NEJMp1113361

8/2/2019 NEJMp1113361

1/3

n engl j med 366;12 nejm.org march 22, 2012

PERSPECTIVE

1075

Slower Growth in Medicare Spending

for harsh, simple-minded cuts set-ting the stage for broad-basedpayment reform. Up until theearly 1980s, Medicare reimbursedhospitals for costs incurred, sub- ject to ceilings. The Tax Equityand Fiscal Responsibility Act of

1982 substantially tightened thoselimits, leaving hospitals with noupside they could not earn aprofit by reducing costs and agrowing downside for thosewhose costs exceeded the limits.The next year, legislation waspassed, with the support of thehospital industry, replacing costreimbursement with the inpatientprospective payment system (IPPS), with rates initially calibrated to

leave Medicare outlays unchanged.Hospitals then had the opportu-nity to reduce costs per admis-

sion by shortening lengths of stayand to earn a positive margin inthe process.

The IPPS is generally viewedas a major policy success: it en-couraged hospitals to seek effi-ciencies, and when they found

those efficiencies, it allowed thefederal government to share inthe savings. Should ACOs andother reforms prove effective,they will provide broader oppor-tunities to increase the efficiencyof delivery beyond shorteninglengths of stay, such as managingchronic disease more effectivelyso as to keep beneficiaries out ofthe hospital in the first place.But our current challenge is more

complex than the one faced inthe early 1980s. Broadening theunit of payment will require

reaching across different types ofproviders and helping to stitchtogether real delivery systems inplaces where now there are none.

Disclosure forms provided by the authorsare available with the full text of this arti-cle at NEJM.org.

From the Center for Studying Health Sys-tem Change, Washington, DC.

This article (10.1056/NEJMp1201853) waspublished on March 7, 2012, at NEJM.org.

1. Martin AB, Lassman D, Washington B,Catlin A. Growth in US health spending re-mained slow in 2010; health share of grossdomestic product was unchanged from2009. Health Aff (Millwood) 2012;31:208-19.2. White C. Why did Medicare spendinggrowth slow down? Health Aff (Millwood)2008;27:793-802.3. Hadley J, Reschovsky J, Corey C, Zucker-

man S. Medicare fees and the volume ofphysicians services. Inquiry 2009-2010;46(4):372-90.Copyright 2012 Massachusetts Medical Society.

The Potential for Cost Savings through Bundled Episode PaymentsDavid M. Cutler, Ph.D., and Kaushik Ghosh, Ph.D.

In the quest to manage the spi-raling cost of U.S. health care,one approach has generated greatinterest. The philosophy behindmuch current policy includingthe Affordable Care Act (ACA) is that aggregating fee-for-servicereimbursement into payments forbroader bundles of care will leadto greater efficiency in the provi-sion of care and thus lower costs.

Under the accountable careorganization model, perhaps thebest-known example of this strat-

egy, medical reimbursements areaggregated to the person-yearlevel. Other programs aggregatereimbursement for episodes ofcare for example, care for aparticular cardiovascular or ortho-pedic condition. The Episode ofCare Payment Demonstration proj-ect, which is authorized by theACA, requires the Centers forMedicare and Medicaid Services

to experiment with bundlingMedicare Part A and Part B pay-ments for inpatient care. It re-sembles the Acute Care Episodepayment program, an ongoingdemonstration program thatbundles Part A and Part B pay-ments for select types of inpa-tient care episodes.

The central issue in any pro-posal for aggregating paymentsis determining at what level ser-vices should be bundled together.Episode-based bundled payments

are easier for individual physi-cians or small physician groupsto manage, since a given physi-cian is often involved in the fullcourse of a care episode. In con-trast, accepting global paymentsfor all of a particular patientscare generally requires a high de-gree of integration among multi-ple physicians. It may be for thisreason that episode-based bun-

dled payments seem to save moremoney than patient-based bundledpayments.1 On the other hand,bundling payments for care epi-sodes does not provide incentivesto reduce the number of episodes.If limiting the number of epi-sodes of care is a major consider-ation in reducing costs, bundlingcare at the patient level would bepreferred.

A central issue, then, is thetradeoff between the relative easeof bundling at the episode level

and the additional savings incen-tives from bundling at the patientlevel. Are a good share of poten-tial savings given up by usingonly episodes for setting bundledpayments?

We set out to estimate the costsavings associated with episode-based and patient-based bundledpayments. For episode-based bun-dled payment, we selected a ran-

The New England Journal of Medicine

Downloaded from nejm.org on March 24, 2012. For personal use only. No other uses without permission.

