The Journal of Arthroplasty 32 (2017) 702e708

Contents lists available at ScienceDirect

The Journal of Arthroplasty

journal homepage: www.arthroplastyjournal .org

Health Policy and Economics

The High Value Healthcare Collaborative: Observational Analyses ofCare Episodes for Hip and Knee Arthroplasty Surgery

William B. Weeks, MD, PhD, MBA a, b, *, William J. Schoellkopf, MS c, Lyle S. Sorensen, MD d,Andrew L. Masica, MD, MSCI e, Robert E. Nesse, MD f, James N. Weinstein, DO, MS a, b, g

a The Dartmouth Institute for Health Policy and Clinical Practice, Lebanon, New Hampshireb Geisel School of Medicine, Hanover, New Hampshirec The High Value Healthcare Collaborative Program Office, Portland, Mained Virginia Mason Medical Center, Seattle, Washingtone Baylor Scott & White Health, Dallas, Texasf Mayo Clinic, Rochester, Minnesotag Dartmouth-Hitchcock Health, Lebanon, New Hampshire

a r t i c l e i n f o

Article history:Received 6 June 2016Received in revised form25 August 2016Accepted 16 September 2016Available online 28 September 2016

Keywords:joint arthroplastyMedicarehealth care costsepisodes of carepostacute care

This project was supported by grant number 1C1CMDepartment of Health and Human Services, CenterServices.

One or more of the authors of this paper have discconflicts of interest, which may include receipt of payminstitutional support, or association with an entity inmay be perceived to have potential conflict of intedisclosure statements refer to http://dx.doi.org/10.101

The research was conducted by the awardee; findingconsistent with or confirmed by the findings ofcontractor.* Reprint requests: William B. Weeks, MD, PhD, M

NH 03768.

http://dx.doi.org/10.1016/j.arth.2016.09.0090883-5403/© 2016 Elsevier Inc. All rights reserved.

a b s t r a c t

Background: Broader use of value-based reimbursement models will require providers to transparentlydemonstrate health care value. We sought to determine and report cost and quality data for episodes ofhip and knee arthroplasty surgery among 13 members of the High Value Healthcare Collaborative(HVHC), a consortium of health care systems interested in improving health care value.Methods: We conducted a retrospective, cross-sectional observational cohort study of 30-day episodes ofcare for hip and knee arthroplasty in fee-for-service Medicare beneficiaries aged 65 or older who had hipor knee osteoarthritis and used 1 of 13 HVHC member systems for uncomplicated primary hip arthro-plasty (N ¼ 8853) or knee arthroplasty (N ¼ 16,434), respectively, in 2012 or 2013. At the system level, wecalculated: per-capita utilization rates; postoperative complication rates; standardized total, acute, andpostacute care Medicare expenditures for 30-day episodes of care; and the modeled impact of reducingepisode expenditures or per-capita utilization rates.Results: Adjusted per-capita utilization rates varied across HVHC systems and postacute care re-imbursements varied more than 3-fold for both types of arthroplasty in both years. Regression analysisconfirmed that total episode and postacute care reimbursements significantly differed across HVHCmembers after considering patient demographic differences. Potential Medicare cost savings weregreatest for knee arthroplasty surgery and when lower total reimbursement targets were achieved.Conclusion: The substantial variation that we found offers opportunities for learning and collaboration tocollectively improve outcomes, reduce costs, and enhance value. Ceteris paribus, reducing per-episodereimbursements would achieve greater Medicare cost savings than reducing per-capita rates.

© 2016 Elsevier Inc. All rights reserved.

S331029-01-00 from the U.S.s for Medicare & Medicaid

losed potential or pertinentent, either direct or indirect,the biomedical field which

rest with this work. For full6/j.arth.2016.09.009.

s to date may or may not bethe independent evaluation

BA, 307 River Road, Lyme,

Although the cost and quality impact of value-based reim-bursement appear to be modest to date [1-3], the Centers forMedicare and Medicaid Services (CMS) is expanding and acceler-ating such efforts, intending to tie 85 percent of Medicare re-imbursements to quality or value and having 30 percent ofMedicare payments in alternative payment models by the end of2016 [4], while working to engage other payers in similar activities[5]. Broader use of these new reimbursement models will requireproviders to engage patients [6], enhance coordination of care [7],and transparently demonstrate value [8-10].

