Impact of Hospital Affiliation with a Flagship Hospital System on Surgical Outcomes

STRUCTURED ABSTRACT

Objective:

To compare general surgery outcomes at flagship systems, flagship hospitals, and flagship hospital affiliates versus matched controls.

Summary Background Data:

It is unknown whether flagship hospitals perform better than flagship hospital affiliates for surgical patients.

Methods:

Using Medicare claims for 2018–2019, we matched patients undergoing inpatient general surgery in flagship system hospitals to controls who underwent the same procedure at hospitals outside the system but within the same region. We defined a “flagship hospital” within each region as the major teaching hospital with the highest patient volume that is also part of a hospital system; its system was labeled a “flagship system.” We performed four main comparisons: patients treated at any flagship system hospital versus hospitals outside the flagship system; flagship hospitals versus hospitals outside the flagship system; flagship hospital affiliates versus hospitals outside the flagship system; and flagship hospitals versus affiliate hospitals. Our primary outcome was 30-day mortality.

Results:

We formed 32,228 closely-matched pairs across 35 regions. Patients at flagship system hospitals (32,228 pairs) had lower 30-day mortality than matched control patients (3.79% versus 4.36%, difference=−0.57% [−0.86%,−0.28%], p<0.001). Similarly, patients at flagship hospitals (15,571/32,228 pairs) had lower mortality than control patients. However, patients at flagship hospital affiliates (16,657/32,228 pairs) had similar mortality to matched controls. Flagship hospitals had lower mortality than affiliate hospitals (difference-in-differences=−1.05% [−1.62%,−0.47%], p<0.001).

Conclusions:

Patients treated at flagship hospitals had significantly lower mortality than those treated at flagship hospital affiliates. Hence, flagship system affiliation does not alone imply better surgical outcomes.

MINI-ABSTRACT

Medicare patients undergoing inpatient general surgery at the flagship hospital of a flagship system had significantly lower 30-day mortality than matched control patients at other hospitals outside the flagship system. However, patients treated at affiliate hospitals within the same flagship system did not have different mortality than their matched controls.

INTRODUCTION

Although primarily driven by financial and regulatory considerations,^{1, 2} hospital system mergers and acquisitions are typically presented to patients and policymakers as beneficial to patient care,^{3, 4} such as through improved care coordination across hospitals^{5, 6} or economies of scale that facilitate process efficiency and investments in advanced technology.^5–7 Perhaps most fundamentally, affiliation with a major regional hospital system is often promoted by these same systems as an opportunity for patients to benefit locally from the same standard of care experienced at their flagship hospital.^{8, 9} This branding association with one of the best hospitals in the region offers affiliates the ability to distinguish themselves from surrounding hospitals.

Indeed, there is widespread patient perception,⁸ perhaps misguided,^10–12 that surgical care provided at hospitals affiliated with a major regional system is no different than care at the system’s flagship hospital, suggesting better surgical care at affiliated hospitals than hospitals outside the flagship system. Prior research has focused on what affiliation offers for hospitals by looking at outcomes and financial performance before and after affiliation.^{5, 13–15} However, a vital question remains: should patients and policymakers expect superior surgical outcomes from hospitals affiliated with major regional systems (i.e., “flagship systems”), particularly hospitals other than the flagship hospital, compared to hospitals outside the flagship system?

To explore this, we performed a matched cohort study in 35 of the nation’s largest hospital referral regions (HRRs)^{16, 17} using Medicare claims data to compare surgical outcomes between patients at: 1) flagship system hospitals (i.e., all hospitals within the preeminent regional academic hospital system) versus hospitals within the same HRR but outside the flagship system; 2) flagship hospitals versus within-HRR hospitals outside the flagship system; and 3) affiliate hospitals in flagship systems versus within-HRR hospitals outside the flagship system.

METHODS

Patient Population

We used Medicare administrative claims data (Inpatient, Outpatient, Carrier/Part B, Skilled Nursing Facility, Home Health Agency, and Durable Medical Equipment files) for all fee-for-service Medicare beneficiaries through the Centers for Medicare and Medicaid Services (CMS) Virtual Research Data Center.¹⁸ We analyzed patients 66 years and above who underwent inpatient general surgery procedures between 2018–2019. These patients were categorized into clinically relevant groups by International Classification of Diseases, Tenth Revision (ICD-10) principal procedure codes (Supplemental Digital Content 1, Section A eTable 1). For patients with multiple procedures, we used their first. We excluded patients if, in a one-year lookback prior to their admission, they either: 1) lacked fee-for-service Medicare claims; 2) did not have complete enrollment in Medicare Parts A and B; or 3) were enrolled in a health maintenance organization at any point.