Copyright 2012 Massachusetts Medical Society. All rights reserved.

8/2/2019 NEJMp1113361

2/3

PERSPECTIVE


dom 5% of the elderly populationin fee-for-service Medicare in 2007,assigning the principal diagnosiscode of every hospital admissionto one of 285 categories in theClinical Classification System, asdetermined by the Agency forHealthcare Research and Quality.We considered a new episode tohave started when a claim was notpreceded by a hospitalization re-lated to the same organ systemwithin 180 days. For example, ahospital admission in March 2007

for myocardial infarction wouldcount as a new episode if therewere no related hospitalizationsduring the prior 6 months. A read-mission for heart failure 2 monthslater would count as part of thesame episode. In all cases, reha-bilitation services would not qual-ify as a new episode.

We then assigned to each epi-sode under consideration all in-patient and outpatient spendingfor the same organ system occur-ring within 180 days of the epi-sodes onset (going into 2008 asnecessary). Under this definition,a person could have multiple over-lapping episodes (for example, amyocardial infarction and a hipfracture within a 6-month period).Some claims, however, were notassigned to an episode for in-stance, physician visits not relat-

ed to a prior hospitalization. Weadjusted for geographic differenc-es in prices using data from theDartmouth Atlas of Health Carefor 2007 Medicare reimbursementsin each hospital referral region.2

Using these criteria, we arrivedat 245 types of episodes. In total,spending on these episodes ac-

counted for just over half of Medi-care spending for this sample in2007. The remainder of spend-ing was for physician and outpa-tient claims not associated with ahospitalization. Spending wasnot uniform across episodes butwas instead concentrated withina relatively small number. The top5 and top 17 episode types (seetable) accounted for one fourthand one half, respectively, of thecosts of the 245 episode types,and three fourths of the spending

on these episodes was accountedfor by the top 43 conditions.The most expensive condition

was osteoarthritis, and most epi-sodes of care for osteoarthritisinvolved elective hip or kneereplacement. Other high-cost epi-sodes included those for cardiovascular conditions and muscu-

Cost Savings through Bundled Episode Payments

Total Cost for the 17 Most Expensive Conditions among Medicare Patients 65 Years of Age or Older.*

RankCCS

Category ConditionTotal Spending

(billions of U.S. $)Share of TotalSpending (%)

CumulativeShare (%)

1 203 Osteoarthritis 7.3 5.7 5.7

2 101 Coronary atherosclerosis and other heart disease 6.5 5.1 10.8

3 226 Fracture of neck of femur (hip) 5.8 4.6 15.4

4 108 Congestive heart failure, nonhypertensive 5.8 4.5 19.95 109 Acute cerebrovascular disease 4.8 3.8 23.7

6 122 Pneumonia (except that caused by tuberculosis or STD) 4.7 3.7 27.4

7 106 Cardiac dysrhythmias 4.4 3.5 30.9

8 100 Acute myocardial infarction 4.4 3.5 34.4

9 237 Complication of device, implant, or graft 3.2 2.5 36.9

10 205 Spondylosis, intervertebral disk disorders, other back problems 3.1 2.5 39.4

11 2 Septicemia (except in labor) 2.7 2.1 41.5

12 127 COPD and bronchiectasis 2.5 2.0 43.5

13 159 Urinary tract infections 2.3 1.8 45.3

14 131 Respiratory failure; insufficiency; or arrest (adult) 2.1 1.6 46.9

15 157 Acute and unspecified renal failure 1.8 1.5 48.4

16 231 Other fractures 1.8 1.4 49.8

17 96 Heart valve disorders 1.6 1.3 51.1

Total spending 64.8

* The sample contains Medicare beneficiaries at least 65 years of age who do not belong to a health maintenance organization andwho are entitled to both Part A and Part B during the month of their first inpatient admission for that episode and all 6 monthsafter their first admission. Costs are calculated for the same organ system within 180 days after the start of the episode. CCS de-notes Clinical Classification System, COPD chronic obstructive pulmonary disease, and STD sexually transmitted disease.




8/2/2019 NEJMp1113361

3/3


PERSPECTIVE

1077

loskeletal conditions (hip and legfractures). This is not surprising,since the usual candidates forbundled payments are episodesof care for hip fractures and jointreplacements3 and for cardiaccare.4 Individual cancers did not

make the top 17 (colon cancerbeing the most costly, at number28), though cancer as a whole isvery costly.

