The Financial Impact of an Avoided Readmission for Teaching and Safety-Net Hospitals Under Medicare’s Hospital Readmission Reduction Program

Abstract

We examined the financial incentives to avoid readmissions under Medicare’s Hospital Readmission Reduction Program (HRRP) for safety-net hospitals (SNHs) and teaching hospitals (THs) compared to other hospitals. Using Medicare’s FY2016 Hospital Compare and readmissions data for 2,465 hospitals, we tested for differential revenue gains for SNHs (n=658) relative to non-SNHs (n=1,807), and for major (n=231) and minor (n=591) THs relative to non-THs (n=1,643). We examined hospital-level factors predicting differences in revenue gains by hospital type. The revenue gains of an avoided readmission were 10–15% greater for major THs compared to non-THs ($18,047 vs. $15,478 for AMI) but no different for SNHs compared to non-SNHs. The greater revenue gains for THs were strongly positively predicted by hospitals’ poor initial readmission performance. We found little evidence that the HRRP creates disincentives for SNHs and THs to invest in readmission reduction efforts, and THs have greater returns from readmissions avoidance than non-THs.

INTRODUCTION

Medicare’s Hospital Readmissions Reduction Program (HRRP), introduced under the Affordable Care Act and implemented in Fiscal Year (FY) 2013, penalizes hospitals with excess 30-day unplanned readmissions for beneficiaries ages 65 and older. Each year, approximately three-quarters of eligible hospitals are penalized for having excess readmissions (KFF 2015; MedPAC 2017). Over the first several years of implementation, HRRP has been credited with reductions in readmissions rates, with declines in readmissions for HRRP targeted as well as non-targeted conditions from 21.5% to 17.8% from 2007 to 2015 (Zuckerman, Sheingold, Orav, Ruhter, and Epstein 2016).

Despite these recent reductions in hospitals readmission rates, progress in reducing readmissions has not been consistent across different types of hospitals. Safety-net hospitals (SNHs) and teaching hospitals (THs)—mostly nonprofit and teaching institutions that serve low-income high-complexity patients—continue to lag behind other hospitals in HRRP performance (Salerno et al. 2017). This is despite the fact that SNHs and THs are also more likely to be penalized for poor readmissions performance, and, when penalized, are more likely to experience larger penalties than other hospitals (Joynt and Jha 2013; Marks, Loehrer, and McCarthy 2013). In 2017, 92% of major teaching hospitals and 89% of hospitals serving large proportions of low-income patients, compared to 80% of all hospitals, will have reduced payments under HRRP (MedPAC 2017). There is disagreement in the literature regarding the extent to which these penalty differences across hospital types reflect unmeasured patient characteristics or hospital quality issues (Krumholz et al. 2017). Although the HRRP program has some special considerations for SNHs and THs (e.g., protection of supplemental Medicare payments), critics of the HRRP program argue that higher readmission penalties for SNHs and THs might be exacerbating financial challenges these hospitals already face and discourage them from making investments in readmission reduction efforts (Axon and Williams 2011; Joynt and Jha 2013; van Walraven, Bennett, Jennings, Austin, and Forster 2011; van Walraven et al. 2011; Berenson, Paulus, and Kalman 2012; Ash et al. 2011; MedPAC 2013). The potential differential impacts of the HRRP penalty structure on SNHs and THs have led to growing concern that, instead of nudging hospitals to improve quality of care, HRRP penalties might be worsening health disparities for low-income Medicare beneficiaries with complex care needs (National Quality Forum 2014; Joynt and Jha 2013, 2013; Axon and Williams 2011). Recent legislative updates to HRRP’s will slightly modify HRRP’s methodology beginning in FY 2019 (P.L. 114–255 2016), which may reduce concerns that HRRP will worsen health disparities. However, this change may reduce the incentive to avoid a marginal hospital readmission for SNHs and THs even while it reduces their total penalty amounts.

When it comes to a hospital’s decision to invest (or not) in readmission reduction efforts, however, the penalty amount alone does not tell the full story. When considering readmission avoidance, hospital decision-makers do not exclusively consider the size of the penalty, but also how much of the penalty amount can be avoided by a gradual reduction in readmissions (Norton, Li, Das, and Chen 2017). Hospital decision-makers may also consider the amount of lost revenue associated with an avoided readmission. Although preventing a repeat hospitalization saves direct care costs, the loss of reimbursement revenue leaves overhead costs for that hospitalization not covered which may mitigate hospital incentives to reduce readmissions. The larger the expected financial gain—accounting for both revenue gained and lost—following an incremental reduction in readmissions, the more incentive a hospital has to invest in readmission reduction efforts.