Defining Hospital Systems

The Agency for Healthcare Research and Quality maintains a database of health systems across the United States, which it defines as including “at least one hospital and at least one group of physicians that provid[e] comprehensive care…who are connected with each other and with the hospital through common ownership or joint management.”¹⁶ Using this list for 2018, we defined “hospital systems” as health systems with at least two acute care hospitals within the same HRR. HRRs are geographic areas that share the same tertiary care referral patterns based on Medicare data, reflecting distinct healthcare markets.¹⁷ HRRs have been widely used to describe healthcare utilization and cost.^{14, 19, 20} Of 306 HRRs in the United States, 35 met our volume (N≥20,000) and system criteria as summarized below.

We defined a “major COTH hospital” as a hospital with a resident-to-bed ratio ≥0.25 – consistent with “major” or “very major” teaching hospitals – that is also a member of the Council of Teaching Hospitals and Health Systems (COTH). Within each HRR, we defined the “flagship hospital” as the largest (i.e., highest combined medical and surgical patient volume) major COTH hospital that also had affiliated hospitals within the same HRR. We defined the “flagship system” as the system that included the flagship hospital. Thus, each HRR was defined to have only one flagship hospital and one flagship system. Patients at all other hospitals within the same HRR but not in the flagship system were labeled as potential controls. Potential controls could therefore come from unaffiliated hospitals, hospitals affiliated with other academic centers within the same HRR, or even other major COTH hospitals within the same HRR but not in the flagship system. We investigate this further in the stability analysis described below. See Supplemental Digital Content 1, Section B and eTables 2 and 3 for further detail.

HRRs with Multiple Major COTH Hospitals

Some HRRs contained several major COTH hospitals. In our primary analysis, patients from all such major COTH hospitals that were not labeled part of the flagship system per our criteria were instead included as potential controls. However, we also performed a stability analysis that removed matched pairs containing control patients admitted to these non-flagship system major COTH hospitals. As will be seen, this stability analysis only strengthened our findings.

Defining Multimorbidity

In our previous work,²¹ we defined multimorbidity for older surgical patients as the presence of at least one cluster of comorbidities – termed qualifying comorbidity sets (QCSs) – confidently associated with at least double the odds of 30-day mortality compared to the typical patient undergoing the same procedure in the same age group. We have since refined our multimorbidity definition²² to be compatible with ICD-10 codes and incorporate functional status indicators, allowing us to identify particularly high-risk patients. These updates were completed prior to the present study and utilized data that did not overlap with this study.

Outcomes

The primary outcome was 30-day mortality. We also examined 90-day mortality and an updated 30-day failure-to-rescue outcome, which represents mortality after in-hospital postoperative complications.^23–26 An updated list of complications used for computing failure-to-rescue is provided in Supplemental Digital Content 1, Section C eTable 4.

Statistical Analysis

Matching Methodology

We used optimal subset matching^27–30 to balance many covariates in an optimal manner³¹ to match patients in flagship system hospitals (“treated” hospitals) to control patients in other hospitals outside the flagship system who underwent the same procedure within the same HRR. This required an exact match for the surgical procedure and, subject to that requirement, picked the closest possible pairing of patients based on patient demographics, socioeconomic status (including dual-eligibility and neighborhood education and poverty levels), presence of a multimorbid QCS, emergent admission status, and risk of death, for a total of 147 risk factors as displayed in Supplemental Digital Content 1, Section D eTable 5. Matching was performed at the patient level. To further strengthen our match quality, we aimed to attain standardized differences <0.1 after matching, more stringent than the conventional <0.2.^{32, 33} Matching was completed before viewing outcomes.³⁴

Comparing Hospitals

We performed four main analyses using the one matched sample we described above, which paired patients in flagship hospital systems to control patients undergoing the same inpatient surgical procedure within the same HRR but at a hospital outside the flagship system. Of note, all of our analyses look at these pairs, sometimes grouping pairs in different ways depending on the question, although who is paired with whom never changes. First, we examined all matched pairs to ask whether patients have lower mortality at any flagship system hospitals than a hospital outside the flagship system. Then, we separated this pool of flagship system hospital matched pairs into flagship hospital matched pairs and affiliate hospital matched pairs. This allowed us to address three additional questions, referred to as analyses 2, 3, and 4: (2) do patients have lower mortality at a flagship hospital than a hospital outside the flagship system; (3) do patients have lower mortality at an affiliate of the flagship hospital (excluding the flagship hospital itself) than a hospital outside the flagship system; and (4) do patients have lower mortality at flagship hospitals versus affiliate hospitals?

