J A C C : H E A R T F A I L U R E V O L . 6 , N O . 5 , 2 0 1 8

ª 2 0 1 8 B Y T H E AM E R I C A N C O L L E G E O F C A R D I O L O G Y F O UN DA T I O N

P U B L I S H E D B Y E L S E V I E R

MINI-FOCUS ISSUE: PAYER ISSUES AND PATIENT ACCESS

CLINICAL RESEARCH

30-Day Episode Payments andHeart Failure Outcomes AmongMedicare Beneficiaries

Rishi K. Wadhera, MD, MPHIL,a,b Karen E. Joynt Maddox, MD, MPH,c Yun Wang, PHD,b,d Changyu Shen, PHD,b

Robert W. Yeh, MD, MSCb

ABSTRACT

ISS

Fro

Su

Me

of

Ins

the

Pro

fro

an

Na

Ca

Ma

OBJECTIVES The purpose of this study was to examine the association of 30-day payments for an episode of

heart failure (HF) care at the hospital level with patient outcomes.

BACKGROUND There is increased focus among policymakers on improving value for HF care, given its rising

prevalence and associated financial burden in the United States; however, little is known about the relationship between

payments and mortality for a 30-day episode of HF care.

METHODS UsingMedicare claimsdata for all fee-for-servicebeneficiarieshospitalized forHFbetweenJuly 1, 2011, andJune30,

2014,we examined the associationbetween30-dayMedicare payments at the hospital level (beginningwith a hospital admission

for HF and across multiple settings following discharge) and patient 30-day mortality using mixed-effect logistic

regression models.

RESULTS We included 1,343,792 patients hospitalized for HF across 2,948 hospitals. Mean hospital-level 30-day

Medicare payments per beneficiary were $15,423 � $1,523. Overall observed mortality in the cohort was 11.3%. Higher

hospital-level 30-day payments were associated with lower patient mortality after adjustment for patient characteristics

(odds ratio per $1,000 increase in payments: 0.961; 95% confidence interval [CI]: 0.954 to 0.967). This relationship was

slightly attenuated after accounting for hospital characteristics and HF volume, but remained significant (odds ratio per

$1,000 increase: 0.968; 95% CI: 0.962 to 0.975). Additional adjustment for potential mediating factors, including

cardiac service capability and post-acute service use, did not significantly affect the relationship.

CONCLUSIONS Higher hospital-level 30-day episode payments were associated with lower patient mortality following

a hospitalization for HF. This has implications for policies that incentivize reduction in payments without considering

value. Further investigation is needed to understand the mechanisms that underlie this relationship.

(J Am Coll Cardiol HF 2018;6:379–87) © 2018 by the American College of Cardiology Foundation.

N 2213-1779/$36.00 https://doi.org/10.1016/j.jchf.2017.11.010

m the aBrigham andWomen’s Hospital Heart & Vascular Center, Harvard Medical School, Boston, Massachusetts; bRichard and

san Smith Center for Outcomes Research in Cardiology, Division of Cardiology, Beth Israel Deaconess Medical and Harvard

dical School, Boston, Massachusetts; cWashington University School of Medicine, Saint Louis, Missouri; and the dDepartment

Biostatistics, Harvard T.H. Chan School of Public Health, Boston, Massachusetts. Dr. Wadhera is supported by National

titutes of Health Training Grant T32HL007604-32, Brigham andWomen’s Hospital, Division of Cardiovascular Medicine, and by

Jerome H. Grossman, MD, Fellowship in Healthcare Delivery Policy at the Harvard Kennedy School’s Healthcare Policy

gram; and he previously served as a consultant for Sanofi and Regeneron. Dr. Joynt Maddox has received research support

m the National Heart, Lung, and Blood Institute (K23HL109177-03); and is a former employee of the U.S. Department of Health

d Human Services, where she continues work on a limited basis as a contractor. Dr. Yeh has received research support from the

tional Heart, Lung and Blood Institute (R01HL136708) and the Richard A. and Susan F. Smith Center for Outcomes Research in

rdiology. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.

nuscript received October 3, 2017; revised manuscript received November 22, 2017, accepted November 26, 2017.

ABBR EV I A T I ON S

AND ACRONYMS

CI = confidence interval

CMS = Centers for Medicare

and Medicaid Services

HF = heart failure

HVBP = Hospital Value-Based

Purchasing Program

PCI = percutaneous coronary

intervention

SNF = skilled nursing facility

Wadhera et al. J A C C : H E A R T F A I L U R E V O L . 6 , N O . 5 , 2 0 1 8

Payments and Heart Failure Outcomes M A Y 2 0 1 8 : 3 7 9 – 8 7

380

T here is increased focus among policy-makers on improving value for heartfailure (HF) care, given the high and

rising prevalence of HF and its associatedfinancial burden in the United States (1). HFis the most common cause of hospitalizationamong Medicare beneficiaries and accountsfor a large portion of total Medicare expendi-tures (2,3); however, mortality rates remainhigh, and improvements have slowed inrecent years (4). To incent ongoing progressin both costs and outcomes of care, the Cen-

ters for Medicare and Medicaid Services (CMS),through themandatory Hospital Value-based Purchas-ing (HVBP) program, financially rewards or penalizeshospitals based on performance on both mortalityand paymentmeasures for HF aswell as other commonconditions (5).

