the relationship between hospital admission rates and rehospitalizations

7/27/2019 The Relationship Between Hospital Admission Rates and Rehospitalizations

1/9

T he n e w e n g l a n d j o u r n a l o f me dicine

n engl j med 365;24 nejm.org december 15, 2011 2287

special article

The Relationship between Hospital

Admission Rates and RehospitalizationsArnold M. Epstein, M.D., Ashish K. Jha, M.D., M.P.H.,

and E. John Orav, Ph.D.

From the Department of Health Policy andManagement, Harvard School of Public

Health (A.M.E., A.K.J.); the Division ofGeneral Medicine, Brigham and Wom-ens Hospital (A.M.E., A.K.J., E.J.O.); andthe Veterans Affairs Boston HealthcareSystem (A.K.J.) all in Boston. Addressreprint requests to Dr. Epstein at the De-partment of Health Policy and Manage-ment, Harvard School of Public Health,677 Huntington Ave., Boston, MA 02115,or at [email protected].

N Engl J Med 2011;365:2287-95.Copyright 2011 Massachusetts Medical Society.

A b s t ra c t

Background

Efforts to reduce hospital readmissions have focused primarily on improving transi-tional care. Yet variation in readmission rates may more closely reflect variation in the

underlying hospitalization rates than differences in the quality of care during andafter discharge.

Methods

We used national Medicare data to calculate, for each local hospital referral region(HRR), the 30-day, 60-day, and 90-day readmission rates among patients dischargedwith congestive heart failure or pneumonia. We also calculated population-based all-cause admission rates among Medicare enrollees in each HRR. We examined the varia-tion in HRR readmission rates that was explained by overall hospitalization ratesversus differences in patients coexisting conditions, quality of discharge planning,physician supply, and bed supply.

Results

HRR readmission rates ranged from 11 to 32% for congestive heart failure and from8 to 27% for pneumonia. In univariate analyses, all-cause admission rates accountedfor the highest proportion of regional variation in readmission rates for congestiveheart failure (28%, 34%, and 37% at 30, 60, and 90 days, respectively); the next high-est proportions were explained by case mix (11%, 15%, and 18%) and the number ofcardiologists per capita (12%, 14%, and 15%). Results for pneumonia were similar,except that the number of pulmonologists per capita accounted for a lower proportionof the variation (6%, 8%, and 7%, respectively). In multivariate analyses, admissionrates accounted for 16 to 24% of the variation for congestive heart failure and 11 to20% for pneumonia; no other factor accounted for more than 6%.

Conclusions

We found a substantial association between regional rates of rehospitalization andoverall admission rates. Programs directed at shared savings from lower utilizationof hospital services might be more successful in reducing readmissions than pro-grams initiated to date. (Funded by the Commonwealth Fund.)

The New England Journal of Medicine

Downloaded from nejm.org by Ratna Putri on September 13, 2013. For personal use only. No other uses without permission.

Copyright 2011 Massachusetts Medical Society. All rights reserved.


2/9

Th e n e w e n g l a n d j o u r n a l o f me dicine

n engl j med 365;24 nejm.org december 15, 20112288

Unplanned readmissions after hos-

pitalization are costly and ref lect subop-timal patient outcomes. Policymakers have

focused on reducing readmissions as a way to bothlower costs and improve outcomes. Evidence ofsuboptimal care at hospital discharge and shortlythereafter1,2 has prompted clinical interventions to

improve discharge planning, ensure timely follow-up, and improve transitional care. Although somesingle-site efforts have reduced readmissions andthe cost of care,3-7 there is little evidence that thesesorts of clinical interventions can be effectively de-livered more broadly.8-10 Our own work has showna weak relationship between publicly reportedmeasures of discharge planning and readmissionrates.11

The severity of illness and the presence or ab-sence of coexisting conditions affect the likelihoodthat an individual patient will be rehospital-

ized.12,13 Readmission rates among black patientsare higher than those among other patients,14,15perhaps because of strained resources, coexistingconditions of greater severity, or poorly coordi-nated care. Yet, whether variations in these fac-tors explain regional variations in readmissionrates is unclear. Variations in the number of phy-sicians or hospital beds in a community mightexplain variations in readmission rates, althoughthe association between these factors and ratesof utilization has been inconsistent.16-19 Finally,the underlying hospitalization rate may explainvariations in readmission rates.20 For example,the likelihood that a patient is admitted for aclinical condition may depend in large part onpractice patterns in the community and mayplay a substantial role in whether a recently dis-charged patient is readmitted. However, there isalmost no empirical evidence to support or re-fute this notion.