However, the financial impact of an avoided readmission—and therefore the incentive to avoid a readmission—may not be the same for all hospitals. Incentives may be larger or smaller due to hospital-specific factors used to adjust estimates of excess readmissions under HRRP—Medicare discharge volume, readmission rates, and patient medical severity. Because these characteristics can vary across SNHs, THs, and other hospitals (Joynt and Jha 2011, 2013; Joynt, Orav, and Jha 2011; MedPAC 2015), certain hospital types may enjoy greater financial benefit from a prevented readmission. (See Appendix A for additional details.) The differential contributions of hospital-specific factors to the overall financial impact are unknown.

In this study, we explore these quality incentives under HRRP. Our first aim is to estimate the extent to which the incentive varies for SNHs and THs compared to other hospitals, and our second aim is to identify the hospital-specific factors that contribute to different incentives across hospitals.

Background

HRRP defines excess readmissions as the difference between a hospital’s predicted and expected unplanned 30-day readmissions among Medicare beneficiaries aged 65 and older during a lagged 3-year period (e.g., 2011–2014 for FY 2016) (CMS 2016; Tilson and Hoffman 2012). Hospitals with a ratio of predicted to expected readmissions—or excess readmissions ratio (EXRR)—that is greater than one receive a penalty adjustment that is applied to the hospital’s DRG payment (excluding supplemental Medicare payments) for every Medicare discharge in a fiscal year. Therefore, small differences of even one or two percentage points in the EXRR may translate into large annual payment reductions (of up to one to two percent of total Medicare reimbursement), where the maximum penalty is up to three percent of total Medicare reimbursement. See Appendix A for additional details.

New Contribution

There is great concern that HRRP will discourage SNH and TH hospitals from engaging in prevention, potentially harming patient care in SNHs and THs (Joynt and Jha 2013; Axon and Williams 2011; National Quality Forum 2014). This is because successfully reducing readmissions can be more difficult for these hospitals due to their more complex patient mix and service to economically-disadvantaged communities, and in particular for hospitals with more limited financial resources (Axon and Williams 2011; Chollett, Barrett, and Lake 2011). Our work offers a tool for policymakers to determine the extent to which the revenue gains of one avoided readmission may differentially, or disproportionally, impact SNHs and THs, and also how hospital characteristics predict these revenue gains—clarifying the policy’s poorly understood incentive structure, with lessons for future policy modification.

METHODS

Design

The study is a secondary data analysis of the FY 2016 readmissions performance data from the FY 2016 Medicare Hospital Compare database (https://data.medicare.gov/Hospital-Compare/Hospital-Readmission-Rates/92ps-fthr/data) linked with FY 2016 data from the Medicare Impact File (comprising data from the March 2015 update of the FY 2014 MedPAR file, the March 2015 update of the Provider Specific File, and FY 2012/FY2013 cost report data) (https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/AcuteInpatientPPS/FY2016-IPPS-Final-Rule-Home-Page-Items/FY2016-IPPS-Final-Rule-Data-Files.html), and the incremental revenue gains data from Yakusheva et al. (2018) (Yakusheva and Hoffman 2018), which was obtained using a mathematical derivation of the HRRP formula. We use the mathematical derivation, and then plug empirical data from the Medicare files into that derived formula, in order to obtain estimates of the comparative incentives across hospital types, and of the relative contributions of various hospital factors to any differences in those incentives across hospital types.

Sample

From the 2,941 hospitals in the FY 2016 Hospital Compare data file, we excluded hospitals without excess readmissions (EXRR<=1) for any of the applicable conditions. Using the remaining 2,465 hospitals with excess readmissions (EXRR>1) for at least one of the applicable conditions, we created five-condition specific hospital samples – acute myocardial infarction (AMI), heart failure (HF), chronic obstructive pulmonary disease (COPD), pneumonia (PN), and total hip/knee arthroplasty (THA/TKA) – each including hospitals with excess readmissions (EXRR>1) for the specific applicable condition. The final analysis samples are: 937 hospitals with excess AMI readmissions, 1,409 hospitals with excess HF readmissions, 1,356 hospitals with excess COPD readmissions, 1,409 hospitals with excess PN readmissions, and 718 with excess THA/TKA readmissions. Most hospitals are included in more than one condition-specific sample.

Measures

Revenue Gains from Avoiding One Readmission.

