

Characteristic	Pre-PPACA (2008-2014) (n = 55,433)	PPACA (2014-2016) (n = 21,690)	CJR (2016-2018) (n = 29,549)	p value
Age at time of transplantation in years	52 ± 14	52 ± 14	53 ± 14	< 0.001
Women	39 (21,498)	39 (8388)	39 (11,435)	0.95
Racea				< 0.001
American Indian or Alaska Native	1 (675)	1 (267)	1 (291)
Asian	5 (2679)	6 (1229)	6 (1720)
Black	31 (17,250)	32 (7036)	32 (9330)
Native Hawaiian or Pacific Islander	1 (490)	1 (236)	1 (290)
Other or Multiracial	< 1 (253)	< 1 (81)	< 1 (120)
White	61 (34,077)	59 (12,839)	60 (17,783)
Hispanic/Latino ethnicitya	16 (8741)	17 (3579)	17 (5020)	< 0.001
Primary cause of ESKD				< 0.001
Diabetes	28 (15,312)	27 (5728)	27 (7803)
Hypertension	23 (12,918)	25 (5459)	28 (8028)
Glomerulonephritis	25 (14,062)	25 (5396)	23 (6833)
Cystic kidney disease	7 (4005)	7 (1458)	8 (2222)
Other urologic	3 (1539)	3 (581)	2 (719)
Other cause	9 (4841)	9 (1961)	9 (2529)
Unknown cause	5 (2519)	4 (940)	3 (944)
Comorbidities
Cancer	3 (1418)	3 (600)	3 (803)	0.63
Smoking	4 (2172)	4 (797)	3 (1001)	< 0.001
Chronic obstructive pulmonary disease	2 (902)	2 (326)	2 (436)	0.03
Congestive heart failure	12 (5952)	10 (2101)	9 (2590)	< 0.001
Peripheral vascular disease	4 (2211)	3 (715)	3 (811)	< 0.001
BMI in kg/m2	28 (24-32)	28 (24-32)	28 (24-32)	< 0.001
Medicaid dual eligibility	45 (25,022)	48 (10,403)	46 (13,480)	< 0.001

Data presented as mean ± SD, % (n), or median (IQR). Percentages were calculated after removing observations with missing values.

^aThe social constructs of race and ethnicity as collected by CMS 2728 (2005-2018 versions) and reported to USRDS were considered as “perceived race” or “perceived ethnicity,” as these questions on the form were completed by the attending physician, head nurse, or social worker involved in the patient’s treatment of kidney disease.

Ascertainment of Characteristics and TJA

We obtained information on patient and recipient characteristics, including comorbid conditions, from the USRDS. Data on diagnoses of TJA were identified using Medicare Severity Diagnosis Related Groups (MS-DRGs) and ICD-9 and ICD-10 codes, as previously described elsewhere [12]. We included TJAs starting from January 1, 2008, and ending on December 31, 2018.

Descriptive Data

There were small differences in demographics and underlying disease for patients undergoing dialysis (Table 1) and transplantation (Table 2), but these differences were unlikely to affect the decision to undergo arthroplasty. Nonetheless, we included these covariables in our regression models to adjust for case mix, as we were interested in understanding how access to care might change for individual patients instead of how rates of TJA might change in response to changing demographic trends.

Receipt of TJA Over Time in Patients Undergoing Kidney Replacement Therapy

For patients receiving dialysis, receipt of TJA was assessed starting from the most recent date of either dialysis initiation or January 1, 2008, until the earliest date of a TJA procedure, death, transplant, end of Medicare primary coverage, change in time era, or administrative censorship on December 31, 2018.

For patients posttransplant, receipt of TJA was assessed starting from date of transplant until the earliest date of a TJA procedure, death, end of Medicare primary coverage, change in time era, or administrative censorship on December 31, 2018.

We compared the rates of TJA over time (PPACA versus pre-PPACA and CJR versus pre-PPACA and PPACA) using mixed-effects Poisson regression, with a random intercept term for each state. Models were adjusted for calendar time, age, gender, BMI, dual eligibility, and comorbid conditions (cancer, chronic obstructive pulmonary disease, congestive heart failure, peripheral vascular disease, and tobacco use).

Ethical Approval

This study was a secondary data analysis of publicly available data and therefore determined to be exempt from review by the NYU Langone Health Institutional Review Board.