SEE PAGE 398

As value-based and alternative payment modelsincreasingly spur hospitals to both reduce spendingand improve outcomes for HF, it is important to un-derstand how payments for an episode of HF care arerelated to clinical outcomes. Prior studies havedemonstrated significant variation in payments asso-ciated with HF hospitalizations in the United States(6–8); however, little is known about whether higherpayments are associated with better outcomes, whichwould at least potentially indicate high value, orwhether higher payments are associated with worseoutcomes, which would indicate poor value. Therelationship is unclear, in part, because little is knownabout how higher payments might translate intoactual care delivery. Higher payments may reflect theoveruse of low-value resources or use of unnecessarypost-acute care, which would not have a positive ef-fect on outcomes. Alternatively, high payments maybe driven by the use of resources and services thatactually improve quality of care and survival. Giventhat for hospitals participating in HVBP, episodepayment metrics together with 30-day mortalitymetrics account for one-half of a total score used byCMS to evaluate performance (and determine finan-cial rewards or penalties), understanding the rela-tionship between these measures for an episode of HFcare is increasingly important and policy-relevant (5).

Therefore, in this study, we aimed to answerseveral questions. First, how do hospitals whosepatients with HF incur high 30-day Medicare episodepayments, reflecting both hospital-based and outpa-tient care in the immediate post-discharge time frame,differ from those with low episode payments? Second,are higher hospital-level 30-day episode payments for

HF care, beginning with a hospital admission for HFand across multiple care settings following discharge,associated with lower patient-level mortality? And ifso, to what extent do differences in hospital charac-teristics, cardiac service capability, and post-acuteservice use explain the association?

METHODS

PATIENT COHORT. Eligible patients were fee-for-service Medicare beneficiaries age 65 years or olderhospitalized at an acute care hospital for HF based onprincipal discharge diagnosis codes from the Inter-national Classification of Diseases-9th Revision fromJuly 1, 2011, through June 30, 2014. Patient charac-teristics included age, sex, and comorbidities thatwere defined based on a validated administrativeclaims model used to profile hospital 30-day mortalitymeasures for HF (9). We applied inclusion criteriapreviously used by CMS for HF mortality rate mea-sures. For patients with more than 1 HF admissionacross the 3-year period, only 1 randomly selectedepisode of care was included (7,9,10).

30-DAY EPISODE PAYMENTS FOR HEART FAILURE.

Hospital-level risk-standardized payments for a30-day episode of HF care (which from this point onwill be referred to as episode payments), a measuremade publicly available by CMS on Hospital Compare,were used to characterize payments across a 30-dayperiod beginning with a hospital admission for HF(7,8). Episode payments reflect variations in hospitalpractice patterns and resource use related to HF careand are calculated by summing total payments asso-ciated with a 30-day episode of HF care for eachbeneficiary, beginning with index admission andacross multiple care settings, services, and supplies(i.e., inpatient, outpatient, skilled nursing facility[SNF], home health, physician/clinical laboratory/ambulance services, durable medical equipment). Inaddition to inpatient payments, post-acute paymentsare attributed to the hospital of index HF admissionbecause care and discharge planning during a hospi-talization influences subsequent expenditures acrosscare settings. Geographic differences and policy ad-justments (i.e., for costs of living, graduate medicaleducation, and disproportionate share hospitals) inMedicare payment rates are accounted for. Addi-tionally, episode payments are risk-adjusted for dif-ferences in patient characteristics and hospital casemix and account for clustering of patients withinhospitals. For patients transferred from 1 acute carehospital to another, total payments across 30 days areattributed to the first hospital where the patient wasadmitted (7). Hospitals in the Inpatient Prospective

J A C C : H E A R T F A I L U R E V O L . 6 , N O . 5 , 2 0 1 8 Wadhera et al.M A Y 2 0 1 8 : 3 7 9 – 8 7 Payments and Heart Failure Outcomes

381

Payment System with too few HF cases (<25) were notincluded in the analysis. Hospital payment perfor-mance is assessed over a rolling 3-year period; here,we used performance data representing July 1, 2011,to June 30, 2014, which were used in the HVBP pro-gram in fiscal year 2016 (7).

We used episode payments because they representhealth care expenditures from a payer’s (Medicare)perspective, and reflect a hospital’s style of practicein terms of the level of use of resources andservices by their patients with HF; for example, somehospitals may tend to discharge more patients withHF to institutional post-acute care than others or toprovide more post-discharge follow-up care. Wefocused on a 30-day episode of HF care because CMSpayment measures through HVBP do so, making thisthe most policy-relevant approach to assessing value.Episode payments were assessed on a hospital levelrather than patient level to examine the effect of apatient population being exposed to varying styles ofcare rather than an individual patient being exposedto more intense care, and to diminish the likelihoodof patient-level confounding (11).

