Is the Centralization of Complex Surgical Procedures an Unintended Spillover Effect of Global Capitation? – Insights from the Maryland Global Budget Revenue Program

Abstract

Objective: 

To determine if global budget revenue (GBR) models incent the centralization of complex surgical care.

Summary Background: 

In 2014, Maryland initiated a statewide GBR model. While prior research has shown improvements in cost and outcomes for surgical care post-GBR implementation, the mechanism remains unclear.

Methods: 

Utilizing state inpatient databases, we compared the proportion of adults undergoing elective complex surgeries (gastrectomy, pneumonectomy/lobectomy, proctectomies, and hip/knee revision) at high-concentration hospitals (HCHs) in Maryland and control states. Annual concentration, per procedure, was defined as hospital volume divided by state volume. HCHs were defined as hospitals with a concentration at least at the 75^th percentile in 2010. We estimated the difference-in-differences (DiD) of the probability of patients undergoing surgery at HCHs before and after GBR implementation.

Findings: 

Our sample included 122,882 surgeries. Following GBR implementation, all procedures were increasingly performed at HCHs in Maryland. States satisfied the parallel trends assumption for the centralization of gastrectomy and pneumonectomy/lobectomy. Post-GBR, patients were more likely to undergo gastrectomy (DiD: 5.5 p.p., 95% CI [2.2, 8.8]) and pneumonectomy/lobectomy (DiD: 12.4 p.p., 95% CI [10.0, 14.8]) at an HCH in Maryland compared with control states. For our hip/knee revision analyses, we assumed persistent counterfactuals and noted a positive DiD post-GBR implementation (DiD: 4.8 p.p., 95% CI [1.3, 8.2]). No conclusion could be drawn for proctectomy due to different pre-GBR trends.

Conclusions: 

GBR implementation is associated with increased centralization for certain complex surgeries. Future research is needed to explore the impact of centralization on patient experience and access.

The United States spent 19.7% of the gross domestic product on health care in 2020.¹ This is almost twice as much as the average OECD country, yet the United states continues to lag in many population health measures.² As a result, policymakers are deeply interested in leveraging alternative payment models (APMs) as a tool for controlling the growth in health care costs and improving care quality.

In 1977, Maryland received a waiver from the Centers for Medicare and Medicaid Services (CMS) that allowed the state’s rate-setting commission to establish a hospital-specific uniform rate for health care services across all payers (ie, Medicare, Medicaid, and commercial patients).³ This waiver granted significant control over their health care spending. Later, when concerns about excessive volume and spending arose, Maryland introduced the global budget revenue (GBR) model in 2014.⁴ Implemented statewide, each hospital’s budget is prospectively set by trending its historical revenue forward and adjusting for factors such as inflation, patient volume shifts, attainment of quality benchmarks, and market trends.^5–7 This policy similarly includes all payers (Medicare, Medicaid, and commercial) and requires the state to meet specific spending (eg, expectation for $330 million or more in Medicare savings over 5 years) and quality targets, including reductions in rates of hospital readmissions and hospital-acquired conditions.⁸ The latter entails participation in the Maryland Hospital-Acquired Conditions Program, which focuses on complications that result from the course of care and treatment delivered and not on the progression of the underlying disease.⁹

Prior research has documented a statistically significant association between Maryland’s implementation of a GBR model and a dampening of hospitalization cost growth and reductions in the rate of avoidable complications and readmissions following major surgery.^10,11 However, there is growing recognition among policymakers and health service researchers that broad-based APMs, such as GBR, might have wider consequences or “spillover effects” than those intended at the onset.^12,13 These spillovers have significant ramifications for the evaluation of the performance of these APMs and the design of future payment reform.¹³ We postulate that the post-GBR period will be associated with an increase in the centralization of care for patients with complex surgical needs. This is because the clinical accountability for cost and outcomes is significantly stronger for hospitals in Maryland relative to other states where the prevailing fee-for-service system reimburses any activity, for example, 30-day readmission or avoidable complication. Consequently, Maryland hospitals are strongly incented by the GBR to allocate effort and resources towards the provision of cost-efficient and high-quality surgical services. In this environment, patients with complex surgical care needs may represent an excessive financial risk for low-volume hospitals and result in steerage towards high-volume hospitals where the necessary infrastructure and processes in place, for example, the use of care redesign pathways, investments in ancillary services and technologies (eg, interventional radiology, advanced imaging, intensive care), use of multidisciplinary treatment teams, and care transition programs.¹⁴

Our aim was to examine the association between Maryland’s implementation of the GBR model and the extent to which complex surgical procedures were centralized, if at all. We focused on a broad range of complex surgical patients and hypothesized that GBR implementation would incentivize Maryland hospitals to reallocate the care for these patients around high-concentration hospitals (HCHs), ie, centralization, as a strategy to reduce cost overruns and improve outcomes.

METHODS

Study Design

This synthetic panel study leveraged a difference-in-differences (DiD) framework to compare the change in centralization for complex surgery before (2010–2013) and after (2014–2017) implementation of the GBR model in Maryland. This study was approved by the Johns Hopkins Hospital Institutional Review Boards and was deemed exempt from informed consent, given our use of publicly available and de-identified data. Our reporting of results followed the Strengthening the Reporting of Observational Studies in Epidemiology reporting guidelines.¹⁵

Data Source

We utilized the Healthcare Cost and Utilization Project state inpatient databases (SIDs) from 2010 to 2017. Given the statewide implementation of the GBR model in Maryland, our comparison population was drawn from New Jersey and New York. These states were selected because they shared a similar record and timing of Medicaid expansion with Maryland (January 1, 2014). This also satisfies the “common shocks” assumption for our DiD analysis, that is, an exogenous event that could potentially influence our outcome of interest but affects both treatment and control groups equally.^16,17 Due to our interest in events associated with the index hospitalization (ie, site of care receipt), the all-payer SIDs provided the most complete and representative study population, capturing 97% of all eligible hospital discharges. Other databases would have been more segmented, that is, focused on public (Medicare, Medicaid) or commercial payers (Optum Clinformatics Data Mart, Health Care Cost Institute, or IBM MarketScan Research Databases) alone.¹⁰

Study Population

The study sample included adult patients (≥18 y) who underwent elective gastrectomy, pneumonectomy/lobectomy, proctectomy, or revision hip/knee arthroplasty between January 1, 2010, and December 31, 2017. These procedures were chosen because they are high cost, associated with considerable morbidity, and frequently the focus of centralization policies.^18–29 Three eligible complex procedures (esophagectomy, cystectomy, and pancreatectomy) were excluded as they were noted to be highly centralized in Maryland before GBR enactment (Supplemental Figure S1, Supplemental Digital Content 1, https://links.lww.com/SLA/E319 and Supplemental Table S1, Supplemental Digital Content 1, https://links.lww.com/SLA/E319).³⁰ Emergency procedures were excluded given their noted association with increased costs and higher risk of major complications.³¹ In addition, the Emergency Medical Treatment and Labor Act (EMTALA) mandates the stabilization and initial treatment for emergency conditions before transfer, potentially confounding our calculation of HCHs and our ability to reliably identify the site of receipt of definitive surgical care.³² We also initially excluded patients who received surgeries in the 10 rural hospitals participating in the Total Patient Revenue (TPR) model, an antecedent program to the GBR, starting in 2010. For these 10 hospitals, the payment policy (ie, implementation of caps on hospital revenue) changed in 2010 instead of 2014. Patients undergoing gastrectomy, pneumonectomy/lobectomy, or proctectomy were identified using the appropriate codes from the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) and the International Classification of Diseases, Tenth Revision, Procedure Coding System (ICD-10-PCS). Patients receiving gastrectomy with morbid obesity (ICD-9-CM 278.01; ICD-10-CM E66.01) as the primary diagnosis were excluded. We did not limit other procedures by diagnosis. Health care Cost and Utilization Project assigned the Diagnosis Related Group (DRG) for each discharge record, regardless of the payer, using the Medicare DRG Grouper algorithm. Patients with hip/knee revision were identified using DRG 466, 467, and 468.

