Interpretations of Integration in Early Accountable Care Organizations

Abstract

Context

It is widely hoped that accountable care organizations (ACOs) will improve health care quality and reduce costs by fostering integration among diverse provider groups. But how do implementers actually envision integration, and what will integration mean in terms of managing the many social identities that ACOs bring together?

Methods

Using the lens of the social identity approach, this qualitative study examined how four nascent ACOs engaged with the concept of integration. During multiday site visits, we conducted interviews (114 managers and physicians), observations, and document reviews.

Findings

In no case was the ACO interpreted as a new, overarching entity uniting disparate groups; rather, each site offered a unique interpretation that flowed from its existing strategies for social-identity management: An independent practice association preserved members’ cherished value of autonomy by emphasizing coordination, not “integration”; a medical group promoted integration within its employed core, but not with affiliates; a hospital, engaging community physicians who mistrusted integrated systems, reimagined integration as an equal partnership; an integrated delivery system advanced its careful journey towards intergroup consensus by presenting the ACO as a cultural, not structural, change.

Conclusions

The ACO appears to be a model flexible enough to work in synchrony with whatever social strategies are most context appropriate, with the potential to promote alignment and functional integration without demanding common identification with a superordinate group. “Soft integration” may be a promising alternative to the vertically integrated model that, though widely assumed to be ideal, has remained unattainable for most organizations.

Accountable care organizations (ACOs) represent the latest attempt to heal a fragmented and increasingly expensive health care system by fostering integration. Yet there is considerable uncertainty about what the ACO model will mean in practice: what degree of integration will ACOs pursue; how do they propose to achieve it; and how will these aspects differ from what has been tried in the past? Integration is a continuous, multidimensional construct encompassing operational, financial, and social connectedness among groups. Foundational work distinguished three types of integration: functional (systemwide coordination of support functions and activities), physician-system (economic and social linkages between physicians and the system), and clinical (“the extent to which patient care services are coordinated across people, functions, activities, processes, and operating units so as to maximize the value of services delivered”; Shortell et al. 1996, 30). This work indicated that clinical integration is paramount and must be supported by physician-system integration. Fundamental to both is the social dimension; both hinge on effective intergroup (provider-system and provider-provider) relationships. We suggest, therefore, that in order to understand what ACOs might offer, it is essential to consider this new model through a social-psychological lens.

The 1990s saw a major push towards vertically integrated systems under single ownership, based on the assumption that, having become part of a single entity, providers across the continuum of care would perceive themselves and behave as such. However, structural change did not create the desired social-psychological change: Despite being nominally part of the same organization, physicians and hospitals continued to see themselves as separate groups with divergent interests, values, and worldviews (Budetti et al. 2002; Fiol, Pratt, and O’Connor 2009). While some integrated systems have achieved exceptional performance, many others have faltered at operational, financial, legal, and—perhaps especially—social hurdles (Crosson and Tollen 2010; Shortell et al. 2000). With the decline of ownership-based vertical integration, attention has increasingly turned to “virtual integration” based on contractual relationships and strategic alliances; yet this, too, has its drawbacks, as it offers less of a social or economic basis for coordinated action (Conrad and Shortell 1996; Robinson and Casalino 1996).

ACOs have been heralded as a way to address “deficient” integration (Berwick 2011, 1) without once again demanding that physicians and hospitals adopt radical structural change. In the ACO model, providers across the continuum of care agree to become accountable for a population of patients: if they meet quality targets while saving money, they share in the savings; if they lose money, they may be responsible for the shortfall. This model, designed to accommodate virtual as well as traditional forms of integration, promises a way to coordinate care in less integrated delivery systems—which, while seldom viewed as the ideal, are recognized as the norm (Shortell and Casalino 2008; Fisher et al. 2007). However, whereas the structural, financial, and technical characteristics of ACOs are increasingly well specified (McClellan et al. 2010), it remains unclear what integration means in the context of this new model and especially what it means in social-psychological terms.

The ACO appears to be a peculiarly social intervention, using social levers—shared accountability and collective incentives—to encourage new relationships among groups. But do ACOs attempt to create the same sorts of intergroup relationships as vertical integration (albeit through different means or on a longer time frame)? If so, the ACO might be understood as an organizational structure that brings disparate groups together by promoting intergroup harmony and collaboration. Or is there something qualitatively different about the social dynamics that ACOs seek to foster? To illuminate this issue, we examined how implementers actually interpret ACOs as a mechanism of integration. As noted earlier, “integration” is a complex construct, and the term can denote a structural, operational, and/or social reality. However, we purposely maintained a loose definition of integration, as we were less interested in the objective meaning of the concept than in its subjective meaning to participants. The study explored to what extent, and in what ways, implementers engaged with this concept when discussing their nascent ACO.

Methods

Theoretical Framework

Understanding the ACO's implications for intergroup relations requires a theory of how groups interact. The literature abounds with discussions of the importance of organizational culture, climate, governance, leadership, and collaboration in general. Yet what is needed is a theory that can account for the social processes that give rise to these phenomena. In contrast to prevailing approaches, which tend to be descriptive rather than explanatory, the social identity approach (SIA, comprising social identity theory; Tajfel and Turner 1979, and its extension, self-categorization theory; Turner et al. 1987), offers such a framework. Focusing on the important role of group memberships (e.g., “I am a physician”) in individuals’ self-concepts, this broad theory addresses multiple dimensions and determinants of group and intergroup dynamics. A recent review illustrated the SIA's potential as a paradigm for integrating diverse conceptual approaches to health care groups (Kreindler et al. 2012a).

