Characteristics of Hospitals That Did and Did Not Join the Bundled Payments for Care Improvement–Advanced Program

Abstract

This study describes the characteristics of hospitals participating vs not participating and remaining vs leaving the 2018 CMS Bundled Payments for Care Improvement–Advanced Program (BPCI-A), a payment reform model that reimburses for a 90-day episode of care rather than for individual services.

The Bundled Payments for Care Improvement (BPCI) initiative¹ and its 2018 successor, BPCI-Advanced (BPCI-A), which continues through 2023, are central to the Centers for Medicare & Medicaid Services’ (CMS’) move toward value-based and alternative payment models.² Under bundled payments, an accountable entity takes responsibility for the costs of an “episode” of care from hospitalization through 90 days after discharge.

Broad participation is necessary if alternative payment models are to have a substantial effect. However, previous reports showed that only 12.1% of eligible hospitals enrolled in BPCI, 20% of hospitals that enrolled dropped out, and 47% of the hospital-condition pairs were discontinued prematurely.³ The BPCI-A program also allows hospitals to drop out without penalty. Therefore, understanding which hospitals joined (and left) BPCI-A is important as policy makers contemplate the design of this and other payment programs.

Methods

We obtained hospital participation in BPCI (2013-2018) and BPCI-A (2018-2019) from CMS and hospital characteristics from the 2014 American Hospital Association Survey and the 2013 Area Resource File. We compared characteristics among BPCI-A participating hospitals vs nonparticipants and among hospitals that dropped out vs those that remained. Among participants in the original BPCI program, we used logistic regression at the hospital-condition pair level to determine whether duration of participation, completion of BPCI, achieving reduced Medicare payments,⁴ or achieving reduced readmission rates were associated with joining BPCI-A for the same condition.

Analyses were conducted using SAS version 9.4 (SAS Institute Inc). Two-tailed P<.05 was considered significant. This study was approved by the Office of Human Research Administration at the Harvard School of Public Health.

Results

There were 832 hospitals that joined BPCI-A (Table 1) for 5868 hospital-condition pairs, compared with 425 hospitals and 3050 pairs in BPCI. Among other characteristics, BPCI-A participants had higher payments and higher readmission rates at baseline; were more often large, for-profit, urban teaching hospitals and system members than nonparticipants; and had higher operating margins and lower proportions of dually enrolled beneficiaries.

Table 1. Participation in and Dropout From the BPCI-A Program.

	BPCI-A Participants (n = 832)	BPCI-A Nonparticipantsa (n = 2686)	P Value	Among Initial BPCI-A Participants			P Value
				Stayers (n = 338)	Partial Dropouts (n = 371)	Full Dropouts (n = 123)
2016 Medicare payments, $b,c	25 151	23 903	<.001	25 413	25 036	24 288	<.001
2016 90-day readmissions, %b	35.1	33.6	<.001	36.5	34.6	32	<.001
Hospital characteristicsd
For profit, %	30.4	22.3	<.001	29.2	25.4	50.9	<.001
Nonprofit, %	65.4	59.6	<.001	68.5	68.5	44.3	<.001
Public, %	4.3	18.1	<.001	2.3	6.1	4.7	.04
Urban location, %	98	89	<.001	98.6	98.3	95.3	.09
Northeast, %	19	14	<.001	16	23.2	14.2	.02
Midwest, %	22.9	22.9	.99	24.6	24	13.2	.04
South, %	38.2	44.8	<.001	34.7	37.8	50.9	.01
West, %	20	18.3	.29	24.6	14.9	21.7	<.001
Member of a system, %	72	47.4	<.001	76.2	68.2	70.8	.06
Teaching (major/minor) hospital, %	45.9	29.8	<.001	46.4	49.2	33	.01
Safety-net hospital, %	9.5	9.4	.92	12	7.7	7.7	.12
ACO hospital, %	5.3	2.5	<.001	4.6	6.3	4.3	.50
Total beds, mean No.	289	186	<.001	301	300	209	<.001
Hospital total margin, mean %	6.4	4.2	.05	6.5	6.4	6.4	.98
Hospital operating margin, mean %	1.8	−4.1	<.001	1.9	1.3	2.7	.65
County-level market characteristics
Household income in 2013, mean of the median, $	53 682	50 616	<.001	54 476	53 617	51 291	.11
Medicare Advantage in 2013, mean %	28.6	25.9	<.001	31.6	26.4	26.5	<.001
Skilled nursing facility total beds in 2013, mean No.	4889	3357	<.001	5500	4764	3299	.02
Rehabilitation hospitals in 2012, mean No.	0.82	0.55	<.001	1.08	0.67	0.48	<.001
Hospital market share, mean proportion	0.37	0.5	<.001	0.31	0.41	0.43	<.001
Hershman-Herfindahl Index, meane	0.13	0.13	.24	0.12	0.13	0.14	.10
Patient characteristics, mean %
Dual Medicare/Medicaid	28.2	30.9	<.001	30.4	26.8	26.3	<.001
Black	10.4	10.4	.98	11.2	10.4	8.2	.11
Hispanic	8.1	5.7	<.001	8.7	5.7	14.1	<.001

Abbreviations: ACO, accountable care organization (defined as participation in the Medicare Shared Savings Program); BPCI-A, Bundled Payments for Care Improvement–Advanced.