In this study, we examined regional variation inreadmission rates among patients initially hospi-talized for congestive heart failure or pneumonia.

We sought to determine the degree to which re-gional variations in readmission rates are ex-plained by variations in population-based rates ofadmission for all medical and surgical conditions,variations in patients coexisting conditions andrace or ethnic group, variations in measures ofdischarge planning, and variations in numbers ofphysicians and beds.

Methods

Data and Study Variables

We used the 2008 Medicare Provider Analysis and

Review (MEDPAR) file to identify all Medicare

beneficiaries 65 years of age or older who were

discharged alive between January 1, 2008, and

June 30, 2008, with a principal diagnosis of con-gestive heart failure (International Classification of

Diseases, 9th Revision [ICD-9] codes 398.91, 404.x1,

404.x3, and 428.0 through 428.9) or pneumonia

(ICD-9 codes 480 through 486). We linked Medi-

care enrollment files to ascertain each patients

residence ZIP Code and assign that person to one

of 306 hospital referral regions (HRRs). Such re-

gions are based on travel for tertiary care and

have been previously described by the Dartmouth

Atlas Project.21

We obtained data from the Dartmouth Atlas on

the number of acute care hospital beds, number ofprimary care physicians, and number of cardiolo-gists and pulmonologists, adjusted for the size ofthe population across HRRs. We used the Amer-ican Hospital Association annual survey to ascer-tain information on each hospitals profit status,number of beds, location (geographic region andurban vs. rural), and membership in the Councilof Teaching Hospitals.

To assess discharge planning, we used datafrom the 2008 Hospital Consumer Assessment ofHealthcare Providers and Systems (HCAHPS) pro-gram. The HCAHPS survey reports results frompatients about the adequacy of their dischargeplanning22 on the basis of two questions: Duringthis hospital stay, did doctors, nurses, or otherhospital staff talk with you about whether youwould have the help you needed when you leftthe hospital? and During this hospital stay, didyou get information in writing about what symp-toms or health problems to look out for after youleft the hospital? For each hospital, the Centersfor Medicare and Medicaid Services (CMS) calcu-

lates the proportion of patients who answered yesto both questions. The sampling of patients andcorrections for mode of survey administration andnonresponse bias have been described previous-ly.23-25 We developed HRR-level estimates for dis-charge planning by weighting hospitals scores byeach hospitals number of discharges of Medicarebeneficiaries during 2008. The 3763 hospitals





3/9

Hospital Admission Rate s and Rehospitalizations


with HCAHPS data account for 96.4% of inpatientcare for congestive heart failure and 93.1% forpneumonia.

We calculated the all-cause admission rate ineach HRR as the ratio of the total number ofmedical and surgical hospital admissions of Medi-care enrollees in a given HRR during 2007 and

2009 to the total number of enrollees residing inthe HRR during those time periods.

Predicted Rates of Readmission as a Measure

of Case Mix

We used MEDPAR data to calculate each patientslikelihood of readmission within 30, 60, and 90days after discharge on the basis of the patientscharacteristics, including age, sex, self-reportedrace and ethnic group, and the presence or absenceof up to 29 coexisting conditions in the Elixhauseradjustment scheme.26 We summed these likeli-

hoods to calculate the mean predicted probabilityof readmission for patients hospitalized in theHRR. To identify the unit of observation, we usedthe same approach as the CMS, including all hos-pitalizations for congestive heart failure (or pneu-monia) during the 6-month period except thosethat occurred within 30 days after an index admis-sion. A total of 91% of patients with congestiveheart failure and 97% of patients with pneumoniahad only one index hospitalization during the pe-riod. The Elixhauser adjustment, initially developedto predict mortality rates,26 has good predictivevalidity for readmissions.11

Statistical Analysis

We examined the variation in rates of readmission

within 30 days after discharge across HRRs and

identified the five most frequent principal diag-

noses on readmission among the HRRs across the

four quartiles of readmission rates. For ease of pre-

sentation, we combined the middle two quartiles.