Assessment of Centralization

To evaluate the degree of centralization for each complex surgical procedure, we calculated the hospital “concentration” each year, defined as the hospital volume divided by the total volume in the state. The hospital concentration was then joined to the patient sample.²⁶ We then used our first observation period (2010) to establish the quartiles for hospital concentration and defined the top quartile (> 75th percentile) as the cut-off for designating HCHs in each subsequent year. This means the cut-off was held constant throughout the succeeding time periods of 2011 to 2017.²⁶ This approach is consistent with published work and allows us to see the changes in the distribution of patients in HCHs, that is, the degree of centralization, with time.²⁶

Statistical Analysis

Demographic and clinical characteristics of the study population in both Maryland and comparison states were described in Table 1. For each surgery, the proportions of patients receiving surgery in HCHs each year was calculated for both Maryland and comparison states (Fig. 1). To evaluate the difference of changes in centralization between Maryland and comparison states, we estimated the DiD in the probability of having surgery done is an HCH using a linear probability model, adjusted for patient age, sex, race/ethnicity (Asian/Pacific Islander, Black, Hispanic, Native American, White, or other), insurance type (Medicare, Medicaid, private insurance, self-pay, no charge, or other), comorbidity, hospital teaching status, health system membership, bed size, and location (rural or urban). These covariates were chosen on account of their established relationships with inter-facility surgical transfers, referrals to high-volume hospitals, and care regionalization.^32,33 Patients with unknown sex, race/ethnicity, or insurance type (< 1%) were excluded from the adjusted analysis. In this model, the DiD estimator is expressed as an interaction term between the dichotomous indicator for Maryland versus comparison states and another dichotomous indicator for the years before GBR (2010–2013) versus after (2014–2017). The parallel trends assumption was examined using the data before GBR implementation.