A long tradition of research has shown that shared social identity is the wellspring of cooperation, mutual influence, and committed action in the service of group goals (Haslam 2004; Kreindler et al. 2012a). Nonshared social identities do not necessarily spell discord; however, identity threat (to the existence, status, distinctiveness, or norms of a valued group) can trigger highly destructive intergroup conflict (Hewstone, Rubin, and Willis 2002). Moreover, even in the absence of outright conflict, intergroup collaboration is challenging because communication does not flow as easily across intergroup boundaries (Haslam 2004).

How, then, can organizations overcome intergroup silos? One option is decategorization: encouraging staff to focus on personal rather than social identity. This approach has shown limited success in the workplace; employees tend to cling to existing social identities or form new ones (Peters, Morton, and Haslam 2010). Even when decategorization succeeds, individualistic approaches do not call forth the same degree of commitment as those that harness the power of social identity (Haslam, Reicher, and Platow 2010) and seem more conducive to fragmentation than integration. A second option, recategorization, promotes a shared, superordinate group as an alternative to subgroups—but to compel highly identified subgroup members to forsake their distinct identities provokes identity threat and resistance (as observed during the structural consolidation of the 1990s; Fiol, Pratt, and O’Connor 2009). The social identity literature appears to have converged on a third option, dual identity, which entails emphasizing both subgroup and superordinate identities. However, determining when, how, and how much to stress different identities is a complex undertaking in a small interprofessional team, let alone an entire delivery system (Hean and Dickinson 2005). Notwithstanding some promising theoretical models of how organizations may enshrine dual identities (Fiol, Pratt, and O’Connor 2009; Haslam, Eggins, and Reynolds 2003), it remains unclear to what extent organizations accomplish this in practice and with what impacts. This study used social identity as a sensitizing concept to investigate how implementers would manage the many identities that ACOs brought together. In particular, we looked at whether the ACO was perceived as a new superordinate group that would create harmony among existing subgroups.

Methodology

This study was part of a broader, ongoing evaluation of four organizations participating in the Brookings-Dartmouth ACO Collaborative: Monarch Healthcare, an independent practice association (IPA); HealthCare Partners (HCP), a combined group practice and IPA; Tucson Medical Center (TMC), a hospital engaging community physicians; and Norton Healthcare, an integrated delivery system (for detailed descriptions, see Carluzzo et al. 2012a, 2012b; Gbemudu et al. 2012a, 2012b; also see Larson et al. 2012; Van Citters et al. 2012). Each pilot ACO had recently signed a letter of agreement with a private payer (a fifth was excluded because it had not confirmed its payer partner). Capitalizing on our access to candid informants from real organizations grappling with the challenges of ACO development, the research afforded a snapshot of prevailing attitudes and strategies at the outset of this process.

Qualitative methods are most appropriate for exploring perspectives and interpretations. We undertook a two- to five-day visit at each site, led by a Dartmouth researcher who focused on the technical/structural aspects of ACO implementation (Bridget Larson) and an independent researcher who focused on social aspects (Sara Kreindler); two health policy fellows (Kathleen Carluzzo, Josette Gbemudu) and one PhD student (Frances Wu) attended two site visits each. We requested access to staff representing a variety of professional roles (specialist/primary care/nonphysician), organizational levels (senior/middle/nonmanager), and attitudes (ACO proponent/skeptic) who could comment on the developing ACO. We conducted individual and group interviews with 114 participants (Monarch 32, HCP 35, TMC 17, Norton 30), including senior and upper-middle managers of the parent organization; physicians who were affiliated (Monarch, TMC) or employed (all sites); and high-level representatives of potential ACO partners (Monarch). Our semistructured interview guide covered topics that were both explicitly social (e.g., perceptions of the ACO and integration, relationships among relevant groups and how the ACO might affect these, physician engagement strategies) and nonsocial (e.g., ACO structure, technical capabilities); both types of questions elicited information pertinent to social dynamics. Interviews typically lasted thirty to sixty minutes (a total of nine to nineteen hours per site). Also, we were participant-observers at regularly scheduled meetings (all sites but Norton) and a specialist forum (Monarch) and reviewed corporate websites and ACO-related documents. Interviews were recorded and transcribed, and extensive field notes taken. We were interested in both the content of participants’ comments and the language they chose.

Analysis followed the constant comparative method (Strauss and Corbin 1998). First, one of us (Sara Kreindler) open-coded the data using codes that reflected literal content; this phase supported the inductive development of a list of codes that were grouped into themes (e.g., “reasons for ACO: values,” “physician engagement: input”). This coding scheme was applied to the data and continuously revised. Next, preliminary findings were considered in the light of the SIA to generate a theory-based coding scheme describing management strategies (e.g., “supporting subgroup identity,” “promoting a common identity”). The theory-based scheme was applied to the full data set by the main coder and to 10 percent of the data by a second, naive coder (Ashley Struthers); disagreements were resolved by consensus. When the consensus involved revising the coding scheme, the revised version was reapplied to all similar instances; this process was reiterated until we arrived at the final version. We coded by hand and used Microsoft Excel to organize the data during later stages of analysis. Both literal and theory-based coding schemes were used to draw interpretations, which were tested against the data (e.g., by searching for disconfirming instances), then discussed by the full team, and checked with key informants.