One alternative payment mechanism is “bundling” paymentsfor an episode of care. Cutler and Ghosh [11] calculated potentialannual Medicare cost savings of $4.7-$29 billion, depending on

W.B. Weeks et al. / The Journal of Arthroplasty 32 (2017) 702e708 703

how many conditions were subject to some maximum bundledpayment amount. Hussey et al [12] identified bundled payments asthe single most promising mechanism for reducing health carecosts, with a potential $5.4 billion reduction in cumulative nationalhealth care expenditures over 10 years. In April, 2016, CMSimplemented its Comprehensive Care for Joint Replacement modelwhich tests bundled payments and quality measurement for anepisode of care associated with hip or knee arthroplasty surgery in67 metropolitan statistical areas, impacting approximately 800hospitals [13].

The High Value Healthcare Collaborative (HVHC) is a con-sortium of health care delivery systems that was formed with thegoal of accelerating the process of improving quality of care whilereducing cost [14]. In 2012, HVHC received a Health Care Inno-vation Award from the Center for Medicare & Medicaid Innova-tion to test the impact of patient engagement interventions on,among other measures, outcomes, and costs for hip and kneearthroplasties. These interventions included shared decision-making and preoperative education for healthy patients. Whileuptake of these interventions varied across HVHC members andover the years of the award, we sought to understand overalltrends during the first 2 years of the award, by examining HVHCmemberespecific patient demographics, utilization rates, andstandardized Medicare reimbursements for 30-day episodes ofcare for, and complication rates following, uncomplicated hiparthroplasty and knee arthroplasty surgeries. This paper reportson that effort.

Materials and Methods

Through the Center for Medicare & Medicaid Innovation award,the HVHC obtained 100 percent of the fee-for-service Medicaredata for Dartmouth-Atlas defined hospital referral regions (HRRs)in which participating HVHC members had a market presence forcalendar years 2012 and 2013. This self-evaluationwas approved byCMS as part of the data use agreement that was required to obtainthe data as well as the institutional review board at DartmouthCollege (the HVHC operates under a centralized institutional re-view board model) (Committee for the Protection of Human Sub-jects [CPHS] # 23,820). We followed CMS suppression rules thatprecluded us from reporting information that was derived fromfewer than 11 patients. Data manipulation and statistical analyseswere performed with the SAS 9.4 software (Cary, NC).

Procedures Examined

We examined cohorts of patients who obtained either of 2procedures that occurred during an index admission: uncompli-cated unilateral hip arthroplasty (Medicare Severity DiagnosisRelated Group [MSDRG] 470, International Classification of Dis-eases, 9th edition, Clinical Modification [ICD-9-CM] procedure code81.51) and uncomplicated unilateral knee arthroplasty (MSDRG470, ICD-9-CM procedure code 81.54). We limited hip and kneearthroplasty analysis to those conducted on patients who had aprimary diagnosis of hip osteoarthritis (ICD-9-CM diagnostic codes715.09, 715.15, 715.25, 715.35, or 715.95) or knee osteoarthritis(ICD-9-CM diagnostic codes 715.09, 715.16, 715.26, 715.36, or715.96), respectively.

Patient Demographics

From Medicare files, we obtained patient age, sex, and race(which we present, in aggregate, as the proportion of patients whowere white). We limited our analysis to fee-for-service Medicarepatients who were Medicare eligible by virtue of age. We used

ICD-9-CM diagnosis codes to calculate Deyo-modified Charlsonscores [15,16]. From Medicare files, we determined whetherpatients were concurrently enrolled in Medicaid. Finally, wecalculated the length of stay for the index admission.

Utilization Rates

Per-capita utilization rates were determined as follows. Bene-ficiaries were attributed to HVHC member systems using thephysician health network methodology that attributes patients to ahospital by identifying their primary care provider (or, if none isavailable, determining the provider from whom they obtain theplurality of care), delineating networks of primary care and spe-cialty care providers, and determining the hospitals to which thoseproviders admit patients [17]. This attribution process was repeatedfor each year. HVHC member hospitals were identified by theirunique Medicare provider number, and, if HVHC members hadmore than 1 hospital, patients attributed to HVHC member hospi-tals were aggregated at the HVHC member level to develop a de-nominator. Surgery counts at HVHC member facilities were thenumerators. Rates were adjusted for age, sex, concurrent enroll-ment in Medicaid, and Charlson score using an indirect adjustmentmethod [18].