TABLE 1 - Patient Characteristics in Maryland and Control States

	Gastrectomy			Pneumonectomy/Lobectomy			Proctectomy			Hip/Knee Revision
	MD	Control	Std. Diff	MD	Control	Std. Diff	MD	Control	Std. Diff	MD	Control	Std. Diff
n, discharges	2708	13776	—	4847	25132	—	3383	20970	—	10124	41942	—
n, hospitals	37	192	—	32	162	—	37	204	—	38	202	—
Patient characteristics, n (%)
Age, mean±SD	61.5±13.3	61.0±14.1	0.03	65.5±11.0	66.5±10.9	0.09	60.2±14.4	61.4±14.5	0.09	64.2±11.6	65.2±11.7	0.09
<50	487 (18.0)	2840 (20.6)	—	355 (7.3)	1624 (6.5)	—	723 (21.4)	4078 (19.4)	—	959 (9.5)	3663 (8.7)	—
50-59	606 (22.4)	3015 (21.9)	—	952 (19.6)	4379 (17.4)	—	869 (25.7)	5139 (24.5)	—	2577 (25.5)	9535 (22.7)	—
60-69	835 (30.8)	3759 (27.3)	—	1692 (34.9)	8375 (33.3)	—	899 (26.6)	5379 (25.7)	—	3262 (32.2)	13283 (31.7)	—
70-79	575 (21.2)	3019 (21.9)	—	1445 (29.8)	8427 (33.5)	—	582 (17.2)	4058 (19.4)	—	2315 (22.9)	10513 (25.1)	—
80+	205 (7.6)	1143 (8.3)	—	403 (8.3)	2327 (9.3)	—	310 (9.2)	2316 (11.0)	—	1011 (10.0)	4948 (11.8)	—
Sex	—	—	0.02	—	—	0.02	—	—	0.02	—	—	0.04
Male	1372 (50.7)	7174 (52.1)	—	2177 (44.9)	11146 (44.3)	—	1667 (49.3)	10098 (48.2)	—	4120 (40.7)	17830 (42.5)	—
Female	1335 (49.3)	6596 (47.9)	—	2669 (55.1)	13980 (55.6)	—	1715 (50.7)	10871 (51.8)	—	6003 (59.3)	24109 (57.5)	—
Unknown	1 (0.0)	6 (0.0)	—	1 (0.0)	6 (0.0)	—	1 (0.0)	1 (0.0)	—	1 (0.0)	3 (0.0)	—
Race/ethnicity	—	—	0.49	—	—	0.47	—	—	0.52	—	—	0.61
Asian/Pacific Islander	117 (4.3)	1142 (8.3)	—	126 (2.6)	1189 (4.7)	—	101 (3.0)	825 (3.9)	—	68 (0.7)	367 (0.9)	—
Black	603 (22.3)	1592 (11.6)	—	911 (18.8)	1793 (7.1)	—	580 (17.1)	1413 (6.7)	—	2532 (25.0)	4708 (11.2)	—
Hispanic	83 (3.1)	1435 (10.4)	—	48 (1.0)	1126 (4.5)	—	80 (2.4)	1474 (7.0)	—	141 (1.4)	2463 (5.9)	—
Native American	5 (0.2)	15 (0.1)	—	7 (0.1)	26 (0.1)	—	4 (0.1)	27 (0.1)	—	13 (0.1)	64 (0.2)	—
Other	69 (2.5)	1207 (8.8)	—	97 (2.0)	1776 (7.1)	—	80 (2.4)	1388 (6.6)	—	113 (1.1)	2315 (5.5)	—
White	1812 (66.9)	8331 (60.5)	—	3590 (74.1)	19143 (76.2)	—	2472 (73.1)	15756 (75.1)	—	6914 (68.3)	31880 (76.0)	—
Unknown	19 (0.7)	54 (0.4)	—	68 (1.4)	79 (0.3)	—	66 (2.0)	87 (0.4)	—	343 (3.4)	145 (0.3)	—
Insurance type	—	—	0.26	—	—	0.04	—	—	0.17	—	—	0.23
Medicare	1163 (42.9)	5489 (39.8)	—	2692 (55.5)	13988 (55.7)	—	1261 (37.3)	8471 (40.4)	—	5321 (52.6)	22720 (54.2)	—
Medicaid	178 (6.6)	1835 (13.3)	—	270 (5.6)	1824 (7.3)	—	249 (7.4)	1852 (8.8)	—	532 (5.3)	2047 (4.9)	—
Private insurance	1271 (46.9)	6083 (44.2)	—	1763 (36.4)	8779 (34.9)	—	1732 (51.2)	10065 (48.0)	—	3779 (37.3)	14053 (33.5)	—
Self-pay	19 (0.7)	185 (1.3)	—	31 (0.6)	221 (0.9)	—	38 (1.1)	300 (1.4)	—	31 (0.3)	399 (1.0)	—
No charge	7 (0.3)	30 (0.2)	—	10 (0.2)	27 (0.1)	—	11 (0.3)	55 (0.3)	—	7 (0.1)	24 (0.1)	—
Other	67 (2.5)	154 (1.1)	—	80 (1.7)	292 (1.2)	—	91 (2.7)	225 (1.1)	—	452 (4.5)	2698 (6.4)	—
Unknown	3 (0.1)	0 (0.0)	—	1 (0.0)	1 (0.0)	—	1 (0.0)	2 (0.0)	—	2 (0.0)	1 (0.0)	—
Elixhauser comorbidity		0.33	—		0.26	—		0.29	—	—	0.22	—
0	165 (6.1)	1563 (11.3)	—	359 (7.4)	2325 (9.3)	—	502 (14.8)	4184 (20.0)	—	986 (9.7)	5546 (13.2)	—
1	424 (15.7)	3085 (22.4)	—	722 (14.9)	5149 (20.5)	—	721 (21.3)	5634 (26.9)	—	2009 (19.8)	9972 (23.8)	—
2	535 (19.8)	3300 (24.0)	—	1025 (21.1)	6081 (24.2)	—	738 (21.8)	4840 (23.1)	—	2318 (22.9)	10782 (25.7)	—
3	574 (21.2)	2525 (18.3)	—	954 (19.7)	5223 (20.8)	—	594 (17.6)	3116 (14.9)	—	2111 (20.9)	7729 (18.4)	—
4	440 (16.2)	1569 (11.4)	—	806 (16.6)	3208 (12.8)	—	383 (11.3)	1667 (7.9)	—	1324 (13.1)	4370 (10.4)	—
5+	570 (21.0)	1734 (12.6)	—	981 (20.2)	3146 (12.5)	—	445 (13.2)	1529 (7.3)	—	1376 (13.6)	3543 (8.4)	—
Hospital characteristics, n (%)
Location	—	—	0.07	—	—	0.12	—	—	0.06			0.08
Rural	5 (0.2)	92 (0.7)	—	0 (0.0)	190 (0.8)	—	42 (1.2)	408 (1.9)	—	113 (1.1)	896 (2.1)
Urban	2703 (99.8)	13684 (99.3)	—	4847 (100)	24942 (99.2)	—	3341 (98.8)	20562 (98.1)	—	10011 (98.9)	41046 (97.9)
Teaching status	—	—	0.15	—	—	0.56	—	—	0.21			0.67
No	1033 (38.1)	4250 (30.9)	—	2949 (60.8)	8557 (34.0)	—	1812 (53.6)	9079 (43.3)	—	8467 (83.6)	22704 (54.1)	—
Yes	1675 (61.9)	9526 (69.1)	—	1898 (39.2)	16575 (66.0)	—	1571 (46.4)	11891 (56.7)	—	1657 (16.4)	19238 (45.9)	—
System membership	—	0.50	—	—	0.27	—	—	0.31	—	—	0.24	—
No	523 (19.3)	5764 (41.8)	—	1482 (30.6)	10900 (43.4)	—	913 (27.0)	8680 (41.4)	—	2711 (26.8)	15884 (37.9)	—
Yes	2185 (80.7)	8012 (58.2)	—	3365 (69.4)	14232 (56.6)	—	2470 (73.0)	12290 (58.6)	—	7413 (73.2)	26058 (62.1)	—
Licensed beds	—	—	0.67	—	—	0.78	—	—	0.64		—	0.89
<250	233 (8.6)	1525 (11.1)	—	704 (14.5)	3027 (12.0)	—	484 (14.3)	2112 (10.1)	—	2818 (27.8)	12032 (28.7)	—
250-499	1106 (40.8)	5304 (38.5)	—	2852 (58.8)	9580 (38.1)	—	1775 (52.5)	9148 (43.6)	—	6593 (65.1)	13803 (32.9)	—
500-749	301 (11.1)	3172 (23.0)	—	396 (8.2)	6200 (24.7)	—	308 (9.1)	4674 (22.3)	—	560 (5.5)	9858 (23.5)	—
750-999	830 (30.6)	1289 (9.4)	—	786 (16.2)	2244 (8.9)	—	731 (21.6)	2050 (9.8)	—	144 (1.4)	2119 (5.1)	—
1000+	238 (8.8)	2486 (18.0)	—	109 (2.2)	4081 (16.2)	—	85 (2.5)	2986 (14.2)	—	9 (0.1)	4130 (9.8)	—

