This comparative effectiveness research study examines data after implementation of the Maryland global budget revenue model to gauge its association with spending and clinical outcomes among Medicare beneficiaries who undergo cancer-directed surgery.
Key Points
Question
Is the Maryland global budget revenue (GBR) model associated with meaningful changes in 30-day episode spending and clinical outcomes among Medicare beneficiaries who undergo cancer-directed surgery?
Findings
In this difference-in-differences analysis, comparison of Maryland Medicare beneficiaries with a matched control group found that GBR was associated with a statistically significant decrease in 30-day readmissions. However, we found no consistent changes in 30-day episode spending, emergency department visits, or all-cause mortality.
Meaning
After 4 years of implementation, we identified no significant association between GBR and changes in spending, emergency department visits, and selected clinical outcomes, although a modest decline in 30-day readmissions was noted.
Abstract
Importance
In 2014, Maryland initiated the global budget revenue (GBR) model, placing caps on total hospital expenditures across all care sites. The GBR program aims to reduce unnecessary utilization while maintaining or improving care quality. To date, there has been limited examination of program effects on cancer care.
Objective
To compare changes in spending, clinical outcomes, and acute care utilization through 4 years of the GBR program among Medicare beneficiaries who undergo cancer-directed surgery in Maryland vs matched control states.
Design, Setting, and Participants
Drawing from a matched pool of hospitals in Maryland (n = 35) and 24 control states with a similar timing of Medicaid expansion (n = 101), we identified Medicare beneficiaries from Maryland and control states who underwent any cancer-directed surgery from 2011 through 2018. Using difference-in-differences analysis, we compared changes in outcomes from before (2011-2013) to after (2015-2018) GBR implementation between patients treated in Maryland and control states. We also performed a subgroup analysis among patients who underwent major surgical procedures that are usually performed in the inpatient setting (cystectomy, esophagectomy, gastrectomy, colorectal resection, nephrectomy, pancreatectomy, and lung resection).
Main Outcomes and Measures
Thirty-day episode spending, mortality, readmissions, and emergency department (ED) visits.
Results
Relative to Medicare beneficiaries undergoing cancer surgery in control states (n = 4737; 3323 [70.1%] female; 571 [12.1%] dual-eligible; mean [SD] age 74.9 [6.5] years), patients in Maryland (n = 20 320; 14 068 [69.2%] female; 1705 [8.4%] dual-eligible; mean [SD] age 74.9 [6.5] years) had a statistically significant reduction of 2.2 percentage points (95% CI, −4.3 to −0.1) in the 30-day readmission rate. We found no statistically significant changes in 30-day spending, mortality, or ED visits. We report no significant results in the subgroup analysis of patients undergoing major surgical procedures.
Conclusions and Relevance
Global budget revenue was not associated with changes in expenditures, ED utilization, or clinical outcomes after cancer-directed surgery through 4 years. There was a modest decline in 30-day readmissions. Specialty-specific definitions of care quality and better alignment across the entire care delivery value chain (ie, physician incentives) may be strategies that could improve delivery of high-value care for beneficiaries undergoing cancer surgery.
Introduction
Cancer accounts for the second highest annual spending among chronic conditions, approximately $173 billion in 2020.1 To address high health care spending, policy makers are replacing traditional fee-for-service payments with alternative payment models that increase clinician and hospital accountability for spending and outcomes.2 For example, the Centers for Medicare & Medicaid Services (CMS) have implemented the Oncology Care Model and proposed the Radiation Oncology and Oncology Care First models.
In 2014, Maryland implemented a global budget revenue (GBR) system, which capped annual hospital payments and mandated reductions in avoidable complications.3 Global budget revenue is one of the most comprehensive alternative payment models in the United States, affecting all nonfederal acute care hospitals in the state. By capping annual revenue, GBR seeks to uncouple reimbursement from volume and offer incentives for high-value health system spending that will reduce health care expenditures and improve outcomes.
A CMS evaluation of the Maryland GBR over the model’s first 4.5 years assessed utilization, Medicare expenditures, and total hospital expenditures.4 On average, total inpatient hospitalizations decreased 7.2% more in Maryland than in the comparison group, and the size of the decrease grew over time, consistent with increasing implementation effectiveness.4 Although inpatient utilization reductions did not result in savings to payers because GBR adjusts charges to ensure that hospitals receive the full global budget amount, outpatient expenditure growth for Medicare was lower in Maryland than in comparison hospitals, leading to total savings for Medicare.4 Postacute care and professional expenditures also grew less in Maryland while emergency department (ED) utilization increased relative to controls. CMS has already signaled its intent to expand the GBR.5
Investigations of the costs and quality associated with oncology care after GBR implementation have been limited. This is salient because the hospital setting is the largest component of overall spending and a key driver of regional variation in cancer treatment.6 Surgical care, our present focus, is anchored to a hospital encounter (ie, inpatient or outpatient procedure) and accounts for 51% of Medicare spending.7 A recent study identified an association between GBR implementation and slower cost growth and reductions in avoidable complications after major surgery in Maryland.8 Our goal was to examine the association of Maryland’s GBR with health care spending and clinical outcomes of surgical care for Medicare beneficiaries with cancer using a difference-in-differences design.