HOSPITAL CHARACTERISTICS. Hospital characteris-tics were obtained from the 2013 American HospitalAssociation annual survey and included hospital size,census region, Joint Commission certification status,location (urban or rural), ownership (for-profit, pri-vate not-for-profit, and public), teaching status (largeteaching vs. others), nurse-to-patient ratio, and car-diac service capability (percutaneous coronary inter-vention [PCI] and coronary artery bypass graftsurgery). Additionally, using Medicare inpatient data,we calculated fee-for-service HF volume as well asthe proportion of patients with HF discharged to anSNF or with homecare services for each hospital.

OUTCOMES. The primary outcome was patient-level30-day mortality, defined as all-cause deaths within30 days of the date of admission.

STATISTICAL ANALYSIS. We classified 30-day HFepisode payments at the hospital-level into low-(<25th percentile), average- (25th to 75th percentile),and high-spending (>75th percentile) for descriptivepurposes, similar to previous studies (12–16). Bivar-iate analyses were performed to compare patient andhospital characteristics across episode payment cat-egories, using the Mantel-Haenszel chi-square test forcategorical variables and the Kruskal-Wallis test forcontinuous variables. To demonstrate geographicvariability, we also mapped county-specific episodepayments for HF care.

To evaluate the association between 30-dayhospital-level episode payments and 30-day

patient-level mortality, we fit a mixed model with alogit link function and random hospital intercepts tomodel mortality as a function of 30-day HF episodepayments ($1,000). We fit the model in the followingsequence: 1) unadjusted; 2) risk-adjusted for patientcharacteristics and comorbidities used by CMS toprofile hospital 30-day mortality measures for HF (9);and 3) additionally adjusted for key hospital charac-teristics (ownership, major teaching hospital, urbanvs. rural hospital, region, and nurse staffing) as wellas HF volume. This allowed us to control for indi-vidual risk factors and institutional characteristicsand make inferences at the patient level regarding theassociation of being exposed to different levels of 30-day payments for HF care and outcomes. Next, weexplored potential mediating variables related tohospital payment including cardiac service capabilityand mean proportion of patients with HF dischargedto SNF or with homecare services.

We performed several additional analyses. First,we examined the relationship between CMShospital-level inpatient payments and patient mor-tality to determine if this relationship paralleled ourprimary analysis. Second, we performed quantileregression to assess the nature of the relationshipbetween episode payments and outcomes (17). Thisallowed us to understand whether the associationbetween episode payments and patient risk ofmortality differed across quantiles of mortality.Third, we conducted an iterative nonparametricbootstrapping analysis with a replacement thatrandomly selected hospitals from the CMS episodepayment dataset with a sample size equal to totalhospitals in the dataset. We then refit the mixedmodels described above. We repeated this process1,000 times to obtain the distribution of odds ratiosfor the association between payments and out-comes. Fourth, to understand whether patients whodied required more or less expensive care (11), weassessed the relationship between patient-levelinpatient payments and 30-day patient-level mor-tality. Finally, we also evaluated the relationshipbetween hospital-level 30-day episode paymentsand 1-year mortality following the indexhospitalization.

Statistical tests were 2-sided at a significance levelof 0.05. Analyses were performed using SAS, version9.4 64-bit (SAS Institute Inc., Cary, North Carolina).

RESULTS

STUDY POPULATION. Our study sample included1,343,792 patients with HF hospitalized from July 1,2011, through June 30, 2014, across 2,948 hospitals

TABLE 1 Hospital Characteristics by 30-Day Episode Payment Categories

Low(n ¼ 735)

Average(n ¼ 1,476)

High(n ¼ 737) p Value

30-day HF spending, $ 13,732 (13,268–14,019) 15,308 (14,841–15,809) 17,207 (16,713–17,853)

Hospital size, beds 95 (50–165) 182 (99–310) 260 (165–406) <0.001

Ownership* 0.05

For-profit 139 (19.0) 261 (17.8) 198 (26.9)

Private nonprofit 409 (56.0) 1,012 (68.9) 490 (66.6)

Public 179 (24.5) 195 (13.3) 48 (6.5)

Large teaching hospital 22 (3.0) 118 (8.0) 103 (14.0) 0.005

Joint Commission accredited 519 (70.6) 1,235 (83.7) 648 (87.9) <0.001

Region† <0.001

Northeast 86 (12.2) 240 (16.5) 145 (19.7)

Midwest 218 (31.0) 247 (33.6) 542 (37.3)

South 262 (37.3) 467 (32.1) 184 (25.0)

West 137 (19.5) 205 (14.1) 137 (21.6)

Rural hospital 270 (36.9) 461 (31.4) 119 (16.2) 0.27

Nurses per 100 patients 4.5 (3.7–5.7) 4.3 (3.6–5.4) 4.2 (3.3–5.0) <0.001

Medicare patients, % 49.4 (43.3–54.8) 48.2 (42.4–54.0) 46.2 (40.4–52.7) <0.001

Medicaid patients, % 17.9 (13.4–24.4) 17.4 (13.0–22.4) 18.0 (12.5–22.3) <0.001

Cardiac services

Cardiac surgery capable 72 (9.8) 581 (39.4) 434 (58.9) <0.001

PCI capable 160 (21.8) 810 (54.9) 514 (69.7) <0.001

HF volume 50 (31–96) 121 (64–211) 178 (101–272) <0.001

Values are median (IQR) or n (%), unless otherwise indicated. *Variable missing for 13 hospitals. †Variable missing for 56 hospitals.

HF ¼ heart failure; IQR ¼ interquartile range; PCI ¼ percutaneous coronary intervention.

Wadhera et al. J A C C : H E A R T F A I L U R E V O L . 6 , N O . 5 , 2 0 1 8

Payments and Heart Failure Outcomes M A Y 2 0 1 8 : 3 7 9 – 8 7

382

with $25 eligible HF cases and episode payment in-formation. Patient demographics were generallysimilar across low, average, and high episode pay-ment hospital groups (Online Table 1). Comorbidities,including the proportion of patients with a history ofHF, coronary artery disease, prior myocardial infarc-tion, stroke, and hypertension, were also similaracross groups.

HOSPITAL-LEVEL THIRTY-DAY EPISODE PAYMENTS. The30-day episode payments for HF care varied nearly2-fold across hospitals (10). Mean 30-day episodepayments were $15,423 � $1,523 (Online Figure 1).Median 30-day episode payments per beneficiary were$13,732 in the low-payment group, $15,308 in theaverage-payment group, and $17,207 in the high-payment group (Table 1). There was evidence ofgeographic variation in 30-day episode payments forHF care across the United States (Figure 1).

HOSPITAL STRUCTURAL CHARACTERISTICS. Hospitals inthe high 30-day episode payment group were larger,tended to be teaching hospitals, were morefrequently Joint Commission accredited, were lesslikely to be public hospitals, and had lower nurse-staffing ratios compared with hospitals in theaverage- and low-payment groups (Table 1). Highepisode payment hospitals also had greater medianMedicare fee-for-service HF volume and were more

likely to provide cardiac services including cardiacsurgery and PCI.

HOSPITAL RESOURCE USE AND DISCHARGE PATTERNS.

Procedure and resource use during and immediatelyfollowing hospitalization varied across episode pay-ment groups (Table 2). Hospitals in the high 30-dayepisode payment group were more likely toperform cardiac catheterizations during initial hos-pitalization. Overall rates of revascularization werelow, but high episode payment hospitals were alsomore likely to perform PCI and coronary arterybypass grafting surgery during initial inpatient stay.Patients cared for in high episode payment hospitalshad longer mean lengths of stay and were less likelyto be transferred to another institution. They weremore likely to be discharged to an SNF or withhomecare services. In addition, observed 30-dayreadmission rates were highest at average-paymentinstitutions compared with low- and high-paymentinstitutions.

30-DAY PAYMENTS AND PATIENT MORTALITY. Observedmortality in the overall population was 11.3%. Greater30-day hospital-level episode payments were associ-ated with lower 30-day patient mortality (unadjustedodds ratio [OR]: 0.975 for every $1,000 increase; 95%confidence interval [CI]: 0.969 to 0.982; p < 0.001)(Figure 2). This relationship became more pronounced

FIGURE 1 Geographic Variation in Hospital-Level 30-Day Episode Payments for Heart Failure Care

J A C C : H E A R T F A I L U R E V O L . 6 , N O . 5 , 2 0 1 8 Wadhera et al.M A Y 2 0 1 8 : 3 7 9 – 8 7 Payments and Heart Failure Outcomes

383

following adjustment for patient characteristics(Table 3, Online Table 2) (OR: 0.961 per $1,000 in-crease; 95% CI: 0.954 to 0.967; p < 0.001). When pa-tient characteristics, hospital characteristics, and HFvolume were accounted for, higher hospital episodepayments remained associated with lower 30-daypatient mortality (OR: 0.968 per $1,000 increase;95% CI: 0.962 to 0.975; p # 0.001). Addition of vari-ables that could potentially mediate the relationshipbetween episode payments and outcomes, includingcardiac service capability and discharge to SNF orhomecare use, did not alter our findings (Table 3).ADDITIONAL ANALYSES. We found that the associ-ation between CMS hospital-level inpatient paymentsand patient mortality was similar to our main analysis(Online Table 3). We also examined whether theobserved relationship between 30-day episode pay-ments and mortality was consistent across all

quantile levels of patient 30-day mortality. We founda predominately linear relationship across the distri-bution, although the association between episodepayments and mortality was larger at higher quantilesof mortality (Online Figure 2). Our iterative boot-strapping simulation demonstrated that the distri-bution of the ORs of the association betweenpayments and outcomes remained significant(Online Figure 3). To assess whether our primaryfindings may have been influenced by the fact thatbeneficiaries that died accrued less care, and there-fore payments, patient-level payments and 30-daypatient-level mortality were also examined. At thepatient level, adjusted odds of 30-day mortality werehigher for every $1,000 increase in payments(Online Table 4), diminishing the likelihood that ourprimary findings were due to patients that diedrequiring less care and payments. This is consistent

TABLE 2 Resource Use, Procedure Use, and Discharge Patterns by 30-Day Episode

Payment Categories

Low(n ¼ 166,052)

Average(n ¼ 699,251)

High(n ¼ 478,489) p Value

Observed 30-day mortality 11.6 11.4 11.0 <0.001

Observed 30-day readmission rates 21.6 23.9 22.6 <0.001

Length of stay (days) 4.6 5.1 5.7 <0.001

Cardiac procedures (initial admission)

Cardiac catheterization 2.3 4.6 6.4 <0.001

PCI 0.3 0.7 1.2 0.002

CABG 0.04 0.18 0.32 0.05

Discharge disposition

Transferred out 3.8 2.2 1.6 <0.001

Discharged home 45.7 41.2 38.9 <0.001

Homecare on discharge 20.9 23.8 24.3 <0.001

Discharged to SNF 19.2 21.7 22.5 <0.001

Values are %, unless otherwise indicated.

