Effects of affiliation network membership on hospital quality and financial performance

Abstract

Objective

To examine the effects of hospital membership in affiliation networks—franchise‐like networks sponsored by high‐quality health systems in which affiliate hospitals pay an annual fee for access to sponsor's operational and clinical resources—on clinical quality, patient experience ratings, and financial performance of affiliates and their competitors.

Data sources

Network membership data from press releases and websites of four sponsors (Mayo Clinic, Cleveland Clinic, MD Anderson, Memorial Sloan Kettering), American Hospital Association's Annual Survey, Centers for Medicare & Medicaid Services' Hospital Compare, and Healthcare Cost Report Information System, all for 2005–2016.

Study design

We used a quasi‐experimental design and estimated hospital‐level regressions with hospital‐fixed effects. Dependent variables were measures of clinical quality, patient experience, and financial performance. Independent variables included an indicator for affiliate versus nonaffiliate and fixed effects for hospital characteristics and year. To analyze effects on competitors, we repeated analyses by comparing hospitals in the same county as an affiliate to nonaffiliated, noncompetitor hospitals.

Data collection

Membership was obtained through press releases and network websites then linked across datasets by name and Medicare's identification number.

Principal findings

Across networks, affiliates (N = 199) experienced insignificant clinical quality changes but increased net income by $38,500 and operating margin by 6.6% (p values = 0.01–0.08) compared to nonaffiliates. Multispecialty affiliates improved on no measures. Cancer‐specific affiliates improved their net income ($96,900) and operating margin (3.6%; p‐values < 0.05). Affiliates' competitors experienced mixed changes in clinical measures relative to hospitals without affiliates in market (p‐value < 0.05) but no financial effects. Affiliation was not associated with patient experience ratings for affiliates nor competitors.

Conclusions

Despite quality‐focused missions, affiliation networks are not guaranteed to improve public measures of quality in affiliated hospitals, although hospitals in these communities improve financially. Future research should assess the conditions and mechanisms by which affiliation improves quality consistently and which forms of quality.

What is known on this topic

• Prominent hospitals known for their high quality of care have been sponsoring “affiliation networks” in which affiliate hospitals pay an annual membership fee that allows access to the clinical expertise and resources of the sponsor.

• Prior research has examined the motivations for joining affiliation networks and found that hospitals that face challenges in delivering quality care to patients with complex conditions and those that are financially vulnerable were more likely to join than not join.

• No research has examined whether joining one of these networks translates into better quality of care and/or financial performance for affiliates, nor possible “spillover” effects on local competitors.

What this study adds

• This study finds that joining an affiliation network is not associated with improved clinical quality or financial performance for affiliates of networks sponsored by multispecialty hospitals but is associated with significant financial improvement for affiliates of networks sponsored by cancer hospitals.

• Neither affiliates nor competitors experience changes in patient experience ratings, although competitors experience changes to their clinical quality, some positive, some negative.

• This study suggests that hospitals seeking to improve quality performance on publicly reported metrics may benefit more from investing in other quality improvement strategies or may need to engage differently with affiliation networks to benefit quality wise.

1. INTRODUCTION

Over the last 30 years, changes in the health care industry catalyzed various forms of hospital networks, from highly bound and regulated mergers and acquisitions (M&As) to loosely coupled and loosely defined alliances and partnerships. ¹ , ² Since passage of the Affordable Care Act in 2010, a novel type of network—affiliation networks—has grown yet has been the focus of little research. These networks comprise a sponsor and affiliated hospitals (affiliates), who apply to be members of the sponsor's affiliation network. Sponsors are high‐status hospitals and health systems that are nationally ranked and known for their quality of care including Mayo Clinic, Cleveland Clinic, Memorial Sloan Kettering, and MD Anderson. These four sponsors have publicized networks, with over 200 hospitals participating across them. Other affiliation networks may exist that are not public or disclosed. Affiliates pay an annual fee for access to sponsor's operational and clinical resources such as virtual consultations with specialists, treatment protocols, research findings, and access to expensive technologies and clinical trials.

In essence, affiliation networks utilize a hub‐and‐spoke franchise model, with the sponsor as the hub who provides resources to its spoke affiliates. These resources are to enhance care quality provided by affiliates, uniformly stated as the primary goal of these networks. While their emphasis on quality is not unique among networks, affiliation networks differ from other networks formed through M&As, accountable care organizations (ACOs), and other prominent organizational forms such as strategic hospital alliances in that they have annual contracts that allow easier decoupling, require an explicit payment to a sponsor, entail unidirectional resource flow, are hospital‐driven not government‐policy driven in their conception and formation, and span a national rather than regional footprint. ³ Affiliates are located across the United States and outside the regional markets of sponsors (see Figure 1). These features also distinguish affiliation networks from multihospital systems, defined by the American Hospital Association (AHA) as two or more hospitals owned, leased, or contract managed by a central organization. Sponsors benefit from affiliations through revenue from membership fees (not publicly disclosed) and increased brand awareness, which may lead to new patient referrals, although this is not a stated purpose.