Methods
This study, along with waiver of consent, was approved by the institutional review board at University of Texas MD Anderson Cancer Center in Houston. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines were used to inform the reporting of this observational study.9
Overview of Maryland’s GBR Program
A detailed overview of the GBR program, which began in 2014, has been published elsewhere.4,10 In brief, GBR is a prospective payment model that entailed a hospital-specific global budget for services across all hospital-based sites (ie, inpatient, outpatient, and ED). It encompassed the 95% of total revenues hospital obtain from government and commercial payers and included the following requirements: (1) limiting per capita hospital spending growth to less than 3.6% per year, (2) quality goals of reducing readmissions and hospital-acquired conditions, (3) expectation of $330 million in total savings to Medicare based on reducing per-beneficiary hospital spending growth relative to national benchmarks, and (4) assurances that hospital savings were not due to inappropriate shifting of services to other sites of care.4 Non–hospital-based services, including physician fees, home health services, post-acute care facilities, and outpatient dialysis centers, were excluded from GBR. Hospital budgets are adjusted annually to account for changes in inflation, population, quality scores, market share, and prospective utilization reduction.4
Study Design
We used a matched difference-in-differences approach to estimate the association of GBR with Medicare payments and clinical outcomes of cancer surgical episodes.7,11,12,13 We first matched Maryland hospitals that provided cancer surgeries to hospitals in control states based on observed hospital characteristics in 2013, the year before GBR implementation. Control states, like Maryland, expanded Medicaid on January 1, 2014: Arizona, Arkansas, California, Colorado, Connecticut, Delaware, District of Columbia, Hawaii, Illinois, Iowa, Kentucky, Massachusetts, Minnesota, Nevada, New Jersey, New Mexico, New York, North Dakota, Ohio, Oregon, Rhode Island, Vermont, Washington, and West Virginia. This approach limited unmeasured confounding due to concurrent Medicaid expansion.
After identifying cancer surgeries performed in matched hospitals between 2011 and 2018, we assessed difference-in-differences estimates for each outcome, comparing the pre-GBR period (2011-2013) with the post-GBR period (2015-2018); we did not study 2014, the first year of GBR implementation when hospitals may have made changes as they transitioned to providing care under this model. In 2019, Maryland embarked on the Total Cost of Care Model, an enhancement of GBR.
Data
This study used Medicare Beneficiary Summary Files and inpatient, physician/supplier, and outpatient files for fee-for-service Medicare beneficiaries between 2010 and 2018. For Maryland, West Virginia, Delaware, and DC, 100% Medicare data were available; for other states, a random 5% sample was available. Hospital characteristics were obtained from the 2013 CMS Impact File, Provider of Services file, and American Hospital Association annual survey.
Hospital Selection and Matching
To identify hospitals providing cancer surgeries in 2013, we identified Medicare beneficiaries residing in Maryland and control states who had a cancer-related surgery in a short-term hospital within the same state as their residence in 2013. Cancer-related surgeries were identified from inpatient and outpatient claims using procedure codes from the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), and Current Procedural Terminology (CPT) codes from the Oncology Care Model and peer-reviewed literature (eTable 4 in the Supplement).8,14 Reconstruction and repair procedures performed without a concomitant cancer-related procedure were excluded along with emergency procedures.
Next, we selected beneficiaries who were 66 years and older and continuously enrolled in parts A and B of fee-for-service Medicare in the 12 months before surgery through 30 days after discharge (or death if they died within 30 days). We excluded beneficiaries whose surgery claims were not reimbursed by Medicare (eTable 1 in the Supplement). We also excluded 10 rural Maryland hospitals that participated in a GBR pilot program (Total Patient Revenue) initiated in 2010,15 leaving 35 Maryland hospitals and 805 hospitals in other states.
Among hospitals, we characterized total beds, resident-to-bed ratio, disproportionate share hospital patient percentage, percent capacity (average daily census/total beds), transfer-adjusted case mix, total inpatient surgical operations, total outpatient surgical operations, nurse staffing ratio, and medical school affiliation. Nurse staffing ratio was defined as in other published work.16 We excluded 28 non-Maryland hospitals with missing data for these variables. We then used greedy nearest neighbor matching to match hospitals in Maryland and control states by the propensity of being a Maryland hospital, estimated by a logistic regression model including all hospital characteristics. Each Maryland hospital was matched to up to 3 unique control hospitals, resulting in 101 matched control hospitals. Two control hospitals were from Delaware and 5 were from West Virginia, where 100% Medicare data were available. The remaining 94 hospitals were in control states for which 5% Medicare data were available.
Patients
Patient age, documented sex, race and ethnicity, and dual-eligibility status (Medicare and Medicaid) at the surgery month were obtained from the Medicare Beneficiary Summary Files. These demographic characteristics were abstracted to account for disparities in cancer access, expenditures, and outcomes. We report race and ethnicity using the categories Asian or Pacific Islander, Hispanic, non-Hispanic Black, non-Hispanic White, and other. Race and ethnicity data in the Medicare Beneficiary Summary files are originally from the Social Security Administration master beneficiary record and then improved by algorithms developed by CMS. The algorithms use information in the Medicare enrollment database, such as language preference and location of residence. The algorithms also use Hispanic and Asian/Pacific Islander surname lists developed by the US Census Bureau.