Complication Rates

Complications that occurred during the index admission or the30-day follow-up period were identified using the methodologyidentified by the Yale New Haven Health Services Corporation/Center for Outcomes Research and Evaluation that underlies thecomplication rate that CMS reports [19]. We used version 3.0,modified to only include a 30-day follow-up period; further,readmissions occurring at least 2 days after discharge were alsoconsidered complications.

Calculation of Price-Standardized Bundled Episode of Care Costs

The HVHC program management office developed a meth-odology that identified the index admissions for these proced-ures and calculated Medicare reimbursements for all care from 3days prior to admission to 30 days after discharge. We dividedtotal costs of the episode into 2 periods: the acute care periodthat consists of the time the patient was admitted during theindex admission and the postacute care period. The acute careperiod included 3 service categories: the index Diagnosis-RelatedGroup payment; part B professional fees that include consulta-tive expenses during the index admission; and outpatient facilityfees, generally associated with care provided in anticipation ofadmission over the 3 days prior to admission. The postacute careperiod included 7 reimbursement service categories: outpatientfacility fees, skilled nursing facility care, inpatient rehabilitationcare, long-term nursing home care, home health care, part Bprofessional fees, and inpatient acute care reimbursement thatincluded costs associated with readmission. In cases wherepostacute care extended past the 30-day period, the associatedcosts were prorated for the relative number of days in thatperiod. Patients who expired during the 30-day period wereremoved from the analysis.

Because organizations in different geographies experiencedifferent Medicare reimbursement rates, we standardized Medi-care reimbursement as follows. For each unit of consumption (forinstance, index Diagnosis-Related Group, day of skilled nursingfacility or rehabilitation hospital, or part B charge), we applied theHealth Partners Total Cost of CareeTotal Care Relative ResourceValue methodology which were standardized to 2014 costs and

Table 1Numbers of Surgeries, Mean Lengths of Stay (LOS), Patient Demographics, and Complication Rates for Uncomplicated Hip Arthroplasty and Knee Arthroplasty Surgeries for 14 HVHC Members in 2012 and 2013.

HVHC Member Uncomplicated Hip Arthroplasty Uncomplicated Knee Arthroplasty

N LOS inDays

Age (y) %Male

%White

%MedicaidEnrolled

MeanCharlsonScore

CompRate (%)

N LOS inDays

Age (y) %Male

%White

%MedicaidEnrolled

MeanCharlsonScore

CompRate (%)

2012 Baylor Health Care System 374 2.94 74.2 34.8 90.6 3.5 2.25 5.3 890 3.06 73.0 35.7 86.5 6.9 2.42 6.1Beth Israel Deaconess Medical Center 61 3.39 74.3 42.6 90.2 S 2.49 S 128 3.35 74.4 28.1 78.1 23.4 2.77 SDartmouth-Hitchcock Health 121 2.75 73.3 37.2 94.2 S 2.04 S 121 2.91 73.5 35.5 99.2 9.9 1.76 SEastern Maine Health System 124 3.02 73.9 35.5 100 27.4 1.94 S 268 3.07 73.5 42.2 98.9 25.4 1.90 4.9Intermountain Healthcare 484 2.80 73.7 39.0 97.9 4.3 1.83 6.6 1146 2.95 73.3 37.1 96.9 3.5 1.84 4.5Maine Health 276 2.28 74.4 33.0 99.3 19.2 1.51 S 555 3.05 74.2 33.5 98.2 24.5 1.93 4.5Mayo Clinic 784 3.03 75.0 46.3 96.7 S 2.08 5.5 1392 3.08 74.5 38.6 95.4 3.5 1.92 4.5North Shore LIJ Health System 523 3.39 75.0 35.6 92.2 5.2 2.83 6.9 765 3.38 74.8 29.8 84.3 8.1 2.92 7.2Providence Health & Services 1130 2.88 73.7 36.4 95.5 5.5 1.61 4.4 1886 2.98 73.6 37.8 89.1 9.2 1.90 3.5Scott & White Healthcare 112 2.88 74.3 41.1 92.9 S 2.06 S 293 3.05 73.0 35.5 89.4 5.8 2.17 SUCLA Health 52 3.63 74.3 38.5 86.5 21.2 2.08 S 91 3.51 73.7 24.2 61.5 30.8 2.60 SUniversity of Iowa Health Care 29 3.03 75.8 41.4 100 S 1.83 S 44 3.00 74.3 27.3 95.5 S 2.00 SVirginia Mason Medical Center 81 2.98 75.3 43.2 95.1 S 1.73 S 163 3.12 74.7 31.9 84.7 15.3 1.77 S