FIGURE 1.

Map of affiliate members and sponsors [Color figure can be viewed at wileyonlinelibrary.com]

Our goal is to examine the effects of affiliation network membership on the performance of affiliates and their local competitors. We assess whether membership improves affiliates' clinical outcomes, patient experiences, and financial performance compared to nonaffiliated hospitals. We also assess whether hospitals in the same market as an affiliate (competitors) change performance in these domains following affiliation entry into their market compared to hospitals in markets without affiliates. Recent research indicates that hospitals that join these networks have more quality‐related and financial improvement needs than those that do not join. ⁴ We find no research that addresses whether those needs are met, that is, whether performance improves, a notable gap given the quality‐focused mission of these networks. Additionally, whether there are spillover effects for hospitals in the same market as affiliates has been unexamined. Research has focused on the impact of M&As or ACOs on participants, ⁵ , ⁶ , ⁷ overlooking potential impact networks can have on nonparticipating, competing hospitals.

Understanding how membership in affiliation networks affects hospital and competitor performance is important for hospitals and policy makers as they decide whether such networks provide value and therefore membership in them should be pursued and encouraged. Research on other hospital networks formed through M&As and ACOs has indicated no or mixed effects of those networks on clinical and financial performance. ³ , ⁷ , ⁸ , ⁹ , ¹⁰ If affiliation networks are similar in having insignificant effects, then industry and organizational resources are better directed to other strategies for quality improvement. However, if affiliation networks yield positive effects, particularly for quality without having adverse financial effects, continued investment in their growth would be prudent. This work provides insight for such decision making.

1.1. Conceptual model

Theoretically, hospitals that join affiliation networks should improve their quality of care (proclaimed goal) as well as their financial health and patient experiences. They are expected to improve their clinical quality by learning from and utilizing the clinical expertise of their high‐performing network sponsors, as management research finds that interorganizational learning is a beneficial outcome of network membership. ¹¹ In affiliation networks, access to expert specialist consults and sharing of treatment protocols and best practices (learning) should lead to faster, more accurate diagnoses, more appropriate treatment, fewer medical errors, and better patient outcomes leading to fewer preventable readmissions and lower mortality rates.

High‐quality care at affiliates, and even simply the perception of this, should also have positive financial impact. Organizations with better reputations attract more clients. ¹² , ¹³ Hospitals may experience greater patient volume from affiliating with high‐status sponsors, including growth of patients with more severe conditions/higher acuity that are now treatable with the greater expertise available, thus leading to higher reimbursement and revenue. Treating more and sicker patients may also increase care quality for all patients due to experience‐based learning from complexity and heterogeneity. ¹⁴ Affiliation may improve finances directly through this patient‐care pathway and through spillover to yield nonpatient care revenue (e.g., donations and grants). Furthermore, affiliates may improve patient experience scores because access to new interventions positively influences how patients view the quality of care they receive. ¹⁵ Research finds strong associations between clinical outcomes and patient experience and between corporate reputation and customer satisfaction, ¹⁶ , ¹⁷ which we expect to hold true for hospitals bolstering reputation and skill through affiliation.

For cancer‐specific affiliation networks, these performance effects are expected not only at the service‐line level but also at the hospital level due to improved patient perceptions of affiliates. Surveys show that patients are more willing to seek care from their local hospital following an affiliation regardless of familiarity with sponsor. ¹⁸ For cancer‐specific affiliates, affiliation may improve overall reputation, leading to higher volumes and patient acuity and thus higher revenue.

Spillover performance effects for hospitals in the same market as affiliates are likely to exist as well. Competing hospitals should be motivated to improve their own performance once another hospital in their market joins an affiliation network so as to maintain their competitive position. As they undertake potentially costly strategies to compete, their care quality may improve while short‐term financial performance declines. If competitors maintain market position, then their finances may remain steady or improve over time. Their patient experience ratings should improve alongside their clinical performance in response to these improvement efforts, through the same mechanisms as for affiliates discussed above. We examine whether these performance outcomes occur as expected for affiliates and competitors.

2. METHODS

2.1. Study design

We used a quasi‐experimental study to examine how hospital membership in an affiliation network affects performance for affiliates and competitors. We estimate the effects of membership by analyzing variation in membership and leveraging variation in time of joining. To be included in this study, hospitals must have participated in the AHA's Annual Survey at least once between 2005 and 2016 and have data posted on the Center for Medicare and Medicaid Services' (CMS) Hospital Compare website. The AHA survey collected information about hospital characteristics and sourced financial performance from CMS' Healthcare Cost Report Information System database (HCRIS). The CMS website provided information on clinical outcomes and patient experience. A total of 5728 of the 7131 (80%) hospitals in the United States had data from both sources. We excluded 343 hospitals because they were not candidates for network membership, given their ownership status or specialties of existing sponsor hospitals. Excluded hospitals included federally owned facilities, psychiatric facilities, long‐term care facilities, children's specialty hospitals, hospitals that treat alcoholism, and hospitals focused on intellectual disabilities, leading to a sample of 5385 hospitals.

During our study period (2005–2016), the number of affiliates grew from one to a total of 203 hospitals from 84 multihospital health systems in one of the four publicized affiliation networks in the United States. The networks are the Mayo Clinic Care Network (38 domestic and 5 international members), the Cleveland Clinic Affiliation Network (23 members), MD Anderson Cancer Network (20 members), and Memorial Sloan Kettering Cancer Alliance (3 members). All network sponsors (Mayo Clinic, Cleveland Clinic, MD Anderson Center, and Memorial Sloan Kettering Cancer Center, respectively) are ranked by U.S. News and World Reports in 2017 in the top seven hospitals nationally in at least one of the following specialties: cancer, cardiology, and diabetes/endocrinology, which are among the most prevalent conditions in the United States. ¹⁹ , ²⁰ , ²¹ Sponsors' superior performance suggests expertise to leverage for enhancing the quality at affiliates. We focused on these four networks because they are prominent and disclosed their membership. Information on which hospitals were members of these networks and when they joined was obtained through sponsor and affiliate network websites and press releases. There were no instances of only some hospitals in a health system becoming members.

Hospitals within health systems were counted separately because their performance was recorded separately in datasets, implying hospital‐level affiliate effects are possible. Of the 203 affiliates, 4 were not analyzed because of aforementioned exclusion, leaving 199 for analysis. Hospitals' affiliation status for each year was assessed through review of network websites, coded as 1 in the year they joined and all years thereafter. Per this review, no hospital exited its affiliation network during our study.

2.2. Primary outcome measures

2.2.1. Clinical outcomes

We analyzed overall clinical and condition‐specific performance for affiliates with networks sponsored by multispecialty systems. To assess overall clinical performance, we used 30‐day all‐cause unplanned readmission rates. For condition‐specific measures, Medicare's 30‐day mortality rates were obtained for six conditions: acute myocardial infarction (AMI), heart failure, pneumonia, chronic obstructive pulmonary disease (COPD), stroke, and coronary artery bypass graft (CABG) surgery. CMS did not begin reporting all measures concurrently, thus the number of hospitals in different analyses varies. Reporting of all‐cause readmission rates began in 2012 while mortality rates for AMI, heart failure, and pneumonia were in 2007, for COPD and stroke, in 2013, and for CABG, in 2014. Rates were multiplied by 100 for readability. Clinical outcomes were not evaluated for cancer network affiliates due to lack of posted cancer‐specific quality measures and inapplicability of available measures.

2.2.2. Patient experience and satisfaction

We used two scores to capture patient experience and satisfaction with care: global hospital rating and willingness to recommend score. ²² These were assessed using the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) Survey and reported on the CMS website starting in 2007. We used the top‐box approach, which captures the percentage of patients who gave ratings of 9 or 10 on a scale of 1–10 for the global rating, and the percentage who responded “would definitely recommend” on the willingness to recommend question. The top‐box approach is used by the US Agency for Healthcare Research and Quality when analyzing CAHPS survey results. ²³

2.2.3. Financial performance

We used two variables to measure financial performance: operating margin and net income per bed. Operating margin is the profitability of a hospital from patient care services (revenue divided by patient care services costs). Net income per bed is defined as revenues less expenses divided by total hospital beds and is a holistic measure of profitability that includes nonpatient care services revenue and expenses. We scaled these measures, so one unit represents $1000 per bed for interpretability.

2.2.4. Overall performance

To capture a hospital's overall performance, we used hospitals' total performance score (TPS), which is calculated by CMS based on performance in four domains: (1) clinical care, (2) patient‐ and caregiver‐centered experience of care/care, (3) safety, and (4) efficiency and cost reduction. The TPS is a component of Medicare's Hospital Value‐Based Purchasing program that was implemented in 2012 when the Affordable Care Act required Medicare payment adjustments for acute inpatient services based on a hospital's TPS. Although domains have remained consistent over time, their weighting has changed, and TPS scores have declined over time nationally. ²⁴

2.3. Secondary outcome measures that might explain changes in primary outcomes

Patient volume. Patient volume was defined as the average daily census, as reported in the AHA Annual Survey.