A cost-distribution analysis forthe 17 most expensive episodesshows that the initial inpatientadmission accounted for 60% ofspending. Subsequent admis-sions (i.e., readmissions relatedto the same organ system) ac-counted for 23%, with separately

billed physician services (i.e.,procedures and evaluation andmanagement) totaling another10% and non-inpatient imagingrepresenting 1%. The remainderof spending was for durable medi-cal equipment, home health care,hospice care, and other items.

By averaging costs for eachcondition in the 306 hospital re-ferral regions, we also found thatspending varies widely from re-gion to region. For each condi-tion, some regions are high-cost, whereas others are low-cost. Totranslate this variation into po-tential dollars saved, we simulat-ed what the spending would be ifcosts in each area in which theaverage was above the 25th per-centile were brought down to the25th-percentile level a featthat might be accomplished if,

for example, Medicare paid a bun-dled rate for episodes and cappedit at the level of the 25th-percen-tile areas. If it did so, the reduc-tion in costs for these 17 con-ditions would be $10 billionannually. If the cap were set atthe 50th-percentile level, the sav-

ings would be $4.7 billion. If suchcaps were applied to all 245 epi-sode types, the annual savingsgained from reducing spendingto the 25th-percentile level wouldtotal $29 billion; reducing it tothe 50th-percentile level would

save $15 billion.To estimate the amounts of

patient-based global payments,we determined the average costsper patient in each of the 306hospital referral regions, againadjusting for regional price differ-ences. Setting the patient-basedglobal payment at the level of aver-age spending in the 25th-percen-tile regions would save $35 billionnationally. If spending were set

at the 50th-percentile level, thesavings would be $18.2 billionnationally.

These results suggest that anepisode-based bundled-paymentsystem could save nearly as muchas a patient-based bundled-pay-ment system. Episode-based bun-dled payments would save 83%of the amount that would be saved with the use of a patient-basedbundling system if the 25th-per-centile standard were used and82% if the 50th-percentile stan-dard were used. The reason forthis similarity is that there can besubstantial heterogeneity in spend-ing for different types of episodeswithin a given region. Some re-gions have high spending for cer-tain conditions even if they havelow spending overall. Episode-based payment encourages effi-

ciency in treating the conditionson which spending is high, re-gardless of whether the region asa whole is low-cost. Patient-basedpayment, by contrast, achieves noadditional savings if the regionas a whole is not high-cost.

In summary, our results sug-

gest that it is possible to achievevery substantial health care sav-ings by moving from a fee-for-service model to bundled pay-ments for episodes of care, whether in a stand-alone pro-gram or as a component of an

overall global-payment model. Theprimary limitation of our analy-sis is that we havent accountedfor heterogeneity in the complex-ity of disease within episode typesthat may affect the average costsin a region. We suspect that such variations in average complexityare small,5 however, so that ourresults are an accurate estimateof cost savings from episode-based bundled payments.

Disclosure forms provided by the authorsare available with the full text of this articleat NEJM.org.

From the Department of Economics, Har-vard University (D.M.C.); and the NationalBureau of Economic Research (D.M.C.,K.G.) both in Cambridge, MA.

1. Nelson L. Lessons from Medicares dem-onstration projects on disease management,care coordination, and value-based payment.Washington, DC: Congressional BudgetOffice, January 2012 (http://www.cbo.gov/

doc.cfm?index=12663).2. Skinner JS, Gottlieb DJ, Carmichael D.A new series of Medicare expenditure mea-sures by hospital referral region: 2003-2008.Lebanon, NH: The Dartmouth Institutefor Health Policy and Clinical Practice, June21, 2011 (http://www.dartmouthatlas.org/downloads/reports/PA_Spending_Report_0611.pdf).3. Sood N, Huckfeldt PJ, Escarce JJ,Grabowski DC, Newhouse JP. Medicaresbundled payment pilot for acute and post-acute care: analysis and recommendationson where to begin. Health Aff (Millwood)2011;30:1708-17.4. Cromwell J, Dayhoff DA, Thoumaian AH.

Cost savings and physician responses toglobal bundled payments for Medicare heartbypass surgery. Health Care Financ Rev 1997;19:41-57.5. Miller DC, Gust C, Dimick JB, BirkmeyerN, Skinner J, Birkmeyer JD. Large variationsin Medicare payments for surgery highlightsavings potential from bundled payment pro-grams. Health Aff (Millwood) 2011;30:2107-15.Copyright 2012 Massachusetts Medical Society.

Cost Savings through Bundled Episode Payments




Documents

NEJMp1113361