The revenue gains from one avoided readmission were calculated per one avoided readmission as the difference between: (a) the amount of annual HRRP penalty saved per an avoided readmission and (b) Medicare reimbursement revenue lost per an avoided readmission. The amount of penalty saved variable represents an estimated dollar amount of the financial gain (or loss) to the hospital from avoiding one readmission for each of the five applicable conditions, based on each hospital’s own HRRP performance data in FY 2016. This amount was derived by Yakusheva et al. (2018) (Yakusheva and Hoffman 2018) using analytical differential calculus methods and include the dollar amount of the HRRP penalty saved from one avoided readmission. The amount of Medicare reimbursement revenue lost is reimbursement forfeited by avoiding the readmission minus costs of resources that would have been utilized caring for the readmitted patient—or base operating Medicare DRG payment plus supplemental Medicare payments minus estimated patient care cost. Supplemental Medicare payments include adjustments for performance in Medicare’s Value-Based Purchasing program, the empirical Operating Disproportionate Share Hospital (DSH) adjustment as well as uncompensated care payments, IME payments, and payments to low-volume hospitals (CMS 2016). We estimated the costs of resources that would have been utilized in patient care (or the patient revenue markup) for an avoided readmission as follows: the Medicare patient care costs equaled 95% of the Medicare payment amount, and 20% of these costs accounted for the hospital’s overhead (i.e., fixed) costs incurred regardless of the number of discharges (administrative overhead, utilities, depreciation, etc.) (MedPAC 2014, 2015). The data also account for the HRRP penalty floor ($0 minimum if EXRR<=1) and ceiling (3% maximum if EXRR>>1) effects, and are adjusted for direct care cost savings from readmission avoidance (Yakusheva and Hoffman 2018).

SNHs and THs.

SNHs were identified as hospitals in the top 25^th percentile of the FY 2016 Operating DSH adjustment (Joynt and Jha 2013). Teaching status was evaluated based on Medicare’s measure of hospital teaching intensity and its resident-to-bed ratio (0 = non-TH, >0 to <0.25 = Minor TH, and 0.25 = Major TH).

Hospital characteristic (FY 2012–14 measurement period).

Number of discharges represents the total number of Medicare discharges (for applicable conditions) (CMS 2016) during the performance measurement period; the number of readmissions is the total number of unplanned 30-day readmissions (CMS 2016); patient mix is the FY 2016 hospital case mix value; the Medicare base patient revenue markup operating DRG payment is defined as the wage index-adjusted standard payment for operating costs, adjusted for the diagnosis-related group (DRG, or the relative costliness of a case), to compensate hospitals for inpatient services provided to the patient (CMS 2016); supplemental Medicare payments include DSH and IME, policy adjustments to compensate hospitals for treating a disproportionate amount of low-income patients (DSH) and higher indirect costs of patient care for hospitals training medical residents (IME) (CMS 2016).

Analysis Plan

We compute means and standard deviations of the hospital characteristics for each of the five condition-specific samples, stratified by SNH and TH status.

Aim 1.

We compute means and standard deviations of the incremental revenue gains from one avoided readmission by SNH and TH status and conduct two-sample t-tests to determine whether the differences between the means are significantly different from zero using a two-tailed α=0.05 and 80% power.

Aim 2.

We examine hospital characteristics (number of Medicare discharges, number of Medicare readmissions, Medicare patient mix, Medicare base operating DRG payment, and supplemental Medicare payments) as predictors of the observed differences in mean revenue gains from one avoided readmission by SNH and TH status. These predictors can be interpreted as factors that explain why incentives vary across SNHs/ non-SNHs and THs/ non-THs. We determine the independent contributions of each of these individual hospital characteristics to the difference in the financial impact of an avoided revenue by SNH and TH status. Hospital characteristics that have a positive independent contribution increase (or amplify) the mean revenue gains differential between SNHs/ THs and other hospitals; on the other hand, hospital characteristics that have a negative independent contribution decrease (or suppress) (MacKinnon, Krull, and Lockwood 2000), and they reduce (or mute) the differential revenue gains of readmission avoidance for SNHs/ THs relative to other hospitals. (See Appendix B for additional details.)

RESULTS

Description of Sample (Table 1 and Table S1 in Digital Supplement C)

Table 1.

Descriptive statistics, by applicable condition and SNH/ TH status.