CABG ¼ coronary artery bypass grafting; IQR ¼ interquartile range; PCI ¼ percutaneous coronary intervention;SNF ¼ skilled nursing facility.

FIGURE 2 Hospita

Wadhera et al. J A C C : H E A R T F A I L U R E V O L . 6 , N O . 5 , 2 0 1 8

Payments and Heart Failure Outcomes M A Y 2 0 1 8 : 3 7 9 – 8 7

384

with prior studies that have shown that care deliveredto Medicare decedents is significantly more expen-sive than survivors during a hospitalization (11).Finally, we found that higher hospital-level 30-dayepisode payments following an HF hospitalizationwere also associated with lower 1-year mortality(Online Table 5).

DISCUSSION

In this study of Medicare payments and mortalityrates among beneficiaries hospitalized for HF, wefound that higher hospital-level Medicare paymentsfor a 30-day episode of HF care, which included both

l-Level 30-Day Episode Payments and 30-Day Patient Mortality

inpatient and post-acute care payments acrossmultiple settings, were associated with lower patient-level 30-day mortality, even after accounting forpatient case mix. Overall, patients admitted for HF tohospitals with 30-day payments 1 SD ($1,523) abovethe mean, compared with hospitals 1 SD below themean, had an associated z1.3% lower mortality. Thisassociation was attenuated, but remained significantafter adjusting for key hospital characteristics and HFvolume. Variation in hospital cardiac service capa-bility and post-acute care use did not mediate theassociation between spending and outcomes. Collec-tively, our findings should not necessarily be inter-preted as causal in nature, but rather, provide insightregarding HF care in the context of CMS value-basedpayment models, particularly given the continuedevolution of HVBP and voluntary cardiac bundledpayments (18).

There has been a broad interest among policy-makers in improving the value of HF care, given thehigh and rising prevalence of HF and its associatedfinancial burden. This interest is reflected in theaddition of HF-specific spending measures, deter-mined by Medicare payments made to hospitals, tothe HVBP, and through alternate payment models(1,19). Hospital characteristics associated with higherlongitudinal payments are not accounted for in CMSpayment models, however, leaving certain in-stitutions susceptible to penalties, even if they pro-vide high-quality care. In fact, early evidence fromthe HVBP demonstrates that some hospital types thattend to have better outcomes, including majorteaching hospitals (20,21), often have high payments,as in our study, and consequently are more likely tobe penalized (22). Conversely, some low-paymentinstitutions that also provide low-quality carereceive financial rewards (22). Whether the HVBPprogram truly incents value, then, is unclear.

Because hospitals are held financially accountablefor both longitudinal spending and outcomesfollowing a hospitalization for HF, it will be importantto understand what resources and services actuallyimprove quality of care. It is possible that hospitalsreceiving higher payments use diagnostic testing,invasive procedures, specialist care, and intensivecare units more frequently, to the benefit of outcomes(23). Alternatively, factors not captured in our study,such as use of education and counseling services,interdisciplinary collaborative ambulatory care, andhome monitoring services, may result in higher pay-ments, but better care quality (24–26). In addition,early and frequent outpatient follow-up and the typeof outpatient care received following discharge (i.e.,primary care provider vs. cardiologist vs. HF clinic),

TABLE 3 Unadjusted and Multivariate Adjusted ORs of

30-Day Patient Mortality for Every $1,000 Increase in

30-Day Hospital-Level Episode Payments

Model* OR (per $1,000 increase) 95% CI

A 0.975 0.969–0.982

B 0.961 0.954–0.967

C 0.972 0.966–0.979

D 0.968 0.962–0.975

E 0.969 0.962–0.977

F 0.969 0.962–0.976

*p < 0.001 for all models.

A¼ unadjusted; B ¼ adjusted for patient characteristics; C ¼ adjusted for patientand hospital characteristics; D ¼ adjusted for patient and hospital characteristicsand HF volume; E ¼ adjusted for patient/hospital characteristics, HF volume, andcardiac service capability; F ¼ adjusted for patient/hospital characteristics,HF volume, cardiac service capability, and discharge disposition (SNF and homehealth care); CI ¼ confidence interval; HF ¼ heart failure; OR ¼ odds ratio.

J A C C : H E A R T F A I L U R E V O L . 6 , N O . 5 , 2 0 1 8 Wadhera et al.M A Y 2 0 1 8 : 3 7 9 – 8 7 Payments and Heart Failure Outcomes

385

may have also mediated the observed relationship(24). Understanding these mechanisms will beessential to ensuring that hospitals do not respond topayment programs, such as the HVBP, by reducingexpenditures in a manner that is ultimatelydetrimental to patient care, which is a potentialunintended consequence. Additionally, identifyinglow-value procedures, resources, and services thatare of little benefit to patient care will also be of equalimportance as hospitals navigate where to reduceexpenditures (27).

Our study found significant differences in institu-tional characteristics across low, average, and highepisode payment groups. Although controlling forthese characteristics in our analysis only somewhatattenuated our findings and does not necessarily ac-count for heterogeneity in care delivered withinhospitals, these patterns may shed light on potentialunderlying mechanisms. For example, high-paymenthospitals were more likely to be large, teaching in-stitutions. Previous studies have shown that admis-sion for HF to major teaching hospitals, comparedwith nonteaching hospitals, is associated with lower30-day mortality (20,21). Similarly, a greater propor-tion of high-payment hospitals were Joint Commis-sion accredited, which has been linked to betterperformance on HF quality measures (28). HF volumeand cardiac service capabilities also varied acrosshospital payment groups, which may reflect differ-ences in institutional experience and expertise incaring for patients with HF. Collectively, thesepatterns suggest that greater capacity and expertisemay result in better outcomes for patients with HF,albeit at higher costs.

We also found that beneficiaries at high-paymentinstitutions had longer average lengths of stay,

which may have resulted in patients being moreclinically optimized at time of discharge. This kind ofreasoning may also be supported by the HospitalReadmission Reduction Program; with its incentivesto reduce readmission rates (29), the program mayhave helped counteract pressures for hospitals todischarge patients ever more quickly and encouragedgreater care coordination. Notably, observed 30-dayreadmission rates were slightly higher in averageepisode payment institutions compared with low-and high-payment groups. In addition, patients athigh-payment institutions were more likely to bedischarged to SNFs and/or with home health services.Although SNF use has been linked at the patient levelwith a higher risk of death, likely because it reflectsgreater severity of illness (30), the use of home healthservices as a supplemental support mechanism hasbeen associated with lower mortality after HF hospi-talization (31,32). We found that the use of these post-acute care services did not explain the relationshipbetween higher payments and better outcomes; thus,it is not clear whether such patterns necessarilyreflect high-value HF care. This is consistent withprior work that has shown that greater high-intensitypost-acute care use in higher-spending regions ofthe United States is not associated with betterquality care (33). Because institutions have a greaterfinancial stake in post-acute care for HF throughshared-savings programs and episode-based paymentmodels, they will be incentivized to ensure appro-priate use of post-acute care services (34).

Prior investigations have focused on the associa-tion of inpatient spending with HF outcomes,although findings have been mixed. For example, astudy of Medicare beneficiaries hospitalized in 2006found that higher inpatient costs of care for HF wereassociated with lower mortality, whereas anotherfound no relationship between risk-adjusted costsand HF outcomes (12,13). Two subsequent analyses inMedicare showed that higher hospital spending in-tensity was associated with better in-hospital HFoutcomes. Both studies used an end-of-life expendi-ture index (spending on hospital and physician ser-vices provided in patients’ last years of life) to assessspending intensity. This measure is informative forcare intensity in general, but may not necessarilyreflect condition-specific treatment decisions for pa-tients not at the end of life (14,16,35,36). Our study isnovel in that it examined payments over a longitu-dinal episode that included both inpatient and post-discharge care across multiple settings, which hasgreater relevance given the movement towardepisode-based reimbursement. In addition, we eval-uated HF-specific payments, rather than hospital

PERSPECTIVES

COMPETENCY IN MEDICAL KNOWLEDGE:

Higher hospital-level 30-day payments following an

admission for HF are associated with a lower

likelihood of patient mortality.

TRANSLATIONAL OUTLOOK: Our findings high-

light the challenges institutions may face as they are

increasingly held financially responsible for both

payments and outcomes associated with an episode of

HF care through value-based and alternative payment

programs. Further research is needed to understand

the mechanisms that mediate the relationship be-

tween higher payments and better outcomes.

Wadhera et al. J A C C : H E A R T F A I L U R E V O L . 6 , N O . 5 , 2 0 1 8

Payments and Heart Failure Outcomes M A Y 2 0 1 8 : 3 7 9 – 8 7

386

costs, thus reflecting use beyond hospitals’ internalcost structure.STUDY LIMITATIONS. Because of the study’s obser-vational nature, we are unable to make inferencesabout causation. We only examined Medicare pa-tients older than age 65 years, and our findings maynot generalize to other populations. Patients mayhave varied in ways that were not captured by thecomorbidities coded in claims data; however, suchunmeasured severity would have most likely biasedour results toward the null. In addition, claims datado not capture clinical information (i.e., vital signs,ejection fraction, biomarkers, cardiac medications)that is of relevance to a HF population. We did notaccount for socioeconomic determinants of health,because payment programs such as HVBP do not,although these factors do play an important role in HFoutcomes (37). Thirty-day episode payments reflectedaggregate data at the hospital level, which may besubject to ecological fallacy, although our outcomewas evaluated at the patient level. We also lackeddetailed clinical data and a decomposition of episodepayment data with which to identify specific mecha-nisms by which higher payments improved outcomes;this represents an important area for future study.Although we defined value as the relationship be-tween payments and outcomes, we were unable tocharacterize quality of life, which is an importantconsideration when evaluating value of care. Becauseour sample size was large, our interpretation of thisanalysis focused on the estimated effect sizes for therelationship between episode payments and out-comes, rather than the extent to which p values weresignificant.