Statistical Analysis

To compare characteristics of patients undergoing kidney replacement therapy across time periods, we used the Pearson chi-square tests for categorical variables, ANOVA for normally distributed continuous variables, and the Kruskal-Wallis test for nonnormally distributed continuous variables. We used complete case analysis to handle missing data as they ranged from 0.0001% to 5.5%. All analyses were performed using Stata 17.0/MP for Linux, with α = 0.05.

Results

Receipt of TJA Over Time in Patients Receiving Dialysis

Accounting for linear temporal trend and patient case mix, there was no evidence of association between policy implementation and receipt of TJA for patients receiving dialysis (Table 3), although the crude incidence of TJA decreased over time (Fig. 2). The adjusted incidence rate ratio (IRR) for TJA was 1.06 (95% confidence interval [CI] 0.98 to 1.14; p = 0.2) comparing PPACA to the previous period and 1.02 (95% CI 0.96 to 1.08; p = 0.6) comparing CJR to the previous periods (Table 3). Data on additional factors associated with total hip and knee arthroplasty in patients receiving dialysis have been provided (Supplemental Table 1; http://links.lww.com/CORR/B417).

Table 3.

Case mix–adjusted changes in the rate of TJA in the dialysis cohort (n = 529,872)

Model	PPACA versus pre-PPACA, adjusted IRR (95% CI)	p value for interactiona	CJR versus pre-PPACA and PPACA, adjusted IRR (95% CI)	p value for interactiona
Overall	1.06 (0.98-1.14)		1.02 (0.96-1.08)
Subgroup analysis
Age at first ESKD service in years		0.47		0.64
< 65	1.19 (0.98-1.44)		1.01 (0.87-1.18)
≥ 65	1.03 (0.95-1.12)		1.02 (0.96-1.10)
BMI in kg/m2b		0.13		0.08
< 40	1.07 (0.99-1.16)		1.00 (0.94-1.07)
≥ 40	0.84 (0.61-1.14)		1.20 (0.94-1.53)

Overall IRRs were adjusted for calendar time, age, sex, BMI, dual eligibility, and comorbid conditions (cancer, chronic obstructive pulmonary disease, congestive heart failure, peripheral vascular disease, and tobacco use). Subgroup IRRs were calendar time, age, sex, dual eligibility, and comorbid conditions.

^aInteraction p value denotes whether the interaction term was statistically significant. The interaction term demonstrates whether the association between the eras and the rate of TJA was different across levels of another variable (that is, age and BMI).

^bCancer and chronic obstructive pulmonary disease were not included in the models because of the small number of patients with these conditions.

Fig. 2.

Incidence of total hip and knee arthroplasty over time. A color image accompanies the online version of this article.

Receipt of TJA Over Time in Patients Who Underwent Transplant

After accounting for the linear temporal trend and recipient case mix, there was no evidence of association between policy implementation and receipt of TJA for patients who had received a kidney transplant (Table 4), although the crude incidence of TJA increased over time (Fig. 2). The adjusted IRR for TJA was 0.82 (95% CI 0.67 to 1.02; p = 0.07) comparing PPACA to the previous period and 1.10 (95% CI 0.94 to 1.28; p = 0.9) comparing CJR to the previous periods (Table 4). Data on additional factors associated with total hip and knee arthroplasty in patients who received kidney transplant have been provided (Supplemental Table 2; http://links.lww.com/CORR/B417).

Table 4.

Case mix–adjusted changes in the rate of TJA in the transplant cohort (n = 106,672)

Model	PPACA versus pre-PPACA, adjusted IRR (95% CI)	p value for interactiona	CJR versus pre-PPACA and PPACA, adjusted IRR (95% CI)	p value for interactiona
Overall	0.82 (0.67-1.02)		1.10 (0.94-1.28)
Subgroup analysis
Age at transplantation in years		0.93		0.38
< 65	0.88 (0.67-1.15)		1.15 (0.94-1.42)
≥ 65	0.77 (0.56-1.05)		0.98 (0.78-1.23)
BMI in kg/m2b		0.06		0.85
< 40	0.82 (0.66-1.01)		1.12 (0.95-1.31)
≥ 40	1.06 (0.26-4.32)		0.54 (0.20-1.47)

Overall IRRs were adjusted for calendar time, age, sex, BMI, dual eligibility, and comorbid conditions (cancer, chronic obstructive pulmonary disease, congestive heart failure, peripheral vascular disease, and tobacco use). Subgroup IRRs were calendar time, age, sex, dual eligibility, and comorbid conditions.