FIGURE 1: Changes in the distribution of patients in HCHs with time. HCH indicates high-concentration hospitals.

When the parallel trends assumption was not met, similar to published work, we attempted to account for differential pre-GBR trends of the proportion of patients receiving care at HCHs between Maryland and comparison states by modifying our DiD model to include an interaction term between the Maryland indicator and a linear year trend.⁷ The counterfactual of this “persistent trends” assumption is that patients in Maryland and comparison states would have followed their separate and different trends throughout the years if the GBR had never been implemented. Our DiD estimates of treatment effect (ie, GBR implementation) are presented as percentage points (p.p.) which can be interpreted as the absolute difference in the change over time in the percent of patients receiving care in HCHs in Maryland relative to control states.

All statistical tests were 2-sided, and p values < 0.05 were considered statistically significant. All analyses were performed with SAS Enterprise Guide version 7.15 (SAS Institute Inc. Cary, NC, USA).

Sensitivity Analyses

To confirm the robustness of our DiD estimates of treatment effect (ie, GBR implementation), we conducted 3 sets of sensitivity analyses. First, we used a 12-month waiting or “washout” period after the GBR model was implemented in Maryland (ie, the excluded calendar year 2014) since meaningful change to clinical practice patterns would take time to develop. This ensures that our estimates are not associated with the immediate post-policy period (“carryover bias”), a time when changes to practice are unlikely to be associated with policy changes. This is consistent with previous studies that have utilized a difference-in-difference framework to study state-level policy interventions, including Maryland’s previous all-payer model or Medicaid expansion following the Affordable Care Act.^10,34,35 Second, we repeated our DiD analysis and included all patients who received complex surgical care at participating hospitals in the TPR program initiated in 2010 in our analytical sample. Lastly, we investigated whether there was any variation in the overall market concentration, symbolized by the state-level Herfindahl Hirschman Index, for our surgical procedures of interest attributable to GBR implementation. This alternative specification of concentration is well represented in the surgical literature and characterizes how market share is distributed across hospital competitors in a market.³⁶

RESULTS

Of the 122,882 total patients, 21,062 were from Maryland and 101,820 were from comparison states. Patient demographics and characteristics categorized by the surgical procedure are shown in Table 1. Our analytic sample was comprised of 16,484 gastrectomies, 29,979 pneumonectomies and lobectomies, 24,353 proctectomies, and 52,066 revision hip/knee arthroplasties. Applying a cut-off of 0.1 to the standardized differences reported in Table 1, our state populations (MD vs. control) were reasonably balanced with respect to age, sex, and hospital location but were largely unbalanced for the race, insurance status, comorbidity burden, hospital size, teaching status, and system membership.