Finally, based on literature suggesting higher quality hospitals have superior outcomes for higher-risk surgical patients^{21, 24, 35–40}, we separated pairs of patients with multimorbidity²² from pairs without multimorbidity and asked whether those with multimorbidity derived a disproportionate mortality benefit in flagship systems compared to those without multimorbidity within each hospital type comparison.

Most of our analyses refer to a difference in two binary survival rates in matched pairs, see Fleiss et al.⁴¹ A difference-in-difference is the difference of two such differences, and its variance is the sum of their two variances, by independence. Tables 3 and 4 report separate multiplicity adjustments to P-values using the Bonferroni-Holm method.^{42, 43}

The difference of two failure-to-rescue rates is considerably more complicated. A failure is a death following a complication. One person in a pair may have a complication when the other does not. As a consequence, the failure rate describes the population of pairs and is not meaningful for a single pair. We therefore created a standard error of the difference in failure rates by bootstrapping (i.e., resampling) whole pairs.⁴⁴

All analyses were completed using SAS version 9.4.⁴⁵ The study was approved by the Children’s Hospital of Philadelphia Institutional Review Board.

RESULTS

Beginning with 37,223 general surgery patients in flagship system hospitals and 97,728 control patients in hospitals not in the flagship system but in the same HRR, we formed 32,228 closely matched pairs across 35 of the largest HRRs. As seen in Table 1 and Supplemental Digital Content 1, eTables 5–8, we achieved excellent matches. All 32 general surgical procedures were matched exactly. All demographic, socioeconomic status, comorbidity, and risk of death variables had absolute standardized differences below 0.1, usually considerably smaller.

Table 1.

Quality of matched pairs comparing patients, before and after matching, in flagship hospital systems to their matched controls^a from other hospitals in the same HRR

	Flagship Systems		Matched Controls		Standardized Difference
Variable	Before	After	After	Before	Before	After
N	37,223	32,228	32,228	97,728	--	--
Demographics (%)
Age on Day of Surgery (mean years)	75.22	75.41	75.63	75.79	−0.08	−0.03
Age 85+ (%)	12.2	13.0	13.0	14.0	−0.05	0.00
Race
White Non-Hispanic	84.0	84.2	85.3	85.3	−0.04	−0.03
Black	9.5	9.2	8.7	7.8	0.06	0.02
Hispanic	0.9	1.0	0.6	1.3	−0.04	0.04
Female	55.8	56.1	56.7	57.1	−0.02	−0.01
Dually eligible	12.1	12.3	12.5	14.4	−0.07	−0.01
High poverty neighborhoodb	7.8	7.9	7.0	8.4	−0.02	0.03
Low education neighborhoodc	10.4	10.4	9.8	13.2	−0.09	0.02
Emergent admission	39.7	42.3	43.9	46.6	−0.14	−0.03
Probability of death on admission (%)	5.0	5.0	5.0	4.9	0.01	0.01
Comorbidities (%) (see Supplementa for full list)
Chronic Pulmonary Diseases	27.2	26.8	27.6	27.3	0.00	−0.02
Diabetes with Complications	26.0	25.2	25.7	26.2	−0.01	−0.01
Heart Failure	25.0	25.0	25.4	25.8	−0.02	−0.01
Protein Calorie Malnutrition	18.7	18.6	17.0	15.2	0.09	0.04
Thrombocytopenia and Other Hematological Disorders	16.1	16.0	14.4	14.7	0.04	0.04
Metastatic Cancers	15.7	15.0	13.9	12.7	0.09	0.03
CKD Stage 4–5 or Dialysis	6.1	6.1	5.7	5.9	0.01	0.02
Functional Status (%)
Home Oxygen Use	3.9	3.8	4.5	4.4	−0.02	−0.03
Home Hospital Bed or Wheelchair Use	2.1	1.9	2.2	2.4	−0.02	−0.01
Multimorbid (%)	60.3	60.5	60.6	59.3	0.02	0.00

^aThe 32 general surgery procedure groups were exactly matched. For a complete list of all 147 matching variables in all matches, see Supplemental Digital Content 1, Section D eTables 5–7.

^bProportion of patients that live in a zip code in which >20% of adults live below the federal poverty line.

^cProportion of patients that live in a zip code in which <80% of adults have a high school diploma.

Abbreviations: CKD, chronic kidney disease; Diff Ave, average difference; HRR, hospital referral region.

Compared to their matched controls, flagship system hospitals were nearly twice as large, were more likely to be teaching hospitals, provide more advanced interventions, and have superior nursing resources (Table 2). However, these differences were largely attributable to the flagship hospitals themselves, whereas affiliate hospitals were far more similar to the matched controls. For instance, the mean number of beds at flagship hospitals was 974 versus 405 for matched controls, compared to 407 at affiliate hospitals versus 350 for matched controls.

Table 2.

Characteristics^a of hospitals in flagship systems, flagship hospitals, and affiliate hospitals in flagship systems compared to their within-HRR matched controls

	All U.S. Hospitals (N=3,407)	Flagship Systems (N=35)	Controls	Flagship Hospitals (N=35)	Controls	Affiliate Hospitals (N=121)	Controls
Study patients (N)	1,687,511	32,228	32,228	15,571	15,571	16,657	16,657
Number of beds (mean)	200.7	681	377	974	405	407	350
Teaching Status
Resident-to-bed ratio	0.09	0.369	0.215	0.593	0.254	0.160	0.179
COTH status (%)	7.2	57.2	24.9	100.0	30.2	17.2	20.0
Hospital Resources (%)
High technology statusb	39.8	79.2	69.7	99.8	72.0	60.0	67.5
Availability of PCIc	47.9	88.7	86.0	100.0	87.6	78.2	84.5
Comprehensive cardiac technologyd	34.6	57.5	60.4	71.2	62.4	44.6	58.6
Nurse-to-bed ratio	1.63	1.78	1.72	2.06	1.79	1.53	1.66
Highest 1/3 (%)	33.3	49.3	47.5	68.0	50.1	31.8	45.1
Middle 1/3 (%)	33.3	34.7	33.7	22.9	32.9	45.7	34.4
Lowest 1/3 (%)	33.3	16.1	18.8	9.1	17.0	22.6	20.5
Nursing skill mix e	0.89	0.965	0.937	0.978	0.939	0.953	0.934
Highest 1/3 (%)	33.3	62.9	48.8	73.8	51.1	52.7	46.6
Middle 1/3 (%)	33.3	30.8	36.9	23.3	35.2	37.8	38.5
Lowest 1/3 (%)	33.3	6.4	14.3	2.9	13.8	9.6	14.9

^aCharacteristics were weighted by the number of patients in each type of hospital.

^bThe proportion of patients in hospitals that perform both open heart surgery and organ transplantation.

^cThe proportion of patients in hospitals that performed at least 10 PCIs during each year of the study.

^dThe proportion of patients in hospitals that have a cardiac catheterization laboratory, a coronary care unit, and provide cardiothoracic surgery services.

^eThe proportion of registered nurses to the total number of registered nurses and licensed practical nurses.

Abbreviations: COTH, Council of Teaching Hospitals and Health Systems; HRR, hospital referral region; PCI, percutaneous coronary intervention.

Flagship System Hospitals versus Hospitals outside the Flagship System

Patients at flagship system hospitals had significantly lower rates of 30-day mortality than matched controls, who had highly similar comorbidities and socioeconomic status and underwent the same procedure within the same HRR (3.79% [flagship system hospitals] versus 4.36% [controls], difference=−0.57% [95% CI −0.86%,−0.28%], p<0.001) (Table 3, Figure 1a). Findings were similar for 90-day mortality (6.62% versus 7.41%, difference=−0.79% [95% CI −1.15%, −0.42%], p<0.001) (Supplemental Digital Content 1, Section E eTable 9).

Table 3.

Rates of 30-day mortality for general surgery patients in all flagship system hospitals, flagship hospitals only, or affiliated hospitals in flagship systems compared to within-HRR matched controls at other hospitals

Flagship System Hospitals versus Matched Controls
	N	Flagship System Hospitals	Matched Controls	Difference in Rates	95% CI	p-value	Adjusted p-value*
All Patients	32,228	3.79%	4.36%	−0.57%	(−0.86%, −0.28%)	<0.001	0.001
With Multimorbidity	19,317	5.79%	6.69%	−0.90%	(−1.36%, −0.44%)	<0.001	0.001
Without Multimorbidity	12,511	0.74%	0.84%	−0.10%	(−0.32%, 0.13%)	0.43	1.000
With-versus-without Multimorbidity (Difference-in-differences)				−0.81%	(−1.31%, −0.30%)	0.002	0.015
Flagship Hospitals versus Matched Controls
	N	Flagship Hospitals	Matched Controls	Difference in Rates	95% CI	p-value	Adjusted p-value*
All Patients	15,571	3.30%	4.41%	−1.11%	(−1.53%, −0.70%)	<0.001	<0.001
With Multimorbidity	9,488	4.93%	6.70%	−1.77%	(−2.42%, −1.12%)	<0.001	<0.001
Without Multimorbidity	5,852	0.70%	0.80%	−0.10%	(−0.43%, 0.23%)	0.59	1.000
With-versus-without Multimorbidity (Difference-in-differences)				−1.67%	(−2.38%, −0.96%)	<0.001	<1.000
Affiliate Hospitals in Flagship Systems versus Matched Controls
	N	Affiliate Hospitals	Matched Controls	Difference in Rates	95% CI	p-value	Adjusted p-value*
All Patients	16,657	4.25%	4.32%	−0.07%	(−0.48%, 0.35%)	0.77	1.000
With Multimorbidity	9,829	6.62%	6.68%	−0.06%	(−0.73%, 0.61%)	0.88	1.000
Without Multimorbidity	6,659	0.78%	0.87%	−0.09%	(−0.41%, 0.23%)	0.63	1.000
With-versus-without Multimorbidity (Difference-in-differences)				0.03%	(−0.70%, 0.75%)	0.94	1.000
Difference-in-Differences: Flagship Hospitals minus Affiliate Hospitals in Flagship Systems
		Flagship Difference	Control Difference	Difference-in-Differences	95% CI	p-value	Adjusted p-value*
All Patients		−0.95%	0.10%	−1.05%	(−1.62%, −0.47%)	<0.001	0.003
With Multimorbidity		−1.69%	0.02%	−1.71%	(−2.63%, −0.79%)	<0.001	0.003
Without Multimorbidity		−0.08%	−0.07%	−0.01%	(−0.45%, 0.43%)	0.96	1.000

Abbreviations: CI, confidence interval; HRR, hospital referral region; N, number.

^*Adjusted P-values use the Bonferroni-Holm method.^{42, 43}

Figure 1:

Kaplan-Meier Survival Plots by Hospital Type versus Matched Controls outside the Flagship System but within the Same HRR for: (a) All Flagship System Hospitals, (b) Flagship Hospitals Only, and (c) Affiliate Hospitals in Flagship Systems

To determine whether care at flagship system hospitals was associated with a differential benefit for general surgery patients with versus without multimorbidity, we compared outcomes for these patients at flagship system hospitals versus other hospitals. Thirty-day mortality was lower for patients with multimorbidity treated at flagship system hospitals versus matched controls, while no significant difference was observed for patients without multimorbidity (Table 3, Figure 2a). A larger mortality reduction was demonstrated for patients with versus without multimorbidity at flagship system hospitals versus matched controls (difference-in-differences=−0.81% [95% CI −1.31%,−0.30%], p=0.002). Similar findings were seen for 90-day mortality (Supplemental Digital Content 1, eTable 9).

Figure 2:

Kaplan-Meier Survival Plots by Hospital Type and Multimorbidity (MM) Status versus Matched Controls outside the Flagship System (FS) but within the Same HRR for: (a) All Flagship System Hospitals, (b) Flagship Hospitals Only, and (c) Affiliate Hospitals in Flagship Systems

No differences were found in rates of in-hospital postoperative complications (Supplemental Digital Content 1, eTable 10). However, rates of 30-day failure-to-rescue were significantly lower in flagship system hospitals compared to matched controls (11.12% versus 12.93%, difference=−1.81% [95% CI −2.65%,−0.96%], p<0.001) (Table 4). A larger reduction in 30-day failure-to-rescue was observed for patients with versus without multimorbidity at flagship system hospitals versus matched controls (difference-in-differences=−1.48% [95% CI −2.96%,0.00%], p<0.05); however, this P-value exceeds 0.05 after multiplicity adjustment.