	AMI		HF		COPD		PNA		THA/TKA
Variable	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.
	Non-SNHs
Number of Medicare discharges	245.18	191.11	381.21	355.01	336.91	261.88	333.15	238.25	477.19	387.95
Number of Medicare readmissions	46.81	34.61	92.22	85.53	75.32	59.47	63.35	45.22	28.95	20.78
Medicare patient mix	1.59	0.22	1.45	0.26	1.46	0.26	1.47	0.27	1.63	0.22
Medicare base operating DRG payment	10,611.40	864.22	6,689.89	575.36	6,165.52	509.50	6,458.47	527.44	11,606.17	837.07
Medicare supplemental payments	1,226.63	1,637.29	1,002.01	1,034.00	922.03	942.33	973.92	1,020.62	1,352.76	2,069.24
Actual readmission rate	20.55	4.48	24.54	2.87	22.34	2.48	19.12	2.46	6.55	1.62
EXRR	1.06	0.05	1.06	0.05	1.06	0.05	1.06	0.04	1.14	0.12
Revenue gains per readmission	16,766.14	7,768.48	9,174.81	3,681.16	8,397.87	3,938.27	10,343.17	4,773.59	58,597.30	26,234.51
Number of observations	597		964		949		982		560
	SNHs
Number of Medicare discharges	242.53	223.45	383.56	303.93	309.62	218.24	290.24	206.58	404.46	331.54
Number of Medicare readmissions	48.46	40.45	98.04	76.92	72.12	51.62	58.56	41.01	25.51	17.15
Medicare patient mix	1.68	0.26	1.58	0.28	1.60	0.28	1.61	0.29	1.72	0.24
Medicare base operating DRG payment	10,924.93	1,064.46	7.042.10	725.12	6,484.28	679.48	6,766.48	690.93	11,872.80	1,171.37
Medicare supplemental payments	3,501.46	2,551.26	2,963.73	2,369.06	2,739.83	2,048.64	2,785.27	2,085.23	3,262.62	2,314.92
Actual readmission rate	21.56	4.21	26.00	3.45	23.54	2.89	20.60	3.22	7.27	2.69
EXRR	1.07	0.05	1.08	0.06	1.06	0.05	1.06	0.05	1.17	0.13
Revenue gains per readmission	16,727.19	7,224.25	9,563.66	3,374.13	8,811.11	3,725.61	10,235.68	4,607.57	57,009.06	25,469.51
Number of observations	340		445		407		427		158
	Non-THs
Number of Medicare discharges	185.39	144.36	298.53	259.23	286.04	224.48	281.06	194.94	403.37	322.33
Number of Medicare readmissions	36.34	25.47	73.39	63.35	63.91	50.76	53.85	37.20	24.53	16.45
Medicare patient mix	1.53	0.18	1.40	0.23	1.40	0.24	1.41	0.24	1.60	0.22
Medicare base operating DRG payment	10,647.40	985.07	6,698.42	631.32	6,170.64	569.99	6,466.53	590.33	11,585.86	960.05
Medicare supplemental payments	1,159.30	1,633.37	1,094.52	1,143.46	1,008.77	1,050.46	1,037.70	1,081.70	1,105.90	2,040.51
Actual readmission rate	21.31	4.83	24.91	3.21	22.36	2.55	19.29	2.63	6.67	1.96
EXRR	1.06	0.05	1.06	1.35	1.05	0.05	1.06	0.04	1.14	0.11
Revenue gains per readmission	15,554.87	7,999.69	8,966.51	3,704.36	8,242.55	3,945.32	9,974.68	4,848.19	57,388.04	26,537.60
Number of observations	481		919		844		895		385
	Minor THs
Number of Medicare discharges	305.95	238.74	531.31	382.40	431.35	276.89	428.81	264.43	530.28	419.46
Number of Medicare readmissions	57.85	42.44	128.80	92.37	97.45	63.64	82.26	49.70	31.78	22.08
Medicare patient mix	1.65	0.21	1.59	0.21	1.61	0.21	1.63	0.22	1.65	0.18
Medicare base operating DRG payment	10,654.44	849.77	6,887.70	619.39	6,311.48	536.63	6,574.81	527.19	11,637.29	808.04
Medicare supplemental payments	1,834.08	1,600.84	1,612.12	1,407.16	1,423.98	1,206.78	1,421.31	1,254.64	1,791.45	1,958.18
Actual readmission rate	20.17	3.93	24.60	2.69	22.78	2.72	19.58	2.90	6.56	1.77
EXRR	1.06	0.05	1.06	0.05	1.06	0.05	1.06	0.05	1.14	0.12
Revenue gains per readmission	17,942.87	7,071.66	9,752.10	3,574.59	9,091.01	3,766,95	11,300.47	4,296.08	58,096.08	26,261.63
Number of observations	291		303		346		321		239
	Major THs
Number of Medicare discharges	306.82	230.89	549.94	452.58	331.81	251.67	320.68	259.28	522.34	433.68
Number of Medicare readmissions	61.26	43.70	139.28	110.70	79.35	59.63	65.37	51.46	34.06	24.87
Medicare patient mix	1.83	0.29	1.76	0.31	1.76	0.31	1.80	0.32	1.84	0.26
Medicare base operating DRG payment	11,076.62	958.32	7,165.68	619.80	6,616.83	605.27	6,909.04	615.38	12,058.36	980.42
Medicare supplemental payments	5,039.10	2,508.66	4,227.09	2,714.07	3,891.70	2,336.16	3,941.55	2,235.02	4,458.64	1,864.97
Actual readmission rate	21.07	3.74	26.10	3.22	24.24	2.58	20.86	3.00	7.20	2.12
EXRR	1.08	0.05	1.09	0.06	1.06	0.05	1.07	0.05	1.18	0.14
Revenue gains per readmission	18,141.59	6,542.62	10,188.42	2,726.61	8,756.02	3,637.46	10,221.95	4,601.33	62,154.88	23,323.47
Number of observations	165		187		166		193		94

Note: SNH = safety-net hospital, TH = teaching hospital, AMI = acute myocardial infarction, HF = heart failure, COPD = chronic obstructive pulmonary disease, PN = pneumonia, THA/TKA = total hip/knee arthroplasty, payment=payment, DRG = diagnosis-related group. Number of Medicare discharges and number of Medicare readmissions are for the specific applicable condition listed in each column.

The sample included 737 SNHs, 698 minor THs, and 282 major THs. Of the 737 SNHs, 224 (30%) were minor THs and 175 (24%) were major THs. Across the condition-specific samples, approximately 30–47% of the hospitals are SNHs and the proportion of minor THs and major THs are approximately 22–33% and 13–15% respectively. Compared to non-SNHs, SNHs tend to have slightly lower discharge volumes, a more complex patient mix, and higher excess readmission ratios (generally one or two percentage points higher). Compared to non-THs, THs tend to have larger discharge volumes, more complex patients, and higher excess readmission ratios (also generally one or two percentage points higher).

Aim 1 (Table 2)

Table 2.

Revenue gains per an avoided readmission, by condition and SNH/ TH status.

	AMI	HF	COPD	PNA	HTA/TKA
	Mean (P-val)	Mean (P-val)	Mean (P-val)	Mean (P-val)	Mean (P-val)
	By SNH status
Non-SNH	16,766.14	9,174.81	8,397.87	10,343.17	58,597.30
	(<0.01)**	(<0.01)**	(<0.01)**	(<0.01)**	(<0.01)**
SNH	16,727.19	9,563.66	8,811.11	10,235.68	57,009.06
	(<0.01)**	(<0.01)**	(<0.01)**	(<0.01)**	(<0.01)**
Difference	−38.95	388.85	413.24	−107.49	−1,588.24
	(0.96)	(0.06)	(0.07)	(0.70)	(0.50)
	By TH status
Non-TH	15,554.87	8,966.51	8,242.55	9,974.68	57,388.04
	(<0.01)**	(<0.01)**	(<0.01)**	(<0.01)**	(<0.01)**
Minor TH	17,942.87	9,752.10	9,091.01	11,300.47	58,096.08
	(<0.01)**	(<0.01)**	(<0.01)**	(<0.01)**	(<0.01)**
Difference	2,388.00	785.59	848.46	1,325.79	708.04
	(<0.01)**	(<0.01)**	(<0.01)**	(<0.01)**	(0.74)
Major TH	18,141.59	10,188.42	8,756.02	10,221.95	62,154.88
	(<0.01)**	(<0.01)**	(<0.01)**	(<0.01)**	(<0.01)**
Difference	2,586.72	1,221.91	513.47	247.27	4,766.84
	(<0.01)**	(<0.01)**	(0.12)	(0.51)	(0.11)

^*= P-val <0.05

^**= P-val <0.01

Note: SNH = safety-net hospital, TH = teaching hospital, AMI = acute myocardial infarction, HF = heart failure, COPD = chronic obstructive pulmonary disease, PN = pneumonia, THA/TKA = total hip/knee arthroplasty.