Figure 1 outlines the annual rates of patients receiving care at an HCH for each of our complex surgical procedures of interest over the entire study period. Overall, we observed increased centralization over time in both Maryland and comparison states for gastrectomy, pneumonectomy/lobectomy, and proctectomy. For hip/knee revision, the centralization in Maryland increased while, in comparison states, it decreased slightly in recent years (Fig. 1 and Table 2). Ensuing analysis of the pre-2014 period revealed significant differences in the baseline trend for proctectomy and hip/knee revision (P<0.0001, P<0.0001, respectively, Table 2).

TABLE 2 - Impact of Maryland Global Budget Revenue Program on Centralization of Complex Surgeries

	% Patients in HCHs, Pre-GBR		% Patients in HCHs, Post-GBR			DID Estimates*
	Maryland	Control States	Maryland	Control States	P Value for Pretrends	DID	95% CI	P
Gastrectomy	45.46	28.39	57.96	30.19	0.0543	5.5 p.p.	2.2, 8.8	0.0011
Pneumonectomy /Lobectomy	28.67	27.40	41.81	27.88	0.9147	12.4 p.p.	10.0, 14.8	< 0.0001
Proctectomy	40.28	22.64	53.13	29.53	< 0.0001	4.1 p.p.	1.3, 6.9	0.0037
Hip/Knee Revision	29.59	22.57	39.22	19.95	< 0.0001	9.2 p.p.	7.5, 10.8	< 0.0001

^*The model adjustment included patient age, sex, race/ethnicity, insurance type, comorbidity, hospital teaching status, system membership, bed size, and location (rural or urban).

Control states include New Jersey, and New York.

p.p. indicates percentage points

Degree of Centralization post-GBR

Following GBR implementation, gastrectomy, pneumonectomy/lobectomy, and hip/knee revision were noted to be increasingly performed at HCHs in Maryland (Fig. 1). When compared with changes in comparison states, this difference was statistically significant for gastrectomy (DiD: 5.5 p.p., 95% CI [2.2, 8.8]) and pneumonectomy/lobectomy (DiD: 12.4 p.p., 95% CI [10.0, 14.8]) (Table 2). Given the existing trends for hip/knee revision, additional analysis assuming persistent counterfactual trends before and after GBR revealed a positive DiD in Maryland after the state’s implementation of GBR (DiD: 4.8 p.p., 95% CI [1.3, 8.2], Table 4). We did not perform such analysis for proctectomy because the pretrends did not appear to be linear. Therefore, no conclusion could be drawn for proctectomy due to differences in pre-GBR trends.

Sensitivity Analyses

In the first set of sensitivity analysis that entailed a 12-month waiting period, we report similar directionality to our primary results. Of note, pneumonectomy/lobectomy was associated with a larger DiD (15.5 p.p., 95% CI [13.0, 18.1], Table 3). Similar increases were also appreciated for hip/knee revision after excluding the year of implementation from the post-GBR period (8.6 p.p., 95% CI [4.3, 12.9], Table 4). Upon inclusion of the TPR hospitals, we similarly identified statistically significant increases in centralization, post-GBR, for gastrectomy (DiD: 6.5 p.p., 95% CI [3.2, 9.7]) and pneumonectomy/lobectomy (DiD: 11.4 p.p., 95% CI [9.1, 13.7]) (Supplemental Table S2, Supplemental Digital Content 1, https://links.lww.com/SLA/E319). Our findings were similar when the waiting period was excluded from the analysis (Supplemental Table S3, Supplemental Digital Content 1, https://links.lww.com/SLA/E319). For hip/knee revision, the DiD estimates derived from analyzing the sample, including the TPR hospitals, were very close to what we observed in the main analysis (Supplemental Table S4, Supplemental Digital Content 1, https://links.lww.com/SLA/E319). Finally, we observed, to varying degrees, increased state-level Herfindahl Hirschman Index across each of our procedures of interest: gastrectomy (8.28 p.p.), pneumonectomy/lobectomy (0.02 p.p.), proctectomy (2.19 p.p.), and hip/knee revision (0.33 p.p.). This suggests that post-GBR, there was less competition in the Maryland health care market compared with the control states due to dominant hospitals in Maryland. Complete results are available in Supplemental Table S5, Supplemental Digital Content 1, https://links.lww.com/SLA/E319.