Patient acuity. Patient acuity was measured by a hospital's case mix index (CMI), reported in HCRIS. CMI ranges from 0 to 5, with higher values indicating more complex and resource‐intensive patient populations.

2.4. Covariates

We included six covariates in our models as potential confounders of the relationship between membership in an affiliation network and performance because of their established relationship to hospital performance: number of hospital beds, ²⁵ number of full‐time physicians and dentists, ²⁶ teaching status, ²⁷ nonprofit status, ²⁸ health system membership, ⁵ and rural location. ²⁹ The last four named were each binary variables.

2.5. Analysis

2.5.1. Hospital characteristics

We calculated descriptive statistics (i.e., mean and standard deviation) for three hospital groups at the beginning (2005) and end of the study period (2016)—nonaffiliated, noncompetitor hospitals, all affiliates regardless of join date, and market competitors of affiliates—to assess their characteristics and trends in hospital performance. We conducted t‐tests (for continuous variables) and chi‐squared tests (for categorical variables) to compare characteristics and performance of affiliates and competitors to each other, and affiliates and competitors to nonaffiliated, noncompetitor hospitals at both time points.

2.5.2. Effect on affiliate performance

To assess the effect of network membership on outcomes, we estimated linear regression models of the following form for each outcome:

$$Y_{\mathit{it}}=\alpha +\beta {\text{AFFIL}}_{\mathit{it}}+\gamma X_{\mathit{it}}+{\tau }_t+{\delta }_i+{\epsilon }_{\mathit{it}},$$ (1)

where Y _it is the outcome of interest, AFFIL _it is an indicator for whether a hospital (i) is a member of an affiliation network in year t, $\beta$ is our measure of interest, X _it is a vector of hospital covariates, ${\tau }_t$ represents year‐fixed effects, and ${\delta }_i$ are hospital‐fixed effects. Hospital‐fixed effects adjust for time‐invariant differences that may result from hospital‐specific factors and examine within‐hospital changes from baseline. Year‐fixed effects adjust for overall trends in outcome variables. We accounted for repeated observations of the same hospital (over several years), generated robust standard errors using the cluster option in Stata v15, and clustered at the hospital level. We first compared performance of all affiliates to all other hospitals in the sample (labeled “nonaffiliates”) and then performed separate analyses for multispecialty and cancer affiliation networks. For affiliates of networks by multispecialty sponsors (i.e., Mayo Clinic and Cleveland Clinic), we compared their performance to non‐multispecialty‐affiliated hospitals. For affiliates of networks by cancer hospitals (i.e., Memorial Sloan Kettering and MD Anderson), we compared performance to non‐cancer‐care‐affiliated hospitals that provided cancer services. All outcomes of interest were evaluated in each comparison and measured at the hospital rather than service line level, per our Conceptual Model and data availability. The sample for each measure varies because not all hospitals met either the CMS volume or case criteria for reporting of that measure in each year.

2.5.3. Effect on competitors

To assess whether membership in an affiliation network affects the performance of competitors, we created an indicator variable COMP _it to denote whether a hospital (i) was located in the same market as an affiliate in year t. Markets were defined at the county level because research on affiliation networks shows competition affects joining at the county but not the hospital referral region (HRR) level. ⁴ However, for sensitivity analysis, we reanalyzed performance effects with HRR as the market. We estimated linear regression models for each outcome variable using a version of Equation (1) in which we replaced the indicator for affiliation (AFFIL _it ) with the COMP _it variable. We excluded hospitals that were affiliates in year t from the comparison group to remove bias, leaving the comparison group as nonaffiliated, noncompetitor hospitals.

2.5.4. Robustness

We tested the robustness of our models and results by creating samples matched to affiliates based on characteristics using propensity scores and repeating analyses using the matched samples as control groups. This approach adjusts for potential confounders that may bias the true effect of joining an affiliation network. We matched hospitals on all covariates as well as location, market competition as calculated by the Herfindal–Hirschman Index, and county‐level Medicare spending per beneficiary. Due to the staggered nature of joining, we created matched groups using data near the beginning (2006) and midpoint of the study period (2011). For each year, we created two matched groups, first, at a 1:4 ratio of affiliates to controls (nearest neighbors with replacement), and second, at a 1:1 ratio (nearest neighbor without replacement) for a total of four matched samples to test the sensitivity of the propensity scores. Results using propensity scores are consistent with the main analysis and not presented. We present results from the full‐samples analysis instead because propensity score matching, despite benefits from matching on several variables, can create imbalances and bias. ³⁰

3. RESULTS

3.1. Summary of hospital sample

Characteristics of the study hospitals are shown in Table 1. At the beginning and end of the study period, there are no clinical differences between affiliates and nonaffiliated, noncompetitor hospitals. Competitors, however, have significantly better mortality rates and worse readmission rates than nonaffiliated, noncompetitor hospitals and affiliates at the beginning and end of the study period though effect sizes are small. Financially, affiliates and competitors have higher net income compared to nonaffiliated, noncompetitor hospitals at the beginning and end but do not differ from them in operating margin. At the end, however, affiliates have higher operating margin than at the start and competitors.

TABLE 1.

Summary of hospital characteristics

	2005 a			2016
Variable	Nonaffiliated, noncompetitor hospitals, mean (SD)	Affiliates, mean (SD)	Competitors, mean (SD)	Nonaffiliated, noncompetitor hospitals, mean (SD)	Affiliates, mean (SD)	Competitors, mean (SD)
Hospital characteristics
Sample size	4033	176	456	4075	195	514
# Hospital beds	144.53 (161.94)	276.48 (225.96) b	266.32 (245.98) b	141.93 (176.09)	282.87 (240.29) b	254.76 (276.69) b
# Full‐time MDs and dentists	10.84 (60.89)	32.25 (67.65) b	27.24 (83.30) b	18.74 (88.41) d	56.23 (142.65) b , d	41.57 (121.06) b , d
Average daily census	96.35 (123.48)	206.39 (181.13) b	195.85 (200.01) b	88.28 (129.60) d	191.23 (186.16) b	171.85 (205.30) b , d
Teaching status	0.05 (0.21)	0.13 (0.33) b , c	0.18 (0.39) b	0.05 (0.18)	0.11 (0.31) b	0.15 (0.35) b
Nonprofit status	0.56 (0.50)	0.88 (0.33) b , c	0.69 (0.46) b	0.56 (0.50)	0.89 (0.32) b	0.62 (0.49) b , c , d
Rurality	0.46 (0.50)	0.24 (0.43) b , c	0.05 (0.22) b	0.43 (0.49) d	0.19 (0.39) b , c	0.03 (0.17) b
System member	0.75 (0.43)	0.88 (0.33) b	0.82 (0.38) b	0.75 (0.43)	0.91 (0.29) b , c	0.81 (0.39) b
Case mix index	1.32 (0.27)	1.46 (0.25) b	1.47 (0.27) b	1.58 (0.35) d	1.72 (0.33) b , d	1.75 (0.39) b , d
Total admissions	6298.88 (8108.18)	14,305.80 (11,861.56) b	12,505.61 (11,778.87) b	5986.23 (8760.89)	13,650.89 (12,476.50) b , c	11,169.49 (12,190.69) b
Clinical performance e (sample size shown in each cell in italics)
All‐cause readmissions	0.16 (0.01) 3710	0.16 (0.01) c 194	0.16 (0.01) b 447	0.15 (0.01) d 3632	0.15 (0.01) c , d 192	0.15 (0.01) b , d 434
30‐day mortality rates
AMI	0.16 (0.01) 3301	0.16 (0.01) 149	0.16 (0.01) b 358	0.14 (0.01) d 1888	0.13 (0.01) d 95	0.13 (0.01) d 186
Heart failure	0.11 (0.01) 3484	0.11 (0.01) c 177	0.11 (0.01) b 410	0.12 (0.02) d 3015	0.12 (0.02) c , d 185	0.11 (0.02) b , d 376
Pneumonia	0.11 (0.02) 3503	0.11 (0.02) c 178	0.11 (0.01) b 421	0.16 (0.02) d 3439	0.16 (0.02) c , d 189	0.15 (0.02) b , d 386
COPD	0.08 (0.01) 3079	0.08 (0.01) c 185	0.08 (0.01) b 391	0.08 (0.01) d 2998	0.08 (0.01) c 184	0.08 (0.01) b , d 375
Stroke	0.15 (0.02) 2318	0.15 (0.02) c 168	0.15 (0.02) b 350	0.15 (0.02) d 2171	0.15 (0.02) d 163	0.14 (0.02) b 334
CABG	0.03 (0.01) 1054	0.03 (0.01) 91	0.03 (0.01) b 173	0.03 (0.01) d 774	0.03 (0.01) 92	0.03 (0.01) 161
Patient experience
Sample size	1751	123	230	2990	98	205
Overall score	0.64 (0.09)	0.65 (0.09) c	0.61 (0.09) b	0.72 (0.08) d	0.73 (0.08) c , d	0.71 (0.11) b , d
Will recommend	0.67 (0.10)	0.69 (0.10) c	0.66 (0.10)	0.72 (0.09) d	0.73 (0.08) b , c , d	0.72 (0.12) d
Financial performance
Sample size	4033	176	456	4075	194	514
Operating margin	−3.39 (48.55)	−0.31 (9.16)	−2.80 (35.29)	−3.89 (30.80)	2.27 (12.87) c , d	−0.37 (22.04)
Net income	$59.2 (19.7)	$124.0 (191) b	$109.0 (323) b	$83.81 (16.6) d	$171.0 (207) b , c , d	$142.0 (209) b
Overall performance	55.26 (14.81) 2344	55.77 (13.49) 162	55.72 (15.05) 364	36.74 (13.07) 2411	34.65 (11.86) 165	34.80 (13.12) 227

Abbreviations: AMI, acute myocardial infarction; CABG, coronary artery bypass graft; COPD, chronic obstructive pulmonary disease.

^a Not all variables have data beginning in 2005. Patient experience ratings, and mortality rates for AMI, heart failure, and pneumonia begin in 2007. All‐cause readmission rates begin in 2012. COPD and stroke mortality rates begin in 2013 and CABG mortality rates begin in 2014.

^b p‐Value < 0.05 for comparisons against the full sample within the same time (i.e., full sample compared to affiliates in 2005 and full sample compared to competitors in 2005).

^c p‐Value < 0.05 for comparisons of affiliates and competitors within the same time (i.e., affiliates vs. competitors at 2005).

^d p‐Value < 0.05 for comparisons across time of same group (i.e., full sample at 2005 compared to full sample at 2016).

^e Values appear the same due to rounding.

3.2. Affiliates' performance

Table 2 presents results for the performance analysis for all affiliates compared to nonaffiliates (Model 1), multispecialty affiliates compared to non‐multispecialty‐affiliated hospitals (Model 2), and cancer‐care affiliates compared to non‐cancer‐care‐affiliated hospitals with cancer programs (Model 3). Across models, affiliation was not associated with changes in clinical performance, patient experience, total performance, patient volume, or patient acuity (p‐value > 0.10). However, the models show effects on finances. Affiliation was associated with significant improvement in net income per bed of $38,500 on average (p = 0.01) and modest improvement in operating margin of 6.6% (p = 0.08) across all affiliates (Model 1). The stratified analysis showed that the financial effects were concentrated in the cancer‐care networks. We found no financial effects for multispecialty affiliates (Model 2), whereas cancer‐care affiliates (Model 3) significantly improved in operating margin (3.58%) and net income ($96,870; p < 0.01). Tables showing all variable coefficients for all models are available upon request.

TABLE 2.

Effect of affiliation networks on performance

		Model 1: All affiliates		Model 2: Multispecialty affiliates		Model 3: Cancer affiliates		Model 4: Competitors a
Primary outcome	Total N (affiliate N)	Coefficient (SE)	p‐Value	Coefficient (SE)	p‐Value	Coefficient (SE)	p‐Value	Coefficient (SE)	p‐Value
Clinical performance
All‐cause readmissions	4551 (197)	NA		−0.02 (0.05)	0.72	NA		−0.10 (0.04)	0.02*
30‐day mortality rates
AMI	4017 (112)	NA		−0.19 (0.12)	0.11	NA		0.16 (0.07)	0.02*
Heart failure	4397 (132)	NA		0.19 (0.11)	0.10	NA		−0.05 (0.05)	0.33
Pneumonia	4530 (133)	NA		0.14 (0.14)	0.34	NA		−0.06 (0.07)	0.37
COPD	3787 (130)	NA		−0.00 (0.11)	0.99	NA		−0.02 (0.05)	0.63
Stroke	2938 (118)	NA		−0.02 (0.15)	0.90	NA		0.14 (0.08)	0.08
CABG	1080 (63)	NA		0.13 (0.14)	0.38	NA		0.03 (0.10)	0.78
Patient experience
Overall score	3832 (169)	0.0004 (0.004)	0.93	−0.001 (0.004)	0.83	0.002 (0.007)	0.84	0.004 (0.003)	0.15
Will recommend	3832 (169)	−0.0005 (0.004)	0.90	−0.0007 (0.004)	0.87	−0.003 (0.007)	0.97	0.004 (0.003)	0.09
Financial performance
Operating margin	5311 (185)	6.60 (3.80)	0.08	6.89 (5.16)	0.18	3.58 (1.31)	<0.01*	−0.25 (2.43)	0.92
Net income	5385 (185)	38.50 (15.68)	0.01*	24.10 (17.93)	0.18	96.87 (36.23)	<0.01*	20.93 (8.89)	0.09
Overall performance
Total performance score	3167 (168)	−1.47 (1.10)	0.18	−1.11 (1.48)	0.45	−1.25 (1.55)	0.42	−0.93 (0.77)	0.22
Intermediate outcomes
Patient volume	5385 (199)	96.29 (127.13)	0.45	89.87 (149.18)	0.55	203.77 (253.63)	0.42	−3.26 (1.26)	0.01*
Patient acuity (CMI)	3618 (165)	0.0006 (0.01)	0.91	−0.001 (0.007)	0.83	0.007 (0.01)	0.52	0.02 (0.006)	<0.01*

Note: Readmissions and mortality rate coefficients have been multiplied by 100 for ease of readability. Sample sizes vary due to differences in reporting years for outcome measures.

