[Skip to Content]
[Skip to Content Landing]
Figure.
Admission-Level Analysis of Differential Changes in Skilled Nursing Facility (SNF) Spending in the Postcontract Period by Accountable Care Organization (ACO) Subgroup and Cohort
Admission-Level Analysis of Differential Changes in Skilled Nursing Facility (SNF) Spending in the Postcontract Period by Accountable Care Organization (ACO) Subgroup and Cohort

Differential changes in SNF spending in the postacute care period are displayed by ACO subgroup, with more negative estimates indicating greater reductions in spending. Displayed subgroup estimates are for the pooled 2013-2014 postcontract period and are provided separately for the 2012 and 2013 entry cohorts of ACOs in the Medicare Shared Savings Program. Error bars indicate 95% CI. Overall, differential reductions in SNF spending for independent physician groups (N = 128; 69 in the 2012 entry cohort, 59 in the 2013 entry cohort; differential change pooled across cohorts and 2013 and 2014, −$114; P = .02) and organizations financially integrated with hospitals (N = 92; 45 in the 2012 entry cohort, 47 in the 2013 entry cohort; differential change, −$67; P = .003) were similar (P = .38 for subgroup difference in estimates).

Table 1.  
Differential Changes in Patient Characteristics From the Precontract Period to 2014 for ACO-Attributed Patients vs Control Groupa
Differential Changes in Patient Characteristics From the Precontract Period to 2014 for ACO-Attributed Patients vs Control Groupa
Table 2.  
Patient-Level Analysis: Acute and Postacute Care Spending and Mortality for 2012 Entry Cohort of MSSP ACOs vs Control Group
Patient-Level Analysis: Acute and Postacute Care Spending and Mortality for 2012 Entry Cohort of MSSP ACOs vs Control Group
Table 3.  
Admission-Level Analysis Among Hospitalized Patients: Postacute Care Spending and Use for 2012 Entry Cohort of MSSP ACOs vs Control Group
Admission-Level Analysis Among Hospitalized Patients: Postacute Care Spending and Use for 2012 Entry Cohort of MSSP ACOs vs Control Group
Table 4.  
SNF Stay-Level Analysis Among Patients Discharged to a SNF: Spending, Length of Stay, Readmissions, and Proportion Discharged to Highly Rated Facilities for 2012 Entry Cohort of MSSP ACOs vs Control Group
SNF Stay-Level Analysis Among Patients Discharged to a SNF: Spending, Length of Stay, Readmissions, and Proportion Discharged to Highly Rated Facilities for 2012 Entry Cohort of MSSP ACOs vs Control Group
1.
Institute of Medicine Committee on Geographic Variation in Health Care Spending and Promotion of High-Value Care.  Variation in Health Care Spending: Target Decision Making, Not Geography. Washington, DC: The National Academies Press; 2013.
2.
Medicare Payment Advisory Commission. A data book: health care spending and the Medicare program. http://www.medpac.gov/docs/default-source/data-book/june-2016-data-book-health-care-spending-and-the-medicare-program.pdf. Published June 2016. Accessed January 10, 2017.
3.
Ackerly  DC, Grabowski  DC.  Post-acute care reform—beyond the ACA.  N Engl J Med. 2014;370(8):689-691.PubMedGoogle ScholarCrossref
4.
McWilliams  JM, Chernew  ME, Landon  BE, Schwartz  AL.  Performance differences in year 1 of pioneer accountable care organizations.  N Engl J Med. 2015;372(20):1927-1936.PubMedGoogle ScholarCrossref
5.
McWilliams  JM, Hatfield  LA, Chernew  ME, Landon  BE, Schwartz  AL.  Early performance of accountable care organizations in Medicare.  N Engl J Med. 2016;374(24):2357-2366.PubMedGoogle ScholarCrossref
6.
Schwartz  AL, Chernew  ME, Landon  BE, McWilliams  JM.  Changes in low-value services in year 1 of the Medicare Pioneer Accountable Care Organization program.  JAMA Intern Med. 2015;175(11):1815-1825.PubMedGoogle ScholarCrossref
7.
Mostashari  F, Sanghavi  D, McClellan  M.  Health reform and physician-led accountable care: the paradox of primary care physician leadership.  JAMA. 2014;311(18):1855-1856.PubMedGoogle ScholarCrossref
8.
McWilliams  JM, Chernew  ME, Zaslavsky  AM, Landon  BE.  Post-acute care and ACOs—who will be accountable?  Health Serv Res. 2013;48(4):1526-1538.PubMedGoogle ScholarCrossref
9.
Colla  CH, Lewis  VA, Bergquist  SL, Shortell  SM.  Accountability across the continuum: the participation of postacute care providers in accountable care organizations.  Health Serv Res. 2016;51(4):1595-1611.PubMedGoogle ScholarCrossref
10.
Colla  CH, Lewis  VA, Tierney  E, Muhlestein  DB.  Hospitals participating in ACOs tend to be large and urban, allowing access to capital and data.  Health Aff (Millwood). 2016;35(3):431-439.PubMedGoogle ScholarCrossref
11.
McWilliams  JM.  Changes in Medicare shared savings program savings from 2013 to 2014.  JAMA. 2016;316(16):1711-1713.PubMedGoogle ScholarCrossref
12.
Lage  DE, Rusinak  D, Carr  D, Grabowski  DC, Ackerly  DC.  Creating a network of high-quality skilled nursing facilities: preliminary data on the postacute care quality improvement experiences of an accountable care organization.  J Am Geriatr Soc. 2015;63(4):804-808.PubMedGoogle ScholarCrossref
13.
McHugh  JP, Trivedi  AN, Zinn  JS, Mor  V.  Post-acute integration strategies in an era of accountability.  J Hosp Adm. 2014;3(6):103-112.PubMedGoogle Scholar
14.
Rahman  M, Foster  AD, Grabowski  DC, Zinn  JS, Mor  V.  Effect of hospital–SNF referral linkages on rehospitalization.  Health Serv Res. 2013;48(6, pt 1):1898-1919.PubMedGoogle ScholarCrossref
15.
Research Data Assistance Center. Shared savings program accountable care organizations (ACO) provider-level RIF. http://www.resdac.org/cms-data/files/ssp-aco-provider-level-rif. Accessed September 26, 2016.
16.
Medicare.gov. Nursing Home Compare. https://www.medicare.gov/NursingHomeCompare/Resources/Downloadable-Database.html. Accessed September 26, 2016.
17.
Zaslavsky  AM, Ayanian  JZ, Zaborski  LB.  The validity of race and ethnicity in enrollment data for Medicare beneficiaries.  Health Serv Res. 2012;47(3, pt 2):1300-1321.PubMedGoogle ScholarCrossref
18.
Agency for Healthcare Research and Quality. Creation of new race-ethnicity codes and socioeconomic status (SES) indicators for Medicare beneficiaries: final report. https://archive.ahrq.gov/research/findings/final-reports/medicareindicators/. Updated August 2012. Accessed January 10, 2017.
19.
Centers for Medicare and Medicaid Services. Chronic Condition Data Warehouse. https://www.ccwdata.org/web/guest/home. Accessed September 26, 2016.
20.
Pope  GC, Kautter  J, Ellis  RP,  et al.  Risk adjustment of Medicare capitation payments using the CMS-HCC model.  Health Care Financ Rev. 2004;25(4):119-141.PubMedGoogle Scholar
21.
Yun  H, Kilgore  ML, Curtis  JR,  et al.  Identifying types of nursing facility stays using Medicare claims data: an algorithm and validation.  Health Serv Outcomes Res Methodol. 2010;10(1):100-110. doi:10.1007/s10742-010-0060-4Google ScholarCrossref
22.
US Census Bureau. 2010 American Community Survey (ACS). http://www.census.gov/programs-surveys/acs/data.html. Updated October 13, 2016. Accessed January 10, 2017.
23.
Binder  DA.  On the variances of asymptotically normal estimators from complex surveys.  Int Stat Rev. 1983;51:279-292. doi:10.2307/1402588Google ScholarCrossref
24.
Mechanic  RE.  When new Medicare payment systems collide.  N Engl J Med. 2016;374(18):1706-1709.PubMedGoogle ScholarCrossref
25.
Centers for Medicare & Medicaid Services, Dept of Health and Human Services. Medicare program; advancing care coordination through episode payment models (EPMs); cardiac rehabilitation incentive payment model; and changes to the comprehensive care for joint replacement model (CJR). Proposed rule. https://s3.amazonaws.com/public-inspection.federalregister.gov/2016-17733.pdf. Accessed September 26, 2016.
26.
Colla  CH, Lewis  VA, Kao  LS, O’Malley  AJ, Chang  CH, Fisher  ES.  Association between Medicare accountable care organization implementation and spending among clinically vulnerable beneficiaries.  JAMA Intern Med. 2016;176(8):1167-1175.PubMedGoogle ScholarCrossref
Original Investigation
Health Care Reform
April 2017