We used chi-square tests and analysis of variance

to compare structural characteristics (e.g., number

of beds) and patient characteristics (e.g., race orethnic group) at hospitals across the four quartiles

of HRR readmission-rate category and admission-

rate category.

We examined potential predictors of readmis-sion, including case mix (the HRR mean predictedprobability of readmission), discharge planning(mean responses to the HCAHPS survey), key

supply-side variables (population-based rates ofprimary care physicians, cardiologists [for theanalysis of congestive heart failure], and pulmon-ologists [for the pneumonia analysis] and numberof hospital beds), and all-cause admission rates.Analysis of variance was used to compare eachpredictor across three categories of HRRs: lowest

quartile of readmission rates, combined middletwo quartiles, and highest quartile of readmis-sion rates.

Next, we used two approaches to assess thedegree to which each predictor explained varia-tions in readmission rates. First, we used the co-efficient of determination (r2) to quantify theamount of variance explained by each predictorindividually with the use of a simple linear regres-sion, with the HRR readmission rate as the depen-dent variable. Second, we built multivariable mod-els that included all predictors and calculated the

reduction in explained variation after sequentialremoval of each predictor from the full model. Thelatter analysis provides an estimate of the amountof additional variance explained by that variableand not explained by any of the other predictorstogether. In these analyses, we included as addi-tional covariates the variables listed in Table 1.

We performed sensitivity analyses. To reducethe chance of exaggerating our estimate of varia-tion because of small numbers in some HRRs, werepeated our analyses of variation with a Bayesianpredicted rate that was calculated with the use ofhierarchical generalized linear models.

Ideally, we would have focused our analyses onunplanned, preventable rehospitalizations. How-ever, there is no consensus on how to identifypreventable hospitalizations. As a second sensitiv-ity analysis, we repeated our models but excludedreadmissions with a primary diagnosis that indi-cated admission for chemotherapy or surgery per-formed on a nonurgent basis.

We were concerned that higher all-cause admis-sion rates may primarily be a function of higher

rates of readmission. We therefore calculated, foreach HRR, the admission rate with the exclusionof all readmissions (i.e., admissions for patientswho were discharged in the prior 30 days). We thencorrelated the all-cause admission rate of eachHRR with the all-cause admission rate that ex-cluded readmissions and found a correlation of0.99, suggesting that variations in readmission





4/9



rates were not causing substantial variations inadmission rates.

Finally, to illustrate the potential reduction in

readmission rates that might be achieved if all-cause admission rates were lowered to the averagein the quintile of regions with the lowest admis-sion rates, we used multivariable regression toquantify the magnitude of the association betweenall-cause admission rates and rates of readmission.We then projected rates of readmission in eachHRR on the assumption that the all-cause admis-sion rates equaled the average admission rate inthe lowest quintile. For example, if the averageadmission rate in that lowest quintile was 10% and

our regression showed that a decrease of 1 per-centage point in the admission rate was associatedwith a reduction of 2 percentage points in the

readmission rate, then we projected that an HRRthat had an initial admission rate of 12% wouldimprove to 10%, therefore lowering its readmissionrate by 4 percentage points.

Results

Characteristics of HRRs

The 306 HRRs included 4432 hospitals and234,477 discharges between January 1 and July30, 2008, with a principal diagnosis of congestive

Table 1. Hospital and Patient Characteristics of Hospital Referral Regions (HRRs), According to the Quartile of 30-DayReadmission Rates after the Index Hospitalization for Congestive Heart Failure.*

Characteristic

Highest Quartileof Readmission Rates

(N = 76)

Middle Two Quartilesof Readmission Rates

(N = 154)

Lowest Quartileof Readmission Rates

(N = 76) P Value

Hospitals

No. of beds %


5/9



heart failure and 4497 hospitals and 237,025 dis-charges with a principal diagnosis of pneumonia.Readmission rates after the index hospitalizationfor congestive heart failure varied by a factor ofalmost three across HRRs, from 11% to 32% (Fig.1) and to a similar degree for pneumonia (8% to27%) (Fig. 2). In analyses that used hierarchical

models, readmission rates ranged from 19 to 31%for congestive heart failure and from 13 to 25% forpneumonia.