Classification was performed by a study coauthor (Y.L.L.) and approved by the entire research team. Cancer type (breast, colorectal, lung, prostate, bladder, female genitourinary, kidney, metastasis, and other) was determined using the primary diagnosis of the surgery claim. CMS Hierarchical Condition Categories were identified using the claims in the year before surgery. We identified institutionalized beneficiaries using claims for nursing facility services in the 90 days before surgery (CPT codes 99304-99310 and 99318). The surgery setting (inpatient or outpatient) was determined by the source of the surgery claim. We calculated absolute standardized differences in patient characteristics between patients from Maryland hospitals and matched control hospitals to assess the balance of patient characteristics. Absolute standardized differences greater than 0.1 were considered unbalanced.17,18
Outcomes
Our main outcome was total Medicare payments during the cancer surgical episode, which included payments for inpatient, outpatient, and part B claims during the surgical admission and the 30 days after hospital discharge, similar to the definition of surgical episodes for other CMS payment models.19 Payments were adjusted for inflation to 2018 US dollar amounts using the consumer price index for medical care.20 We also investigated all-cause readmissions, ED visits, and all-cause mortality in the 30 days after discharge.21,22
Pre-GBR Baseline Trends (2011-2013)
For each outcome measure, before estimating the difference-in-differences, we examined the patient-level trends across Maryland and control hospitals during the pre-GBR baseline period to test the parallel-trend assumption (eMethods in the Supplement). We did not observe any statistically different baseline trends between the 2 groups of patients in the pre-GBR period (eFigures 1 and 2 in the Supplement).
Statistical Analysis
We applied a linear regression model that included hospital and year fixed effects to estimate the difference-in-differences of each outcome measure (eMethods in the Supplement).23 We included year fixed effects to control for the unobserved confounding over time. Hospital fixed effects were included to account for patient clustering within hospitals and unobserved hospital confounders. We modeled the natural logarithm of total Medicare payment during the cancer surgical episode due to right-sided skewedness. Two-sided significance testing was performed at the 5% level. Analyses were performed with SAS Enterprise Guide version 7.1 (SAS Institute) at the CMS Virtual Research Data Center.
Sensitivity Analyses
We conducted additional analyses among patients who underwent major cancer surgery in the inpatient setting, including cystectomy, esophagectomy, gastrectomy, colectomy, rectum resection, nephrectomy, pancreatectomy, or lung resection. These procedures are associated with high morbidity, readmission rates, and expenditures24,25,26,27 and are likely to be a specific focus of hospital care redesign initiatives intended to reduce spending and improve outcomes. For these analyses, we also assessed discharge to an institutional post-acute care setting (skilled nursing facility or inpatient rehabilitation facility) vs home based on the discharge setting on the claim. This analysis excluded patients who died (n = 148) or were discharged to another hospital (short-term, long-term, federal, critical access hospital, n = 248), hospice (n = 34), or other care facility (intermediate care facility, Medicaid certified nursing home, other health care institution, n = 41). We additionally excluded 19 patients to ensure Maryland hospitals and their matched controls had data from the same years.
Results
Hospital Matching and Study Population
Table 1 shows hospital characteristics in Maryland and control states, before and after propensity matching. Most factors were balanced after matching, except for the resident-to-bed ratio, which was slightly greater than the prespecified 0.1 threshold. Our primary analysis included 20 320 patients from the 35 Maryland hospitals and 4737 patients from the 101 matched control hospitals who had a cancer surgery between January 2011 and November 2018. Selected demographic characteristics, across Maryland and control states, are described in Table 2 and eTable 2 in the Supplement. Compared with patients in the matched control hospitals, patients in Maryland hospitals were less likely to be non-Hispanic white and dual eligible. Maryland had a lower proportion of patients with breast, colorectal, or prostate cancer and a slightly higher proportion of patients with female genitourinary or metastasis. Maryland patients were more likely to have inpatient surgeries. The prevalence of chronic obstructive pulmonary disease was lower in Maryland.
Table 1. Hospital Characteristics Before and After Propensity Match.
| Hospital characteristic in 2013 | Mean (SD) | Absolute standardized difference | |||
|---|---|---|---|---|---|
| Maryland hospitals (n = 35) | Control hospitals, before match (n = 777) | Control hospitals, after match (n = 101)a | Before match | After match | |
| Ln(total beds) | 5.46 (0.66) | 5.35 (0.72) | 5.42 (0.60) | 0.167 | 0.078 |
| Resident-to-bed ratio, per 100 beds | 7.45 (12.45) | 11.01 (20.38) | 5.98 (12.47) | 0.211 | 0.118 |
| Disproportionate share hospital patient percentage | 27.63 (15.31) | 28.03 (14.97) | 27.75 (16.97) | 0.026 | 0.008 |
| Percent capacity (average daily census/total beds) | 71.92 (5.46) | 61.21 (14.65) | 71.40 (12.11) | 0.969 | 0.055 |
| Transfer-adjusted case mix | 1.47 (0.21) | 1.56 (0.24) | 1.48 (0.16) | 0.404 | 0.045 |
| Ln(total inpatient surgical operations) | 8.13 (0.83) | 7.92 (0.88) | 8.12 (0.69) | 0.241 | 0.013 |
| Ln(total outpatient surgical operations) | 8.81 (0.71) | 8.58 (0.73) | 8.80 (0.66) | 0.320 | 0.014 |
| Nurse staffing ratio (RN productive hours per adjusted patient day)b | 4.61 (2.76) | 5.63 (2.48) | 4.69 (2.07) | 0.389 | 0.034 |
| Medical school affiliation, No. (%) | 15 (42.9) | 360 (46.3) | 41 (40.6) | 0.070 | 0.046 |
Among the 35 Maryland hospitals, each of 31 was matched to 3 control hospitals and each of the remaining 4 was matched to 2 control hospitals. Among the 101 matched control hospitals, 2 were from Delaware and 5 were from West Virginia, where 100% Medicare data were available. The remaining 94 hospitals were located in control states where 5% data were available.
It was assumed that 85% of the hours were productive.
Table 2. Patient Characteristics in Maryland Hospitals and Their Matched Hospitals in Control States.
| Patient characteristic | Main cohort | Major surgery subgroup | ||||
|---|---|---|---|---|---|---|
| No. (%) | Standardized difference (95% CI) | No. (%) | Standardized difference (95% CI) | |||
| Maryland (n = 20 320) | Control states (n = 4737) | Maryland (n = 6858) | Control states (n = 1686) | |||
| Age at surgery, mean (SD), y | 74.9 (6.5) | 74.9 (6.5) | 0.01 (−0.03 to 0.04) | 75.4 (6.3) | 75.3 (6.4) | 0.01 (−0.04 to 0.06) |
| Age at surgery, y | 0.04 (0.01 to 0.07) | 0.06 (0.00 to 0.11) | ||||
| 66-69 | 5087 (25.0) | 1193 (25.2) | 1440 (21.0) | 374 (22.2) | ||
| 70-74 | 5897 (29.0) | 1385 (29.2) | 1991 (29.0) | 487 (28.9) | ||
| 75-79 | 4397 (21.6) | 1032 (21.8) | 1629 (23.8) | 378 (22.4) | ||
| 80-84 | 2980 (14.7) | 665 (14.0) | 1159 (16.9) | 284 (16.8) | ||
| ≥85 | 1959 (9.6) | 462 (9.8) | 639 (9.3) | 163 (9.7) | ||
| Sex | −0.02 (−0.05 to 0.01) | 0.07 (0.01 to 0.12) | ||||
| Male | 6252 (30.8) | 1414 (29.9) | 3310 (48.3) | 869 (51.5) | ||
| Female | 14 068 (69.2) | 3323 (70.1) | 3548 (51.7) | 817 (48.5) | ||
| Race and ethnicity | 0.48 (0.45 to 0.52) | 0.48 (0.42 to 0.53) | ||||
| Asian and Pacific Islander | 440 (2.2) | 58 (1.2) | 166 (2.4) | 30 (1.8) | ||
| Hispanic | 304 (1.5) | 97 (2.0) | 100 (1.5) | 35 (2.1) | ||
| Non-Hispanic Black | 4106 (20.2) | 245 (5.2) | 1247 (18.2) | 69 (4.1) | ||
| Non-Hispanic White | 15 142 (74.5) | 4265 (90.0) | 5249 (76.5) | 1541 (91.4) | ||
| Other | 328 (1.6) | 72 (1.5) | 96 (1.4) | 11 (0.7) | ||
| Dual eligible at the month of surgery discharge | −0.12 (−0.15 to −0.09) | −0.14 (−0.19 to −0.09) | ||||
| No | 18 615 (91.6) | 4166 (87.9) | 6303 (91.9) | 1479 (87.7) | ||
| Yes | 1705 (8.4) | 571 (12.1) | 555 (8.1) | 207 (12.3) | ||
| Cancer diagnosis | 0.23 (0.20 to 0.26) | 0.28 (0.23 to 0.34) | ||||
| Breasta | 8106 (39.9) | 2119 (44.7) | - | - | ||
| Colorectal | 2711 (13.3) | 778 (16.4) | 2334 (34.0) | 737 (43.7) | ||
| Lung | 2096 (10.3) | 444 (9.4) | 1726 (25.2) | 405 (24.0) | ||
| Prostatea | 1237 (6.1) | 347 (7.3) | - | - | ||
| Bladder | 792 (3.9) | 172 (3.6) | 520 (7.6) | 116 (6.9) | ||
| Female genitourinary | 1389 (6.8) | 255 (5.4) | 149 (2.2) | 13 (0.8) | ||
| Kidney | 993 (4.9) | 248 (5.2) | 801 (11.7) | 226 (13.4) | ||
| Metastasis | 640 (3.1) | 88 (1.9) | 261 (3.8) | 37 (2.2) | ||
| Other | 2356 (11.6) | 286 (6.0) | 1067 (15.6) | 152 (9.0) | ||
| Chronic obstructive pulmonary disease | −0.10 (−0.13 to −0.07) | −0.16 (−0.21 to −0.11) | ||||
| No | 16 125 (79.4) | 3557 (75.1) | 4835 (70.5) | 1061 (62.9) | ||
| Yes | 4195 (20.6) | 1180 (24.9) | 2023 (29.5) | 625 (37.1) | ||
| Index surgery setting | −0.12 (−0.15 to −0.09) | |||||
| Inpatient | 12 210 (60.1) | 2566 (54.2) | ||||
| Outpatient | 8110 (39.9) | 2171 (45.8) | ||||
| Surgery type | 0.33 (0.27 to 0.38) | |||||
| Colectomy | 2640 (38.5) | 790 (46.9) | ||||
| Cystectomy | 316 (4.6) | 60 (3.6) | ||||
| Esophagectomy or gastrectomy | 302 (4.4) | 46 (2.7) | ||||
| Lung resection | 1899 (27.7) | 441 (26.2) | ||||
| Nephrectomy | 1024 (14.9) | 280 (16.6) | ||||
| Pancreatectomy | 438 (6.4) | 41 (2.4) | ||||
| Multiple surgeries | 239 (3.5) | 28 (1.7) | ||||
Because of small numbers of patients, those in the major surgery subgroup who had breast or prostate cancer were consolidated with the group of patients with other cancers.