Table 4.

Rates of 30-day failure-to-rescue for general surgery patients in all flagship system hospitals, flagship hospitals only, or affiliated hospitals in flagship systems compared to within-HRR matched controls at other hospitals

Flagship System Hospitals versus Matched Controls
	N	Flagship System Hospitals	Matched Controls	Difference in Rates	95% CI	p-value	Adjusted p-value*
All Patients	32,228	11.12%	12.93%	−1.81%	(−2.65%, −0.96%)	<0.001	<0.001
With Multimorbidity	19,317	13.47%	15.61%	−2.14%	(−3.19%, −1.08%)	<0.001	<0.001
Without Multimorbidity	12,511	3.63%	4.28%	−0.65%	(−1.69%, 0.38%)	0.22	1.000
With-versus-without Multimorbidity (Difference-in-differences)				−1.48%	(−2.96%, 0.00%)	<0.05	0.398
Flagship Hospitals versus Matched Controls
	N	Flagship Hospitals	Matched Controls	Difference in Rates	95% CI	p-value	Adjusted p-value*
All Patients	15,571	9.28%	12.75%	−3.47%	(−4.63%, −2.32%)	<0.001	<0.001
With Multimorbidity	9,488	11.10%	15.40%	−4.30%	(−5.72%, −2.89%)	<0.001	<0.001
Without Multimorbidity	5,852	3.32%	3.94%	−0.62%	(−2.05%, 0.81%)	0.39	1.000
With-versus-without Multimorbidity (Difference-in-differences)				−3.68%	(−5.69%, −1.67%)	<0.001	0.003
Affiliate Hospitals in Flagship Systems versus Matched Controls
	N	Affiliate Hospitals	Matched Controls	Difference in Rates	95% CI	p-value	Adjusted p-value*
All Patients	16,657	13.00%	13.10%	−0.10%	(−1.36%, 1.16%)	0.87	1.000
With Multimorbidity	9,829	15.91%	15.81%	0.10%	(−1.48%, 1.68%)	0.90	1.000
Without Multimorbidity	6,659	3.93%	4.60%	−0.67%	(−2.20%, 0.86%)	0.39	1.000
With-versus-without Multimorbidity (Difference-in-differences)				0.77%	(−1.43%, 2.97%)	0.49	1.000
Difference-in-Differences: Flagship Hospitals minus Affiliate Hospitals in Flagship Systems
		Flagship Difference	Control Difference	Difference-in-Differences	95% CI	p-value	Adjusted p-value*
Overall		−3.72%	−0.35%	−3.37%	(−5.08%, −1.67%)	<0.001	0.001
With Multimorbidity		−4.81%	−0.41%	−4.40%	(−6.52%, −2.28%)	<0.001	<0.001
Without Multimorbidity		−0.61%	−0.66%	0.05%	(−2.04%, 2.14%)	0.96	1.000

Abbreviations: CI, confidence interval; HRR, hospital referral region; N, number.

^*Adjusted P-values use the Bonferroni-Holm method.^{42, 43}

Flagship Hospitals versus Hospitals outside the Flagship System

General surgery patients at flagship hospitals had lower rates of 30-day and 90-day mortality than matched controls outside the flagship system undergoing the same procedure within the same HRR (30-day: 3.30% [flagship hospitals] versus 4.41% [controls], difference= −1.11% [95% CI −1.53%,−0.70%], p<0.001; 90-day: 6.17% versus 7.62%, difference=−1.45% [95% CI −1.98%,−0.92%], p<0.001) (Table 3, Figure 1b, Supplemental Digital Content 1, eTable 9). Patients with multimorbidity treated at flagship hospitals had lower 30-day mortality rates than matched controls (4.93% versus 6.70%, difference=−1.77% [95% CI −2.42%,−1.12%], p<0.001), while no difference was observed for patients without multimorbidity (Figure 2b). A larger mortality reduction was noted for patients with versus without multimorbidity at flagship hospitals versus matched controls (difference-in-differences=−1.67% [95% CI −2.38%,−0.96%], p<0.001). Findings were similar for 90-day mortality.

