This Figure depicts the hospitals entering Maryland’s global budget program in 2014 and the Maryland counties included in our analyses.
These plots show unadjusted annual rates of hospital and primary care utilization among fee-for-service Medicare beneficiaries residing in the 8 Maryland counties where hospitals received global budgets in 2014 vs the matched control counties. The error bars represent 95% confidence intervals for the point estimates in a given year and are calculated using standard errors clustered by county. The vertical lines dividing each panel distinguish between the preintervention (left) and postintervention (right) periods in Maryland. Hospital stays include inpatient admissions and observation stays. See the notes to Table 2 and the eAppendix in the Supplement for additional information about the outcomes reported.
eAppendix. Background and additional details
eTable 1. Maryland hospitals entering the Global Budget Revenue program in 2014
eTable 2. County-level matching variables
eTable 3. Matched control counties
eTable 4. Area-level characteristics of Maryland and matched control counties
eTable 5. Checks of the representativeness of the weighted intervention sample
eTable 6. Average differences, and differential changes, in Maryland vs. control population characteristics
eTable 7. Differential changes from the pre-intervention period (2009-13) through 2014, comparing Medicare beneficiaries in the intervention Maryland versus control counties
eTable 8. States represented in alternate control group
eTable 9. Difference-in-differences estimates using the alternate control group
eFigure 1. Placebo test results
eFigure 2. Analyses of supplemental measures
eTable 10. Percent of hospital stays due to inpatient admissions vs observation stays by year
Customize your JAMA Network experience by selecting one or more topics from the list below.
Roberts ET, McWilliams JM, Hatfield LA, et al. Changes in Health Care Use Associated With the Introduction of Hospital Global Budgets in Maryland. JAMA Intern Med. 2018;178(2):260–268. doi:10.1001/jamainternmed.2017.7455
Was Maryland’s hospital global budget program associated with changes in patients’ use of hospital and primary care after 2 years?
Comparing Medicare beneficiaries in Maryland with a matched control group, we found no consistent changes in annual hospital stays (defined as admissions and observation stays), 30-day return hospital stays, emergency department visits, hospital outpatient department utilization, or visits with primary care physicians.
After 2 years, Maryland’s global budget program was not associated with changes in hospital or primary care use that were clearly attributable to the program.
In 2014, the State of Maryland placed the majority of its hospitals under all-payer global budgets for inpatient, hospital outpatient, and emergency department care. Goals of the program included reducing unnecessary hospital utilization and encouraging greater use of primary care.
To compare changes in hospital and primary care use through the first 2 years of Maryland’s hospital global budget program among fee-for-service Medicare beneficiaries in Maryland vs matched control areas.
Design, Setting, and Participants
We matched 8 Maryland counties (94 967 beneficiaries) with hospitals in the program to 27 non-Maryland control counties (206 389 beneficiaries). Using difference-in-differences analysis, we compared changes in hospital and primary care use in Maryland vs the control counties from before (2009-2013) to after (2014-2015) the payment change, using 2 different assumptions. First, we assumed that preintervention differences between Maryland and the control counties would have remained constant past 2014 had Maryland not implemented global budgets (parallel trend assumption). Second, we assumed that differences in preintervention trends would have continued without the payment change (differential trend assumption).
Main Outcomes and Measures
Hospital stays (defined as admissions and observation stays); return hospital stays within 30 days of a prior hospital stay; emergency department visits that did not result in admission; price-standardized hospital outpatient department (HOPD) utilization; and visits with primary care physicians (overall and within 7 days of a hospital stay).