Revenue gains from readmission avoidance are larger for THs relative to non-THs. Relative to non-THs, the revenue gains from avoiding one readmission is $2,388 (p<0.01) larger for minor THs and $2,587 (p<0.01) larger for major THs in the AMI sample; $786 (p<0.01) larger for minor THs and $1,222 (p<0.01) larger for major THs in the HF sample; $848 (p<0.01) larger for minor THs in the COPD sample; and $1,326 (p<0.01) larger for minor THs in the PN sample. Revenue gains differentials by SNH status are small and non-significant.

Aim 2 (Figures S3A–S5B)

Examination of the role of hospital characteristics on the revenue gains differential between SNHs/ THs and other hospitals reveals patterns of hypothesized amplification due to readmissions performance, and offsetting suppression due to supplemental Medicare payments (see Digital Supplement C, Figures S3A.−3.E, S4.A–4.E and S5.A–5.E). Supplemental Medicare payments are a significant suppressor of the revenue gains differential in several models (Figures S3.B, S3.D, S5.B, S5.D) (illustrated by the leftward-extending bars for supplemental payments, indicating an increase in the size of the negative independent contribution of supplemental payments).

Hypothesized hospital-specific factors for the revenue gains (amounts and directions of the independent contributions of the factors) were similar in the analysis with SNHs—meaning that the positive and negative independent contributions of hospital characteristics canceled one another out, resulting in no overall differences in the estimated revenue gains between SNHs and non-SNHs.

DISCUSSION

In this analysis, we estimated the revenue gains from avoiding one readmission across for SNHs and THs, compared to other hospitals, under Medicare’s HRRP program. To our knowledge, this is the first comparison of financial incentives for incremental readmission reduction across hospital types. We found differential revenue gains across hospital types and then specific hospital factors that are predictors of the revenue gains, with implications for policymakers. First, we found little evidence that the HRRP creates disincentives for either SNHs or THs to invest in readmission reduction efforts. On the contrary, we found that THs have greater financial returns from incremental readmissions avoidance than non-THs. We observed impacts that were 10–15% greater for major THs compared to non-THs and no difference in impacts by SNH status. That is, the financial benefit from an avoided readmission—given current relative readmission performance and supplemental Medicare payments of THs—is considerably higher than for non-THs. Accounting for these and other factors, the financial benefit for SNHs is not notably different relative to non-SNHs, however. Second, we also found that poor readmission performance amplifies and supplemental Medicare payments mute hospital incentives to invest in readmission prevention efforts. These findings have implications for decisions to invest in prevention by SNHs and THs and their low-income patient populations, and therefore for future changes (or potential changes) to HRRP methodology.

Overall revenue gains from avoiding one readmission and incentives to avoid readmissions across hospital types

Our first finding indicates that financial incentives for THs and SNHs to invest in readmission prevention efforts are as strong or stronger than those of other hospitals. Critics have expressed concern that disproportionate penalties for SNHs and THs under HRRP will create further inequity in access to and quality of care. Specifically, the concern was that, perversely, the HRRP penalties would impede the ability of SNHs and THs to invest in high-intensity discharge planning programs to reduce readmissions, further exacerbating disparities in care for disadvantaged patients in hospitals typically caring for them (Joynt and Jha 2013; Axon and Williams 2011; National Quality Forum 2014). Critics have also suggested that HRRP incentives are generally too small to encourage hospitals to avoid readmission (Berenson, Paulus, and Kalman 2012; Carey and Lin 2016) and suggested that the quality of care for the patient populations SNHs and THs treat—low-income patients—could be affected if HRRP continues to disproportionately penalize those hospitals (Joynt and Jha 2013; Singh, Lin, Kuo, Nattinger, and Goodwin 2014; Barnett, Hsu, and McWilliams 2015; Herrin et al. 2015). However, this study suggests that HRRP financial incentives for hospitals are sizeable for all hospitals, and particularly so for THs given current performance levels among those hospitals. Our results are also consistent with recent trends in national readmissions, as SNHs and non-SNHs (with their non-differential financial incentives) have had similarly improving performances (Salerno et al. 2017). Moreover, overall readmission rates have decreased by an average of 0.35 percentage points per year following the introduction of HRRP in 2013, suggesting that providers are responding to the policy (MedPAC 2018).

