Level of Reconciliation Payments by Safety-Net Hospital Status Under the First Year of the Comprehensive Care for Joint Replacement Program

Abstract

This study uses data from the Comprehensive Care for Joint Replacement program to assess the level of Medicare reconciliation payments and examine these levels by hospital size and teaching hospital status to understand hospital performance.

In April 2016, Medicare introduced the Comprehensive Care for Joint Replacement (CJR) program, its first mandatory bundled payment program, for patients undergoing hip or knee replacements. Approximately 800 hospitals in 67 metropolitan statistical areas were mandated to participate. The CJR holds hospitals accountable for the cost and quality of care of the entire episode of care that starts with the hospital admission and continues until 90 days after hospital discharge. If total episode spending is below the target rate and quality thresholds are met, hospitals receive reconciliation payments proportional to the difference between total episode spending and target rate. Starting in 2017, hospitals were required to repay Medicare if spending exceeded the target.

The target rate in the first 3 years of the CJR program is a weighted mean of each hospital’s historical per-episode payment rates and regional per-episode payment rates. Therefore, hospitals that had historical rates above regional rates had to decrease per-episode spending more to receive reconciliation payments than did hospitals that had rates below regional means. This payment structure might penalize hospitals caring for patients of low socioeconomic status (ie, safety-net hospitals [SNHs]). Episode spending for these patients tends to be high because they have higher complication rates and are more likely to use institutional postacute care for recovery.^1,2,3 Therefore, we assessed the level of reconciliation payments by SNH status in the first year of the program. We also examined reconciliation payment levels by hospital size and teaching hospital status to understand different hospitals’ performance under the CJR program.

Methods

We used the publicly available list of the CJR-participating hospitals and reconciliation payments of 2016 to classify hospitals as receiving high (exceeding the median per-episode reconciliation payment), low (at or below median), or no reconciliation payments (Figure). We also used the provider of services and specific files to identify SNH status (those in the top quartile of disproportionate share hospital index), size of hospitals, and major teaching hospital status. The Oregon Health & Science University’s institutional review board approved this study and waived informed consent. Claims data were not deidentified and were provided by the Centers for Medicare & Medicaid Services as a limited data set.

Figure. Distribution of Per-Episode Reconciliation Payments in 2016 Among Hospitals That Earned Payments.

Per-episode reconciliation payments were adjusted for wage index differentials.

There were 799 CJR program participating hospitals in 67 metropolitan statistical areas. We eliminated 47 specialized hospitals that had performed no hip or knee replacement operations (eg, alcohol and drug addiction rehabilitation centers, cardiac specialty hospitals, or children’s hospitals). We also excluded 6 hospitals with insufficient data.

We compared hospital characteristics across levels of reconciliation payments. We also conducted multinomial logistic regressions to calculate the relative risk ratio of receiving high vs no payments and low vs no payments. Two-tailed P < .05 was considered to be statistically significant.

Results

Among the 746 hospitals participating in the CJR program, 381 (51.1%) earned reconciliation payments (Table). High reconciliation payments were more common in non-SNHs (137 of 518 [26.4%]) than in SNHs (53 of 228 [23.2%]), in large hospitals (43 of 110 [39.1%]) than in small hospitals (78 of 383 [20.4%]), and in major teaching hospitals (50 of 141 [35.4%]) than in non–major teaching hospitals (140 of 605 [23.1%]). Results from multinomial logistic regressions were consistent.

Table. The Comprehensive Care for Joint Replacement Program Participating Hospital Characteristics by Reconciliation Payment Group^a.

Characteristic	High Payment (n = 190)b			Low Payment (n = 191)c			No Payment (Reference Outcome), No. (%) (n = 365)d
	No. (%)	Relative Risk Ratio (95% CI)	P Value	No. (%)	Relative Risk Ratio (95% CI)	P Value
Safety-net hospital
Yes	53 (23.2)	0.65 (0.44-0.95)	.03	39 (17.1)	0.43 (0.29-0.65)	<.001	136 (59.6)
No	137 (26.4)	1 [Reference]	NA	152 (29.3)	1 [Reference]	NA	229 (44.2)
Size of hospital, beds
Large (≥400)	43 (39.1)	2.77 (1.68-4.57)	<.001	26 (23.6)	1.32 (0.76-2.28)	.32	41 (37.3)
Medium (200-399)	69 (27.3)	1.54 (1.04-2.29)	.03	66 (26.1)	1.16 (0.79-1.71)	.44	118 (46.6)
Small (<200)	78 (20.4)	1 [Reference]	NA	99 (25.8)	1 [Reference]	NA	206 (53.8)
Teaching hospital
Majore	50 (35.5)	1.71 (1.12-2.61)	.01	28 (19.9)	0.82 (0.51-1.34)	.43	63 (44.7)
Not major	140 (23.1)	1 [Reference]	NA	163 (26.9)	1 [Reference]	NA	302 (49.9)

Abbreviation: NA, not applicable.

^aThe mean (SD) per-episode reconciliation payment, adjusted for wage index differentials, was $1455 ($282) for hospitals in the high-payment group and $801 ($370) in the low-payment group.

^bComposed of hospitals that received a per-episode (wage index adjusted) payment that exceeded the median per-episode payment ($1183) in year 1 of the CJR program.

^cComposed of hospitals that received a payment at or below the median.

^dComposed of hospitals that did not receive a payment.

^eMajor teaching hospitals are hospitals with membership in the Council of Teaching Hospitals of the Association of American Medical Colleges.

Discussion

The SNHs were less likely to receive high reconciliation payments. Large hospitals and major teaching hospitals were more likely to receive higher reconciliation payments. Large and teaching hospitals are, in general, known to have higher readmission rates,⁴ but this may not necessarily be true for patients undergoing hip or knee replacement. Institutional resources may play an important role in meeting the CJR targets.

This study has limitations. First, results represent first-year payments and may not be generalizable to later years when the target rates for each hospital will no longer account for hospital historical payment rates. Second, in 2018, Medicare made the CJR program voluntary in 33 metropolitan statistical areas and began covering outpatient knee replacements. It is unknown how these changes will affect hospital performance under the CJR program.

Evidence suggests that bundled payments effectively reduce health expenditures.⁵ However, our findings suggest heterogeneous effects of the CJR program across different hospital types. Particularly, the SNHs were less likely to receive high reconciliation payments, indicating that the CJR program might disproportionately benefit hospitals that serve patients with higher socioeconomic status. In the interest of promoting equity in health care, future modifications of the CJR program might consider ways to counteract these potential imbalances.