We matched 8 Maryland counties with hospitals in the program (94 967 beneficiaries; 41.8% male; mean [SD] age, 72.3 [12.2] years) to 27 non-Maryland control counties (206 389 beneficiaries; 42.8% male; mean [SD] age, 71.7 [12.5] years). Assuming parallel trends, we estimated a differential change in Maryland of −0.47 annual hospital stays per 100 beneficiaries (95% CI, −1.65 to 0.72; P = .43) from the preintervention period (2009-2013) to 2015, but assuming differential trends, we estimated a differential change in Maryland of −1.24 stays per 100 beneficiaries (95% CI, −2.46 to −0.02; P = .047). Assuming parallel trends, we found a significant increase in primary care visits (+10.6 annual visits/100 beneficiaries; 95% CI, 4.6 to 16.6 annual visits/100 beneficiaries; P = .001), but assuming differential trends, we found no change (−0.8 visits/100 beneficiaries; 95% CI, −10.6 to 9.0 visits/100 beneficiaries; P = .87). Comparing estimates with both trend assumptions, we found no consistent changes in emergency department visits, return hospital stays, HOPD use, or posthospitalization primary care visits associated with Maryland’s program.
Conclusions and Relevance
We did not find consistent evidence that Maryland’s hospital global budget program was associated with reductions in hospital use or increases in primary care visits among fee-for-service Medicare beneficiaries after 2 years. Evaluations over longer periods should be pursued.
Public and private payers in the United States are implementing alternative models to fee-for-service payment in an effort to control the growth of health care spending and to improve outcomes.1-11 One alternative payment model, introduced by the state of Maryland, gives each of the state’s hospitals an annual global budget for inpatient, hospital outpatient, and emergency department care. By severing the link between volume and revenue, Maryland’s hospital budget program creates financial incentives for hospitals to deter expensive hospital utilization, including admissions and emergency department visits, and to expand other services, such as primary care.12,13 Among alternative payment models, Maryland’s policy is unique because it places care in each hospital under a budget and because it encompasses all payers, including Medicare, Medicaid, and commercial insurers.14 The Centers for Medicare and Medicaid Services (CMS) has expressed interest in implementing hospital global budgets elsewhere in the United States15 and is developing a similar program for rural hospitals in Pennsylvania.16
Maryland’s global budget program was established to meet performance goals of a waiver with CMS.17 This waiver, which was enacted on January 1, 2014, requires Maryland to limit hospital spending growth and to meet certain quality targets, including a reduction of readmissions.14 The CMS has reported that, since 2014, the rate of hospital spending growth has slowed in Maryland, putting the state on track to meet requirements of its waiver.18 By setting budgets for hospitals, Maryland can largely control hospital spending in the state. What remains unclear is how hospitals met their budgets. Under the structure of Maryland’s program, hospitals can lower spending by reducing hospital utilization and enhancing primary care—as policymakers intended—or by reducing their prices.12,13
Numerous reports have described efforts by Maryland hospitals to deter hospitalizations by opening new primary care clinics and expanding outpatient services for individuals with chronic conditions.19-22 One report found that hospitalization and readmission rates declined among Medicare beneficiaries in Maryland in the first year of the program,23 while a CMS-funded evaluation conducted among Medicare beneficiaries found reductions in hospital admissions, increases in emergency department visits, and no changes in primary care associated with the program.24
We examined hospital and primary care use through the second year (2015) of Maryland’s global budget program using a difference-in-differences analysis that compared changes in utilization among Medicare beneficiaries in Maryland with changes among similar out-of-state beneficiaries.
Maryland’s global budget program builds on the state’s hospital rate-setting system, which was created in the 1970s and establishes a set of regulated rates at which hospitals bill all payers. This system exempts Maryland hospitals from Medicare’s national inpatient and outpatient Prospective Payment Systems.25-27 Like Medicare’s national hospital payment programs, Maryland’s prior system paid hospitals per Diagnosis Related Group (for inpatient care) or per service (outpatient care) without constraining total revenue. Thus, hospitals could increase their revenues by providing more care.
Maryland’s new payment system gives each nonfederal acute care hospital in the state an annual all-payer budget for inpatient, emergency department, and hospital outpatient department (HOPD) services. By constraining hospitals’ revenues, the program counteracts incentives to provide more care in the hospital and enables Maryland to meet spending targets of its waiver—specifically, limiting all-payer hospital spending growth to less than 3.58% per year and generating $330 million in Medicare savings over 5 years (based on the difference between Medicare hospital spending growth in Maryland and all other states).14 Global budget agreements were negotiated and finalized for all 36 hospitals eligible for the program by July 2014, although the budgets were retroactive to January 1 of the year.24 (These 36 hospitals joined 10 hospitals in rural Maryland that remained on global budgets through an earlier pilot program,28 which is not examined in this study.)