TABLE 3 - Impact of Maryland Global Budget Revenue Program on Centralization of Complex Surgeries, Excluding 2014 Data From the Post Period

	% Patients in HCHs, Pre-GBR		% Patients in HCHs, Post-GBR		DID Estimates*
	Maryland	Control States	Maryland	Control States	DID	95% CI	P
Gastrectomy	45.46	28.39	56.28	29.63	4.3 p.p.	0.9, 7.7	0.0139
Pneumonectomy /Lobectomy	28.67	27.40	44.53	27.50	15.5 p.p.	13.0, 18.1	< 0.0001
Proctectomy	40.28	22.64	57.96	28.54	8.8 p.p.	5.7, 11.8	< 0.0001
Hip/Knee Revision	29.59	22.57	42.24	20.37	10.6 p.p.	8.8, 12.4	< 0.0001

^*The model adjustment included patient age, sex, race/ethnicity, insurance type, comorbidity, hospital teaching status, system membership, bed size, and location (rural or urban).

Control states include New Jersey, and New York.

p.p. indicates percentage points.

TABLE 4 - Impact of Maryland Global Budget Revenue Program on Centralization of Hip/knee Revision, Assuming Persistent Trends

	DID Estimates*
2014 Data in The Post Period	DID	95% CI	P
Yes	4.8 p.p.	1.3, 8.2	0.0065
No	8.6 p.p.	4.3, 12.9	< 0.0001

^*The model adjustment included patient age, sex, race, insurance type, comorbidity, hospital teaching status, system membership, bed size, and location (rural or urban).

p.p. indicates percentage points.

DISCUSSION

In the present study, we found increasing centralization for gastrectomy, pneumonectomy/lobectomy, and hip/knee revisions in Maryland, relative to comparison states, after the implementation of the GBR model in 2014. Similar results were also reported following the application of a 12-month waiting period (ie, 2014 calendar year) and the inclusion of hospital participants in the TPR program. These sensitivity analyses confirmed the robustness of our primary results. Our examination of the association between the GBR model and the centralization of proctectomy was limited due to a violation of the parallel trends’ assumption. Finally, we documented an increase in hospital market concentration for gastrectomy, pneumonectomy/lobectomy, proctectomy, and hip/knee revisions in Maryland following GBR onset. Since the present health care climate is increasingly focused on provider accountability for care quality and costs, these spillover effects of the GBR, namely increased centralization and reduced market competitiveness for complex surgical procedures, will be of great interest to payers, health system leadership, and surgical practitioners. These results highlight the critical need to systematically evaluate broad-based APMs through the lens of different conditions and specialties. Of note, the CMS-mandated evaluation of the GBR program adopted a general medicine focus.

Given the highly complex nature of the organization, financing, and delivery of US health care, it is no surprise that broad-based policy reforms such as the GBR would have consequences, desirable and undesirable, beyond those originally planned.¹² In Maryland, increased centralization of complex surgical care might be partially explained by changed incentive structures. Rather than risk penalties for not achieving quality targets (eg, 30-day readmissions or iatrogenic complications) or cost overruns, low-volume Maryland hospitals may elect to refer complex surgeries to tertiary centers with greater clinical experience, appropriate clinical workflows (ERAS protocols, care pathways) and supporting infrastructure (care transition programs, patient-specific education, multidisciplinary teams, data analytics). These processes and structural elements can reduce the costs associated with complications and extended hospitalizations while maintaining quality targets. Although there have been no public announcements about referral partnerships for complex surgical care in Maryland, a recently published qualitative study about the perspectives of health system leaders on GBR implementation described a positive shift in patient volume at many academic medical centers as they sought to expand services for patients seeking specialized care from other markets.³⁷

Studies consistently find that outcomes for complex surgical procedures improve when performed at high-volume centers. For gastrectomy, high hospital volume is associated with lower postoperative mortality and shorter length of stay.^38–40 One study of 3,711 gastrectomy patients in New York found that patients undergoing surgery at hospitals in the highest-volume quartile had absolute risk-adjusted mortality lower than those at hospitals in the lowest-volume quartile.³⁸ Surgeon volume alone did not make up for the difference. Patients undergoing surgery by above-median-volume surgeons at below-median-volume hospitals had significantly poorer mortality. This may potentially be due to differences in hospital infrastructure or ancillary services.