^aNonparticipants include hospitals that participate in the Inpatient Prospective Payment System and were therefore eligible to participate in BPCI.

^bFor the 28 inpatient bundles included in BPCI-A.

^c2016 payments include inpatient, readmission, skilled nursing, inpatient rehabilitation, long-term care hospitals, home health, and durable medical equipment.

^dFifteen BPCI-A participating hospitals were not able to be linked to the American Hospital Association files and are therefore not included in the hospital characteristics section.

^eThe Hershman-Herfindahl Index is calculated as the sum of squares of the market share for hospitals in a market and reflects the competitiveness of a market, with lower numbers indicating more competition.

Among initial BPCI-A participants, 123 (14.8%) dropped out of inpatient bundles fully and 371 (44.6%) dropped at least 1 bundle (Table 1). Hospitals that dropped out were more often for profit, smaller, and located in areas with a lower supply of skilled nursing and inpatient rehabilitation facilities compared with hospitals that remained in the program.

Of the 2175 hospital-condition pairs from BPCI for conditions included in BPCI-A, 35.1% were reenrolled in BPCI-A (same hospital enrolled for the same condition) (Table 2). Pairs that stayed in the program until its completion were more likely to join BPCI-A (41.0% vs 27.0% for dropouts; difference, 14.0%; 95% CI, 9.5%-18.4%; P < .001). There was no relationship between duration of participation, saving at least 2%, or reducing readmission rates under BPCI and participation in BPCI-A.

Table 2. Association of Duration, Dropout, and Performance Under the Bundled Payments for Care Improvement With Participation in the Bundled Payments for Care Improvement–Advanced Program^a.

	No. of Hospital-Condition Pairs	Reenrolling, %	Difference, % (95% CI)	P Value
Overall	2175	35.1
Participated for longer than median duration	948	36.6	2.5 (−2.0 to 7.1)	.28
Participated for median duration (5 quarters) or less	1227	34.0
Stayed until program completion	1265	41.0	14.0 (9.5 to 18.4)	<.001
Did not stay until program completion	910	27.0
Saved at least 2% from baseline (mean, −$4779)	1277	36.4	3.2 (−1.1 to 7.5)	.14
Did not save at least 2% from baseline (mean, +$3479)	898	33.3
Reduced readmission rates from baseline	1052	34.3	−1.7 (−5.8 to 2.4)	.41
Did not reduce readmission rates from baseline	1123	36.0

^aAnalysis was conducted at the hospital-condition pair level. Models included all listed covariates. The 6-month baseline period for each pair was defined as the time from 9 months prior to 3 months prior to its start date. Conditions for which hospitals’ standardized inflation-adjusted Medicare payments for hospitalization plus 90 days after discharge declined from baseline to 2016 by at least 2% were considered to have saved because in general, hospitals achieving this level of savings were eligible to share in those savings. Payments include inpatient, readmission, skilled nursing, inpatient rehabilitation, long-term care hospitals, home health, and durable medical equipment. Conditions for which hospitals’ 90-day readmission rates declined between baseline and 2016 were considered to have reduced readmission rates. Payments and readmission rates were adjusted for medical comorbidities using the Centers for Medicare & Medicaid Services Chronic Conditions Warehouse.

Discussion

In 2018-2019, there were more hospitals participating in BPCI-A than participated in BPCI, although BPCI-A hospitals were larger and more well resourced than nonparticipants. Hospitals that dropped out early from BPCI-A were smaller and located in areas with lower levels of postacute care supply. Hospitals that stayed in BPCI were particularly likely to reenroll in BPCI-A.

The broader enrollment seen in BPCI-A thus far may reflect financial incentives created under the Medicare Access and CHIP Reauthorization Act to participate in alternative payment models,⁵ which make these programs more attractive. It is also possible that hospital leaders believe that such models will be mandatory in coming years and participating now may improve their ability to manage effectively in the future.

There are limitations to this study. Only hospital participation in BPCI and BPCI-A was examined, and results may not generalize to physician group practice participants. Because CMS does not release target price information, savings estimates reflect changes in payments only and not penalties or rewards under BPCI.

In summary, participation in BPCI-A has exceeded that in BPCI. However, different strategies may be needed to attract a more diverse hospital population to participate persistently, including smaller hospitals and those with less favorable finances at baseline.

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