Hospital budgets are determined from several factors, including historical utilization, regulated prices, and operating costs.29 While hospitals continue to bill per admission or per service, they are now expected to raise or lower their prices throughout the year—as much as a 5% increase or reduction from initial regulated levels—to target an annual amount of total revenue. Hospitals that reduce utilization increase their prices (and presumably their margins), while hospitals with excess volume must lower their prices. This gives hospitals an incentive to reduce utilization subject to a budget and to refer patients for care in other settings, including independent primary care practices.12,13
To remove the incentive to divert patients to other hospitals, Maryland’s program includes provisions to adjust hospitals’ budgets based on shifts in volume that do not reflect improvements in care.30 Consistent with performance incentives established in Medicare’s national Hospital Value Based Purchasing and Hospital Readmissions Reduction Programs, Maryland also adjusts hospitals’ budgets for quality of care, including readmission rates.31-34 Payments for physician services are not included in the budgets. (See the eAppendix in the Supplement for additional details of Maryland’s program.)
We examined changes in hospital and primary care use in the Medicare population because the high chronic disease burden and rate of hospital use among those in this population make it likely to be affected by the payment change. Owing to the use of deidentified patient data, the Harvard institutional review board determined that the study did not involve human subjects and was thus exempt from review. We analyzed 2009 through 2015 Medicare Part A and Part B claims for a random 20% sample of beneficiaries enrolled in fee-for-service Medicare for at least 1 full year or until death. To assess established diagnoses, we limited the sample in each study year to beneficiaries with continuous fee-for-service enrollment in the prior year.
We included all beneficiaries living in 8 Maryland counties where the hospitals received global budgets beginning in 2014 (Figure 1).12,35 This 8-county region included 32 of the 36 hospitals entering the 2014 program. The 32 hospitals accounted for 88% of admissions from Medicare beneficiaries residing in this 8-county region. Hospitals in the 4 excluded counties served primarily rural communities, including some with seasonally variable tourist populations (eAppendix in the Supplement). We conducted analyses on an intention-to-treat basis, considering all beneficiaries in the 8 intervention counties to be exposed to the program regardless of where they received care.
To select a control group, we used coarsened exact matching36 to match each of the 8 intervention counties to out-of-state control counties with similar numbers of hospitals, hospital beds, and physicians per capita; poverty rates; proportions of African American residents; and population densities (eAppendix in the Supplement). Preliminary analyses indicated that these county characteristics were predictive of hospitalization trends in the fee-for-service Medicare population.
From Medicare enrollment data, we determined beneficiaries’ age, sex, race and ethnicity, whether disability was the original reason for Medicare eligibility, receipt of Medicaid benefits (ie, dual eligibility), and the presence of end-stage renal disease. From the Chronic Conditions Data Warehouse (CCW), we assessed each beneficiary’s accumulated burden of 27 chronic conditions prior to each study year. Using enrollment information and diagnoses reported on prior-year claims, we calculated each beneficiary’s Hierarchical Condition Category (HCC) risk score, which is used for risk adjustment in Medicare programs.37 Finally, we incorporated data from the Census Bureau’s American Community Survey to measure the poverty rate and proportion of residents with less than a high school education in the Zip Code Tabulation Areas (ZCTAs) where beneficiaries resided.