Research for revision joint arthroplasty is largely consistent. High-volume hospitals and surgeons are associated with improved outcomes, including reduced complication and readmission rates, following hip/knee revisions.^18,21 Perhaps more notable, however, is the data describing disparities in travel burden and access in a hypothetical scenario wherein patients change New York state hospitals for their revision hip/knee surgery. The study of 37,147 revision total joint arthroplasty patients identified increased travel distance for Hispanic patients who transferred care to the nearest higher volume institution from the lowest 50% by volume hospitals (OR 12.3) relative to nonHispanic counterparts. Similar statistically significant findings were also observed for revision total joint arthroplasty patients from counties with low median income and rural counties.¹⁹

Therefore, it is important to monitor and understand the impact of centralization on care access, patient experience, and prevailing health care disparities. First, a willingness and ability to travel to HCH for complex surgical care may not be broadly applicable to all patients.⁴¹ Increased travel distance related to care receipt at an HCH could result in significant caregiver work-life disruption, out-of-pocket costs (lodging, parking, reduction in work hours), and financial toxicity.⁴² This also has important implications for the timely management of postoperative complications and the conduct of surveillance visits in the case of cancer-related surgery.⁴² While high-volume centers are associated with improved surgical outcomes, vulnerable patient populations, including those who are Black or Medicaid beneficiaries, are less likely to have access to these hospitals.²⁰ These patients are more likely to use local low-volume centers even when there is a nearby high-volume center. Policies that encourage centralization, directly or indirectly, should include provisions to increase facilitators of and reduce barriers to patients' access to HCH. Future studies should evaluate the impact of GBR-incented complex surgical care centralization and related variation in health care access and patient-reported outcome measures that characterize patient experience, quality-of-life, and financial toxicity.⁴³

The US has undergone significant provider consolidation in recent decades, with up to 80% of hospital markets being highly concentrated according to federal regulatory guidelines.⁴⁴ Our observation of an anticompetitive effect on the functioning of Maryland markets for complex surgical procedures confirms our primary analysis. Previously described concerns about the impact of market consolidation on the prices of health services and care quality are less relevant in Maryland due to the all-payer hospital cap and documented improvement in clinical outcomes, respectively.⁴⁵

Finally, it is important to understand Maryland’s unique regulatory context before the GBR. The state’s Health Services Cost Review Commission (HSCRC) has had significant authority over health care spending since it received its CMS waiver in 1974, specifically the ability to set all-payer rates for health services until the initiation of the GBR in 2014.⁴⁶ Health care leaders have attributed the bulk of Maryland’s success under the GBR, thus far, to an emphasis on clear expectations, frequent communications, hospital autonomy, and high-quality data.³⁷ Other states should be mindful of these themes as they consider similar payment models.

Limitations

This study has several methodological limitations that merit attention. First, our findings are based on observational data, so we cannot definitively establish causality. While we used a DiD design to reduce any potential confounding or bias, our preferred specification choice would have been to use data from a matched pool of hospitals in Maryland and comparison states. Unfortunately, this was precluded by our state-level analysis and informed our decision to limit the control states to those with a similar record and timing of Medicaid expansion (ie, NY/NJ). This conservative approach was intended to heighten causal inference and limit estimator bias. Second, our analysis was limited to the first few years (2014–2017) following the implementation of the GBR model. Long-term trends may differ from our findings with respect to centralization. Lastly, we used ICD-9 and ICD-10 codes to identify diagnoses at admission and surgical procedures. Although it is unlikely that this would affect Maryland alone, it is possible that the transition from ICD-9 to ICD-10 codes introduced biases or imprecision to our outcomes.

CONCLUSIONS

This study found that Maryland’s transition to GBR may be associated with increased centralization for certain complex surgical procedures. More data is needed to evaluate the long-term impact of GBR on centralization and related changes in patient access, clinical outcomes, and costs. This research is significant given that other states are considering similar global budget models and that the U.S. has a pressing need for meaningful payment and delivery reform.⁴⁷ Lastly, attempts at understanding the full impact of population-based APMs should include a systematic evaluation for unintended consequences and potential spillover effects in surgical populations.¹² This will allow policymakers to optimize the design of these initiatives for societal benefit and patient welfare.