Rates of 30-day failure-to-rescue were lower at flagship hospitals relative to matched controls (9.28% versus 12.75%, difference=−3.47% [95% CI −4.63%,−2.32%], p<0.001) (Table 4). Again, this improvement was concentrated in patients with multimorbidity at flagship hospitals. Rates of in-hospital postoperative complications were similar but statistically higher in flagship hospitals versus matched controls (35.12% versus 33.83%, difference=1.28% [95% CI 0.30%,2.27%], p=0.01), although no disproportionate difference was seen for patients with versus without multimorbidity (Supplemental Digital Content 1, eTable 10).

Flagship Hospital Affiliates versus Hospitals outside the Flagship System

Unlike for flagship systems as a whole or flagship hospitals specifically, general surgery patients at affiliate hospitals in flagship systems did not have significantly different rates of 30-day or 90-day mortality compared to control patients receiving the same procedure at within-HRR hospitals outside the flagship system (30-day: 4.25% [affiliate hospitals] versus 4.32% [controls], difference=−0.07% [95% CI −0.48%,0.35%], p=0.77; 90-day: 7.04% versus 7.21%, difference=−0.17% [95% CI −0.68%,0.34%], p=0.53) (Table 3, Figure 1c, Supplemental Digital Content 1, eTable 9). Also unlike flagship systems or flagship hospitals, no differences were observed for patients with or without multimorbidity at affiliate hospitals in flagship systems versus matched controls (Figure 2c). Similar findings were observed for 30-day failure-to-rescue and in-hospital postoperative complications (Table 4, Supplemental Digital Content 1, eTable 10).

Flagship Hospitals versus Affiliate Hospitals

To further compare surgical outcomes at flagship hospitals versus affiliate hospitals, we performed a difference-in-differences analysis comparing the performance of flagship hospitals versus their controls to that of affiliate hospitals versus their controls. This allowed us to compare flagship hospitals to affiliates. Rates of 30-day mortality and 30-day failure-to-rescue were significantly lower at flagship hospitals versus controls compared to affiliate hospitals versus controls (30-day mortality difference-in-differences=−1.05% [95% CI −1.62%,−0.47%], p<0.001; 30-day failure-to-rescue difference-in-differences=−3.37% [95% CI −5.08%,−1.67%], p<0.001) (Table 3, Table 4). These findings were seen for patients with multimorbidity but not those without multimorbidity. No significant difference-in-differences were observed for in-hospital complications (Supplemental Digital Content 1, eTable 10).

HRRs with Multiple Major COTH Hospitals: Stability Analysis

As aforementioned, some HRRs contained several major COTH hospitals (e.g., Boston, MA) while others had only one (e.g., Charlotte, NC). In an HRR with several COTH hospitals in different systems, a matched control patient may have come from a COTH hospital. Including major COTH controls could blunt the primary findings by comparing outcomes of patients within the flagship system to those of control patients treated at major COTH hospitals. We performed a stability analysis that removed matched pairs containing control patients admitted to these non-flagship system major COTH hospitals, thereby examining the subset of matched pairs in which control patients were not from a major COTH hospital outside the flagship system. The stability analysis removed 18.7% of pairs (6,024/32,228 matched pairs), with no pairs removed in 14 of 35 HRRs in the study. In the stability analysis balance table (eTable 8), as expected, the resident-to-bed ratio from all controls after excluding COTH controls declined slightly (to 0.101 versus 0.173 prior to exclusions).

In the stability analysis, our main findings were unchanged: we still found that flagship hospital patient outcomes were better than stability controls, patients at affiliate hospitals in the flagship system fared no better than controls, and the flagship hospital mortality benefit was entirely due to improved mortality for patients with multmorbidity (Supplemental Digital Content 1, eTable 8, eTable 11). Our conclusions were therefore very stable, and this seems to reflect the fact that 14 HRRs had only one qualifying COTH system, and even when there was a second COTH system, most patients in the HRR were not treated there.