2013 Baylor Health Care System 384 2.76 73.4 41.4 91.7 S 2.33 6.3 845 2.98 72.9 34.6 87.5 5.2 2.29 3.6Beth Israel Deaconess Medical Center 93 3.29 74.8 34.4 87.1 S 3.03 S 178 3.30 74.1 29.2 83.1 12.9 2.68 9.0Dartmouth-Hitchcock Health 165 2.28 73.6 37.6 95.8 6.7 1.49 S 153 3.09 73.3 37.9 96.1 9.2 1.99 SEastern Maine Health System 192 2.85 73.5 29.7 97.9 20.8 2.02 7.3 361 2.85 72.6 39.3 96.4 23.5 1.96 4.7Intermountain Healthcare 550 2.61 73.9 39.3 97.1 2.7 1.84 6.4 1425 2.86 72.8 37.0 94.7 2.7 1.79 3.1Maine Health 366 2.01 74.4 34.7 98.9 15.0 1.71 4.4 506 3.01 74.4 38.7 97.0 19.8 1.81 4.2Mayo Clinic 827 2.92 74.6 41.7 96.0 2.4 1.81 2.8 1457 2.93 74.1 41.5 97.2 2.1 1.84 3.4North Shore LIJ Health System 555 3.37 75.1 35.1 86.8 4.9 2.82 4.0 848 3.37 74.4 32.1 82.7 9.9 2.92 6.5Providence Health & Services 1270 2.67 73.6 38.3 93.1 5.4 1.69 3.9 2272 2.82 73.2 38.2 88.7 8.4 1.85 3.7Scott & White Healthcare 123 2.72 73.3 39.8 91.1 S 2.15 S 359 2.92 73.2 32.3 85.5 5.8 2.21 4.7UCLA Health 52 3.50 75.6 44.2 78.8 S 2.04 S 73 3.33 74.5 34.2 68.5 23.3 2.58 SUniversity of Iowa Health Care 33 3.06 76.5 36.4 93.9 S 2.21 S 47 2.87 71.4 27.7 95.7 0.0 2.02 SVirginia Mason Medical Center 92 2.67 75.2 41.3 93.5 S 1.96 S 168 2.58 74.7 35.7 83.9 10.1 1.55 S

S means that data are suppressed.HVHC, High Value Healthcare Collaborative; LIJ, Long Island Jewish; UCLA, University of California at Los Angeles.

W.B.W

eekset

al./The

Journalof

Arthroplasty

32(2017)

702e708

704

Table

2Crudean

dAdjusted

Per-Cap

itaRates

ofUnco

mplic

ated

Hip

andKnee

ArthroplastySu

rgeriesfor14

HVHCMem

bers

in20

12an

d20

13.

HVHCMem

ber

2012

2013

Hip

ArthroplastyN

Knee

ArthroplastyN

Attribu

tedPo

pulation

Adjusted

Joint

Arthroplasty

Rateper

1000

Hip

ArthroplastyN

Knee

ArthroplastyN

Attribu

tedPo

pulation

Adjusted

Joint

Arthroplasty

Rateper

1000

Hip

Knee

Hip

Knee

Bay

lorHea

lthCareSy

stem

374

890

77,193

4.85

11.22

384

845

75,547

5.06

10.90

BethIsrael

Dea

conessMed

ical

Cen

ter

6112

826

,829

2.36

4.86

9317

827

,722

3.43

6.49

Dartm

outh-H

itch

cock

Hea

lth

121

121

14,937

7.61

7.88

165

153

15,636

9.99

9.50

EasternMaineHea

lthSy

stem

124

268

22,092

5.29

11.72

192

361

21,701

8.44

16.13

Interm

ountain

Hea

lthcare

484

1146

67,854

6.75

16.55

550

1425

75,019

6.96

18.50

MaineHea

lth

276

555

49,528

5.35

11.29

366

506

48,936

7.28

10.43

May

oClin

ic78

413

9263

,472

12.13

22.45

827

1457

64,079

12.75

23.22

North

ShoreLIJHea

lthSy

stem

523

765

141,31

83.84

5.48

555

848

143,44

14.03

6.03

Prov

iden

ceHea

lth&

Services

1130

1886

202,06

15.64

9.40

1270

2272

208,51

26.13

10.94

Scott&

WhiteHea

lthcare

112

293

26,123

4.24

10.73

123

359

25,461

4.77

13.59

UCLA

Hea

lth

5291

13,300

4.48

7.09

5273

13,770

4.17

5.51

University

ofIowaHea

lthCare

2944

5128

5.37

8.31

3347

5664

5.57

7.99

Virginia

Mason

Med

ical

Cen

ter

8116

311

,917

7.01

14.03

9216

813

,087

7.16

13.08

HVHC,H

ighValueHea

lthcare

Collabo

rative

;LIJ,Lo

ngIslandJewish;UCLA

,University

ofCalifornia

atLo

sAnge

les.

W.B. Weeks et al. / The Journal of Arthroplasty 32 (2017) 702e708 705

calibrated to Medicare allowed amounts, to generate a standard-ized reimbursement for all service categories [20].

Regression Analysis

We performed a mixed models regression analysis in which weincluded patient demographics, Charlson score, concurrentMedicaid enrollment status, year, and HVHC member to determinewhether, after adjustment for these factors, there were statisticallysignificant differences in acute or postacute Medicare re-imbursements across HVHC members. To account for the correla-tion of observations within HRR and hospital, we included randomintercepts for both HRR and hospital. For postacute cost models,indicator variables were added to account for the different post-acute care categories and to account for surgeries that had nopostacute care costs. For each model, Wald type III tests wereperformed to test for a significant difference between HVHCmembers and P < .05 was considered statistically significant.

Modeling

We sought to determine whether HVHC members should pri-oritize reducing per-capita utilization rates or episode costs of care.Therefore, we modeled the overall cost impact should, all HVHCmembers that were above the 25th, 50th, or 75th percentile per-formance levels on either of these measures, achieve 25th, 50th, or75th percentile levels of performance. We calculated per-capitacost savings as follows: for each type of joint arthroplasty and foreach system, we multiplied baseline utilization rate by per-capitatotal expenditure. To calculate per-capita cost savings should uti-lization rates decrease to 25th, 50th, or 75th percentile levels, forsystems with performance above those levels, we multipliedbaseline per-capita total expenditures by the utilization rate at thespecified percentile. Similarly, to calculate per-capita cost savingsshould per-capita total expenditures decrease to 25th, 50th, or 75thpercentile levels, for systems with performance above those levels,we multiplied utilization rates by the per-capita expenditures atthe specified percentile. We calculated potential cost savings bysubtracting themodeled per-capita expenditures for the 13 systemsfrom the baseline per-capita expenditures for the 13 systems andpresent results as a percentage cost savings from baseline.

Results

Variation in Numbers of Surgeries, Lengths of Stay, PatientCharacteristics, and Complication Rates

We found substantial variation in the volume and length of stayof procedures as well as in patient characteristics across HVHCsystems (Table 1). For example, for hip surgery, across years andHVHC member sites, the volume of procedures varied from 29 to1270; the mean length of stay, from 2.01 to 3.64; age, from 73.3 to76.5; proportion of patients who were male, from 29.7% to 46.3%;proportion of patients who were white, from 35.3% to 100%; pro-portion of patients who were concurrently enrolled in Medicaid,from 2.4% to 27.4%; mean Charlson score from 1.49 to 3.03; andproportion of patients who experienced complications, from 2.8%to 7.3%. Variation in patient characteristics across HVHC memberswas similar for knee and hip arthroplasty surgery.

Variation in Per-Capita Utilization Rates Across HVHC Members andAcross Years

Adjusted utilization rates for HVHC members varied morewidely in 2012 (from 2.36 to 12.13 per 1000 for hip arthroplasty

Fig. 1. For uncomplicated hip arthroplasty (A) and knee arthroplasty (B) surgery in 2013 and 2012 in 13 HVHC member health care systems, variation in standardized Medicareexpenditures for acute and postacute portions of 30-day episodes of care, and their components in 2014 dollars. SNF, skilled nursing facility; Rehab, inpatient rehabilitation hospital;Inpatient, inpatient hospitalization from readmission; DME, durable medical equipment; HVHC, High Value Healthcare Collaborative; LIJ, Long Island Jewish; UCLA, University ofCalifornia at Los Angeles.

W.B. Weeks et al. / The Journal of Arthroplasty 32 (2017) 702e708706

(5.1-fold variation) and from 4.86 to 22.45 per 1000 for kneearthroplasty (4.6 fold variation)) than they did in 2013 (from 3.43 to12.75 per 1000 for hip arthroplasty (3.7-fold variation) and from5.51 to 23.22 per 1000 for knee arthroplasty (4.2-fold variation))(Table 2). Between 2012 and 2013, per-capita hip surgery arthro-plasty rates decreased for the University of California at Los Angeles(UCLA) but increased for all other HVHC systems; during the sameperiod, per-capita knee arthroplasty surgery rates decreased for 5of the 13 HVHC systems examined.

Variation in Acute and Postacute Reimbursements in 30-DayEpisodes for HVHC Members

Mean acute care costs for hip and knee arthroplasty were similaracross HVHCmember sites, with acute care reimbursements for hiparthroplasty varying from $18,877-$19,874 in 2013 and from$18,573-$19,379 in 2012 and that for knee arthroplasty varyingfrom $18,945-$19,805 in 2013 and from $18,520-$19,406 in 2012(Fig. 1). Postacute care costs varied much more across HVHC sites.For hip arthroplasty, postacute care costs varied from $4021-$12,220 (3.03-fold) in 2013 and from $3888-$13,162 (3.39 fold) in2012; for knee arthroplasty, they varied from $3798-$13,577 (3.58-fold) in 2013 and from $4526-$14,421 (3.19-fold) in 2012.

The makeup of postacute care costs differed considerably acrossHVHC members. In general, higher expenditure systems seemed toutilize rehabilitation hospitals or skilled nursing facilities consid-erably more than lower expenditure systems; seemingly, healthsystems that used considerable skilled nursing facility or rehabili-tation hospital resources did so for both procedures.

Regression Analysis Results

The results of our regression analysis revealed that, afteradjusting for patient age, sex, Medicaid eligibility status, race, andCharlson score, there were statistically significant differencesamongHVHCmembers in postacute and total episode expenditures(Table 3).

Modeling Results

Achievement of either lower per-capita joint arthroplasty utili-zation or reimbursement rates would reduce Medicare expendi-tures related to this procedure. Comparatively, lowerreimbursement rates would generate greater Medicare cost savingsthanwould lower utilization rates (Fig. 2). Furthermore, movementfrom 50th to 25th percentile utilization or reimbursement ratesachieves substantially greater potential cost savings than move-ment from 75th to 50th percentile rates.

Discussion

We examined variation in patient demographics, complicationrates, per-capita utilization rates, and episode of care costs for ep-isodes of total hip arthroplasty and total knee arthroplasty in 13HVHC member system hospitals. We found that HVHC systemsprovided care to patients with different characteristics and haddifferent lengths of stays, complication rates, per-capita utilizationrates, and costs of care. Across HVHC members, adjusted per-capitautilization rates differed by 3- to 5-fold; while rates increased

Fig. 1. (continued).

W.B. Weeks et al. / The Journal of Arthroplasty 32 (2017) 702e708 707

between 2012 and 2013, variation in rates decreased somewhatduring that period. After adjusting for differences in patient char-acteristics, lengths of stay, complication rates, and acute care costswere not statistically significantly different across HVHC members;however, postacute care costs and total episode costs were. Thevery substantial differences in postacute care costs that we foundappeared to be attributable to differing practice patterns thatappeared to be system specific.

These findings confirmed that much of the variability in totalepisode costs is attributable to differences in postacute care costs[21]. Our analysis suggests that organizational administrative andclinical practice patterns may need to change for these differencesto diminish. To the degree that hospitals have academic or financialrelationships with postacute care settings, it may be difficult totransform those practice patterns. Clearly, bundling of payments for

Table 3Results of the Regression Analysis That Determined Whether HVHC Member WasStatistically Significantly Associated With Differences in Particular Outcomes Mea-sures, for Hip and Knee Arthroplasty in 2012 and 2013.