We used the Medicare Provider Analysis and Review (MedPAR) and outpatient claims files to construct several measures of hospital utilization, which we assessed annually for beneficiaries. First, given recent and substantial increases in the provision of observation-unit care in lieu of short-stay admissions,38,39 we constructed a combined measure of admissions and observation stays, which we term hospital stays. Second, because Maryland’s policy gives hospitals additional incentives to avoid readmissions,32,40 we assessed the proportion of patients’ annual admissions and observation stays that were followed by a second hospital stay within 30 days (we term these 30-day return hospital stays). In subanalyses, we assessed changes in inpatient admissions and observation stays separately (reported in the eAppendix in the Supplement). Third, using the outpatient claims file, we assessed visits to hospital emergency departments that did not lead to admission. Fourth, to summarize patients’ use of the wide range of services in HOPDs, we applied national procedure-code level prices to HOPD claims to construct a price-standardized measure of HOPD utilization. Because hospital prices were subject to change under Maryland’s program, we measured prices from hospitals outside of the state (eAppendix in the Supplement).
To analyze changes in utilization of primary care, we used the carrier (physician/supplier) file to assess beneficiaries’ annual visits with primary care physicians in HOPDs or independent physician offices and the proportion of hospital stays followed by a primary care visit within 7 days of discharge or observation.41
We used a difference-in-differences design to assess changes in utilization from the baseline period (pooling 2009 through 2013) to the first (2014) and second (2015) years of Maryland’s program, comparing the health care utilization of Medicare beneficiaries in the 8 intervention vs 27 control counties.
We conducted 2 sets of analyses based on different assumptions. First, we assessed differential changes in Maryland from the preintervention to the postintervention periods, assuming that differences in outcomes between Maryland and the control group would have remained constant past 2014 had Maryland not implemented global budgets. This approach, which is standard in difference-in-difference studies, relies on the assumption of parallel preintervention trends in Maryland and the control group.42 However, it could produce biased estimates if preintervention trends differ. Therefore, in a second analysis, we assumed that differences between Maryland and the control group would have continued to change at the preintervention rate in the absence of Maryland’s program—analogous to an approach used in other analyses of Maryland’s reform.24,29 This approach yields an estimate of changes in Maryland vs the control group, net of changes that would be expected from a continuation of the groups’ preintervention trends past 2014.
Estimates of differential changes were obtained from linear regression models with the patient-year as the unit of analysis and were adjusted for patients’ health status in the prior year (including established chronic conditions from the CCW), demographic characteristics, ZCTA-level poverty and education, and geography. Models were estimated with propensity-score weights that balanced observed characteristics of beneficiaries in Maryland and the control counties in each year. (Additional details of our analyses are provided in the eAppendix in the Supplement.)
We conducted 2 sensitivity analyses. First, to assess whether our estimates were sensitive to the control group used and to other policy changes coinciding with Maryland’s reform, we compared Maryland with urban areas of states that also implemented the Affordable Care Act’s Medicaid expansion in early 2014 (eTable 8 in the Supplement).43
Second, we built on the first sensitivity analysis by conducting a placebo analysis.44 To conduct this analysis, we iteratively assigned an indicator of treatment status to each control state that expanded Medicaid in early 2014 and estimated differential changes in this “placebo” state vs the remaining control states. To assess whether changes in Maryland were within or outside the range of changes among other states, we plotted our findings from the first sensitivity analysis against the range of placebo estimates (eAppendix in the Supplement).
In 2014, our study sample included 94 697 fee-for-service Medicare beneficiaries in the 8 intervention counties in Maryland and 206 389 beneficiaries in the matched control counties. Before weighting, beneficiaries in Maryland were slightly older and more likely to be African American than patients in the control counties but were less likely to have qualified for Medicare because of a disability or have dual Medicare and Medicaid eligibility (Table 1). After weighting to balance observable patient characteristics, we found that the distribution of categorical variables and means of continuous variables were identical in both groups.
Adjusted preintervention trends in Maryland differed appreciably from those of the control group for 30-day return hospital stays, nonadmitted emergency department visits, price-standardized HOPD utilization, and primary care visits (in aggregate and following a hospital stay; Table 2 and Figure 2). From 2009 through 2012, we found comparable changes in hospital stays in Maryland and the control group, followed by a divergence in the groups’ trends from 2012 to 2013.
There were no statistically significant year-1 (2014) differential changes in any outcomes we assessed using either trend assumption (eTable 7 in the Supplement).