The 76 HRRs in the highest quartile of read-mission rates for congestive heart failure had, onaverage, a readmission rate of 28%, whereas thosein the lowest quartile had an average readmissionrate of 20%. HRRs with a high readmission ratewere more likely to have medium-size or largehospitals, be located in the Northeast, and havefor-profit, private nonprofit, urban, and teachinghospitals (Table 1). They also had higher propor-

tions of women, higher proportions of black pa-tients and Hispanic patients, and lower mortalityrates. HRR characteristics with respect to readmis-sion rates for pneumonia are provided in Table 1in the Supplementary Appendix, available with thefull text of this article at NEJM.org. Analogousdata for HRRs stratified according to hospital ad-mission rates are provided in Tables 2A and 2B inthe Supplementary Appendix.

The most common principal diagnoses for thepatients readmitted in high-readmission HRRswere nearly identical to the principal diagnoses forpatients readmitted in low-readmission HRRs (Ta-bles 3A and 3B in the Supplementary Appendix).

Predictors of Rehospitalization

We found that the case mix (as reflected by the

predicted probability of readmission) was similar

across the groups of HRRs, although the small

overall difference among HRRs was significant

(Table 2). HRRs with the highest readmission rates

had lower performance on the HCAHPS discharge-

planning metrics (e.g., for congestive heart failure,

78.7% of patients received discharge planning inthe highest quartile of readmission rates vs. 81.6%

in the lowest quartile; P


6/9



potential diminished substantially in multivari-able models.

The results for pneumonia were qualitativelysimilar, except that the number of pulmonologistsper 100,000 population had a somewhat smallereffect on the rate of readmission after pneumoniathan did the number of cardiologists on the rate ofreadmission after congestive heart failure (Table3). The univariate correlations between all-causeadmission rates and readmission rates rangedfrom 48 to 59%, with multivariate correlations

ranging from 32 to 45%.

Sensitivity Analyses

When we examined the association of predictor

variables with readmission rates on the basis of

Bayesian hierarchical models, we found patterns

that were qualitatively similar to those described

above (Tables 4A and 4B in the Supplementary

Appendix). In analyses that excluded readmissions

for chemotherapy and nonurgent surgery, the re-

sults were again similar (Tables 5A and 5B in the

Supplementary Appendix).

Potential Effect of Reduced All-Cause

Admission Rates on Readmission Rates

On the basis of our data from the f irst 6 months

of 2008, we estimate that each year there are ap-

proximately 115,568 readmissions within 30 days

after discharge among patients initially hospital-

ized for congestive heart failure and 84,854 read-

missions among those initially hospitalized for

pneumonia. We found that if all-cause admissionrates for the HRRs in the upper quintiles of hospi-

tal utilization were reduced to the rate in the lowest

quintile, the readmission rate for congestive heart

failure would be reduced from 24.6% to 21.2%,

eliminating approximately 16,166 (14.0%) of the

readmissions for Medicare beneficiaries. For pneu-

monia, the readmission rate would be reduced

from 17.9% to 15.5%, eliminating 11,434 (13.5%)

of the readmissions.

Table 2. Predictors of Readmission within 30 Days after the Index Hospitalization for Congestive Heart Failure or Pneumonia,According to HRR Quartile of Readmission Rates.*

Predictor

Highest Quartileof Readmission

Rates(N = 76)

Middle Two Quartilesof Readmission

Rates(N = 154)

Lowest Quartileof Readmission

Rates(N = 76) P Value

Congestive heart failure

Case mix: predicted probability of readmission % 24.8 24.6 24.4


7/9



Discussion

Our study has three notable findings: First, con-

sistent with the results of previous studies, there

were large variations in rates of readmission across

HRRs, although the conditions for which patients

were readmitted were similar in regions with high

rates and those with low rates. Second, although

greater severity of coexisting conditions was associ-

ated with higher regional rates of readmission,

differences in coexisting conditions were probablynot the primary cause of regional differences in

rates. Finally, all-cause admission rates were strong

predictors of regional differences in readmission,

with univariate correlations ranging from 52 to

61% (depending on the time period) for conges-

tive heart failure and from 48 to 59% for pneumo-

nia. Correlations in multivariate analyses ranged

from 40 to 48% for congestive heart failure and 32

to 45% for pneumonia.