On average, SNHs and especially THs have strong incentives to avoid readmissions. While this study does not address return on investment for readmission prevention—or, the marginal benefit of an avoided readmission net of the costs of readmission prevention—Yakusheva et al. (2018) (Yakusheva and Hoffman 2018) establish a strong case for investment in prevention, indicating generally positive net returns on investment in readmission prevention (the revenue gains of an avoided readmission net of prevention program costs) for most hospitals. With conservative assumptions regarding program costs of $130-$325 per patient and program effectiveness (of 5–10% reductions in readmissions), they estimate net earnings from investment in readmission prevention program of $25,000-$174,000 for AMI patients and $88,000-$400,000 for knee and hip surgery patients over a three-year period following the year in which the readmission reduction occurred. When stratified by SNH and THs status, they also found that SNHs had similar net earnings to non-SNHs ($30,000-$70,000 higher aggregate net earnings from a year-long implementation), while THs had somewhat higher expected net earnings relative to non-THs (approximately $100,000–300,000 higher for minor THs and approximately $200,000-$400,000 higher for major teaching hospitals).

The business case may be weakened by other factors, however. To the extent that providers are not responding to the extent that might be expected, given the large observed financial incentives, several explanations are possible. First, Quality Improvements (QI) efforts have variable results (Mitchell, Weigel, Laurens, Martin, and Jack 2017) that depend on the skills of QI staff (Leppin et al. 2014; Coleman and Williams 2007). Under-resourced hospitals, such as SNHs and THs, may be disadvantaged compared to other hospitals in readmission QI. A recent national analysis of transitional care implementation in U.S. hospitals found heterogeneity in care transitions efforts, including sites with either no formal transitional care approaches or ‘patchwork’ approaches with duplication in staff efforts resulting gaps in care; a lack of engagement with community partners (skilled nursing facilities, home health agencies, health departments, senior centers) was also commonly observed (Scott et al. 2017). If hospitals perceive investment in QI efforts as unable to reliably prevent readmissions, their engagement with such efforts may be restricted. HRRP’s penalty criteria may also counteract hospitals’ financial incentives, because all readmissions—and not just potentially preventable readmissions (PPR) that may be less common for complex patient populations treated by SNHs and THs—are penalized. PPRs have been consistently declining since HRRP implementation (MedPAC 2017, 2018). The technical complexity of HRRP might also serve as a disincentive (Norton, Li, Das, and Chen 2017), particularly for hospitals that lack the resources to disentangle the program’s technical components. The complex present exercise deriving the statutory formula suggests that providers may be unaware of exact hospital-specific incentive amounts. Finally, although readmission rates continue to decline over time (MedPAC 2018), hospitals may be reaching a readmissions “floor” consistent with medical practice and patient characteristics. Absent new initiatives, such as readmission penalties for post-acute providers or greater attention to social determinants of health and medical utilization, providers may not further shift readmission rates.

Additional changes that may increase HRRP incentives for SNHs and THs

Our second finding provides ideas for strengthening the incentives for quality improvement. We observed two primary independent predictors of the difference in the revenue gains for SNHs and THs, relative to other hospital types. Beyond lessons for policymakers from our first finding on relative revenue gains across hospital types, this finding sheds light on how specific characteristics of hospitals influence the full revenue gains of an avoided readmission. First, findings from our second aim show that poor readmission performance serves as a significant positive predictor, amplifying incentives to reduce readmissions SNH and TH hospitals relative to other hospitals. This is likely due to the fact that under the current “no carrot, all stick” approach, a better performing hospital has a reduced incentive to prevent readmission as its performance improves, in essence creating separate performance corridors with varying levels of financial incentive. For hospitals that are performing poorly, such as many SNHs and THs at present, the likelihood of hitting the “penalty floor” is low—therefore, hospitals in this poor-performance corridor generally receive full financial credit for avoiding a readmission. However, hospitals in the high-performing corridor face a much higher likelihood of reducing readmissions to or below the penalty floor and, consequently, having part or all of the costly readmission reduction effort receive no HRRP penalty credit to offset the costs. Without a “carrot,” those hospitals have limited incentives to avoid a readmission. At present, the penalty floor effect is relatively beneficial for SNHs and THs, due to their poor initial performance under HRRP. As their performance improves (or as other hospitals’ performance worsens) relative to the national mean performance, incentives for SNHs and THs to improve care will diminish.