In the preintervention period (2009-2013), annual hospital utilization in Maryland averaged 40.9 stays per 100 beneficiaries. When we assumed hospital use would have changed comparably in Maryland and the control group without global budgets (ie, assuming parallel trends), we estimated a differential change in Maryland of −0.47 annual hospital stays per 100 beneficiaries from the preintervention period to 2015 (95% CI, −1.65 to 0.72; P = .43; Figure 2 and Table 2), or −1.1% of the state’s preintervention average. However, assuming that diverging preintervention trends would have continued past 2014 without the global budget program, we estimated a differential change in Maryland of −1.24 annual hospital stays per 100 beneficiaries (95% CI, −2.46 to −0.02; P = .047), or −3.0% of Maryland’s baseline.
During the preintervention period, the rate of 30-day return hospital stays in Maryland averaged 21.5% per year. Assuming parallel trends, return hospital stays declined differentially in Maryland from the preintervention period to 2015 (differential change: −1.0 percentage points per year; 95% CI, −2.0 to −0.1; P = .03; Table 2), or −4.7% of the state’s preintervention rate. However, assuming that preexisting differences in trends would have continued, we found a differential change of −0.6 percentage points (95% CI, −2.2 to 1.1; P = .50), constituting a 2.8% decrease from Maryland’s baseline.
Assuming parallel trends, we found a differential change of 0.2 annual emergency visits per 100 beneficiaries in Maryland (95% CI, −1.8 to 2.2; P = .82; Table 2), or 0.5% of the state’s preintervention average of 38.5 visits per 100 beneficiaries, but assuming differential trends, we found a differential change of −1.2 annual visits per 100 beneficiaries (95% CI, −2.8 to 0.3; P = .11), or −3.1% of the state’s baseline.
Estimates of differential changes in primary care visits varied based on whether we assumed parallel trends (differential change in Maryland: +10.6 annual visits/100 beneficiaries; 95% CI, 4.6 to 16.6; P = .001; Table 2) or differential trends (differential change in Maryland: −0.8 visits/100 beneficiaries; 95% CI, −10.6 to 9.0; P = .87). While not statistically significant, estimates of differential changes in price-standardized HOPD use also differed when we assumed parallel vs differential trends (−$56 vs +$40, respectively). There were no consistent differential changes in postdischarge primary care comparing the 2 trend assumptions.
Among beneficiaries in the 8 Maryland counties, the proportion of admissions to the 32 global budget hospitals, vs out-of-state or other Maryland hospitals, did not change substantially from the preintervention period (88.4%) to the postintervention period (87.6%).
Our results were substantively unchanged when we compared Maryland with an alternate control group of Medicaid expansion states (eTable 9 in the Supplement). Changes in Maryland vs other Medicaid expansion states were within the range of changes estimated in placebo analyses (eFigure 1 in the Supplement).
This study examined changes in hospital and primary care use associated with the 2014 introduction of global budgets in Maryland hospitals. We conducted difference-in-differences analyses that compared changes in utilization among Medicare beneficiaries in Maryland with changes in a matched out-of-state control group using 2 sets of assumptions. First, we assumed that changes in Maryland and the control group would have been identical past 2014 (parallel trends). Second, we assumed changes between the groups would have differed based on their preintervention trends (differential trends). Because preintervention tends differed, our results were sensitive to assumptions about whether these trend differences would have continued in the absence of Maryland’s reform. For example, assuming parallel trends, we would conclude that Maryland’s program was associated with a significant differential reduction in 30-day return hospital stays and a differential increase in primary care visits, but assuming prereform trends would continue, we found a much smaller change in return hospital stays and no change in primary care visits. These differences in estimates indicate substantial uncertainty about whether changes after 2014 constituted effects of Maryland’s global budget program or reflected a continuation of prereform state trends.
We continued to find inconsistent results when we compared Maryland with an alternative control group of Medicaid expansion states. Furthermore, placebo analyses indicated that changes in Maryland were well within the range of changes in other states. Together, these findings provide no clear evidence that Maryland hospitals met their budgets by reducing hospital utilization or enhancing primary care beyond changes that would have been expected in the absence of global budgets.