CONCLUSIONS

In a cohort of Medicare beneficiaries hospitalized forHF across a 3-year period, higher 30-day episode

payments at the hospital level for HF care wereassociated with lower patient mortality. Emergingpayment models seek to incentivize high-value HFcare, but our findings illustrate the challenges in-stitutions may face in balancing expenditures andclinical outcomes because these models hold practi-tioners and hospitals financially accountable for boththese elements of care. Further research is needed toidentify interventions that improve outcomes, bothduring and following hospitalization, to ensure thathospitals continue to invest in procedures, services,and resources that improve survival, while cuttingwasteful use.

ADDRESS FOR CORRESPONDENCE: Dr. Robert W.Yeh, Smith Center for Outcomes Research in Cardi-ology, Beth Israel Deaconess Medical Center, 185 Pil-grim Road, Boston, Massachusetts 02215. E-mail:ryeh@bidmc.harvard.edu.

RE F E RENCE S

1. Heidenreich PA, Albert NM, Allen LA, et al.Forecasting the impact of heart failure in theUnited States: a policy statement from theAmerican Heart Association. Circ Heart Fail 2013;6:606–19.

2. Fitch K, Pelizzari P, Pyenson B. The High Cost ofHeart Failure for the Medicare Population: AnActuarial Cost Analysis. Denver, CO: Milliman,2015.

3. Fitch K, Pelizzari PM, Pyenson B. Inpatient uti-lization and costs for medicare fee-for-servicebeneficiaries with heart failure. Am Health DrugBenefits 2016;9:96–104.

4. Joynt KE, Orav EJ, Zheng J, Jha AK. Publicreporting of mortality rates for hospitalized

medicare patients and trends in mortality for re-ported conditions. Ann Intern Med 2016;165:153–60.

5. Hospital Value-Based Purchasing. Centers forMedicare and Medicaid Services, 2017. Available at:https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/Value-Based-Programs/HVBP/Hospital-Value-Based-Purchasing.html. Accessed December 5, 2017.

6. Ziaeian B, Sharma PP, Yu TC, Johnson KW,Fonarow GC. Factors associated with variations inhospital expenditures for acute heart failure in theUnited States. Am Heart J 2015;169:282–9.

7. 2015 Condition-Specific Measure Updates andSpecifications Report Hospital-Level 30-Day Risk-Standardized Payment Measures. Yale New Haven

Health Services Corporation/Center for OutcomesResearch & Evaluation (YNHHSC/CORE), 2015.Available at: https://www.qualitynet.org/dcs/ContentServer?c¼Page&pagename¼QnetPublic%2FPage%2FQnetHomepage&cid¼1120143435383.Accessed December 5, 2017.

8. Payment and Value of Care - Hospital. HospitalCompare Datasets. Center for Medicare andMedicaid, 2017. Available at: https://data.medicare.gov/Hospital-Compare/Payment-and-value-of-care-Hospital/c7us-v4mf/data. AccessedDecember 5, 2017.

9. Krumholz HM, Wang Y, Mattera JA, et al. Anadministrative claims model suitable for profilinghospital performance based on 30-day mortality

J A C C : H E A R T F A I L U R E V O L . 6 , N O . 5 , 2 0 1 8 Wadhera et al.M A Y 2 0 1 8 : 3 7 9 – 8 7 Payments and Heart Failure Outcomes

387

rates among patients with heart failure. Circula-tion 2006;113:1693–701.

10. 2015 Condition-specific measures updates andspecifications report hospital-level: 30-day riskstandardized mortality measures. Yale New HavenHealth Services Corporation, Center for OutcomesResearch and Evaluation for the Centers forMedicare and Medicaid Services. Baltimore: Cen-ters for Medicare and Medicaid Services, 2015.Available at: http://www.qualitynet.org/dcs/ContentServer?c¼Page&pagename¼QnetPublic%2FPage%2FQnetTier3&cid¼1228774398696. AccessedDecember 5, 2017.

11. Cubanski J, Neuman T, Griffin S, Damico A.Medicare Spending at the End of Life: A Snapshot ofBeneficiariesWhoDied in 2014 and theCost of TheirCare. Kaiser Family Foundation: Kaiser FamilyFoundation, 2016. Available at: https://www.kff.org/medicare/issue-brief/medicare-spending-at-the-end-of-life/. Accessed December 5, 2017.

12. Chen LM, Jha AK, Guterman S, Ridgway AB,Orav EJ, Epstein AM. Hospital cost of care, qualityof care, and readmission rates: penny wise andpound foolish? Arch Intern Med 2010;170:340–6.

13. Jha AK, Orav EJ, Dobson A, Book RA,Epstein AM. Measuring efficiency: the associationof hospital costs and quality of care. Health Aff(Millwood) 2009;28:897–906.