Results and Discussion

This article focuses on the way participants understood integration and the ACO. Findings on each site's intergroup landscape and preferred social identity–management strategies are only summarized here but are presented fully in a companion article (Kreindler et al. 2012b). We would note, however, that each organization applied its strategies not only to generate buy-in for the ACO per se but also to engage providers in interventions to improve quality and reduce costs.

Certain commonalities appeared in the way participants at all sites discussed ACOs. The majority expressed enthusiasm about the model's potential to improve quality while reducing costs. Leaders explaining their decision to pursue an ACO mentioned both values and financial considerations but emphasized the former over the latter:

• “You’ve gotta have the right vision and values, you gotta be doing it for the right reason … it's a good business case for the organization as a business, but that's not why we’re doing it; we’re doing it because it's the greater good.” (Monarch senior manager—physician)

• “It's all about the patient; it's all about quality—and we adopted some philosophies here that weren't necessarily good [for] business but were definitely the right thing to do.” (TMC senior manager)

Participants at all sites expected the ACO to “align incentives” by rewarding all providers for the “value, not volume,” of care, enhance care coordination, and improve data collection and sharing through health information technology (HIT). However, aligned incentives and data sharing among groups do not necessarily imply a particular kind of social relationship. In fact, the discourse about how the ACO might affect intergroup relationships varied markedly from site to site. We next look closely at the social context, perspectives on integration, and discourse around the ACO at each of the four sites.

Monarch

Monarch Healthcare was an independent practice association (IPA) in southern California. Monarch's core business was representing independent physicians (with an emphasis on primary care), although it also maintained a small group of employed physicians. With a fairly advanced care-coordination and HIT infrastructure, Monarch planned to use the ACO to extend its existing care-management activities to preferred provider organization (PPO) patients.

The Group Landscape

Both within and outside the context of the ACO, Monarch placed great importance on engaging its primary care physicians. Physicians were understood to identify strongly as physicians and less strongly, if at all, with Monarch. Two primary engagement strategies were in evidence (for details, see Kreindler et al. 2012b). First, managers strove tirelessly to “build that relationship” through outreach, two-way communication, and helpfulness. Second, the organization demonstrated support for physician identity by promoting physician leaders, highlighting Monarch's own physician-led nature, and endorsing physician norms, including the norm of autonomy. Monarch's website represented it as an organization through which physicians could express and fulfill their medical identity (“Our IPA is run by physicians with a personal history of practicing quality medicine, and a long track record of making it a success for doctors who share the same philosophy of excellence”). Managers also—but less frequently—mentioned influencing physician behavior through individualistic rewards such as pay-for-performance. They had also begun to engage specialists in the ACO, using similar strategies (with somewhat more emphasis on individualistic messages). Cautious negotiations were also under way with certain hospitals, which might later decide either to join Monarch's ACO or to form their own (“it's one of these things of ‘Let's be best friends, and oh, by the way we’re gonna be competing against each other at some point’”).

Perspectives on Integration

IPA physicians overwhelmingly portrayed integrated systems as a threat to autonomy, which was viewed as a crucial aspect of physician identity. As one physician declared, “We’re so fiercely independent”—representing independence as not merely a personal preference but a group norm. To IPA physicians, becoming an employee of Kaiser (the most prominent local integrated system) or, worse yet, a hospital, meant departing from the prototype of what a physician should be.

• “I don't wanna work for Kaiser … I don’t! I think it's essential that an organization like Monarch, which is an independent practice organization, shows the world that this whole ACO concept can be done well … where the physicians are not employees, and you don't just go punch a time clock, and you have to do what the guy above you says … or the day of the independent physician is gonna be gone, and that's not the way I wanna practice medicine.”

• “My previous group sold to [a hospital], and a lot of doctors made a lot of money, but … it's like now they’re slave labor, and these people have basically sold their souls.”

While some primary care physicians endorsed “communication integration” or “integration of records,” only one (an employed physician) spoke favorably of integrated systems, and none applauded hospital-run integrated systems (“they just ruined practices and doctors’ lives”).

Two capitated specialists praised integrated systems—although it is noteworthy that the chief benefit they mentioned was the relatively soft feature of communication (not, say, standardization, which implies shared group membership and uniformity); what is more, they described their own views as unusual or even “radical.” However, specialists echoed primary care physicians’ characterization of hospitals as a devalued outgroup. One specialist at a physician forum dubbed hospitals “the largest pigs at the trough … [what happens] when the trough is a little bit smaller but the pigs are still in charge?” The epithet “pigs at the trough” was subsequently taken up by other specialists. Several specialists told us that although they would prefer a specialist-led ACO (or no ACO at all), they would sooner work with Monarch's physician-led ACO than one led by a hospital (“for doctors, the hospitals would be the worst”).