Abbreviations: AMI, acute myocardial infarction; CABG, coronary artery bypass graft; CMI, case mix index; COPD, chronic obstructive pulmonary disease; SE, standard error.

^a Competition is often assessed using the Herfindal–Hirschman Index (HHI). We included HHI as a time‐varying confounder in the first round of these analyses. It was not a significant predictor in any and did not have a significant effect on the results, so we removed it from the models underlying these results to preserve parsimonious models.

^* p‐Value < 0.05.

3.3. Competitors' performance

Analyses of competitors' performance (Table 2, Model 4) indicated mixed effects of a hospital becoming an affiliation network member on local competitors' clinical outcomes. Relative to hospitals in markets without an affiliate, affiliates' competitors' all‐cause readmission rates decreased significantly (p < 0.05), while their mortality rate for AMI increased (p < 0.05) but did not change for other conditions. Patient experience scores also did not change (p > 0.05). However, competitor net income increased marginally (p < 0.10) while operating margin did not change. Their patient volume decreased (p < 0.05), while their patient acuity increased (p < 0.01). Results with market defined as the HRR mirrored county‐level results.

4. DISCUSSION

This study is the first to assess the effect of hospital membership in an affiliation network on organizational performance. The findings indicate that affiliation networks have mixed effects on the performance of affiliates and their competitors, which is consistent with research that finds no to mixed effects of other network forms such as M&As and ACOs. ⁶ , ⁷ , ³¹ Our results are consistent with results from other studies in which affiliates experienced heterogeneous financial effects and no change in clinical quality and patient experience, despite networks' emphasis on quality care. Notably, affiliates' competitors experienced significant positive and negative changes in clinical measures.

4.1. Affiliates' performance

There are at least three possible explanations for no detected quality improvement in networks sponsored by multispecialty hospitals. The first is that providers at affiliates are not using the resources provided by sponsors (e.g., e‐consults and best practices) at the regularity or intensity needed for effect. This may be due to ignorance about availability, difficulty in accessing the resources, or reluctance to change practice, well‐documented barriers to adoption of quality‐improving innovations. ³² Also, quality measures for affiliates prior to joining indicate that their quality scores were commensurate with the national average, leaving less incentive to engage and change practice. Sponsors may select hospitals that already have high quality so as to not detract from their own reputation by associating with low‐quality hospitals, and they may not be providing the level of services needed for meaningful improvement within this group. These not‐underperforming hospitals may be satisfied with reputation gains instead. Another explanation is that the quality measures studied did not capture important improvements. Readmission and mortality rates may not capture faster or more accurate diagnoses due to affiliation, both of which improve care. A third explanation is that the effects from multispecialty networks are too diffuse (i.e., too few cases per specialty) to detect. Future research should evaluate these explanations and those for financial experiences.

Financially, our results showed positive effects across all affiliates; however, when stratified by specialty, only cancer‐network affiliates retained those effects, multispecialty affiliates did not. One explanation is self‐selection. Cancer‐network affiliates have clearly prioritized their cancer program and may be more likely to engage with sponsor resources to realize benefits compared to multispecialty affiliates who may not have a similarly defined or focused goal for affiliation. Higher engagement (potentially yielding higher‐quality cancer care that we were unable to document) combined with higher profitability of cancer services may result in higher revenue. Financial improvement did not result from reputation or learning spillover, which often arise from greater volume and complexity, as cross‐specialty patient volume and acuity did not change. ³³ That multispecialty hospital networks did not experience significant financial improvement may reflect dispersed rather than concentrated network use, limiting quality improvement and financial performance in turn.

Although improved financial performance is not a described goal of affiliation networks, that they are associated with such improvement for cancer affiliates is notable, given the financial distress that many hospitals face, which can adversely affect quality long term. ³⁴ Forming other specialty, affiliation networks may serve this additional improvement goal, whether they make that a primary, disclosed goal. Notably though, affiliates' mean financial performance across network foci was similar or better than competitors and the full sample from the beginning, also suggesting selection of better performers, consistent with research showing wealthier join. ⁴

Patient experience ratings did not change for affiliates, which is consistent with the lack of clinical and volume effects. Without improvement in clinical quality, it becomes less surprising that patient experience ratings remain unchanged as well. Lack of improvement in ratings for affiliates suggests that hospitals do not improve their evaluations simply by joining an affiliation network; reasons remain unclear and should be explored further.