^aInteraction p value denotes whether the interaction term is statistically significant. The interaction term demonstrates whether the association between the eras and the rate of TJA was different across levels of another variable (that is, age and BMI).

^bCancer and chronic obstructive pulmonary disease were not included in the models because of the small number of patients with these conditions.

Discussion

The implementation of new payment models has created financial incentives for health systems and physicians to reduce costs through improved efficiency and decreased complications. The intent of the models was to improve care quality and value and to encourage care providers to improve their patient’s health prior to surgery. Patients receiving kidney replacement therapy—whether through dialysis or transplant—present a unique consideration, as the nature of their disease makes these patients expensive to a bundle payment model even with optimal risk mitigation through transplantation. Despite the implementation of new payment models intended to address the high cost of these procedures, we observed no association between payment model changes and receipt of TJA in this population.

Limitations

Our work had several limitations. First, as this was a registry-based analysis, we do not know whether the demand for TJA in patients undergoing kidney replacement therapy remained constant over time; that is, we do not know whether the number of patients who needed TJA increased or decreased over the study period. However, we are unaware of a mechanism that would cause demand to change over time, particularly with the adjustment for patient case mix. Second, it is possible our findings are confounded by the overall changes in the healthcare system landscape or other policy implementations that coincided with CJR and PPACA. We attempted to minimize this risk by adjusting for temporal trends as well as case mix. Finally, our study uses US claims data for TJA ascertainment, which could be incomplete. This is unlikely to threaten our results, given that patients who were missing TJA claims would have biased our findings to the null.

We were concerned that TJA access would decrease for patients receiving dialysis because of new bundled payment policies. We demonstrated that the incidence rate of TJA decreased for patients receiving dialysis from 2008 to 2018. However, after adjustment for linear temporal trend and patient case mix, there was no evidence of association between policy implementation and TJA in this population contrary to our hypothesis.

We had hypothesized that TJA access would decrease after implementation of bundled payments for patients who had received a kidney transplant. While there was an increase in the incidence rate for transplant recipients from 2008 to 2018, after adjustment for linear temporal trend and patient case mix, there was no evidence of association between policy implementation and TJA.

Discussion of Key Findings

After adjustment for relevant covariates (linear temporal trend and patient case mix), there was no evidence of association between policy implementation and TJA; specifically, there was no change in TJA rate. Our findings of opposite temporal trends in TJA for patients receiving dialysis versus kidney transplant are important. These trends were not abrupt increases associated with the policy changes, but rather gradual changes over time. During the observed time period, national studies and systematic reviews demonstrated that recipients of a kidney transplant who undergo TJA have improved outcomes compared with patients receiving dialysis [2, 6, 8, 14, 16, 17]. Additionally, dialysis has been associated with higher odds of inpatient mortality (6.6- or 3.3-fold) and complications (1.5- or 1.8-fold) among patients with ESKD who undergo hip or knee arthroplasty, respectively [16]. Taken together, the observed trends suggest that physicians may consider patients after transplant as more acceptable risk and patients on dialysis as less acceptable risk, although it is difficult to establish causality.

This is an important finding in the context of the larger discussion about access to TJA and concerns that surgeons will use strict cutoffs to determine eligibility for TJA, thus limiting care for complex patients. Our motivation for the study was based on concern that patients who had received a kidney transplant would have decreased access because of their increased complexity, even though transplantation is the best mechanism to mitigate morbidity and mortality from TJA in patients with ESKD. Were this the case, it would be a strong argument to reconsider the current structure of bundled payments for these patients. Instead, we found an increasing rate of TJA for these patients, suggesting that surgeons recognize that while these patients are medically complex, their kidney transplant decreases their risk of complication, and they are therefore indicated for TJA despite higher costs to a bundle.

While we were surprised that access was maintained, we have lingering concerns that the increased rate of TJA for patients with a kidney transplant may be dependent on socioeconomic status [1, 15]. In future studies, we plan to investigate the relationship between socioeconomic status and receipt of TJA in patients receiving kidney replacement therapy.

Conclusion

Our analysis found no change in the rate of TJA among patients receiving dialysis or kidney transplant after the implementation of PPACA or CJR. Given the complex medical needs present in these patients, the absence of adverse changes on access to TJA in this population mitigates concerns previously raised regarding potential negative impacts of PPACA and CJR.