Monarch managers, too, asserted the superiority of physician-led models, although their depiction of hospitals was more diplomatic (“So the approach we are taking is to try and make it a win-win all around, because it's not the hospitals’ fault per se … we want to help them as they’re transitioning to a more efficient model”). They did not discuss integrated systems and almost never used the word integration except to describe specific initiatives like data integration and integrated care teams, the latter being geographically based teams of nonphysician providers that assisted, but did not necessarily “integrate,” local physicians (“trying to support them … ’cause they are the physician, and we’re there behind them”).

Discourse about the ACO

There was relatively little discussion of what changes the ACO might bring, practically or socially. Leaders explained that “we’ve done it already in the managed care venue” and detailed the technical capabilities that would soon be extended to the non–managed care population. Monarch participants did not seem to conceive of the ACO as a social mechanism. When asked how the ACO would affect relationships among various groups, they typically spoke about existing relationships and engagement strategies. Only one manager mentioned any direct effect the ACO might have on relationships, specifically the promotion of pragmatic cooperation between primary and specialty care (“once they realize that the communication is such an important piece of success and their reimbursement is aligned with that success … they will realize the benefit of improving that communication”).

Monarch planned to include physicians and, where possible, hospitals, in the governance of the ACO. However, the idea that the ACO might create a sense of common identity—or even commonality—among participating groups was conspicuously absent, especially among Monarch leaders. When we asked about this directly, participants seemed to find the question difficult to understand. One senior manager responded that hospitals could formally become “part of our ACO” if they could “show to us that they absolutely are on board and have changed their model sufficiently”—in other words, the ACO was expected to retain Monarch's identity, not a new superordinate identity. Another cited a “shared vision” as a prerequisite—but not an outcome—for ACOs. Likewise, there was no intimation that ACO participation would make physicians feel more united with Monarch or with one another. In fact, the only time we heard a “social” account of ACOs at this site was when external partners used the rhetoric of collectivism in an appeal for fairness toward their own group.

• “If [specialists] feel that someone else is benefiting from their work, they’re gonna be resentful and it's not gonna work well. Almost like you need … the team spirit like you might have on a college football team or something.” (specialist)

Discussion

Monarch's success as an IPA depended on ensuring the engagement of primary care physicians who prized their independence. Accordingly, leaders eschewed any attempt to unite members behind a corporate identity, emphasizing instead how being part of Monarch enhanced physician identity. In an extension of this strategy, they avoided presenting the ACO as a superordinate group. Although Monarch made many efforts to promote positive intergroup relationships, the ACO was not expected to contribute directly to intergroup harmony and was certainly not envisioned as a new entity that would unite disparate groups. With its emphasis on improving care management and electronic connectivity, Monarch pursued functional, but not structural or social, integration. This “soft integration” approach helped Monarch maintain a delicate balance: promoting coordination while protecting physician identities.

HealthCare Partners

HealthCare Partners (HCP), located around Los Angeles, California, combined a medical group practice of employed physicians (300 PCPs, 100 hospitalists, 250 specialists) with an IPA. HCP had a three-decade history of care coordination, with a very advanced HIT and care-coordination infrastructure. It served mostly managed care patients but planned to use the ACO to extend its activities to more PPO patients.

The Group Landscape

Although employed and IPA physicians accounted for roughly equal proportions of patient contacts, HCP was clearly a group model first (for details, see Kreindler et al. 2012b). Managers described employed physicians as both highly identified (“wearing HealthCare Partners uniforms”) and “compliant,” whereas IPA physicians’ lower identification with HCP was treated as a fact of life (“they’re IPAs for a reason”). With employed physicians, managers reported relying on a common organizational identity and culture to stimulate engagement (“it's really the culture that runs that”; “this is the expectation of all of us”); with the IPA, they took a more individualistic and transactional approach (“they are much more customers”). HCP engaged in relationship-building with both groups, stressing two-way communication. Like Monarch, HCP was engaged in pragmatic cooperation with selected hospital partners; relationships with hospitals were described as “good” but also “challenging.”

Perspectives on Integration

HCP's group model was described as highly integrated. Unlike at Monarch, managers likened their model to Kaiser's (“we’re a virtual Kaiser”; “Kaiser without walls”). However, both they and physicians noted the advantages of being slightly less integrated (“we’re less bureaucratic than Kaiser”; “Kaiser almost got it right”; “a happy medium between … private practice and Kaiser”). Integration beyond the group model—that is, with IPA physicians or hospitals—did not seem to be a goal; when it was discussed at all, it was portrayed as unfeasible.

• “Because they [Kaiser] have ownership, they have the luxury of saying … ‘thou shalt do it this way.’ You can't do that in a contractual relationship; you have to say, how do we do this so that it's a win-win, how am I gonna benefit you, how am I going to benefit me … it's a much more complex, almost entity-by-entity kind of decision.” (senior manager)

HCP did not arrange for us to meet with IPA physicians or hospitals, which corroborated our sense that these groups were seen as less central to the organization and its ACO.

Discourse about the ACO

Many participants described HCP as “already” an ACO (“When Health Affairs first started writing about it, [they] actually included us as an example of what they thought an ACO ought to look like, so … [we’ve] always been one”). The ACO was not expected to require significant practice change of employed physicians (“I mean, they’re kind of with the program in the first place”; “I don't see any challenges”). When asked how the ACO would affect relationships beyond the group model, managers talked about business relationships, specifically, consolidation of HCP's business among fewer IPA physicians and hospitals.

• “I think we’re gonna look at that and say, wait a second, we know that you’re in six or seven IPAs, we know that you’re in one or two—I’m gonna do more business with you because I want more loyalty from you relative to treating our patients, and I know you’re over here in six or seven—well, just take our name off your list, go deal with the other folks.”

However, there was no evidence that the ACO was expected to move intergroup relationships beyond pragmatic cooperation, much less that it was envisioned as a unifying “superorganization.” A senior manager (R) highlighted the continuing separateness of the ACO's component groups:

• R: If you think of the variety of self-interests involved, and how do you negotiate each individual self-interest—that's hard. So when I say [the ACO is a] decent [mechanism], well maybe I can deal with one, two, or three. I can't deal with all of them.

• Interviewer: Do you think there's potential in the ACO process to get beyond those self-interests in some way?

• R: You know, when we fall on the sword for God and country, that's tough. When you ask the guy to fall on the sword for health care reform, I would say—you know the answer.

Discussion

In this mature organization, the discourse around integration suggested a state of equilibrium: HCP's group model was already as integrated as leaders wanted it to be, and there was no push to achieve greater or faster integration with other entities. It stands to reason, then, that the ACO was not characterized as an integration vehicle or expected to function as social “glue.” HCP continued to promote a strong common identity within its group model but practiced a transactional approach to manage the IPA and engage external partners. The ACO was presented as neither a new structural entity nor a means of achieving intergroup harmony but simply as part of HCP's existing, gradual strategy of expansion.

Tucson Medical Center

Tucson Medical Center (TMC), a nonprofit hospital, was engaging community physicians in developing the Southern Arizona ACO, which was to cover PPO and Medicare Advantage patients; it was the only one of the four sites to establish a new legal entity. Care-coordination and HIT infrastructure was fairly well developed within the hospital and a few physician practices, but implementers were working to extend it systemwide.

The Group Landscape

While TMC maintained a small employed practice, its ACO development relied on the engagement of nonemployed physicians. Participants reported that since taking office three years earlier, TMC's new administration had greatly improved physician-hospital relationships. According to senior managers, their engagement efforts had begun with intensive relationship-building and evolved into intergroup partnership (“in order to get [physicians’] trust and respect you have to be able to treat them as equals, as partners, not as customers”). TMC fostered partnership through power sharing, as reflected in its plans to give physicians majority representation in the governance of the ACO (with only 20% for the hospital).

Perspectives on Integration

Participants recounted how TMC's failed experiment with vertical integration during the 1990s had left “ghosts and long memories.” The integrated system was said to have provoked “physician rebellion” by “forcing” them to adopt certain practices and financial arrangements, putting cost cutting ahead of clinical values (“it was managed costs, not managed care”) and favoring insurance companies (“under capitation, the winner was designed from the beginning to be the insurance company, and everybody else lost”). When participants spoke in favor of “integration,” they stressed that it meant collaboration, not control.

• “I’m not giving a speech about how we have to merge—I don't believe that at all; but I am saying … that integration and collaboration has to occur at a clinical level.” (senior manager)

• “The collaborative and collegial approach and the integration of different parts and pieces … is really where we’re learning day by day.” (primary care physician)

• “ … create integration without having to employ all the doctors.” (specialist manager)

While participants acknowledged the difficulties of a virtually integrated or nonemployed model, they argued that TMC's model fostered more genuine teamwork.

• “When I go around and listen to how people are saying, what the relationships are with physicians, we are so far ahead … and we don't employ docs … When you employ physicians, you get a lot of passive-aggressive behavior.” (senior manager—physician)

• “[Other hospitals are] trying to build their own network by gobbling up practices … They’re gonna have a heck of a challenge. Just cause you put everyone in the same uniform doesn't make ’em a team.” (specialist manager)

Discourse about the ACO

More than at any other site, participants at TMC discussed the ACO in social terms.

• “I think the ACO brings everybody together in one group to do the right thing.” (manager)

• “ … physicians working cooperatively together to improve outcomes and decrease costs and cost share.” (primary care physician)

• “[The ACO] incentivizes us to work together. Right now everyone is incentivized for themselves.” (nurse practitioner)

• “ … establishing kind of that collaborative environment where we can work together the way we haven't in so very long.” (senior manager)

However, senior managers described the ACO as not an overarching superstructure but a “limited purpose organization” whose “only goal for existence is to track and distribute savings.” Technological and care-management infrastructure was to be provided by a separate management services organization that contracted with the ACO; this bipartite structure would allow TMC to administer the former while keeping the latter physician directed. TMC leaders emphasized that the ACO would not subsume participating groups but represented “what independents could do collaboratively.” A primary care physician agreed:

• “It's not like the hospital is a big umbrella and they’re the one that's gonna pull the pins … the doctors [are] independent, they still retain their autonomy, they are affiliated with TMC as their hospital of choice.”

Likewise, when discussing the ACO's impact on physician relationships, participants frequently alluded to “conversations,” implying a freely entered, nondirective interaction among equal partners (“when doctors talk to doctors, good things happen”; “people are gonna have to sit down and have conversations about how to meet the metrics”).

Participants also stressed that the ACO would be controlled by and reflect the values of health care providers, not insurance companies. When asked how the ACO would differ from TMC's previous experience with managed care, a physician replied, “The ACO is classic, 100 percent managed care, but let's let us manage it and not somebody else.” Other physicians echoed, “It puts the management back where it belongs” and “we’re taking back what we lost.” TMC leaders insisted that the ACO must instantiate physician values (“I think the key ingredient is gonna be that the physicians … believe that this is the right thing to do for their patients”).

Discussion

TMC faced the challenge of engaging independent physicians whose negative experience with an integrated system had left them wary of the potential threat to their autonomy and identity. Accordingly, leaders carefully differentiated the ACO from the earlier version of integration, portraying the ACO as a vehicle for enabling disparate groups to come together while retaining their separate identities. Although TMC's ACO (unlike Monarch's or HCP's) was clearly intended to promote intergroup harmony, it was designed not as a superstructure but as a circumscribed joint venture between equal partners—minimizing the risk that it would be perceived as a superordinate group. By emphasizing the ACO's “soft integration” approach and its foundation in clinical values, implementers sought to banish the specter of the despised integrated system.

Norton Healthcare

Norton Healthcare, an integrated delivery system in Louisville, Kentucky, comprised five nonprofit hospitals, a cancer institute, and a network of physician practices. Its care-coordination and HIT infrastructure existed primarily within the hospitals, but it had begun systemwide implementation of a common electronic health record. The patient population for the ACO consisted of Norton's and the insurance company's self-insured employees.

The Group Landscape

Norton's primary engagement strategy was to promote dual identity: that is, advancing “systemness” while respecting the distinctness of individual hospitals and physician practices (“trying to create that uniqueness at each of the facility levels but … [a] one-system standard in practice across the enterprise”). Managers described being sensitive to the many identities within the organization and engaging members of each subgroup when designing change.

• “There is individual practice, there is unit practice, there is department practice, there is service line practice, there is facility practice … to understand those levels and to embrace those levels and to use those levels to your advantage because they can be used to your advantage effectively.”

They said they maintained “more collaborative types of relationships versus employment relationships” with employed physicians and also affirmed the importance of engaging nonemployed physicians, who made up 75 percent of Norton's medical workforce. They also reported increasing efforts to engaging physicians early on and to nurture physician leaders.

The principle of balancing subgroup and systemwide identities was enshrined in Norton's matrix structure, in which service lines cut across hospitals. Participants acknowledged the challenges of a matrix system, noting the slowness and complexity of building the consensus needed to achieve change (“it's sort of like the United Nations getting work done here”). However, they also expressed pride in the structure, affirming that it gave rise to true collaboration and, ultimately, higher-quality decisions (“it forces collaboration, and it forces communication that otherwise might not happen, so there are very few instances where we get down the road and look back and say how the hell did that happen”).

Perspectives on Integration

Participants described Norton's ongoing “journey” toward integration (“we’re still on that journey of becoming integrated but where we were ten years ago to where we are [now] is significantly different”; “we’ve become more integrated but still not fully”; “we’re integrated but we’re not integrated”; “it's been a journey for us”). Managers described “integration” in positive terms that highlighted unity and teamwork (“understanding that we are a whole, not pieces and parts”; “I think it's a symphony, OK? … Integration means that I not only care about what I am doing, I’m caring about what you’re doing.”). However, they stipulated that they had no intention to “integrate such that people lose their autonomy and their individuality,” become “monolithic,” or “micromanage” physician practices. Similarly, although a few participants intimated that a vertically integrated structure would be easier to work with than a matrix, none actively endorsed a change. There was a consensus that despite its limitations, the matrix structure “seems to work for us.”

Discourse about the ACO

Many participants expected the ACO to promote intergroup collaboration and teamwork.

• “ … being more cooperative among and between our entities.” (senior manager)

• “ … we’re hoping we can all communicate and read basically from the same page.” (specialist)

• “ … more team collaboration on the care of that patient … when we’re all one team, the team should figure out who should be addressing these particular issues.” (manager)

However, leaders insisted that the ACO was not an “entity”:

• “When someone says ‘accountable care organization,’ everybody thinks that's an entity … when in reality, I look at it as how our organization provides accountable care.”

• “Leave off the ‘organization’ word at this point: we’re moving toward accountable care.”

Instead, they referred to the ACO as a “communication vehicle,” a “tactic,” a “selling point” for improvement initiatives, and a means of changing people's “perspective” or “mind-set.” Even the lone manager who said the ACO would provide “structure” clarified that the ACO was not a separate structure but, rather, an ordered way to promote a “collective understanding of ‘here's what we’re aiming for, here's what we’re driving toward.’” Managers portrayed the ACO as a means to advance existing goals and downplayed the centrality of the model itself.

• “It's one philosophy or one method by which we can accomplish that.”

• “The ACO is a tactic to be providing that high-quality cost-effective care. I don't necessarily think it's a strategy. … Our strategy has always been to be a high-quality, cost-effective, low-cost provider.”

Participants also described Norton's approach to the ACO as cautious (“we’ve hedged our bets”; “a great way to get your feet wet”).

Some managers differentiated ACOs from 1990s HMOs (health maintenance organizations), describing the former as “more flexible,” grounded in “better information,” and compatible with physicians’ control over their own practice (“you’re in it, too … it's not somebody else doing this to you”). However, unlike at TMC, HMOs were not a major focus of concern.

Discussion

Norton's challenge was to create systemness without threatening strong subgroup identities. To do so, leaders perfected the strategy of dual identity, respecting diversity while building consensus through voluntary collaboration. In keeping with this delicate approach, they avoided a “hard,” structural definition of the ACO, instead using “soft” language that highlighted shared vision and team spirit. Unlike any other site, Norton was clearly using the ACO to advance a vision of an integrated system—but as a rhetorical rallying point, not a new organizational structure. By making the ACO about cultural, not structural, change, leaders ensured that it furthered—and did not disrupt—Norton's gradual journey toward integration.

An Example of Soft Integration

At several sites (TMC, Monarch, Monarch's two partner hospitals, and, to some extent, Norton), we heard about a model that, while distinct from the ACO, illuminated the emerging theme of soft integration. This model involved combining a strong primary care foundation with capitated groups of specialists, who had significant power over managing the service through either contracts (Monarch) or service-line agreements (hospitals). Where such agreements had been implemented, they were said to align incentives among providers, removing incentives to offer too much or too little care. In terms of social implications, they were reported to promote self-regulation through peer influence (“it's self-regulating … they self-police”; “they can influence each other's behavior”). At TMC, service-line agreements were also credited with helping specialist-hospital relationships progress from pragmatic cooperation to partnership.

• “The discussion really is targeted at improving the quality of care, improving the patient satisfaction, improving the process … it's about everything but money.” (specialist manager)

• “The conversation's a different level of conversation when you’re talking about how do we manage this together. And it's not you, or it's not me—it's us.” (senior manager)

One California hospital leader, who identified strongly as a primary care physician, explicitly presented such arrangements as an alternative to the type of “integration” that overrides subgroup identities. Whereas the vertical/virtual integration debate can be waged on economic grounds alone (cf. Robinson and Casalino 1996), this participant stressed the social dimension.

• “We believe that if you can focus on the outcomes and drive collaboration without forcing integration … you actually can get the same or better outcomes … And the reason I believe in that is because I’ve been an independently practicing physician for most of my career. So the question is, how can you harness what's best about independent doctors and independent entities like hospitals … instead of trying to destroy that or break it down.”

General Discussion

In examining what ACOs meant for integration, we discovered that “integration” signified something different at each site. This variability was not haphazard but reflected each organization's unique social identity landscape (see table 1). At Monarch, “integration” meant sacrificing independent physicians’ identity, but “coordination” remained a legitimate objective. At HCP, integration was part of the group model's culture but did not make sense beyond this highly cohesive core. At TMC, integration was reimagined as collaboration, not amalgamation. At Norton, integration meant a slow and careful journey that should result in a mosaic, not a monolith. Leaders responded to local perceptions by both designing and defining the ACO to fit the prevailing conception of “good integration” and contradict the prevailing conception of “bad integration.” Across the sites, accounts of undesirable integration—whether the object of opprobrium was vertical integration, managed care, employment, or hospital-run systems—centered on threats to subgroup identity and autonomy (which may itself be a key element of identity content, particularly for physicians; cf. Doolin 2002). Each system's existing level of integration conditioned what leaders could say and do without provoking identity threat: at Monarch, this meant not talking about integration at all; at HCP, promoting integration within the group model only; at TMC and Norton, advocating particular kinds of integration while (especially at TMC) repudiating others. Each organization accomplished a balancing act—whether between the primacy of physician identity and the need for coordination (Monarch); a highly integrated core and a much more loosely affiliated network (HCP); partnership and independence (TMC); or systemness and subgroup uniqueness (Norton). Rather than offer the ACO as a new identity object—which could have upset the balance—leaders represented ACOs in a manner tailored to their social context, accentuating superordinate and/or subgroup identities, as appropriate. In no case was the ACO interpreted as a new, overarching entity that would unify disparate groups; rather, each site offered a unique interpretation that flowed from existing strategies for social-identity management.

TABLE 1.

Summary of Findings

	Monarch	HCP	TMC	Norton
Current structure	Independent practice association (IPA)	Group practice plus IPA	Hospital engaging community physicians	Integrated delivery system (matrix)
Key challenge	Promoting coordination without threatening independent physicians’ identity.	“Different worlds” of strongly identified group-model physicians and weakly identified IPA.	Rebuilding physicians’ trust through equal partnership after failure of past integrated system.	Achieving integration through a slow process of bringing diverse groups into consensus.
Key engagement strategy	Support subgroup identity.	Group practice: promote common identity. IPA: pragmatic cooperation.	Intergroup partnership.	Promote dual identity.
Interpretation of ACO	Coordination, not integration: no new superstructure and no push for intergroup harmony.	Business as usual: keep high integration and strong common identity in group model only.	Promoting harmony via circumscribed joint venture between partners—not a new “umbrella.”	ACO supports integration but as a rhetorical rallying point, not a structural entity.
Why?	Invoking common ACO identity would create threat, so continue to be organization through which members express their physician identity.	Fostering unitary identity and culture strengthens highly integrated core but wouldn't work for loosely affiliated IPA and external partners.	Distinguish ACO from oppressive integrated system by emphasizing partners’ independence, voluntary collaboration, and physician power.	Declaring a new structure might threaten careful balance among groups, so soft-pedal ACO and incorporate into existing journey to integration.