A new policy change will also result in changes to the incentives of SNHs and THs, as they become more subject to the penalty floor. Under the 21^st Century Cures Act (P.L. 114–255 2016), beginning in FY 2019 CMS will be required to compare the performance of hospitals to hospitals with similar proportions of dually eligible patient populations. Because dual eligibility appears correlated with readmission performance, this could create lower relative performance standards for SNHs and THs compared to other hospitals—in turn resulting in more expected readmissions and lower excess readmissions, and therefore greater likelihood of hitting the penalty floor. To calibrate HRRP standards so that incentives are more similar across performance corridors—which in turn would address the change in incentives that could occur in FY 2019 for SNHs and THs—policymakers may wish to increase the incentives for incremental readmission reduction through a “carrot and stick” approach, with rewards for better performance. This approach might also mitigate concerns that, as national readmission performance improves, (1) it may become more difficult for hospitals to surpass the national average performance level and (2) penalty sizes will increase due to HRRP’s implicit “penalty multiplier” (a characteristic of the HRRP methodology observed by MedPAC, whereby penalty amounts are inversely related to the national average adjusted readmission rate, which means that as national readmission rates decrease HRRP penalty amounts increase) (MedPAC 2013). Redistributing penalties collected from below-average performers as bonuses to above-average performers is currently done under Medicare’s Shared Savings Program and Value-Based Purchasing Program.

A second hospital characteristic that explains differing financial incentives across hospital types involves Medicare supplemental payments, which suppress incentives for SNHs and THs. Our analysis shows that supplemental Medicare payments serve as a suppressor, i.e., relative to other hospitals, they mute the additional incentive for SNHs and THs to reduce readmissions. Therefore, the business case for these hospitals might also be specifically strengthened by introducing a sharper “stick” into HRRP, by applying the HRRP penalty to supplemental Medicare payments. Broadening the HRRP penalty to include the supplemental Medicare payments would increase the HRRP financial incentives for SNHs and THs and potentially improve care for low-income patients seen at those hospitals. This approach is currently used in another Medicare pay-for-performance program, the Hospital Acquired Conditions program (CMS 2016). However, this approach might be controversial as it may raise concerns about the adequacy of funding that can affect quality of care for low-income patients treated at SNHs and THs. DSH payments are intended to compensate hospitals for higher estimated relative patient costs for DSH compared to other hospitals. IME payment is considered to reflect inefficiencies associated with training of medical residents (Dalton, Norton, and Kilpatrick 2001). These adjustments have garnered attention for being greater than is empirically justified (MedPAC 2007, 2014; Nguyen and Sheingold 2011; Dalton, Norton, and Kilpatrick 2001; Dobson et al. 2013), but concerns remain that changes to the payments could be harmful to SNHs and THs (Coughlin, Holahan, Caswell, and McGrath 2014). However, the implementation of additional penalties for these hospitals could be considered appropriate if estimated differences in readmission performance are due to hospital quality, rather than unmeasured patient factors (Krumholz et al. 2017).

Our study had several limitations. First, our estimates may not apply for hospitals with larger reductions in readmissions, because we used a derivation of the HRRP formula to estimate the effect of an incremental change in readmissions (Yakusheva and Hoffman 2018). The differences in the revenue gains of an incremental reduction across SNHs, THs, and other hospitals are meant to serve only as guides for hospitals and policymakers. Second, the assumptions we used to estimate lost patient revenue markup may underestimate the size of the financial incentive for prevention for SNHs and THs. If direct patient care costs for SNHs and THs are <80% (less than for other hospitals), we will have underestimated the lost marginal patient revenue for an incremental readmission reduction—resulting in smaller suppression effects, and a larger overall incentive. Although the 80% figure comes from MedPAC analyses discharges (MedPAC 2015), this percentage might be considered low. In our companion manuscript (Yakusheva and Hoffman 2018), we conducted a sensitivity analysis using a higher percentage of (up to 50%) from fixed costs, but did not observed any differences in our results. Lost patient revenue markup represents a small proportion of the total revenue gains of an incremental readmission reduction, thereby reducing the impact of these design decisions. Accordingly, our results are conservative given the assumptions we used. Finally, we do not have intervention cost estimates specific to SNHs and teaching hospitals, meaning that direct program costs may be more difficult to absorb for under-resourced hospitals, like SNHs and THs, potentially weakening their business case for investing in readmissions.

Conclusion

Although they are penalized more often, SNHs and THs have similar or greater financial incentives than other hospitals to invest in readmission prevention under HRRP. THs have greater financial returns from incremental readmissions avoidance than non-THs. Overall financial returns for SNHs and THs are substantial. While poor readmission performance justly amplifies incentives to reduce readmissions for poorly performing hospitals, supplemental Medicare payments offset these incentives, potentially contributing to deepening healthcare disparities. As SNHs and THs improve their performance, their incentive to avoid readmissions will decrease. Policymakers may wish to examine HRRP methodology to ensure that incentives are maximized across performance levels to ensure SNH and TH participation in readmission prevention efforts.