Changes in Postacute Care in the Medicare Shared Savings Program

Author Affiliations
  • 1Department of Health Care Policy, Harvard Medical School, Boston, Massachusetts
  • 2Division of General Internal Medicine and Primary Care, Brigham and Women’s Hospital, Boston, Massachusetts
  • 3Division of Cardiovascular Medicine, Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts
  • 4Department of Health Policy, Vanderbilt University School of Medicine, Nashville, Tennessee
JAMA Intern Med. 2017;177(4):518-526. doi:10.1001/jamainternmed.2016.9115
Key Points

Question  What changes in postacute care have been associated with the Medicare Shared Savings Program?

Findings  A study using fee-for-service Medicare claims found that, for accountable care organizations entering in 2012, participation in the Medicare Shared Savings Program was associated with a 9% differential reduction in postacute spending by 2014, driven by reductions in discharges to facilities, length of facility stays, and acute inpatient care. Reductions were smaller for later entrants and similar for accountable care organizations with and without financial ties to hospitals.

Meaning  Payment models that place hospitals at risk for postacute spending are not the only viable strategy to curb excessive postacute care; therefore, accountable care organizations’ incentives to achieve postacute savings should not be weakened.

Abstract

Importance  Postacute care is thought to be a major source of wasteful spending. The extent to which accountable care organizations (ACOs) can limit postacute care spending has implications for the importance and design of other payment models that include postacute care.

Objective  To assess changes in postacute care spending and use of postacute care associated with provider participation as ACOs in the Medicare Shared Savings Program (MSSP) and the pathways by which they occurred.

Design, Setting, and Participants  With the use of fee-for-service Medicare claims from a random 20% sample of beneficiaries with 25 544 650 patient-years, 8 395 426 hospital admissions, and 1 595 352 stays in skilled nursing facilities (SNFs) from January 1, 2009, to December 31, 2014, difference-in-difference comparisons of beneficiaries served by ACOs with beneficiaries served by local non-ACO health care professionals (control group) were performed before vs after entry into the MSSP. Differential changes were estimated separately for cohorts of ACOs entering the MSSP in 2012, 2013, and 2014.

Exposures  Patient attribution to an ACO in the MSSP.

Main Outcomes and Measures  Postacute spending, discharge to a facility, length of SNF stays, readmissions, use of highly rated SNFs, and mortality, adjusted for patient characteristics.

Results  For the 2012 cohort of 114 ACOs, participation in the MSSP was associated with an overall reduction in postacute spending (differential change in 2014 for ACOs vs control group, −$106 per beneficiary [95% CI, –$176 to –$35], or −9.0% of the precontract unadjusted mean of $1172; P = .003) that was driven by differential reductions in acute inpatient care, discharges to facilities rather than home (−0.6 percentage points [95% CI, –1.1 to 0.0], or −2.7% of the unadjusted precontract mean of 22.6%; P = .03), and length of SNF stays (−0.60 days per stay [95% CI, –0.99 to –0.22], or −2.2% of the precontract unadjusted mean of 27.07 days; P = .002). Reductions in use of SNFs and length of stay were largely due to within-hospital or within-SNF changes in care specifically for ACO patients. Participation in the MSSP was associated with smaller significant reductions in SNF spending in 2014 for the 2013 ACO cohort (–$27 per beneficiary [95% CI, –$49 to –$6], or –3.3% of the precontract unadjusted mean of $813; P = .01) but not in the 2013 or 2014 cohort’s first year of participation (–$13 per beneficiary [95% CI, –$33 to $6]; P = .19; and $4 per beneficiary [95% CI, –$15 to $24]; P = .66). Estimates were similar for ACOs with and without financial ties to hospitals. Participation in the MSSP was not associated with significant changes in 30-day readmissions, use of highly rated SNFs, or mortality.

Conclusions and Relevance  Participation in the MSSP has been associated with significant reductions in postacute spending without ostensible deterioration in quality of care. Spending reductions were more consistent with clinicians working within hospitals and SNFs to influence care for ACO patients than with hospital-wide initiatives by ACOs or use of preferred SNFs.

Introduction

Spending on postacute care accounts for much of the wide geographic variation in Medicare spending and has more than doubled since 2001.1,2 The rapid growth in spending on postacute care has been driven by increasing use of care in skilled nursing facilities (SNFs),2 which are the most common type of postacute facility used and are reimbursed by Medicare on a per diem basis. Excessive use of postacute SNF care is thought to be a major source of wasteful spending and a natural target for health care professionals participating in new payment models that reward them for reducing spending below a budget, such as the Medicare accountable care organization (ACO) programs and Bundled Payments for Care Improvement initiative.3

Through the first full year of operation, evaluations of the Medicare ACO programs—including the Pioneer model and Medicare Shared Savings Program (MSSP)—found proportionately larger savings in spending on SNFs than in other categories, suggesting that ACOs are indeed focusing on postacute care in their strategies to reduce spending.4,5 These early gains are consistent with expectations that wasteful spending is easier to reduce when it is more abundant and with the stronger incentives that ACOs have to limit use of services they do not provide relative to services they do provide.4-7 Specifically, because most ACOs do not include SNFs in their organizations or ACO contract,8,9 shared-savings bonuses achieved by lowering spending on SNF care are typically not offset by foregone fee-for-service profits.

The early reductions in SNF spending also suggest that organizations can successfully influence care in facilities they do not own.4,5,10,11 For example, ACOs can direct patients to a preferred network of more efficient SNFs and can employ, or partner with, clinicians and case managers to follow the care of patients in SNFs.12-14 Research to date, however, has not demonstrated the extent of savings in postacute care in the ACO programs after 2013, the source of these savings, or the strategies used to achieve them.

Methods
Study Overview

In the MSSP, ACOs receive shared-savings bonuses if they hold total Medicare spending sufficiently below a budget (or benchmark) and meet minimum performance standards. Using Medicare claims from January 1, 2009, to December 31, 2014, we conducted difference-in-differences comparisons to estimate changes in spending on postacute care and use of postacute care associated with participation in the MSSP. Specifically, we compared beneficiaries served by MSSP ACOs with beneficiaries in ACO service areas served by nonparticipating health care professionals (control group) before vs after the start of ACO contracts. We separately examined each cohort of ACOs entering the MSSP in 2012, 2013, and 2014, as previous evaluations have demonstrated that savings differ by year of entry.5,11

We conducted 4 sets of analyses. First, to obtain overall population estimates, we analyzed inpatient, postacute care facility, and home health spending among all beneficiaries, with patient-years as the units of analysis. Second, to eliminate changes in the use of postacute care due to changes in hospital admission rates, we analyzed use of postacute care and postacute care spending conditional on hospitalization, with admissions as the units of analysis. Third, to isolate spending reductions owing to shorter length of SNF stays, we analyzed SNF spending and length of stay conditional on discharge to a SNF, with SNF stays as the units of analysis. Fourth, we analyzed 30-day readmissions, use of highly rated SNFs, and mortality to assess associated changes in quality of care.

We then conducted additional analyses to further characterize mechanisms for changes in postacute care. Specifically, we tested whether changes in postacute care for hospitalized ACO patients were the result of hospital-wide changes or changes in care selectively affecting ACO patients within hospitals. In a related subgroup analysis, we compared results between ACOs with and without financial ties to hospitals. Although independent physician groups without such ties might still influence discharge planning for their patients within hospitals,10 ACOs that are financially integrated with hospitals are more likely to have the additional option of implementing hospital-wide strategies. Finally, we assessed the extent to which changes in SNF spending and length of stay were due to the use of different SNFs vs changes in care for ACO patients within SNFs. The study was approved by the Harvard Medical School Committee on Human Studies and the Privacy Board of the Centers for Medicare & Medicaid Services.

Study Population

For each study year from 2009 to 2014, we analyzed data from Medicare claims and enrollment files for a random 20% sample of fee-for-service beneficiaries continuously enrolled in Parts A and B in that year (while alive for decedents) and in the previous year (to assess preexisting conditions). Using previously described methods5 and the Centers for Medicare & Medicaid Services ACO Provider-level Research Identifiable File,15 which defines each ACO as a collection of provider taxpayer identification numbers (TINs) and Centers for Medicare & Medicaid Services Certification Numbers (for safety-net providers),5,15 we attributed each beneficiary in each study year to the ACO or non-ACO taxpayer identification number accounting for the most allowed charges for office visits with a primary care physician during that year (eAppendix in the Supplement). In sensitivity analyses, we modified definitions of ACOs to address potential bias from changes in constituent taxpayer identification numbers and physicians over time (eAppendix in the Supplement). We excluded beneficiaries without a visit with a primary care physician. As this was a secondary data analysis of claims, no primary data were collected and there was no intervention requiring patient consent. Beneficiary-level identifying information was not included in the data files.

Study Variables
Acute and Postacute Care Spending and Use

For analyses of all beneficiaries in our study sample, we assessed annual per-beneficiary Medicare spending for inpatient care, care in postacute facilities (overall and by facility type), and home health care initiated in the community (outpatient) vs care after a hospitalization or postacute facility stay (postacute). Facility types included SNFs, inpatient rehabilitation facilities, and long-term care hospitals.

For admission-level analyses, we assessed the following variables for the postacute period: discharge to a facility vs home, spending for facility stays and number of covered days spent in a facility (by facility type), and spending for home health care and number of days of home health care. We defined the postacute period to begin at hospital discharge and end at the conclusion of any episode of postacute care (inclusive of transfers between facilities and of home health care following hospital or postacute facility stays), at the start of another hospitalization if the patient was hospitalized while receiving postacute care, or at discharge if no postacute care was provided. Admissions with no stay at a postacute facility or home health care were given values of zero for measures of postacute care spending and use during the postacute period. In addition, we determined the diagnosis related group payment weight, inpatient spending, and length of stay for each admission to test whether reductions in admissions achieved by ACOs led to changes in case mix among patients admitted in the postcontract period. For analyses of postacute SNF stays, we assessed spending per stay and length of stay.

Quality of Care

We examined 3 measures to detect potential changes in quality of care associated with participation in the MSSP. First, we assessed annual mortality in the beneficiary-level analysis of the full study sample. Second, for both admission-level and SNF stay–level analyses, we assessed readmissions within 30 days of hospital discharge. We examined readmissions because they are included as a quality measure in ACO contracts and because reductions in readmission rates could mechanically lengthen postacute facility stays and thus influence interpretation of results. Third, for SNF stay–level analyses, we determined the star rating for the SNF to which beneficiaries were discharged from publicly available Nursing Home Compare data.16 The star rating is a composite of ratings from health inspections, 16 quality measures, and staffing, blending both short and long stays. We used Nursing Home Compare data from each study year to create an indicator for being discharged to a 4-star or 5-star (highly rated) facility.

Patient Characteristics

From Master Beneficiary Summary Files for each study year, we assessed age, sex, race/ethnicity,17,18 Medicaid coverage, disability as the original reason for Medicare eligibility, and end-stage renal disease. From the Chronic Conditions Data Warehouse, which draws from diagnoses since 1999 to describe accumulated disease burden,19 we assessed the presence of each of 27 conditions (acute myocardial infarction, Alzheimer disease, Alzheimer disease and related disorders or senile dementia, anemia, asthma, atrial fibrillation, benign prostatic hyperplasia, chronic kidney disease, chronic obstructive pulmonary disease, depression, diabetes, heart failure, hip or pelvic fracture, hyperlipidemia, hypertension, hypothyroidism, ischemic heart disease, osteoporosis, rheumatoid arthritis or osteoarthritis, stroke or transient ischemic attack, breast cancer, colorectal cancer, endometrial cancer, lung cancer, prostate cancer, cataracts, and glaucoma) by the start of each study year. From diagnoses in the preceding year of claims, we calculated a Hierarchical Condition Categories risk score for each beneficiary in each study year.20 We determined whether beneficiaries were long-term residents of a nursing home in the prior year using a validated claims-based algorithm.21 Finally, from US Census Bureau data,22 we assessed area-level sociodemographic characteristics.

ACO Subgroup Comparison

For a prespecified subgroup comparison, we used Centers for Medicare & Medicaid Services descriptions and information on organizations’ websites to categorize ACOs as financially integrated with hospitals (eg, integrated delivery systems and physician-hospital organizations) vs independent physician groups, some of which partner with hospitals in ACO contracts but are not owned by hospitals or health systems. Because baseline levels of efficiency have been predictive of ACO savings,4,5,11 we adjusted this subgroup comparison for differences in spending reductions associated with baseline SNF use (eAppendix in the Supplement).

Statistical Analysis

We fit linear regression models predicting each dependent variable as a function of ACO fixed effects (omitting the control group as the reference group), fixed effects for the beneficiary’s hospital referral region (HRR) of residence, year fixed effects, interactions between HRR and year fixed effects, the patient characteristics listed in Table 1, and indicators for each ACO cohort in each postcontract year. For each ACO cohort, the last set of indicators estimated changes from the precontract period to each postcontract year for beneficiaries attributed to ACOs that differed from local concurrent changes for beneficiaries attributed to non-ACO health care professionals (the estimate of interest). Interactions between these indicators and ACO subgroups were added for subgroup comparisons. The ACO fixed effects adjusted for precontract differences between each ACO and the control group, as well as for changes in the distribution of beneficiaries across ACOs. The fixed effects for HRR by year combinations adjusted for geographical differences between the ACO and control groups and for HRR-specific changes in the use of acute or postacute care in the control group.

Thus, we used precontract levels for ACOs and local concurrent changes from the precontract to postcontract period in the control group to establish counterfactuals that would be expected in the absence of participation in the MSSP, and we estimated differences from this expectation (ie, the differential change or the change attributable to MSSP participation). The precontract period in the model differed by ACO cohort: 2009 to 2011 for the 2012 cohort, 2009 to 2012 for the 2013 cohort, and 2009 to 2013 for the 2014 cohort. Models of admissions and SNF stays additionally included diagnosis related group fixed effects. We used robust variance estimators to account for clustering within ACOs (for the ACO group) or HRRs (for the control group).23

We conducted admission-level analyses with and without fixed effects for each hospital by year combination to determine the extent to which differential changes in postacute care were due to within-hospital changes in care specifically for ACO patients vs hospital-wide changes affecting postacute care for all admitted patients (by comparing estimates from models with and without the fixed effects). We similarly conducted SNF stay–level analyses with and without SNF fixed effects to determine the extent to which differential changes in SNF spending and length of stay were due to discharges of ACO patients to different facilities vs changes in care specifically for ACO patients within SNFs.

As prespecified in our study protocol, we took several measures to limit the number of statistical tests that were conducted for supporting inferences. First, we focused on measures of SNF use and spending as our main outcomes and considered spending on SNF care to be the primary outcome, because such spending was the main source of savings in postacute care in Medicare ACO programs in their first full year of operation.4,5 Second, we conducted admission-level analyses only for ACO cohorts with significant differential changes in SNF spending in beneficiary-level analyses, and we conducted SNF stay–level analyses only for ACO cohorts with significant differential changes in admission-level analyses. Third, we considered tests of differential changes in 2014 as primary, given evidence of growing savings in the MSSP from 2013 and 2014.11 We present 2013 estimates for descriptive purposes, except that we pooled estimates across 2013 and 2014 in subgroup comparisons to enhance precision. We treated 2012 as a transition year for the 2012 ACO cohort, effectively excluding it from both the precontract and postcontract periods, because ACOs in this cohort entered in April or July rather than January.

We conducted additional analyses to explore potential sources of bias. First, we compared trends in each outcome between each ACO cohort and the control group during the precontract period. Similar precontract trends would support our assumption that changes from the precontract to postcontract periods would have been similar for the ACO and control groups in the absence of the ACO programs. Second, we tested whether patient characteristics and inpatient admission characteristics differentially changed from the precontract to postcontract period in the ACO groups relative to the control group.

Results

Beneficiary-level analyses included 25 544 650 patient-years from 2009 to 2014, with approximately 19% of beneficiaries attributed to ACOs annually on average. Admission-level analyses included 8 395 426 admissions and SNF stay–level analyses included 1 595 352 SNF stays. In analyses adjusted only for geography, differences in patient characteristics between ACO-attributed beneficiaries and the control group in the precontract period were mostly small (eTable 1 in the Supplement) and changed minimally from the precontract period to 2014 (Table 1). In analyses of admissions and SNF stays, differential changes in patient characteristics, diagnosis related group payment weights, spending per admission, and admission length of stay also were minimal (Table 1). Adjusted precontract trends in all dependent variables were similar for ACO-attributed beneficiaries and the control group, as were adjusted precontract means for most measures (Tables 2, 3, and 4 and eTables 2 and 3 in the Supplement).

Beneficiary-Level Analysis
2012 Entry Cohort

In comparisons of ACOs entering the MSSP in 2012 with the control group, participation in the MSSP was associated with reductions in total annual inpatient spending (differential change in 2014 for ACOs vs control group, −$77 per beneficiary [95% CI, –$132 to –$22], or −2.3% of the precontract unadjusted mean of $3400; P = .006) and spending on postacute care in facilities (−$106 per beneficiary [95% CI, –$176 to –$35], or −9.0% of the precontract unadjusted mean of $1172; P = .003), with the latter driven mostly by reduced spending on SNF care (Table 2). Participation in the MSSP was also associated with a differential reduction in spending in 2014 for home health care in the outpatient setting (−$16 per beneficiary [95% CI, –$31 to –$1], or −3.1% of the precontract unadjusted mean of $522; P = .03) but not in the postacute care setting ($2 per beneficiary [95% CI, –$2 to $6], or 1.9% of the precontract unadjusted mean of $108; P = .31). In 2013, differential changes for ACOs in the 2012 entry cohort in inpatient spending and SNF spending were smaller than differential changes in 2014 (Table 2).

2013 and 2014 Entry Cohorts

For ACOs entering the MSSP in 2013 (eTable 2 in the Supplement), participation in the MSSP was associated with a significant reduction in SNF spending in 2014 (differential change, −$27 per beneficiary [95% CI, –$49 to –$6], or −3.3% of the precontract unadjusted mean of $813; P = .01) but not with inpatient spending (−$33 per beneficiary [95% CI, –$92 to $27], or −1.0% of the precontract unadjusted mean of $3400; P = .28) and with no significant differential changes in 2013. For the 2014 entry cohort, participation in the MSSP was not associated with a significant change in either inpatient spending (−$8 per beneficiary [95% CI, –$60 to $45], or −0.2% of the precontract unadjusted mean of $3400; P = .78) or SNF spending ($4 per beneficiary [95% CI, –$15 to $24], or 0.5% of the precontract unadjusted mean of $813; P = .66) in 2014, so admission-level and SNF stay–level analyses were not pursued.

Admission-Level Analysis
2012 Entry Cohort

For the 2012 entry cohort of ACOs, participation in the MSSP was associated with significant reductions in 2014 in the proportion discharged to a postacute care facility (differential change in 2014, −0.6 percentage points [95% CI, –1.1 to 0.0], or −2.7% of the precontract unadjusted mean of 22.6%; P = .03), spending on SNF care in the postacute period (−$124 per admission [95% CI, –$200 to –$48], or −5.6% of the precontract unadjusted mean of $2207; P = .002), days in a SNF during the postacute care period (−0.22 per admission [95% CI, –0.41 to –0.04], or −4.1% of the precontract unadjusted mean of 5.34 days; P = .02), and postacute use of long-term care hospitals (Table 3). Estimates were largely unchanged after adjustment for hospital-level means in the year of admission, indicating that reductions in SNF use were due to different treatment of ACO-attributed patients within hospitals.

2013 ACO Entry Cohort

For the 2013 MSSP cohort (eTable 3 in the Supplement), the differential change in 2014 in postacute care SNF spending was not statistically significant (−$50 per admission [95% CI, –$102 to $3], or −2.3% of the precontract unadjusted mean of $2207; P = .07), so SNF stay-level analyses were not pursued.

SNF Stay-Level Analysis

For the 2012 entry cohort of ACOs (Table 4), participation in the MSSP was associated with significant reductions in spending per SNF stay (differential change in 2014, −$393 per stay [95% CI, –$558 to –$229], or −3.5% of the precontract unadjusted mean of $11 206; P < .001) and length of SNF stay (−0.60 days per stay [95% CI, –0.99 to –0.22], or −2.2% of the precontract unadjusted mean of 27.07; P = .002). Adjustment for the SNF to which patients were discharged reduced these estimates by 23% to 25%, indicating that most of these reductions were due to changes in care specifically for ACO patients within SNFs rather than shifts to different SNFs.

Subgroup Analysis

Estimated reductions in inpatient and SNF spending associated with participation in the MSSP were consistently greater in the 2012 and 2013 entry cohorts for independent physician groups than for ACOs that were financially integrated with hospitals, but these differences were not statistically significant (Figure and eFigure in the Supplement).

Quality of Care

Participation in the MSSP was not associated with differential changes in mortality (Table 2 and eTable 2 in the Supplement), 30-day readmissions (Tables 3 and 4 and eTable 3 in the Supplement), or the proportion of patients discharged to 4-star or 5-star SNFs (Table 4).

Discussion

Spending on postacute SNF care for patients of ACOs entering the MSSP in 2012 or 2013 was significantly reduced in 2014 relative to local concurrent changes among patients served by nonparticipating health care professionals. Consistent with changes in total Medicare spending associated with participation in the MSSP,11 reductions in SNF spending grew with longer ACO participation, and later entrants required more time to achieve reductions than did early entrants. Reductions in spending on SNF care for ACO patients were due in part to lower use of inpatient care, in part to lower rates of discharge to SNFs, and in part to shorter length of SNF stays. Adjustment for the hospital or SNF to which patients were admitted revealed that reductions in SNF use and length of stay for ACO patients were largely due to within-hospital changes in care specifically for ACO patients, as opposed to hospital-wide efforts, and to changes in care within SNFs for ACO patients, as opposed to use of different SNFs. Spending reductions in postacute care were achieved without ostensibly compromising or improving the quality of care for ACO patients, based on mortality, readmissions, and use of highly rated SNFs.

Reductions in spending on SNF care associated with participation in the MSSP were statistically similar for independent physician groups and organizations financially integrated with hospitals, with higher estimates for independent groups contributing to their greater overall savings, as previously described.5,11 On one hand, financial integration with hospitals offers ACOs potential economies of scale for discharge planning and forming preferred SNF networks. On the other hand, hospitals in ACOs have weak incentives to pursue hospital-wide strategies to limit wasteful postacute care because patients attributed under an organization’s ACO contract typically account for only a small proportion of admissions to the organization’s hospitals. Moreover, compared with hospital-integrated ACOs, those that are independent physician groups report a similar capacity for influencing inpatient care—and thus decisions about postacute care at discharge—presumably by staffing admissions directly or using hospitalist groups or care managers to influence their patients’ inpatient care.10 Our results are consistent with ACOs using similar strategies to shorten length of SNF stays, such as employing or partnering with clinicians to follow the care of patients in SNFs.

Thus, although hospitals may ostensibly be better positioned to influence postacute care than are physician groups, they may have limited financial incentives to develop strategies that differ from those available to physician groups.7 Incentives for hospitals in ACOs may change as bundled payment initiatives expand, but our results suggest that financial integration with hospitals is not necessary for ACOs to achieve sizable savings in the postacute care setting. These findings in turn provide valuable information for clinicians considering employment changes to meet the challenges of payment reform and for practices trying to gauge their potential reach of influence and success under new payment models.

Our study also has important implications for Medicare payment policy and antitrust law enforcement. Under current and proposed payment programs that bundle acute and postacute care (eg, the Bundled Payments for Care Improvement initiative), hospitals keep any savings within the bundle for patients attributed to an ACO participating in the MSSP.24,25 Because ACOs may be able to achieve postacute care savings independently, however, our results suggest that ACOs should keep some, if not all, of the savings in cases of overlap with bundled payment programs—ideally to the extent that savings are attributable to ACO efforts. More generally, bundled payment programs that place hospitals at risk for postacute care spending may not be the only viable strategy to curb excessive postacute care. Our findings also challenge claims by leaders of health care provider organizations that mergers and acquisitions involving acute care or postacute care facilities are necessary for achieving more efficient postacute care under new payment models. More generally, the ability of ACOs to curb postacute care spending exemplifies the potential for health care professionals to influence care across multiple settings without necessarily establishing common ownership over the full continuum of care, as few ACOs own SNFs.

Limitations

Our study has several limitations. First, the MSSP is a voluntary program and ACOs likely differ from nonparticipating providers. Postacute care spending levels and trends were similar for ACO and non-ACO providers during the precontract period, however, and we would not expect ACOs to reduce their use of postacute care without an incentive to do so. Second, reductions in SNF use and length of stay for ACO patients conditional on hospital or SNF admission could have been due to differential changes in the case mix resulting from reductions in hospital or SNF admissions. However, we found no evidence of differential changes in patient characteristics, consistent with hospital and SNF admission rates decreasing in proportion to patients’ risk of admission.26 Third, our analysis did not support inferences about substitution of postacute care settings. For example, our finding of lower use of postacute care facilities for ACO patients with no change in use of postacute home health care could be due to both a substitution of postacute home care for facility care and an accompanying reduction in both postacute and outpatient home care for other patients, but our results reflect only the net effects.

Conclusions

Participation in the MSSP has been associated with significant reductions in postacute care spending without ostensible changes in quality, suggesting gains in the value of health care. Postacute care spending reductions were more consistent with efforts by clinicians working within hospitals and SNFs to influence care for ACO patients than with hospital-wide initiatives by ACOs or use of preferred SNFs. Understanding such early successes can support regulatory policy that enhances rather than inhibits the effectiveness of payment and delivery system reforms.

Back to top
Article Information

Accepted for Publication: October 2, 2016.

Corresponding Author: J. Michael McWilliams, MD, PhD, Department of Health Care Policy, Harvard Medical School, 180 Longwood Ave, Boston, MA 02115 (mcwilliams@hcp.med.harvard.edu).

Published Online: February 13, 2017. doi:10.1001/jamainternmed.2016.9115

Author Contributions: Dr McWilliams had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: McWilliams, Chernew, Grabowski.

Acquisition, analysis, or interpretation of data: McWilliams, Gilstrap, Stevenson, Huskamp.

Drafting of the manuscript: McWilliams, Stevenson.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: McWilliams, Stevenson, Grabowski.

Obtained funding: McWilliams, Chernew.

Administrative, technical, or material support: Gilstrap, Huskamp.

Study supervision: McWilliams.

Conflict of Interest Disclosures: Dr McWilliams reported having served as an expert witness for the Federal Trade Commission and serving as a consultant to Abt Associates for an evaluation of the Accountable Care Organization Investment Model. Dr Grabowski reported serving on the Scientific Advisory Board for NaviHealth. No other disclosures were reported.

Funding/Support: This study was supported by grants from the Laura and John Arnold Foundation and grant P01 AG032952 from the National Institute on Aging of the National Institutes of Health.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the Laura and John Arnold Foundation or the National Institutes of Health.

Additional Contributions: Pasha Hamed, MA, Department of Health Care Policy, Harvard Medical School, provided statistical programming support. Jesse B. Dalton, MA, Department of Health Care Policy, Harvard Medical School, provided research assistance. They were supported by the grants that supported this study and received no additional compensation for their work on this study.

References
1.
Institute of Medicine Committee on Geographic Variation in Health Care Spending and Promotion of High-Value Care.  Variation in Health Care Spending: Target Decision Making, Not Geography. Washington, DC: The National Academies Press; 2013.
2.
Medicare Payment Advisory Commission. A data book: health care spending and the Medicare program. http://www.medpac.gov/docs/default-source/data-book/june-2016-data-book-health-care-spending-and-the-medicare-program.pdf. Published June 2016. Accessed January 10, 2017.
3.
Ackerly  DC, Grabowski  DC.  Post-acute care reform—beyond the ACA.  N Engl J Med. 2014;370(8):689-691.PubMedGoogle ScholarCrossref
4.
McWilliams  JM, Chernew  ME, Landon  BE, Schwartz  AL.  Performance differences in year 1 of pioneer accountable care organizations.  N Engl J Med. 2015;372(20):1927-1936.PubMedGoogle ScholarCrossref
5.
McWilliams  JM, Hatfield  LA, Chernew  ME, Landon  BE, Schwartz  AL.  Early performance of accountable care organizations in Medicare.  N Engl J Med. 2016;374(24):2357-2366.PubMedGoogle ScholarCrossref
6.
Schwartz  AL, Chernew  ME, Landon  BE, McWilliams  JM.  Changes in low-value services in year 1 of the Medicare Pioneer Accountable Care Organization program.  JAMA Intern Med. 2015;175(11):1815-1825.PubMedGoogle ScholarCrossref
7.
Mostashari  F, Sanghavi  D, McClellan  M.  Health reform and physician-led accountable care: the paradox of primary care physician leadership.  JAMA. 2014;311(18):1855-1856.PubMedGoogle ScholarCrossref
8.
McWilliams  JM, Chernew  ME, Zaslavsky  AM, Landon  BE.  Post-acute care and ACOs—who will be accountable?  Health Serv Res. 2013;48(4):1526-1538.PubMedGoogle ScholarCrossref
9.
Colla  CH, Lewis  VA, Bergquist  SL, Shortell  SM.  Accountability across the continuum: the participation of postacute care providers in accountable care organizations.  Health Serv Res. 2016;51(4):1595-1611.PubMedGoogle ScholarCrossref
10.
Colla  CH, Lewis  VA, Tierney  E, Muhlestein  DB.  Hospitals participating in ACOs tend to be large and urban, allowing access to capital and data.  Health Aff (Millwood). 2016;35(3):431-439.PubMedGoogle ScholarCrossref
11.
McWilliams  JM.  Changes in Medicare shared savings program savings from 2013 to 2014.  JAMA. 2016;316(16):1711-1713.PubMedGoogle ScholarCrossref
12.
Lage  DE, Rusinak  D, Carr  D, Grabowski  DC, Ackerly  DC.  Creating a network of high-quality skilled nursing facilities: preliminary data on the postacute care quality improvement experiences of an accountable care organization.  J Am Geriatr Soc. 2015;63(4):804-808.PubMedGoogle ScholarCrossref
13.
McHugh  JP, Trivedi  AN, Zinn  JS, Mor  V.  Post-acute integration strategies in an era of accountability.  J Hosp Adm. 2014;3(6):103-112.PubMedGoogle Scholar
14.
Rahman  M, Foster  AD, Grabowski  DC, Zinn  JS, Mor  V.  Effect of hospital–SNF referral linkages on rehospitalization.  Health Serv Res. 2013;48(6, pt 1):1898-1919.PubMedGoogle ScholarCrossref
15.
Research Data Assistance Center. Shared savings program accountable care organizations (ACO) provider-level RIF. http://www.resdac.org/cms-data/files/ssp-aco-provider-level-rif. Accessed September 26, 2016.
16.
Medicare.gov. Nursing Home Compare. https://www.medicare.gov/NursingHomeCompare/Resources/Downloadable-Database.html. Accessed September 26, 2016.
17.
Zaslavsky  AM, Ayanian  JZ, Zaborski  LB.  The validity of race and ethnicity in enrollment data for Medicare beneficiaries.  Health Serv Res. 2012;47(3, pt 2):1300-1321.PubMedGoogle ScholarCrossref
18.
Agency for Healthcare Research and Quality. Creation of new race-ethnicity codes and socioeconomic status (SES) indicators for Medicare beneficiaries: final report. https://archive.ahrq.gov/research/findings/final-reports/medicareindicators/. Updated August 2012. Accessed January 10, 2017.
19.
Centers for Medicare and Medicaid Services. Chronic Condition Data Warehouse. https://www.ccwdata.org/web/guest/home. Accessed September 26, 2016.
20.
Pope  GC, Kautter  J, Ellis  RP,  et al.  Risk adjustment of Medicare capitation payments using the CMS-HCC model.  Health Care Financ Rev. 2004;25(4):119-141.PubMedGoogle Scholar
21.
Yun  H, Kilgore  ML, Curtis  JR,  et al.  Identifying types of nursing facility stays using Medicare claims data: an algorithm and validation.  Health Serv Outcomes Res Methodol. 2010;10(1):100-110. doi:10.1007/s10742-010-0060-4Google ScholarCrossref
22.
US Census Bureau. 2010 American Community Survey (ACS). http://www.census.gov/programs-surveys/acs/data.html. Updated October 13, 2016. Accessed January 10, 2017.
23.
Binder  DA.  On the variances of asymptotically normal estimators from complex surveys.  Int Stat Rev. 1983;51:279-292. doi:10.2307/1402588Google ScholarCrossref
24.
Mechanic  RE.  When new Medicare payment systems collide.  N Engl J Med. 2016;374(18):1706-1709.PubMedGoogle ScholarCrossref
25.
Centers for Medicare & Medicaid Services, Dept of Health and Human Services. Medicare program; advancing care coordination through episode payment models (EPMs); cardiac rehabilitation incentive payment model; and changes to the comprehensive care for joint replacement model (CJR). Proposed rule. https://s3.amazonaws.com/public-inspection.federalregister.gov/2016-17733.pdf. Accessed September 26, 2016.
26.
Colla  CH, Lewis  VA, Kao  LS, O’Malley  AJ, Chang  CH, Fisher  ES.  Association between Medicare accountable care organization implementation and spending among clinically vulnerable beneficiaries.  JAMA Intern Med. 2016;176(8):1167-1175.PubMedGoogle ScholarCrossref
×