Participants’ insistence that the ACO would not produce a superordinate identity is all the more notable when viewed in the context of two recent studies. In one, executives undertaking a hospital merger stressed the need to replace separate organizational identities with a new, common identity (Clark et al. 2010). In the other, clinicians defined interdisciplinary “integration” as structural unification that subsumed professional identities (Boon et al. 2009). The ACO, in contrast, appears to represent a flexible model with the potential to promote alignment and coordination without demanding that participants share a sense of identification with a superordinate group. Indeed, ACOs may offer a type of “integration for the rest of us”—an alternative to the vertically integrated model that is widely held to be ideal yet has not proved attainable for most organizations. “Soft integration” approaches may be a promising option for systems where formal integration is seen as undesirable or unfeasible—“soft integration” meaning not only the structural fact of virtual integration, but the deployment of social strategies that link coordination to identity(ies) other than that of a superordinate group. These findings are consistent with ACO proponents’ initial focus on enhancing coordination in virtually integrated systems (Fisher et al. 2007). Importantly, they also clarify how ACOs might go about achieving this aim and why they might be able to succeed where past approaches have failed.

Findings underscore the need for ACO developers to consider the social, tailoring their implementation strategies to the existing degree of intergroup closeness (see Kreindler et al. 2012b). Policymakers should continue to encourage ACO formation within nonintegrated systems in particular, ensuring that the current focus on “readiness” does not dissuade the very types of organizations for which the model was designed. Advocates might point out how ACOs differ (socially) from familiar integration models, stressing that structural unification is neither a necessary precondition nor even a necessary goal. To facilitate ACO development, organizations with limited care-coordination infrastructure might be “twinned” with those that are technologically more advanced but socially similar, as change-management approaches that work in one intergroup context may be ineffective or even counterproductive in another. As well, evaluations of ACO implementation should examine intergroup dynamics.

This research has several limitations. Our sample of sites was restricted to four early-adopting organizations taking part in a pilot and may not be representative of ACOs in general. Our sample of interviewees was also subject to bias; as participants were selected by the organizations, the range of views to which we were exposed depended on senior managers’ goodwill and comfort level. We were able to obtain a good sampling of the views of leaders but only a suggestion of those of other groups (sufficient for triangulation but not for firm conclusions about subgroup perspectives). Two additional sources of bias were that (1) managers were present during some interviews with nonmanagers and (2) participants knew we were associated with Brookings-Dartmouth, and may have edited their comments to conform to “our” position, censored themselves for fear of saying something inaccurate, or failed to articulate assumptions that they believed were obvious to an ACO insider (as the visits progressed, we tried to mitigate the latter problem by inviting participants to pretend we knew nothing about ACOs). We continuously reflected on how these sources of bias may have affected the data and tried to identify gaps in our understanding. When drawing interpretations, we kept in mind that we had access to the representations that organizational leaders (and, to some extent, others) offered publicly, but not to a full picture of organizational dynamics or attitudes. We also reflected on how our own expectations might have colored our interactions and inferences. The unexpected finding that each site revealed its own “signature” interpretation of ACOs suggests that our results were not merely a product of our preconceptions. Finally, this study did not examine all ACO-related intergroup relationships; relationships with payers and patients were out of scope, and participants focused on physicians, physician organizations, and hospitals. Other groups across the continuum of care may become more central as ACO development progresses; we would note, however, that physician organization and physician-hospital relations are widely seen as the core issues for health care integration (Crosson and Tollen 2010).

Although we can state that the four sites have successfully formed ACOs and implemented various initiatives to coordinate care, the most important measure of success—whether these (and other) ACOs do improve quality and reduce costs—can be taken only in the future. It will then be essential to assess the impact of social-identity management strategies on these outcomes. We will also be able to ask, Is some minimum level of integration or intergroup harmony requisite for successful ACO formation? Are governance models that support physician identity (i.e., physician- or collaboratively-led rather than hospital-led) more likely to flourish? To what extent, and how, might social identities evolve as ACOs mature? Such questions demand ongoing, mixed-methods research. It would also be valuable to examine interpretations of integration in other ACOs and in traditional integrated systems.

Conclusion

We set out to investigate how ACOs are understood as a model of integration. What we discovered is that ACOs are not necessarily about “integration” at all. Rather, they may be about offering levers (primarily financial and technical) for improving care coordination without disrupting the local form of social organization. The integrated-system experiment of the 1990s sought to improve coordination by fundamentally altering social organization—often with disastrous results. In contrast, the ACO model appears flexible enough to be used in synchrony with whatever social strategies are most suitable for the unique context. It can be implemented with hard integration, soft integration, or both in different parts of an organization. Thus, although it is very compatible with a focus on collaboration and social cohesion, its defining feature is not what it does socially but what it does not attempt to do. Universally, what the ACO appears to offer is not so much the promotion of intergroup harmony as the avoidance of intergroup chaos.