Change in Total Medicare Payment Post-GBR
Total Medicare spending for surgical episodes was higher in Maryland relative to control states in the pre-GBR period (Table 3) and decreased in both Maryland and control states in the post-GBR period. The declines in per-episode payments did not differ statistically between patients treated in Maryland and control states (difference-in-differences, $646; 95% CI, −$208 to $1542).
Table 3. Association of Maryland Global Budget Revenue Program With Total Medicare Payments, Readmissions, Emergency Department Visits, and Mortality Through 30 Days After Discharge.
| Pre-GBR period | Post-GBR period | DID estimatesa | ||||
|---|---|---|---|---|---|---|
| Maryland | Control states | Maryland | Control states | DID estimate (95% CI), pp | P value | |
| All surgeries | ||||||
| Total Medicare payments, mean (SD), $ | 31 720 (36 494) | 20 676 (26 467) | 28 312 (31 123) | 16 583 (21 087) | 646 (−208 to 1542) | .14 |
| Any all-cause readmission, % | 13.5 | 8.9 | 10.4 | 7.6 | −2.2 (−4.3 to −0.1) | .04 |
| Any ED visit, % | 16.1 | 13.9 | 14.6 | 12.6 | −1.0 (−3.4 to 1.4) | .42 |
| Death, % | 2.5 | 2.1 | 1.6 | 1.3 | −0.4 (−1.4 to 0.5) | .36 |
| Major cancer surgeries | ||||||
| Total Medicare payments, mean (SD), $ | 47 137 (42 465) | 34 923 (35 789) | 41 955 (36 313) | 28 179 (29 048) | 1471 (−1140.86 to 4307.42) | .28 |
| Any all-cause readmission, % | 20.1 | 15.6 | 17.4 | 13.2 | −0.6 (−5.1 to 4.0) | .80 |
| Any ED visit, % | 22.1 | 20.4 | 22.1 | 18.9 | 1.3 (−3.6 to 6.2) | .59 |
| Death, % | 4.6 | 3.8 | 2.8 | 2.8 | −1.3 (−3.5 to 0.9) | .24 |
| Discharged to SNF or IRF, %b | 14.3 | 17.2 | 13.7 | 14.3 | 2.1 (−2.1 to 6.3) | .33 |
Abbreviations: DID, difference-in-differences; ED, emergency department; GBR, global budget revenue; IRF, inpatient rehabilitation facility; pp, percentage points; SNF, skilled nursing facility.
Models for all surgeries were adjusted for hospital fixed effect, year fixed effect, and patient characteristics, including race and ethnicity, dual eligibility at surgery month, cancer type, surgery setting (inpatient or outpatient), and chronic obstructive pulmonary disease. Models for major cancer surgeries were adjusted for hospital fixed effect, year fixed effect, race and ethnicity, dual eligibility at surgery month, surgery type, and chronic obstructive pulmonary disease.
Patients who were not discharged home, to SNF, or to IRF were excluded, leaving 8054 patients (6431 from Maryland; 1623 from control hospitals) in the analysis. Excluded patients were those who died (n = 148) or were discharged to another hospital (short-term, long-term, federal, critical access hospital, n = 248), hospice (n = 34), or other care facilities (intermediate care facility, Medicaid certified nursing home, other health care institution, n = 41). We additionally excluded 19 patients to ensure Maryland hospitals and their matched controls had data from the same years.
Changes in Postoperative Outcomes Post-GBR
Thirty-day postoperative mortality and all-cause readmission were more frequent in Maryland than in the control states in the pre-GBR period and decreased in both groups in the post-GBR period (Table 3). In difference-in-differences analyses, GBR implementation was statistically associated with a decrease of 2.2 percentage points (95% CI, −4.3 to −0.1) in all-cause readmissions. Global budget revenue was not statistically associated with ED visits or mortality (Table 3).