DISCUSSION

Hospital systems have often said that their mergers and acquisitions – which have accelerated in recent years^{1, 14, 16, 46, 47} – are associated with higher quality due to improved care coordination and economies of scale.^1–7 Prior literature has suggested that, despite evidence of variation in surgical outcomes within highly rated hospital systems,¹⁰ many if not most patients expect hospitals from the same system to offer the same standard of care, regardless of whether they are at a flagship hospital or its local affiliate.⁸

To study this, we performed a large, carefully matched study across 35 of the largest HRRs controlling for 32 general surgical procedures, 55 comorbidities and functional status indicators, multimorbidity status, emergent admission status, and sociodemographic variables including age, sex, race, dual-eligibility, and neighborhood education and poverty levels. We found that Medicare patients undergoing inpatient general surgery at flagship system hospitals had lower rates of 30-day and 90-day mortality compared to their matched controls at hospitals outside the flagship system who underwent the same procedure within the same HRR. However, these mortality differences were driven almost entirely by flagship hospitals – the major “brand-name” hospital in each flagship system – and almost entirely by lower mortality for older patients with multimorbidity in those hospitals. By contrast, no mortality difference was observed at affiliate hospitals of flagship systems, and difference-in-differences analysis confirmed that the mortality difference between flagship hospitals and their controls was significantly larger than that between affiliate hospitals and their controls. Similar results were seen for 30-day failure-to-rescue, a surgical quality indicator.^23–26 Therefore, while flagship hospitals exhibited superior surgical outcomes versus their matched controls, affiliates of these flagship hospitals did not.

Additionally, we found that older patients with multimorbidity²² undergoing surgery at flagship hospitals had lower 30-day mortality, 90-day mortality, and 30-day failure-to-rescue rates than matched controls at other hospitals, whereas this was not true for patients without multimorbidity. These disproportionate benefits for older patients with multimorbidity at flagship hospitals are consistent with existing literature demonstrating that higher-quality hospitals are associated with superior outcomes for high-risk patients.^{21, 22, 24, 37–40} On the other hand, we did not find any differential benefits for patients with multimorbidity at affiliate hospitals relative to their controls. We further demonstrated that older patients with multimorbidity appeared to have significantly improved outcomes at flagship hospitals relative to their controls compared to those at affiliate hospitals in the same flagship system relative to their controls.

Our analysis builds on prior work examining surgical outcomes in major hospital systems. Using Medicare data for patients undergoing colectomy, coronary artery bypass graft, or hip replacement in 16 highly rated hospital systems, Sheetz et al. uncovered wide variation in surgical outcomes among affiliated hospitals in the same system while also noting that outcomes were not consistently better at flagship versus affiliate hospitals.¹⁰ In contrast, research examining outcomes after complex cancer treatment at top-ranked cancer hospitals versus their affiliates revealed superior survival at the top-ranked hospitals.^{11, 12} Prior analyses have also suggested that rates of mortality and readmissions for all inpatients do not improve after a hospital is acquired, while patient experience may actually worsen.¹³ Our work extends these analyses by comparing flagship hospitals and affiliate hospitals to within-HRR matched control patients at other hospitals across the breadth of general surgery while also examining whether differential patterns are observed for high-risk patients.

These findings are relevant for patients and policymakers. Patients should not expect superior quality of general surgical care at affiliate hospitals of major regional systems over hospitals outside these systems based solely on their affiliation. Additionally, some patients – especially older patients with multimorbidity – may be better served at the flagship hospital itself rather than at its affiliates or hospitals outside the flagship system in the same region.

Our study had limitations. We examined 35 of the largest HRRs in the United States (out of 306 HRRs). Each included HRR had a flagship system, including one flagship hospital plus its affiliates, and their controls. The hospitals in these 35 HRRs had higher numbers of beds, resident-to-bed ratios, and high technology capabilities compared to the average hospital across the nation, suggesting that the larger HRRs we examined contained better-resourced hospitals than the typical hospital. Also, the study used only fee-for-service Medicare claims. Additionally, some information on chronic conditions may be inconsistently recorded across hospitals. We partially addressed this limitation by using a one-year lookback to obtain information on chronic conditions from both inpatient claims and claims from physician offices and CMS outpatient files, which should have reduced this issue. Further, our definition of multimorbidity²² and our matched analyses incorporated several forms of objective information, such as functional status indicators obtained from the CMS Durable Medical Equipment files.

In conclusion, we found that while flagship system hospitals offered superior outcomes for patients undergoing inpatient general surgery procedures compared to matched controls outside the flagship system but in the same HRR, those benefits were driven almost entirely by flagship hospitals themselves and concentrated in older patients with multimorbidity. In contrast, affiliates of these flagship hospitals did not offer any significant outcomes benefits over controls. Thus, hospital affiliation to a flagship system does not alone assure better surgical outcomes.