4.2. Competitors' performance

There were more changes in performance for affiliates' competitors than for affiliates though only in partial support of our expectations. Readmission rates improved for competitors though the effect size of 0.001% is small compared to the national median of 15.5%. Additionally, AMI mortality rates increased by 0.0016%. This effect size is also small (the national median is 14%) and may be linked to significantly higher acuity and, therefore, more difficult to treat patients, which we find. Competitors may actively seek higher acuity patients, and the higher associated reimbursement, as part of their competitive strategy. Alternatively, their improvement could reflect already occurring improvement efforts that may be more effective than affiliation to improve population health. Affiliation may have galvanized such effort further so as not to widen the performance gap that might otherwise occur if affiliate performance improved. Information on hospital strategies in each market is needed to determine whether the observed effects are a response to affiliation or reflect another reason for improvement, a subject for future research.

Competitors did not experience significant negative financial effects with market entry of an affiliation, despite decreasing patient volume. The decrease may reflect patients seeking care at higher‐performing hospitals including now more visible network sponsors. Nevertheless, the stability in finances alongside insignificant change in affiliates' patient volume indicates that competitors retain their competitive position despite affiliate market entry.

4.3. Limitations and future research

The relatively small number of affiliated hospitals is a potential limitation. Additionally, our analysis has focused on publicized networks. Our results may not generalize to nonpublicized affiliation networks (if they exist), their affiliates, and future affiliates of studied networks. Hidden affiliation networks may present a threat to our analysis if they are truly of the same form as the four studied. Another limitation is the number of clinical measures studied. Publicly available quality measures for the study period are limited and other measures are proprietary. Although the measures used for our multispecialty affiliates are highly regarded as indicators of clinical quality and used as the basis for pay‐for‐performance programs such as hospital value–based purchasing, the available measures may not be sufficient to capture the effects for affiliation networks where clinical benefits may be more nuanced. In addition, Hospital Compare measures are Medicare‐specific and may not reflect results for younger or commercially insured patients, who may benefit more from the network. Ideally, analysis of cancer care clinical quality for such focused networks would examine standardized care measures, but they were unavailable. As cancer measures and other applicable clinical measures become more widely available, they should be analyzed. This analysis should furthermore examine whether affiliation serves as an effective growth and quality strategy for outpatient cancer care that is not detected using inpatient comparison.

Future research should examine process and outcome measures. We focused on outcome measures to be most inclusive in how quality improvements could be captured. Performance on these measures is difficult to improve. Process measures would offer more insight into network effects. We could not evaluate process changes because which guidelines, protocols, and other resources were shared within the networks are undisclosed, preventing selection of relevant process measures. As details about affiliation networks increase, process analysis may become possible. The financial measures we used also may not capture the full extent of potential effects due to their holistic nature and being subject to varied accounting practices; however, they are commonly used as measures of financial health.

We relied on network websites to determine hospitals' annual membership status. Although sponsors likely benefit little from maintaining a divorced, nonpaying member on their websites, start dates may be more reliably captured on websites than end dates. Nondocumented exits, if they occurred, would bias against detecting an effect, especially small effects. A final limitation is the inability to isolate the conditions and mechanisms underlying the results. Our conceptual model offered theory‐based mechanisms. Future qualitative work with hospital administrators might help uncover how and why affiliates and competitors are not realizing expected quality benefits including their expectations for joining the network.

5. CONCLUSION

Affiliation networks have the potential to enhance quality and financial performance for affiliates and stimulate improvement efforts in their local competitors. Their growth suggests that hospitals will pay for the privilege of affiliating with a high‐status sponsor in return for better quality and/or financial performance. Our results indicate, however, that these networks currently do not affect clinical quality and patient care experiences substantially, despite the quality‐focused mission. Neither do patients appear to be swayed by reputational effects of affiliation, given stability in affiliates' patient volumes. Cancer but not multispecialty affiliates benefit financially, though whether that is due to selection and/or affiliation remains unclear. Competitor hospitals experience no financial improvements and mixed changes in quality. Thus, affiliations may be legitimizing and sometimes beneficial without improving quality. Other strategies appear needed to achieve that goal, at least in multispecialty‐sponsored networks.

Jin B, Nembhard IM. Effects of affiliation network membership on hospital quality and financial performance. Health Serv Res. 2022;57(2):248-258. doi: 10.1111/1475-6773.13876