Sensitivity Analyses
Among 6858 patients in Maryland and 1686 patients in control states who underwent major cancer surgery, surgical episode payments, mortality, and readmission rates declined between the pre-GBR and post-GBR periods (Table 2 and eTable 3 in the Supplement). We did not observe any statistically significant associations between GBR implementation and these outcomes (Table 3), nor was GBR implementation statistically associated with a change in likelihood of discharge to a skilled nursing or inpatient rehabilitation facility vs home among patients undergoing major cancer surgery (Table 3).
Discussion
In this analysis of health care spending and perioperative outcomes for Medicare beneficiaries who underwent cancer-directed surgery, 30-day all-cause readmissions declined by 2.2 percentage points more in Maryland than in control states after GBR implementation. Changes in 30-day episode expenditures, ED visits, and all-cause mortality identified did not differ between Maryland and control states. These results should be contextualized within the comprehensive program evaluation that estimated approximately $975 million in total Medicare savings and reductions in utilization and spending.4 Specifically, they highlight the importance of systematically investigating the effectiveness of broad-based policy interventions in terms of their value to different stakeholder groups.
Despite the null findings with respect to expenditures, mortality, and ED visits, hospitals appear to be partially responding to the quality targets of GBR by reducing all-cause readmissions among Medicare beneficiaries undergoing cancer surgeries. It is possible that the absence of an association between GBR implementation and expenditures related to cancer surgery was due to the attention hospitals and states gave to other conditions and procedures. The state-funded implementation programs for GBR focused on chronic medical conditions such as chronic obstructive pulmonary disease, congestive heart failure, diabetes, and some high-volume surgical procedures such as joint replacements. Health care administrators may not have viewed surgical oncology as an important opportunity for savings with respect to utilization management and care redesign, perhaps because of the complexity and heterogeneity of these services.
Another explanation may be a misalignment of incentives across the physician groups that interact with the Maryland GBR during surgical care episodes. Specifically, payments for physician services are not included in GBR revenue caps, limiting the program’s influence on surgeon behavior.28 Surgeon practices may be sensitive to motivational factors that include but are not limited to financial rewards.29,30 In 2017, GBR introduced an amendment called the Care Redesign Program, which was designed to engage hospital- and community-based clinicians in the task of improving inpatient care and care transitions respectively. Furthermore, in 2019, GBR was succeeded by the Total Cost of Care model, which limits per capita growth in total expenditures, including physician payments, for Maryland Medicare fee-for-service beneficiaries and also included a Care Redesign Program track centered on episodic care such as surgery.4 These features may engender more accountability among surgeons, although cancer-specific episodes have yet to be identified or implemented. We propose that surgical cancer care be considered as a future expansion area.
Further, it is possible that GBR may not have achieved savings in cancer surgery–related expenditures because of the significant downward trends in surgical expenditures observed in Maryland and control states over this period. Because GBR aimed to achieve savings by halting the growth of health care expenditures, the policy may have little effect on costs in health care sectors with stable or declining expenditures. Future research is needed to better elucidate trends in Medicare spending for surgical oncology.
Our results also call to attention to the need to iterate GBR with a broader suite of meaningful, condition-specific quality metrics. Although GBR applies to all inpatient care, the associated quality targets are more relevant to chronic conditions other than cancer. High-quality cancer care is characterized by care coordination, symptom management, patient-physician communication, palliative care services, and reduced treatment intensity at the end of life.31 For surgical oncology specifically, structural metrics, such as use of multidisciplinary care teams and attainment of case volume thresholds for complex procedures, may be more appropriate indicators of quality.32
Limitations
This study has several limitations. First is the challenge in finding a comparison group of patients with cancer who have parallel baseline period trends given the statewide implementation of GBR and Maryland’s unique all-payer rate setting, which has affected Medicare, Medicaid, and commercial insurance plans since 1974. Various strategies have been used by researchers to mitigate this challenge, such as county-level matching, propensity-score weighting, and matching on hospital-level characteristics.10,15,28,33 We performed hospital-level matching because surgical procedures of interest typically occur within the hospital setting, the locus of GBR-incented care redesign, and to reduce potential bias from the unmeasured cluster-level confounding factors typical in multilevel medical data (eg, geography and treating physicians).34,35 It is possible that unobserved confounding might still bias our results despite evidence for nondifferential baseline trends.
Second, our analysis leveraged Medicare fee-for-service claims, although GBR included all payers in the state. It is unclear if our findings for Medicare beneficiaries undergoing cancer surgery are generalizable to all Maryland patients with cancer who undergo surgery.28 Recent work leveraging the all-payer state inpatient database to assess the association of GBR with complex surgical procedures reported reduced health expenditures and improved clinical outcomes in Maryland relative to control states.8 Interestingly, that study reported a shift toward more commercially insured (and younger) patients with less comorbidity in Maryland post-GBR implementation. This shift strongly suggests favorable patient selection (“cherry picking”), which would not be detected in our Medicare-only claims analysis. Furthermore, the article by Aliu et al8 focused on hospital-based care only rather than 30-day episodes of care because the State Inpatient Database does not include patient identifiers. Our 30-day time horizon is more robust because it encompasses all sites of care and is aligned with the current framework for surgical care in alternative payment models.
Third, GBR overlapped with the implementation of some Affordable Care Act provisions, which could also introduce unmeasured confounding. Fourth, although Medicare data have been used extensively in health services research, coding errors and irregularities are possible.
Conclusions
In recent months, there has been considerable interest in global budgets in light of the economic resiliency it provided to Maryland hospitals amid the COVID-19 pandemic.36 Our results reveal that GBR was not associated with any meaningful variation in expenditures and most clinical outcomes for Maryland patients undergoing cancer-directed surgery. These findings underscore the need for more nuanced examination and future monitoring of broad-based payment models across various population types, including individuals with serious illnesses, such as cancer, that require hospital-based treatments.
eMethods.
eTable 1. Selection of cancer surgery episodes in 2013 in Maryland and control states
eTable 2. Patient characteristics in the main cohort (all inpatient and outpatient cancer surgical episodes from Maryland and matched control hospitals)
eTable 3. Characteristics of patients receiving a major cancer surgery
eFigure 1. Annual difference-in-differences estimates: main cohort
eFigure 2. Annual difference-in-differences estimates: inpatient major cancer surgery subcohort
eTable 4. Surgery codes for identifying cancer surgeries
eReferences
References
- 1.Mariotto AB, Yabroff KR, Shao Y, Feuer EJ, Brown ML. Projections of the cost of cancer care in the United States: 2010-2020. J Natl Cancer Inst. 2011;103(2):117-128. doi: 10.1093/jnci/djq495 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Burwell SM. Setting value-based payment goals: HHS efforts to improve U.S. health care. N Engl J Med. 2015;372(10):897-899. doi: 10.1056/NEJMp1500445 [DOI] [PubMed] [Google Scholar]
- 3.Rajkumar R, Patel A, Murphy K, et al. Maryland’s all-payer approach to delivery-system reform. N Engl J Med. 2014;370(6):493-495. doi: 10.1056/NEJMp1314868 [DOI] [PubMed] [Google Scholar]
- 4.Centers for Medicare & Medicaid Services . Evaluation of the Maryland all-payer model: volume 1: final report. Published November 2019. https://downloads.cms.gov/files/md-allpayer-finalevalrpt.pdf
- 5.Centers for Medicare & Medicaid Services . Regional budget payment concept. Updated May 4, 2021. https://innovation.cms.gov/innovation-models/regional-budget-payment
- 6.Brooks GA, Li L, Uno H, Hassett MJ, Landon BE, Schrag D. Acute hospital care is the chief driver of regional spending variation in Medicare patients with advanced cancer. Health Aff (Millwood). 2014;33(10):1793-1800. doi: 10.1377/hlthaff.2014.0280 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Kaye DR, Luckenbaugh AN, Oerline M, et al. Understanding the costs associated with surgical care delivery in the Medicare population. Ann Surg. 2020;271(1):23-28. doi: 10.1097/SLA.0000000000003165 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Aliu O, Lee AWP, Efron JE, Higgins RSD, Butler CE, Offodile AC II. Assessment of costs and care quality associated with major surgical procedures after implementation of Maryland’s capitated budget model. JAMA Netw Open. 2021;4(9):e2126619. doi: 10.1001/jamanetworkopen.2021.26619 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP; STROBE Initiative . The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Ann Intern Med. 2007;147(8):573-577. doi: 10.7326/0003-4819-147-8-200710160-00010 [DOI] [PubMed] [Google Scholar]
- 10.Morrison M, Haber S, Beil H, Giuriceo K, Sapra K. Impacts of Maryland’s global budgets on Medicare and commercial spending and utilization. Med Care Res Rev. 2021;78(6):725-735. doi: 10.1177/1077558720954693 [DOI] [PubMed] [Google Scholar]
- 11.Chhabra KR, Ghaferi AA, Yang J, Thumma JR, Dimick JB, Tsai TC. Relationship between health care spending and clinical outcomes in bariatric surgery: implications for Medicare bundled payments. Ann Surg. 2022;275(2):356-362. doi: 10.1097/SLA.0000000000003979 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Offodile AC II, Mehtsun W, Stimson CJ, Aloia T. An overview of bundled payments for surgical oncologists: origins, progress to date, terminology, and future directions. Ann Surg Oncol. 2019;26(1):3-7. doi: 10.1245/s10434-018-7037-x [DOI] [PubMed] [Google Scholar]
- 13.Ellimoottil C, Ryan AM, Hou H, Dupree JM, Hallstrom B, Miller DC. Implications of the definition of an episode of care used in the comprehensive care for joint replacement model. JAMA Surg. 2017;152(1):49-54. doi: 10.1001/jamasurg.2016.3098 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Center for Medicare & Medicaid Innovation . Evaluation of the oncology care model: performance periods 1-5. Published January 2021. https://innovation.cms.gov/data-and-reports/2021/ocm-evaluation-pp1-5
- 15.Roberts ET, Hatfield LA, McWilliams JM, et al. Changes in hospital utilization three years into Maryland’s global budget program for rural hospitals. Health Aff (Millwood). 2018;37(4):644-653. doi: 10.1377/hlthaff.2018.0112 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Spetz J, Donaldson N, Aydin C, Brown DS. How many nurses per patient? measurements of nurse staffing in health services research. Health Serv Res. 2008;43(5 pt 1):1674-1692. doi: 10.1111/j.1475-6773.2008.00850.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res. 2011;46(3):399-424. doi: 10.1080/00273171.2011.568786 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Austin PC. Using the standardized difference to compare the prevalence of a binary variable between two groups in observational research. Commun Stat Simul Comput. 2009; 38(6): 1228-1234. doi: 10.1080/03610910902859574 [DOI] [Google Scholar]
- 19.Dummit LA, Kahvecioglu D, Marrufo G, et al. Association between hospital participation in a Medicare bundled payment initiative and payments and quality outcomes for lower extremity joint replacement episodes. JAMA. 2016;316(12):1267-1278. doi: 10.1001/jama.2016.12717 [DOI] [PubMed] [Google Scholar]
- 20.Bureau of Labor Statistics . Measuring price change in the CPI: medical care. Modified February 7, 2022. https://www.bls.gov/cpi/factsheets/medical-care.htm
- 21.Centers for Medicare & Medicaid Services . 2016. All-cause hospital-wide measure updates and specifications report: version 5.0.