The quality of transitional care is in need ofsubstantial improvement1: many elderly patientsfail to receive any ambulatory care before readmis-sion,12 and only a minority of primary care physi-cians report receiving key discharge informationabout their recently hospitalized patients.1 Suchfindings have prompted policymakers to focus ontransitions in care between the hospital and theambulatory setting as a way to reduce readmis-sions. Although these efforts probably prevent

some readmissions, our findings underscore theimportance of the general use of hospital care andsuggest that policy initiatives such as creating ashared savings program with an accountable careorganization might be effective in lowering ratesof readmission as well. These initiatives should bewatched closely to see whether they could be animportant complement to targeted clinical inter-ventions directed at transitional care.

We found that more severe coexisting condi-

Table 3. Amount of Variance across HRRs in Readmission Rates after Index Hospitalization for Congestive Heart Failure or PneumoniaExplained by Different Predictors.*

Predictor30 Days after

Index Discharge60 Days after

Index Discharge90 Days after

Index Discharge

UnivariateAnalysis

MultivariateAnalysis

UnivariateAnalysis


UnivariateAnalysis


percent of variance in readmission rates explained

Congestive heart failure

Case mix 11.0 2.6 15.0 2.4 17.8 3.1

Discharge planning 10.5 1.0 12.9 1.0 12.9 1.0

HRR-level supply variables

PCPs per 100,000 population 0.1 0.5 0.3 0.7 0.4 0.6

Cardiologists per 100,000 population 11.6 0.8 13.6 1.0 15.4 1.9

Hospital beds per 1000 population 5.6 0.7 8.1 0.2 8.9 0.4

All-cause admission rate 27.5 16.0 33.5 20.4 37.4 23.6

Pneumonia

Case mix 11.2 1.2 10.3 2.2 11.4 2.5

Discharge planning 10.0 4.0 11.4 5.2 11.9 6.0

HRR-level supply variables

PCPs per 100,000 population 0.5 1.9 0.1 1.2 0.1 2.8

Pulmonologists per 100,000 population 5.7 0.9 7.7 0.0 7.3 0.1

Hospital beds per 1000 population 3.3 0.7 5.2 1.0 6.0 1.5

All-cause admission rate 22.8 10.5 31.3 16.2 35.0 20.3

* The multivariate model includes case mix, discharge planning, number of primary care physicians per 100,000 population, number of spe-cialists (cardiologists or pulmonologists) per 100,000 population, number of hospital beds per 1000 population, all-cause admission rate,hospital size (


8/9



tions, lower performance on discharge planning,and greater numbers of cardiovascular specialistswere associated with higher rates of readmissionfor congestive heart failure. These relationshipsmay be evidence of greater clinical need, lowerquality of transitional care, and induced demand,respectively. The relationship between readmis-

sions for pneumonia and the number of pulmo-nary specialists was not as strong as that betweenreadmissions for congestive heart failure and thenumber of cardiologists. Perhaps cardiologists aremore likely to provide care for patients hospital-ized for congestive heart failure than are pulmo-nary specialists for patients hospitalized for pneu-monia.

The existing literature on rehospitalization fo-cuses primarily on the likelihood of readmissionfor individual patients or at particular hospitalsrather than on the reasons for interregional dif-

ferences. For example, patient characteristics, in-cluding age, sex, socioeconomic status, severity ofillness, coexisting conditions, and race, help ex-plain differences in rehospitalization rates.15,20,27-36Prior studies have shown that the proportion ofpatients who are members of minority groups,15effectiveness of discharge planning,6,7,10,11,14,37-45case volume,46,47 and number of physicians orhospital beds in the community16-19 are associ-ated with rates of rehospitalization at the hospitallevel, although the literature is inconsistent.11,15,47-51