Surgery Measure P Value

Hip arthroplasty Length of stay .13Complications .33Acute care reimbursements .12Postacute care reimbursements .020Total episode reimbursements <.0001

Knee arthroplasty Length of stay .29Complications .16Acute care reimbursements .34Postacute care reimbursements .0004Total episode reimbursements .0001

HVHC, High Value Healthcare Collaborative.

an episode of care would align incentives to reduce health carecosts, perhaps by providing equivalent care in less expensive set-tings. Reducing utilization of facilities and programs that arecurrently profitable will be difficult but necessary. It will beimportant to monitor utilization as total per-capita re-imbursements fall to ensure that neither care access nor treatmentthresholds do not change [22].

We found that reducing total per-episode reimbursementswould have a greater overall cost impact than reducing per-capitautilization rates. This is somewhat counterintuitive, as reducingutilization rates have the potential to avoid the entire cost of an

Fig. 2. Modeling of percentage per-capita cost savings from baseline should HVHCmembers achieve 25th, 50th, or 75th percentile performance in per-capita utilizationor total reimbursement rates.

W.B. Weeks et al. / The Journal of Arthroplasty 32 (2017) 702e708708

episode of care. However, particularly when understanding thatpatients who do not obtain hip or knee arthroplasty continue tohave considerable costs associated with the ongoing medicalmanagement of hip or knee osteoarthritis [23], reduction of per-capita episode costs appears to be a dominant strategy forreducing overall per-capita cost of care for these joint arthroplastysurgeries. While our results differ from a national examination ofthe impact of utilization rate reductions or price reductions inprostatectomy, coronary artery bypass grafting, or hip arthroplastysurgery on overall Medicare costs [24], this is perhaps because ouranalysis was limited to participating HVHC hospitals: it seemslikely that different baselines would promote differing strategiesfor reducing Medicare expenditures.

Our study has several limitations. First, we used an internalmethod to determine episode costs; use of alternative algorithmsmight change results. Second, our process eliminated direct pa-tient care costs. While these may be limited by statute, theyshould not be ignored in policymaking decisions. Third, weexamined only older Medicare fee-for-service beneficiaries; otherpopulations may have different results. Fourth, we in our analysisof potential Medicare cost savings, we modeled utilization rateand episode cost reductions to 75th, 50th, and 25th performancepercentile levels within HVHC members. While this range seemsreasonableda large proportion of patients eligible for hip andknee arthroplasty do not want it [25], surgeon preference forperforming knee arthroplasty surgery helps explain regional dif-ferences in utilization rates [26], and implementation of shareddecision-making can dramatically reduce utilization rates for hipand knee arthroplasty procedures [23]ddifferent results wouldbe generated had we modeled different performance achieve-ment levels. Fifth, we examined only the first 2 years of the award;we may find performance differences in later self-analyses, afterthe patient engagement interventions or other organizationalimprovement initiatives were more widely adopted. Sixth, it ispossible that in advanced integrated networks associated withlarge referral centers, the attribution model that we used maygenerate higher per-capita utilization rates due to referral bias.Future work examining the stability of the attribution model overtime and across different network types is warranted. Finally,Medicare suppression rules precluded our ability to analyzecomplication rates in all the 13 HVHC health care systems westudied.

Conclusions

The substantial variation that we found warrants ongoinginvestigation and action. Health care systems that rely heavily onmore expensive postacute care health care settings, like rehabili-tation hospitals, might consider substituting lower cost settings,like home health care, to reduce costs; within fully integratedsystems, such changes in care processes may require shedding as-sets. Furthermore, payers should take steps to ensure that costsavings generated through reduced episode costs are not offset byincreased utilization rates wherein patient-reported measures ofpain and function are not significantly impacted. Finally, our workdemonstrates that there is ample room for reduction of variation inpostacute care costs for the procedures we examined; health care

systems should collaborate to learn more efficient ways to managecare processes from one another.

References

1. Kruse GB, Polsky D, Stuart EA, et al. The impact of hospital pay-for-performanceon hospital and Medicare costs. Health Serv Res 2012;47:2118.

