Participation and Dropout in the Bundled Payments for Care Improvement Initiative

Abstract

This study describes hospital participation and dropout in the Centers for Medicare & Medicaid Innovation’s Bundled Payments for Care Improvement Initiative, which allows hospitals and practices to retain savings if they meet quality targets.

The Centers for Medicare & Medicaid Innovation launched the Bundled Payments for Care Improvement (BPCI) Initiative in 2013, a voluntary alternative payment model that holds participating hospitals, practices, or facilities accountable for quality and costs in 30-, 60-, or 90-day episodes of care. Participants can join for as many or as few of 48 eligible conditions as they wish and drop out without penalty. If cost targets are achieved, participants keep a portion of the savings; if cost targets are exceeded, participants reimburse Medicare a portion of the difference.

To our knowledge, no published data characterize participation or dropout from the risk-bearing phase of this program. Such information is important because voluntary alternative payment models are a key element in Medicare’s strategy to improve quality and costs of care.

Methods

We evaluated model 2, which includes inpatient and post–acute spending and is the track selected by more than 99% of hospitals in BCPI. Hospitals could enroll for any of 48 conditions beginning in October 2013. We obtained quarterly public data sets from Medicare covering January 2014 through January 2017, which list participating hospitals and their planned end dates. Dropouts were defined as hospitals that initiated participation but were absent from participant lists prior to their planned end date. Hospital characteristics were obtained from the 2014 American Hospital Association Annual Survey Database.

Wilcoxon, χ², and t tests were used to compare participating hospitals with nonparticipating hospitals. Adjusted odds ratios (aORs) of dropout were calculated for hospital-condition pairs using logistic regression with clustering by hospital. Kaplan-Meier calculations were used to estimate time to dropout, censoring hospitals still active as of January 2017.

Analyses were conducted using SAS (SAS Institute), version 9.4. Two-tailed P values less than .05 were considered statistically significant.

Results

As of January 2017, 422 hospitals had signed up for BPCI model 2 (12.0% of 3523 eligible hospitals) for a mean of 7.2 conditions (SD, 9.6), yielding 3042 hospital-condition pairs (number of hospitals × number of conditions per hospital). Participating hospitals were more often nonprofit, urban, members of a system, and teaching hospitals, and had more beds and better operating margins (Table); participants were less likely to be safety-net hospitals.

Table. Hospital Participation and Dropout in the Bundled Payments for Care Improvement Initiative, 2013-2017.

Hospital Characteristicsa	No. (%)		P Value	Odds Ratio of Dropout by Hospital-Condition Pair (95% CI)	P Value
	Participating Hospitals (n = 422)	Nonparticipating Hospitals (n = 3099)
No. of conditions per participating hospital, median (IQR)	3 (1-12)	NA
Hospital type
For profit, No. (%)	58 (13.7)	774 (25.0)	<.001	2.17 (0.58 to 8.19)	.02
Nonprofit, No. (%)	341 (80.8)	1794 (57.9)		1.15 (0.32 to 4.16)
Public, No. (%)	23 (5.5)	533 (17.2)		1.00 [Reference]
Urban location, No. (%)	418 (99.1)	2769 (89.3)	<.001	0.77 (0.51 to 1.14)	.19
Member of a health system, No. (%)	293 (69.4)	1407 (45.4)	<.001	0.97 (0.64 to 1.47)	.89
ACO hospital, No. (%)	20 (4.7)	100 (3.2)	.12	3.70 (1.19 to 11.1)	.02
Teaching hospital (major or minor), No. (%)	213 (50.5)	846 (27.3)	<.001	1.23 (0.87 to 1.75)	.24
Safety-net hospital (top decile DSH index), No. (%)	29 (7.0)	301 (10.1)	.04	1.64 (0.76 to 3.57)	.21
Total beds, mean (SD)b	342 (254)	191 (186)	<.001	1.00 (1.00 to 1.00)	.46
% Annual Medicare days, mean (SD)b	49.9 (11.5)	49.9 (15.7)	.99	1.00 (0.98 to 1.01)	.59
Hospital total margin %, median (IQR)	6.4 (1.8 to 10.5)	4.4 (−1.4 to 10.5)	<.001	0.998 (0.984 to 1.012)b	.78
Hospital operating margin %, median (IQR)	0.0 (−5.2 to 6.3)	−1.9 (−10.0 to 5.9)	<.001	0.999 (0.998 to 1.000)b	.01