Our study has several limitations. Our assess-ment of discharge planning comes from patientssurvey responses and may not capture all impor-tant aspects of discharge planning and care co-ordination. Ideally, we would have focused ouranalyses on unplanned, preventable rehospital-

izations, but there is no consensus on how toidentify this group. Some readmissions may evenreflect more efficient care (e.g., earlier dischargewith a slightly increased risk of readmission), al-though we were unable to identify and excludethem. Our primary predictor was the use of hos-pital services, and we think this measure largely

reflects the propensity to hospitalize a patient.However, there may be elements of socioeconomicstatus, clinical severity of illness, and poor-qualitycare that contribute as well, and we could notquantify the contributions of the different compo-nents. When we examined the effect of case mixon readmissions, our data on coexisting conditionswere limited to information contained on claims;with more detailed information on disease sever-ity and functional status, case mix might have hada greater effect.

In summary, we found a substantial association

between regional rates of rehospitalization andoverall admission rates. Although most interven-tions designed to reduce readmissions thus farhave focused on better disease management andthe coordination of care, our results underscorethe importance of policy efforts directed at reduc-ing the general incentives to use hospital ser-vices. Programs that provide shared savings withan accountable clinical entity or that structurepayment incentives so that they are closer to thatof capitation may be more successful in reducingreadmissions.

Supported by the Commonwealth Fund.Disclosure forms provided by the authors are available with

the full text of this art icle at NEJM.org.We thank Steve Jencks, M.D., for his helpful comments on an

earlier draft of the manuscript; and Jie Zheng, Ph.D., for dataanalysis and programming assistance.

References

1. Bodenheimer T. Coordinating care a perilous journey through the health caresystem. N Engl J Med 2008;358:1064-71.2. Coleman EA, Berenson RA. Lost intransition: challenges and opportunitiesfor improving the quality of transitionalcare. Ann Intern Med 2004;141:533-6.3. Lorig KR, Sobel DS, Stewart AL, et al.Evidence suggesting that a chronic diseaseself-management program can improvehealth status while reducing hospitaliza-tion: a randomized trial. Med Care 1999;37:5-14.4. Naylor MD, Brooten D, Campbell R, etal. Comprehensive discharge planning andhome follow-up of hospitalized elders: arandomized clinical trial. JAMA 1999;281:613-20.

5. Naylor MD, Brooten DA, CampbellRL, Maislin G, McCauley KM, SchwartzJS. Transitional care of older adults hospi-talized with heart failure: a randomized,controlled trial. J Am Geriatr Soc 2004;52:675-84. [Erratum, J Am Geriatr Soc2004;52:1228.]6. Rich MW, Beckham V, Wittenberg C,Leven CL, Freedland KE, Carney RM. Amultidisciplinary intervention to preventthe readmission of elderly patients withcongestive heart failure. N Engl J Med1995;333:1190-5.7. Phillips CO, Wright SM, Kern DE,Singa RM, Shepperd S, Rubin HR. Com-prehensive discharge planning with post-discharge support for older patients withcongestive heart failure: a meta-analysis.

JAMA 2004;291:1358-67. [Erratum, JAMA2004;292:1022.]8. Peikes D, Chen A, Schore J, Brown R.Effects of care coordination on hospital-ization, quality of care, and health careexpenditures among Medicare beneficia-ries: 15 randomized trials. JAMA2009;301:603-18.9. An analysis of the literature on dis-ease management programs. In: Holtz-Eakin D. Letter to the Honorable DonNickels. Washington, DC: CongressionalBudget Office, October 13, 2004 (http://www.cbo.gov/ftpdocs/59xx/doc5909/10-13-DiseaseMngmnt.pdf).10. McAlister FA, Lawson FM, Teo KK,Armstrong PW. A systematic review ofrandomized tria ls of disease management





9/9



programs in heart failure. Am J Med2001;110:378-84.11. Jha AK, Orav EJ, Epstein AM. Publicreporting of discharge planning and ratesof readmissions. N Engl J Med 2009;361:2637-45.12. Jencks SF, Williams MV, Coleman EA.Rehospitalizations among patients in theMedicare fee-for-service program. N Engl