2. Colla CH, Wennberg DE, Meara E, et al. Spending differences associated with theMedicare physician Group practice demonstration. JAMA 2012;308:1015.

3. McWilliams JM, Landon BE, Chernew ME. Changes in health care spending andquality for Medicare beneficiaries associated with a commercial ACO contract.JAMA 2013;310:829.

4. Burwell SM. Setting value-based payment goals - HHS efforts to improve UShealth care. N Engl J Med 2015;372:897.

5. Rajkumar R, Conway PH, Tavenner M. CMSeengaging multiple payers in pay-ment reform. JAMA 2014;311:1967.

6. Allen H, Nobel JJ, Burton WN. Making health care reformwork: where physicianand employer interests converge. JAMA 2012;308:2465.

7. Press MJ, Rajkumar R, Conway PH. Medicare's new bundled payments: design,strategy, and evolution. JAMA 2016;315:131.

8. Wennberg JE, Fisher ES, Skinner JS, et al. Extending the P4P agenda, part 2: howMedicare can reduce waste and improve the care of the chronically ill. HealthAff (Millwood) 2007;26:1575.

9. Wennberg JE, O'Connor AM, Collins ED, et al. Extending the P4P agenda, part 1:how Medicare can improve patient decision making and reduce unnecessarycare. Health Aff (Millwood) 2007;26:1564.

10. Ginsburg PB. Achieving health care cost containment through provider pay-ment reform that engages patients and providers. Health Aff (Millwood)2013;32:929.

11. Cutler DM, Ghosh K. The potential for cost savings through bundled episodepayments. N Engl J Med 2012;366:1075.

12. Hussey PS, Eibner C, Ridgely MS, et al. Controlling US health care spending-separating promising from unpromising approaches. N Engl J Med 2009;361:2109.

13. Comprehensive care for joint replacement model. 2015. https://innovation.cms.gov/initiatives/cjr [accessed 25.11.15].

14. Weinstein JN, Nesse RE, James BC, et al. The High Value Healthcare Collabora-tive. A national collaborative to improve value in healthcare delivery: structure,process, challenges, and early lessons learned. Health Aff (Millwood) 2012;31:1329.

15. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognosticcomorbidity in longitudinal studies: development and validation. J Chronic Dis1987;40:373.

16. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for usewith ICD-9-CM administrative databases. J Clin Epidemiol 1992;45:613.

17. Bynum JP, Bernal-Delgado E, Gottlieb D, et al. Assigning ambulatory patientsand their physicians to hospitals: a method for obtaining population-basedprovider performance measurements. Health Serv Res 2007;42:45.

18. The Dartmouth Atlas of Health Care: indirect adjustment. http://www.dartmouthatlas.org/downloads/methods/indirect_adjustment.pdf [accessed16.12.15].

19. Center for Medicare and Medicaid Services. Yale New Haven Health Services Corpo-ration Center for Outcomes Research and EvaluationTechnical Expert Panel report oncomplications for total knee replacement V3.0. 2014. https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/HospitalQualityInits/Measure-Methodology.html [accessed 15.04.16].

20. Total Care Relative Resource Value. https://www.healthpartners.com/ucm/groups/public/@hp/@public/documents/documents/cntrb_039627.pdf; [accessed 25.11.15].

21. Chandra A, Dalton MA, Holmes J. Large increases in spending on postacute carein Medicare point to the potential for cost savings in these settings. Health Aff(Millwood) 2013;32:864.

22. Weeks WB, Rauh SS, Wadsworth EB, et al. The unintended consequences ofbundled payments. Ann Intern Med 2013;158:62.

23. Arterburn D, Wellman R, Westbrook E, et al. Introducing decision aids at GroupHealth was linked to sharply lower hip and knee surgery rates and costs. HealthAff (Millwood) 2012;31:2094.

24. Chen LM, Norton EC, Birkmeyer JD. Comparing approaches to controllingMedicare spending. J Hosp Med 2016;11:215.

25. Hawker GA, Wright JG, Coyte PC, et al. Determining the need for hip and kneearthroplasty: the role of clinical severity and patients' preferences. Med Care2001;39:206.

26. Wright JG, Hawker GA, Bombardier C, et al. Physician enthusiasm as anexplanation for area variation in the utilization of knee replacement surgery.Med Care 1999;37:946.