Abbreviations: ACO, accountable care organization (defined as participating in the Medicare Shared Savings Program); DSH, disproportionate share hospital; IQR, interquartile range.

^aParticipants and nonparticipants are characterized at the hospital level. Dropouts are characterized at the hospital-condition pair level because hospitals could drop out for some or all of their selected conditions. Total margins are calculated as net income/total revenue and operating margins as (net patient revenue − total operating expenses)/net patient revenue. P values for participants vs nonparticipants are from χ² tests for categorical variables, t tests for normally distributed continuous variables, and Wilcoxon tests for highly skewed continuous variables. The P value for hospital profit status is from a global χ² test. P values for dropout regressions are clustered by hospital, and the P value for hospital profit status is from a likelihood ratio test. Regressions control for market characteristics including population, median household income, Medicare Advantage penetration, post–acute care supply, market share, and market competitiveness.

^bOdds per 1 bed change in total beds and per 1% change in annual Medicare days, total margin, or operating margin.

Eighty-eight hospitals dropped out fully, 150 dropped out partially (for ≥1 condition), and 184 continued without change. Of the hospital-condition pairs, 1387 dropped out (Figure), ranging by condition from 24% to 83%. Hospital-condition pair dropout rates were 11.4% (95% CI, 10.3%-12.6%) by 6 months, 28.4% (95% CI, 26.8%-30.1%) by 12 months, 39.9% (95% CI, 38.2%-41.7%) by 18 months, and 47.0% (95% CI, 45.0%-48.9%) by 24 months after enrollment. Dropout varied by hospital type. For example, unadjusted dropout rates were 58.3% for hospital-condition pairs from for-profit hospitals (aOR, 2.17 [95% CI, 0.58-8.19]), 42.5% for nonprofit hospitals (aOR, 1.15 [95% CI, 0.32-4.16]), and 39.1% for public hospitals (reference group, P = .02) (Table). Higher operating margins were associated with lower odds of dropout.

Figure. Dropouts of Hospital-Condition Pairs Over Time From the Bundled Payments for Care Improvement (BPCI) Initiative, 2013-2017.

The BPCI Initiative was launched in 2013. The graph presents hospital-condition pairs because hospitals could join for 1 condition in 1 quarter and another in subsequent quarters, and similarly could drop out for each condition independently.

^aEach quarter began on the first day of the month.

Discussion

Only 12% of eligible hospitals signed up for BPCI, and dropout was 47%. Differences between participants and nonparticipants, which confirm prior reports, and the high rate of dropout suggest that voluntary models may not have as much potential as hoped to improve quality and reduce costs across the diverse US health care landscape. One potential solution, Medicare’s mandatory bundling program, was criticized for imposing undue burden on clinicians, and the cardiac component of the mandatory program was recently cancelled. Other potential solutions include providing greater rewards for participation, which Medicare is pursuing under its new Quality Payment Program, or altering target cost amounts to make it easier for hospitals to achieve savings.

This study has limitations. Only hospital participants were analyzed, and patterns may differ for physician practices. Program performance was not examined because of the recent nature of the program. Why hospitals left the program is unknown and represents an important area for future research.

Patterns of participation and dropout in the BPCI program suggest that for voluntary alternative payment models to have a broad effect on quality and costs of health care, barriers to participation and strategies for retention need to be addressed.

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