J Med 2009;360:1418-28. [Erratum, N EnglJ Med 2011;364:1582.]13. Smith DM, Giobbie-Hurder A, Wein-berger M, et al. Predicting non-electivehospital readmissions: a multi-site study.J Clin Epidemiol 2000;53:1113-8.14. Hernandez AF, Greiner MA, FonarowGC, et al. Relationship between early phy-sician follow-up and 30-day readmissionamong Medicare beneficiaries hospitalizedfor hear t failure. JAMA 2010;303:1716-22.15. Joynt KE, Orav EJ, Jha AK. Thirty-dayreadmission rates for Medicare beneficia-ries by race and site of care. JAMA 2011;305:675-81.16. Supply-sensitive care. Lebanon, NH:

Dartmouth Atlas of Health Care, 2011(http://www.dartmouthatlas.org/keyissues/issue.aspx?con=2937).17. Fisher ES, Wennberg JE, Stukel TA, etal. Associations among hospital capacity,utilization, and mortality of US Medicarebeneficiaries, controlling for sociodemo-graphic factors. Health Serv Res 2000;34:1351-62.18. Wennberg JE, Fisher ES, Skinner JS.Geography and the debate over Medicarereform. Health Aff (Millwood) 2002;SupplWeb Exclusives:W96-W114.19. Zuckerman S, Waidmann T, BerensonR, Hadley J. Clarifying sources of geo-graphic differences in Medicare spend-

ing. N Engl J Med 2010;363:54-62.20. Fisher ES, Wennberg JE, Stukel TA,Sharp SM. Hospital readmission rates forcohorts of Medicare beneficiaries in Bos-ton and New Haven. N Engl J Med 1994;331:989-95.21. The Center for the Evaluative ClinicalSciences, Dartmouth Medical School. TheDartmouth atlas of health care. Chicago:American Hospital Publishing, 1998.22. Jha AK, Orav EJ, Zheng J, Epstein AM.Patients perception of hospital care inthe United States. N Engl J Med 2008;359:1921-31.23. Elliott MN, Edwards C, Angeles J,Hambarsoomians K, Hays RD. Patternsof unit and item nonresponse in theCAHPS Hospital Survey. Health Serv Res2005;40:2096-119.24. OMalley AJ, Zaslavsky AM, ElliottMN, Zaborski L, Cleary PD. Case-mix ad-justment of the CAHPS Hospital Survey.Health Ser v Res 2005;40:2162-81.25. OMalley AJ, Zaslavsky AM, Hays RD,Hepner KA, Keller S, Cleary PD. Explor-atory factor analyses of the CAHPS Hospi-tal Pilot Survey responses across and with-

in medical, surgical, and obstetric services.Health Ser v Res 2005;40:2078-95.26. Elixhauser A, Steiner C, Harris DR,Coffey RM. Comorbidity measures for usewith administrat ive data. Med Care 1998;36:8-27.27. Deswal A, Petersen NJ, Souchek J,Ashton CM, Wray NP. Impact of race onhealth care utilization and outcomes in

veterans with congestive heart failure. J AmColl Cardiol 2004;43:778-84.28. Krumholz HM, Parent EM, Tu N, et al.Readmission after hospitalization forcongestive heart failure among Medicarebeneficiaries. Arch Intern Med 1997;157:99-104.29. Krumholz HM, Chen YT, Wang Y,Vaccarino V, Radford MJ, Horwitz RI. Pre-dictors of readmission among elderly sur-vivors of admission with heart failure. AmHeart J 2000;139:72-7.30. Philbin EF, DiSalvo TG. Prediction ofhospital readmission for heart failure: de-velopment of a simple risk score based onadministrative data. J Am Coll Cardiol

1999;33:1560-6.31. Rathore SS, Masoudi FA, Wang Y, etal. Socioeconomic status, treatment, andoutcomes among elderly patients hospi-talized with heart failure: findings fromthe National Heart Failure Project. AmHeart J 2006;152:371-8.32. Philbin EF, Dec GW, Jenkins PL,DiSalvo TG. Socioeconomic status as anindependent risk factor for hospital read-mission for heart failure. Am J Cardiol2001;87:1367-71.33. Lindenauer PK, Bernheim SM, GradyJN, et al. The performance of US hospitalsas reflected in risk-standardized 30-daymortalit y and readmission rates for Medi-

care beneficiaries with pneumonia. J HospMed 2010;5(6):E12-E18.34. Chin MH, Goldman L. Correlates ofearly hospital readmission or death in pa-tients with congestive heart failure. AmJ Cardiol 1997;79:1640-4.35. Felker GM, Leimberger JD, Califf RM,et al. Risk stratif ication after hospitaliza-tion for decompensated heart failure.J Card Fail 2004;10:460-6.36. Yamokoski LM, Hasselblad V, MoserDK, et al. Prediction of rehospitalizationand death in severe heart fa ilure by physi-cians and nurses of the ESCAPE trial.J Card Fail 2007;13:8-13.37. Ashton CM, Del Junco DJ, Souchek J,Wray NP, Mansyur CL. The associationbetween the quality of inpatient care andearly readmission: a meta-analysis of theevidence. Med Care 1997;35:1044-59.38. Gwadry-Sridhar FH, Flintoft V, LeeDS, Lee H, Guyatt GH. A systematic re-view and meta-analysis of studies com-paring readmission rates and mortalityrates in patients with heart failure. ArchIntern Med 2004;164:2315-20.39. Lapp JM, Muhlestein JB, Lapp DL, et

al. Improvements in 1-year cardiovascularclinical outcomes associated with a hos-pital-based discharge medication pro-gram. Ann Intern Med 2004;141:446-53.40. Jovicic A, Holroyd-Leduc JM, StrausSE. Effects of self-management interven-tion on health outcomes of patients withheart failure: a systematic review of ran-domized controlled trials. BMC Cardio-

vasc Disord 2006;6:43.41. Gonseth J, Guallar-Castilln P, Bane-gas JR, Rodrguez-Artalejo F. The effective-ness of disease management programmesin reducing hospital re-admission in olderpatients with heart failure: a systematicreview and meta-analysis of published re-ports. Eur Heart J 2004;25:1570-95.42. Schneider JK, Hornberger S, Booker J,Davis A, Kralicek R. A medication dis-charge planning program: measuring theeffect on readmissions. Clin Nurs Res1993;2:41-53.43. Stewart S, Pearson S, Horowitz JD. Ef-fects of a home-based intervention amongpatients with congestive heart failure dis-

charged from acute hospital care. ArchIntern Med 1998;158:1067-72.44. Koelling TM, Johnson ML, Cody RJ,Aaronson KD. Discharge education im-proves clinical outcomes in patients withchronic heart failure. Circulation 2005;111:179-85.45. Krumholz HM, Amatruda J, SmithGL, et al. Randomized trial of an educa-tion and support intervention to preventreadmission of patients with heart fail-ure. J Am Coll Cardiol 2002;39:83-9.46. Joynt KE, Orav EJ, Jha AK. The asso-ciation between hospital volume and pro-cesses, outcomes, and costs of care forcongestive heart failure. Ann Intern Med

2011;154:94-102.47. Auerbach AD, Hilton JF, Maselli J, Pe-kow PS, Rothberg MB, Lindenauer PK. Shopfor quality or volume? Volume, quality, andoutcomes of coronary artery bypass sur-gery. Ann Intern Med 2009;150:696-704.48. Auerbach AD, Maselli J, Carter J, Pe-kow PS, Lindenauer PK. The relationshipbetween case volume, care quality, andoutcomes of complex cancer surgery. J AmColl Surg 2010;211:601-8.49. Krumholz HM, Merrill AR, SchoneEM, et al. Patterns of hospital perfor-mance in acute myocardial infarction andheart failure 30-day mortality and read-mission. Circ Cardiovasc Qual Outcomes2009;2:407-13.50. Joynt KE, Jha AK. Who has higher re-admission rates for heart failure, andwhy? Implications for ef forts to improvecare using financial incentives. Circ Car-diovasc Qual Outcomes 2011;4:53-9.51. Hummel SL, Pauli NP, Krumholz HM,et al. Thirty-day outcomes in Medicare pa-tients with heart fai lure at heart transplantcenters. Circ Heart Fail 2010;3:244-52.Copyright 2011 Massachusetts Medical Society.



the relationship between hospital admission rates and rehospitalizations

Documents