14. Romley JA, Jena AB, Goldman DP. Hospitalspending and inpatient mortality: evidence fromCalifornia: an observational study. Ann Intern Med2011;154:160–7.

15. Romley JA, Jena AB, O’Leary JF, Goldman DP.Spending and mortality in US acute care hospitals.Am J Manag Care 2013;19:e46–54.

16. Stukel TA, Fisher ES, Alter DA, et al. Associa-tion of hospital spending intensity with mortalityand readmission rates in Ontario hospitals. JAMA2012;307:1037–45.

17. Chen CL. An Introduction to Quantile RegressionandtheQUANTREGProcedure.Cary,NC:SAS InstituteInc. Available at http://www2.sas.com/proceedings/sugi30/213-30.pdf. Accessed August 10, 2017.

18. Ellison A. CMS will cancel major bundledpayment initiatives. Becker’s Hospital Review,2017. Available at: https://www.beckershospitalreview.com/finance/cms-will-cancel-major-bundled-payment-initiatives.html. AccessedDecember 5, 2017.

19. Bundled Payments for Care Improvement(BPCI) Initiative: General Information. Centers forMedicare and Medicaid Services, 2017. Availableat: https://innovation.cms.gov/initiatives/bundled-payments/. Accessed December 5, 2017.

20. Burke LG, Frakt AB, Khullar D, Orav EJ, Jha AK.Association between teaching status and mortalityin US hospitals. JAMA 2017;317:2105–13.

21. Shahian DM, Liu X, Meyer GS, Torchiana DF,Normand SL. Hospital teaching intensity andmortality for acute myocardial infarction, heartfailure, and pneumonia. Med Care 2014;52:38–46.

22. Das A, Norton EC, Miller DC, Ryan AM,Birkmeyer JD, Chen LM. Adding a spending metricto medicare’s value-based purchasing programrewarded low-quality hospitals. Health Aff (Mill-wood) 2016;35:898–906.

23. Barnato AE, Chang CC, Farrell MH, Lave JR,Roberts MS, Angus DC. Is survival better at hos-pitals with higher “end-of-life” treatment in-tensity? Med Care 2010;48:125–32.

24. Hernandez AF, Greiner MA, Fonarow GC, et al.Relationship between early physician follow-upand 30-day readmission among Medicare benefi-ciaries hospitalized for heart failure. JAMA 2010;303:1716–22.

25. Jurgens CY, Goodlin S, Dolansky M, et al. Heartfailure management in skilled nursing facilities: ascientific statement from the American HeartAssociation and the Heart Failure Society ofAmerica. Circ Heart Fail 2015;8:655–87.

26. Hindricks G, Taborsky M, Glikson M, et al.Implant-based multiparameter telemonitoringof patients with heart failure (IN-TIME): arandomised controlled trial. Lancet 2014;384:583–90.

27. Schwartz AL, Landon BE, Elshaug AG,Chernew ME, McWilliams JM. Measuring low-value care in Medicare. JAMA Intern Med 2014;174:1067–76.

28. Schmaltz SP, Williams SC, Chassin MR,Loeb JM, Wachter RM. Hospital performancetrends on national quality measures and the as-sociation with Joint Commission accreditation.J Hosp Med 2011;6:454–61.

29. Readmissions Reduction Program (HRRP).Centers for Medicare and Medicaid Services, 2016.

Available at: https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/AcuteInpatientPPS/Readmissions-Reduction-Program.html. Acce-ssed December 5, 2017.

30. Allen LA, Hernandez AF, Peterson ED, et al.Discharge to a skilled nursing facility and subse-quent clinical outcomes among older patientshospitalized for heart failure. Circ Heart Fail 2011;4:293–300.

31. Feltner C, Jones CD, Cene CW, et al. Transi-tional care interventions to prevent readmissionsfor persons with heart failure: a systematic reviewand meta-analysis. Ann Intern Med 2014;160:774–84.

32. Jones CD, Bowles KH, Richard A, Boxer RS,Masoudi FA. High-value home health care forpatients with heart failure: an opportunity tooptimize transitions from hospital to home.Circ Cardiovasc Qual Outcomes 2017;10:e003676.

33. Newhouse JP, Garber AM. Geographic varia-tion in health care spending in the United States:insights from an Institute of Medicine report.JAMA 2013;310:1227–8.

34. Mechanic R. Post-acute care–the next frontierfor controlling Medicare spending. N Engl J Med2014;370:692–4.

35. Fisher ES, Wennberg DE, Stukel TA,Gottlieb DJ, Lucas FL, Pinder EL. The implicationsof regional variations in Medicare spending. Part 2:health outcomes and satisfaction with care. AnnIntern Med 2003;138:288–98.

36. Doyle JJ Jr. Returns to local-area healthcarespending: evidence from health shocks to patientsfar from home. Am Econ J Appl Econ 2011;3:221–43.

37. Joynt KE, De Lew N, Sheingold SH,Conway PH, Goodrich K, Epstein AM. ShouldMedicare value-based purchasing take socialrisk into account? N Engl J Med 2017;376:510–3.

KEY WORDS heart failure, mortality,outcomes, payments, spending, value

APPENDIX For the supplemental tables andfigures, please see the online version of thispaper.