- 22.Centers for Medicare & Medicaid Services . 2019. Hospital-wide readmission measure updates and specifications report: version 8.0.
- 23.Schmitt M. Do hospital mergers reduce costs? J Health Econ. 2017;52:74-94. doi: 10.1016/j.jhealeco.2017.01.007 [DOI] [PubMed] [Google Scholar]
- 24.Birkmeyer JD, Siewers AE, Finlayson EVA, et al. Hospital volume and surgical mortality in the United States. N Engl J Med. 2002;346(15):1128-1137. doi: 10.1056/NEJMsa012337 [DOI] [PubMed] [Google Scholar]
- 25.Ho V, Town RJ, Heslin MJ. Regionalization versus competition in complex cancer surgery. Health Econ Policy Law. 2007;2(Pt 1):51-71. doi: 10.1017/S1744133106006256 [DOI] [PubMed] [Google Scholar]
- 26.Ivatury SJ, Wong SL. Regionalization, readmissions, and repercussions of major cancer surgery. JAMA Surg. 2018;153(8):727. doi: 10.1001/jamasurg.2018.0403 [DOI] [PubMed] [Google Scholar]
- 27.Hoag JR, Resio BJ, Monsalve AF, et al. Differential safety between top-ranked cancer hospitals and their affiliates for complex cancer surgery. JAMA Netw Open. 2019;2(4):e191912. doi: 10.1001/jamanetworkopen.2019.1912 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.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 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Leitch KK, Walker PM. Surgeon compensation and motivation. Arch Surg. 2000;135(6):708-712. doi: 10.1001/archsurg.135.6.708 [DOI] [PubMed] [Google Scholar]
- 30.Offodile AC II, Sen AP, Holtsmith S, et al. Harnessing behavioral economics principles to promote better surgeon accountability for operating room cost: a prospective study. J Am Coll Surg. 2020;230(4):585-593. doi: 10.1016/j.jamcollsurg.2019.12.013 [DOI] [PubMed] [Google Scholar]
- 31.Levit LA, Balogh EP, Nass SJ, Ganz PA, eds. Delivering High-Quality Cancer Care: Charting a New Course for a System in Crisis. National Academies Press; 2013. doi: 10.17226/18359 [DOI] [PubMed] [Google Scholar]
- 32.Matula SR, Mercado C, Ko CY, Tomlinson JS. Quality of care in surgical oncology. Cancer Control. 2009;16(4):303-311. doi: 10.1177/107327480901600404 [DOI] [PubMed] [Google Scholar]
- 33.Beil H, Haber SG, Giuriceo K, et al. Maryland’s global hospital budgets: impacts on Medicare cost and utilization for the first 3 years. Med Care. 2019;57(6):417-424. doi: 10.1097/MLR.0000000000001118 [DOI] [PubMed] [Google Scholar]
- 34.Goldstein H, Browne W, Rasbash J. Multilevel modelling of medical data. Stat Med. 2002;21(21):3291-3315. doi: 10.1002/sim.1264 [DOI] [PubMed] [Google Scholar]
- 35.Ali MS, Prieto-Alhambra D, Lopes LC, et al. Propensity score methods in health technology assessment: principles, extended applications, and recent advances. Front Pharmacol. 2019;10:973. doi: 10.3389/fphar.2019.00973 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Levy JF, Ippolito BN, Jain A. Hospital revenue under Maryland’s total cost of care model during the COVID-19 pandemic, March-July 2020. JAMA. 2021;325(4):398-400. doi: 10.1001/jama.2020.22149 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods.
eTable 1. Selection of cancer surgery episodes in 2013 in Maryland and control states
eTable 2. Patient characteristics in the main cohort (all inpatient and outpatient cancer surgical episodes from Maryland and matched control hospitals)
eTable 3. Characteristics of patients receiving a major cancer surgery
eFigure 1. Annual difference-in-differences estimates: main cohort
eFigure 2. Annual difference-in-differences estimates: inpatient major cancer surgery subcohort
eTable 4. Surgery codes for identifying cancer surgeries
eReferences