There are several possible explanations for why we did not find strong evidence of population-level changes in hospital or primary care use associated with Maryland’s program. First, because hospitals’ budgets did not include payments to physicians for care provided in the hospital or in community settings, the program could have had limited influence on physician behavior. Other alternative payment models, including Accountable Care Organizations (ACOs), place physicians or both physicians and hospitals at risk for patients’ spending and outcomes of care,2,4,6 whereas only hospitals bore risk in Maryland’s program. While Maryland initially proposed to share a proportion of hospitals’ savings with physicians to promote incentive alignment, the CMS did not approve these plans in the state’s original payment model. In updates to the model, Maryland plans to expand the scope of budgets to include physicians and to establish ACO-like organizations to manage Medicare beneficiaries’ inpatient and outpatient spending.45
Second, although Maryland’s program established annual budgets for hospitals, hospitals were still paid on a per-admission or per-service basis. Hospitals that lowered utilization were expected to raise their prices to receive their budgeted revenue, but since price increases were generally limited to less than 5%, the program’s structure may have dampened incentives to substantially reduce volume.
Third, while Maryland’s program officially began on January 1, 2014, it took several months for hospitals to finalize their global budget agreements, such that our analyses may not encompass 2 full years of exposure to the program’s incentives. Furthermore, interviews with hospital staff revealed that some hospitals experienced initial difficulty aligning physicians and staff with the payment model and implementing new care management programs for patients.29 Thus, more time may be needed to observe impacts of Maryland’s reform on utilization.
This study had several limitations. First, hospitals’ budgets included patients with all types of insurance, but our analyses focused only on Medicare beneficiaries. Because Medicare beneficiaries have high rates of hospital use, we assumed their care was most likely affected by the program. However, impacts of the program could been more evident in younger populations for whom care in the community (as opposed to the hospital) may have been more appropriate. Second, because Maryland’s policy was implemented statewide, we were unable to make comparisons to a within-state control group. To select a control group, we identified counties outside of Maryland with similar characteristics, and used propensity-score weighting to balance observed patient characteristics. Nevertheless, preintervention trends differed in Maryland vs our control group on several outcomes, making it difficult to isolate impacts of Maryland’s global budget program from trend differences predating its implementation. Our findings highlight the importance of using quasiexperimental approaches with control populations to evaluate policy effects and of carefully considering preintervention trends when drawing conclusions so that results can be interpreted with appropriate caution. Finally, our estimates could have been biased by unobserved changes in the Medicare population, the payment environment, and other factors affecting demand for hospital care in Maryland vs the control group.
In conclusion, 2 years after the introduction of hospital global budgets in Maryland, we found no consistent evidence that changes in hospital or primary care could be reliably attributed to the program. Further monitoring is needed to assess how hospitals adapt to this payment model over time.
Corresponding Author: Ateev Mehrotra, MD, MPH, Department of Health Care Policy, Harvard Medical School, 180 Longwood Ave, Boston, MA 02115-5899 (email@example.com).
Accepted for Publication: October 29, 2017.
Published Online: January 16, 2018. doi:10.1001/jamainternmed.2017.7455
Author Contributions: Dr Roberts 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: All authors.
Acquisition, analysis, or interpretation of data: Roberts, McWilliams, Hatfield, Chernew, Gilstrap, Mehrotra.
Drafting of the manuscript: Roberts, McWilliams, Hatfield, Chernew, Gilstrap, Mehrotra.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Roberts, McWilliams, Hatfield, Chernew, Gilstrap, Mehrotra.
Obtained funding: Roberts, Mehrotra.
Administrative, technical, or material support: Gilstrap.
Study supervision: McWilliams, Mehrotra.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by grants from the Commonwealth Fund and the National Institutes of Health (P01 AG032952), with additional financial support provided by the Marshall J. Seidman Center for Studies in Health Economics and Health Care Policy at Harvard Medical School.
Role of the Funder/Sponsor: The funding sources played no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Create a personal account or sign in to: