Accountable care organizations (ACOs) are faced with a fundamental challenge: They are accountable for outcomes that are only partly under their control. Specifically, ACOs, which are networks of providers that are jointly responsible for the costs and quality of care for defined populations of patients (Shortell, Sehgal, et al., 2015; Shortell, Wu, Lewis, Colla, & Fisher, 2014), are accountable for outcomes that are significantly shaped by choices of patients who may or may not seek or comply with care recommendations and who are free to self-refer for services outside the ACO (Ginsburg, 2014). In other words, ACOs are interdependent with—in the sense that their performances partly depend on—the choices of their patients. We call these “interdependencies related to patient choice.”
The challenge is not insignificant: Evidence suggests that more than two thirds of all specialty office visits under Medicare ACOs occur outside of those ACOs (McWilliams, Chernew, Dalton, & Landon, 2014). A study of one urban Medicare ACO found that 90% of beneficiaries had out-of-ACO expenditures, which represented a significant portion of total expenditures (Han et al., 2016). Furthermore, out-of-ACO “leakage” is most pronounced among high-cost patients with multiple chronic conditions (Han et al., 2016; McWilliams et al., 2014). In the absence of benefit and incentive structures to keep patients within ACO networks and compliant with evidence-based care (DeCamp & Lehmann, 2015; Ginsburg, 2014; Sinaiko & Rosenthal, 2010), it is unclear how ACOs can effectively cope with interdependencies related to patient choice.
One early report finds that ACOs may be attempting to do so by investing in patient activation and engagement programs (Shortell, Sehgal, et al., 2015), which have been associated with improved health outcomes and controlled costs (Greene, Hibbard, Sacks, Overton, & Parrotta, 2015; Hibbard & Greene, 2013). In addition, ACOs appear to be using care management programs, usually aimed at overseeing the care of high-risk subgroups, to control outcomes (Rundall, Wu, Lewis, Schoenherr, & Shortell, 2016). Beyond these efforts, however, we know little about how ACOs are attempting to cope with interdependencies related to patient choice or what their strategies reveal about the nature of such interdependencies. Lacking this knowledge, researchers may be unable to explain ACO performance, because they will not know what actions contributed to performance, and managers may misunderstand or be ill-equipped for coping with interdependencies related to patient choice. To address this gap, we examined these interdependencies and coping strategies through 2-year qualitative case studies of four ACOs.
A key insight of open-system perspectives in organization theory is that organizations are interdependent with their environments in the sense that their performances depend not only on internal operations but also on supply and demand fluctuations generated by the actions of external suppliers, competitors, regulators, and customers (Fennell & Alexander, 1987; Pfeffer & Salancik, 2003; Thompson, 1967). Regarding interdependencies, Thompson (1967) distinguishes between constraints and contingencies. Constraints are environmental actions that are sufficiently stable or patterned such that they may be anticipated and accommodated through organizational design. Contingencies, on the other hand, are environmental actions that are emergent, not patterned, and therefore must be flexibly accommodated as they arise.
To manage environmental interdependencies and moderate their undesirable effects on organizational performance, organizations develop various boundary-spanning practices, which are recurring activities aimed at improving the organization’s protection from, influence over, or knowledge of its environment (Thompson, 1967). For example, an organization may buffer against environmental interdependencies by altering internal structures and processes to improve its ability to accommodate fluctuations in supply and demand (Fennell & Alexander, 1987; Pfeffer & Salancik, 2003; Thompson, 1967). By contrast, an organization may attempt to smooth undesirable environmental fluctuations by intervening directly in the environment to influence other actors and create conditions that are more predictable and favorable for the organization (e.g., stimulating demand for services; Thompson, 1967). An organization may also try to improve its power in the environment by establishing connections (e.g., mergers, alliances) with others that have complementary capabilities (Fennell & Alexander, 1987). Finally, because uncertainties regarding the environment can weaken an organization’s power vis-à-vis others in the environment (Pfeffer & Salancik, 2003), an organization may attempt to improve its knowledge of the environment so that it may better adapt (Thompson, 1967).
The literature on boundary-spanning developed to explain how organizations cope with interdependencies created by supply and demand fluctuations. Although consequential for ACOs, such interdependencies are not our focus here. Rather, we are concerned with the unique interdependencies that ACOs confront precisely because of increased accountabilities for outcomes under arrangements wherein patients are free to access services outside the ACO and may or may not seek or comply with care recommendations from those within the ACO (Ginsburg, 2014). The nature of these interdependencies related to patient choice is underexplored; therefore, we do not know how general theories of environmental interdependence apply to ACOs.
Existing research would suggest that interdependencies related to patient choice may be problematic because of three common behavioral patterns that can impact outcomes for which ACOs are accountable. Underuse of evidence-based preventive or chronic care services may unnecessarily increase the prevalence or intensity of illness resulting in higher costs and poorer health outcomes (McGlynn et al., 2003). Overuse of resource-intensive services—where the potential for harm outweighs proven benefit or less-intensive services of equal or better efficacy are available—also risks increasing costs and diminishing quality and health outcomes unnecessarily (Hoffman & Pearson, 2009; Korenstein, Falk, Howell, Bishop, & Keyhani, 2012). Finally, leakage of ACO patients to providers outside of the ACO, including through self-referral, may lead to care that is inefficient, inappropriate, or of poor quality but over which the ACO has no direct influence (Han et al., 2016; McWilliams et al., 2014). Under fee-for-service arrangements, these three behavior patterns had little negative effect on most providers; however, under ACO models, they pose significant threats to performance (Fisher et al., 2009).
Given that ACOs are accountable for outcomes that are shaped by interdependencies related to patient choice, we would expect ACOs to engage in boundary-spanning practices aimed at mitigating the negative impact of these interdependencies on organizational performance. It is unclear, however, how boundary-spanning practices that were developed to address supply and demand fluctuations translate to the unique interdependencies that arise from overuse, underuse, and leakage. An analysis of ACO boundary-spanning is important for both managers and researchers. The former, faced with the prospect of increased accountabilities, need to understand the nature of interdependencies related to patient choice and the range of solutions available for coping. Similarly, the latter need to know what practices ACOs are using to cope with interdependencies, as well as how these practices are presumed to improve control in order to inform evaluation efforts. The purpose of our research was to identify, conceptualize, and categorize ACO efforts to cope with interdependencies related to patient choice.
The work reported here is one line of analysis in a larger project focused on the development and evolution of full financial risk ACO models for commercially insured and Medicaid populations (Hilligoss, Song, & McAlearney, 2016; McAlearney, Song, & Hilligoss, 2016). We conducted four qualitative organizational case studies of a diverse set of ACOs. Consistent with prevailing definitions of ACOs (e.g., Shortell et al., 2014), we selected organizations that (a) involved groups of providers that (b) collectively assumed accountability for both total costs and quality of care for (c) defined populations. The sample (see Table 1) varied in terms of the type of populations served (i.e., pediatric or general), organizational structure (i.e., primary care group or integrated delivery system), and payer (i.e., Medicare, Medicaid, commercial). The organizations were located in four different markets in five states. We selected this highly diverse sample but focused on commonalities across sites under the reasoning that findings common to a small heterogeneous sample are likely more transferable to other instances of the phenomena than are findings common to a small homogenous sample (Patton, 2002).
Because the ACO concept was new and evolving, especially in the commercial space, we elected to visit each site twice for 2- to 3-day visits, with visits separated by a year (summers: 2013 and 2014). During these visits, we conducted hour-long interviews with a total of 89 informants, including executives (e.g., CEO, CFO, vice presidents), managers (e.g., care coordination manager, practice managers), and physicians. (Table 1 provides a breakdown of informants by site.) Because our purpose was to understand how ACOs were attempting to cope with interdependencies related to patient choice, we focused on the types of informants who were tasked with designing and implementing solutions for these interdependencies. We asked open-ended questions about general topics (e.g., market in which ACO formed), as well as questions targeted to the informant’s area of expertise (e.g., we asked a care coordination manager how care managers interact with patients). Our approach during the second round of site visits was essentially the same but focused on understanding how the ACOs were evolving, checking our early interpretations with our informants and gathering data on topics that were either unidentified or underexplored during the first set of visits. The interviews were recorded and transcribed verbatim.
We generated broad a priori codes from the interview guide (e.g., care management, market context) to serve as an index on our database. Multiple research team members individually coded a subset of transcripts by hand on paper and then compared results. We then resolved discrepancies through consensus and refined the coding dictionary. Three research assistants then coded all transcripts using Atlas.ti (Scientific Software Development, 2008). One author led the remainder of the analyses. The other authors reviewed the findings and checked these against their own interpretations. Because all authors were engaged in data collection at all four sites in both years, we each had deep familiarity with the informants and ACOs. We resolved any disagreements over interpretations through consensus.
Next, we read through reports of data coded with a select set of broad codes relevant to interdependencies related to patient choice (e.g., “Attribution,” “Care Management,” “Market Context,” “Patient Outreach,” “Patient Population”) and sorted coded material first by ACO and then, emergently, into categories that represented boundary-spanning practices: recurring actions that ACOs were undertaking to cope with interdependencies related to patient choice. That is, we looked for statements that either explicitly or implicitly indicated actions that were being taken to address underuse, overuse, or leakage. This resulted in 15 boundary-spanning practices.
To understand why the ACOs were using these particular practices and what these practices revealed about the nature of interdependencies, we next analyzed connections among the practices. We wrote analytic memos about each practice to improve our understanding of the distinct purposes they appeared to serve. This allowed us to identify ways that specific practices were related (e.g., several practices appeared to address issues of service availability). It also helped us sharpen distinctions between the practices, such as refining the labels we used for them. This round of analysis resulted in five groupings of boundary-spanning practices. Each grouping represented a category of factors that appeared to play an important role in these interdependencies (e.g., interactions with patients, ACO knowledge of the environment). Each category of factors could be understood as creating conditions contributing to the ACO being dependent on—and thus its performance impacted by—some external action. We then examined the relationships between each category of factors and the boundary-spanning practices contained within it, asking how the practices could be interpreted as strategic attempts to impact the interdependency represented by the category. We identified 11 strategic goals (e.g., increase the convenience of existing services), each of which involved one or more boundary-spanning practice. Table 2 provides an overview of linkages between these concepts.
Throughout the analyses, we required that a theme (i.e., boundary-spanning practice, category of interdependency factors, or strategic goal) appear in at least two different interviews from the same ACO to be included. This ensured we avoided building theory from a single individual’s perspective. In addition, we refined our themes continually through constant comparison of various segments of coded material (Glaser & Strauss, 1967), incorporating insights from discrepant data and returning to transcripts to read additional details for context. The Ohio State University Behavioral and Social Sciences Institutional Review Board approved this research.
Our analyses resulted in a process theory that explains how boundary-spanning practices are intended to enable ACOs to cope with interdependencies related to patient choice. We have organized our findings around five broad categories of interdependency factors. These were aspects of the ACO or its environment that informants indicated—explicitly or implicitly—were consequential in shaping the actions of patients and, in some cases, providers outside of the ACO (hereafter, “non-ACO providers”) in ways that impacted overuse, underuse, leakage, or some combination of these. Under each category of interdependency factor, we present one or more strategic goals aimed at altering interdependencies in ways that improved the ACO’s ability to influence patient actions or other environmental drivers of key outcomes (i.e., costs, quality of care, health). Finally, under each strategic goal, we present the boundary-spanning practices (e.g., recurring patterns of action) that enacted that goal through efforts to understand, cope with, or alter specific aspects of an interdependency between the ACO and its environment. Table 2 provides an overview. Supplemental Digital Content 1 (Table 3, Supplemental Digital Content 1, http://links.lww.com/HCMR/A24) supplies additional supporting quotes.
Availability of Needed Services
Our analyses suggest that ACO leaders believed that patient choices were shaped in part by the relative availability of needed services. We identified three strategic goals that the ACOs were using to improve the convenience, reach, and types of services available.
Increase the convenience of existing services
The first strategic goal aimed at coping with interdependencies that arise from the relative availability of needed services was to increase the convenience of existing services. Echoing a theme we heard from informants at all four ACOs, one executive said, “We need to understand the threats surrounding consumer choice and not put our head in the sand over convenience being just as important to consumers as quality.” To achieve this goal, the ACOs engaged in the practice of expanding primary care hours, accessibility, and appointment options. Specifically, they were attempting to expand office hours, allow more same-day visits, and staff call centers with nurses who could advise patients after hours and direct them to appropriate levels of care. The ACOs were improving convenience as a way of reducing leakage and overuse, particularly of emergency departments (EDs). Informants at three ACOs, however, noted that habits of primary care physician (PCP) staff who “don’t try to squeeze that patient in, [but] instead say ‘Go to the ER’” (Executive) hindered this practice. Similarly, a manager at a different ACO pointed to a barrier in the mindset of some PCPs who feel, “they’re a physician: you need to come to them; these are their hours.”
Increase the reach of the ACO into underserved communities
The second strategic goal aimed at coping with interdependencies that arise from the relative availability of needed services was to increase the reach of the ACO into underserved urban or rural communities. Three of the ACOs were using the practice of providing care beyond the clinical setting through innovative, nontraditional mechanisms or in non-health care settings in order to exert greater influence over patient populations who were “too far away” (Manager) to be reached through traditional health care delivery approaches. For example, one pediatric ACO had been using a variety of tactics including distributing inhalers through the public schools to address asthma rates, making contraception and prenatal care available through a special clinic connected with the public schools to combat high rates of infant mortality and teen pregnancy and sending a mobile unit to rural areas to provide preventive care, including immunizations.
We still have mobile units going down several times a week…to schools in probably…five or six different counties, mainly places where there are not primary care providers, or at least not pediatricians who are there. Working closely with schools in those areas. (Manager)
The other pediatric ACO was exploring the use of mobile clinics and community health workers and had a pilot program using telemedicine to provide “house calls” for medically complex, chronically ill patients. One of the adult ACOs was piloting a community health worker program that used lay health educators to “go out to the homes or other places patients might be found” to do “asthma education for families that show a lot of high utilization” (Manager). Thus, this practice appeared to be aimed primarily at reducing underuse.
Augment the ACO with additional health services
The third strategic goal aimed at coping with interdependencies that arise from the relative availability of needed services was to augment the ACO with additional needed services. Two boundary-spanning practices were used to under this strategy. First, one ACO did not include any postacute services (e.g., skilled nursing facilities [SNFs], home health care), whereas another ACO did not include any specialty physician, hospital, or postacute care services. To compensate, these two ACOs engaged in the practice of developing and maintaining connections with preferred non-ACO providers. These ACOs established formal contracts with a select set of key non-ACO providers to ensure that services across the full continuum of care were available for their populations. In exchange for increased volumes of referrals and, in some cases, higher reimbursements, the ACOs expected non-ACO providers to cooperate with ACO cost and quality goals. One executive, who was also a physician, explained his ACO’s process:
If I spend a month banging heads with a [non-ACO] specialty group who are unwilling to look at guidelines, look at recent literature, unwilling to partner with us in ways that we think are necessary…we may take them off the specialty panel. If I have a [non-ACO] group that is willing to…embrace [our] philosophy…we will look at them to potentially add. (Executive)
Through this ongoing practice of developing and maintaining both formal and informal connections, ACO leaders were attempting to align non-ACO providers with ACO goals but without fully involving those non-ACO providers in the risks and rewards of the ACO. In essence, this practice represented an attempt to lessen ACO dependence created by leakage by increasing ACO influence under conditions where leakage was unavoidable.
Second, all four ACOs reportedly had stable provider networks; however, some informants noted points in their ACOs’ histories, as well as a few recent instances, in which they engaged in the practice of adding new providers to the ACO to fill important gaps. For example, informants at both pediatric ACOs mentioned efforts to recruit or establish pediatric practices in select rural counties. Similarly, the ACO that included no specialty physician services hired a cardiologist, a gastroenterologist, and a psychologist in the time between our two visits. Although different rationales were offered for each of these hires, the important point is that the ACOs attempted to address underuse, overuse, and leakage arising from the relative availability of services through targeted expansions of their own organizations.
Interactions With Patients
Our analyses also suggest that ACO leaders believed that patient choices were shaped in part by the frequency, efficiency, and quality of interactions between them and the ACO. We identified three strategic goals, each enacted by a single boundary-spanning practice, that ACOs were using to improve patient interactions and, thereby, reduce environmental dependencies.
Increase the frequency of interactions with patients
The first strategic goal aimed at coping with interdependencies that arise from insufficient or limited interactions with patients was to increase the frequency of those interactions. All four ACOs engaged in the practice of communicating and reminding patients, establishing and using a variety of channels to connect with patients and encourage them to keep appointments and comply with recommended care. The ACOs used a variety of approaches for reminding patients about appointments, including phone calls, mailings, text messages, e-mail, and Web portals. All four ACOs used practice staff and care coordinators to reach out and remind underutilizers when they were due for preventive care services (e.g., immunizations, mammograms). Although providers have long reminded patients of upcoming appointments, the practice described here goes beyond those traditional efforts. A physician at one ACO explained how this practice, instantiated as part of the ACO, represented a change in the way providers viewed some patients and an increased level of effort to influence them.
…in the old paradigm, we’d just tell [patients] what to do, and if they didn’t do it, they were just a bad patient. And under this paradigm, we’re reaching out to the patient. So the disease registry is the trigger to chase the patient and get the labs done, get them into the physicians’ offices and try to promote higher-quality care, more preventative care. (Physician)
A manager at a different ACO said that this practice entailed, “a lot of encouraging” and “a lot of cheerleading and getting [patients] to make their appointments.”
Improve the efficiency of interactions with patients
The second strategic goal aimed at coping with interdependencies that arise from insufficient interactions with patients was to improve the efficiency of patient interactions. To make the most of each patient encounter, the ACOs were addressing gaps-in-care at the point of care. That is, when patients were interacting with the ACO, such as while in the PCP’s office, the ED, or the hospital for one concern, ACO providers would simultaneously attempt to address underuse of preventive or appropriate chronic care either through directly providing that care or arranging follow-up appointments. This practice entailed reshaping PCP routines and building alerts into the electronic health record (EHR), as a care management administrator explained:
…whenever the patient calls for a flu shot or a prescription refill, it’s right there in their chart. I put it so it’s an alert…that pops up whenever you open their chart. And it will say, “Needs colonoscopy.” And so the doctor, when they’re doing the refill or answering a message for something else, for, you know, a sinus infection, they can see that and have a reminder also at that time. “You’re also due for a colonoscopy. Maybe think about that….” (Manager)
Enhance the quality of interactions with patients
The third strategic goal aimed at coping with interdependencies that arise from insufficient interactions with patients was to enhance the quality of those interactions, primarily through the general practice of improving the efficiencies and effectiveness of PCP processes. All four ACOs had an individual or a team that worked with PCPs and their staffs to improve workflows and reassign tasks among staff so that clinicians could work “at the height of their license,” according to several informants. At three ACOs, this practice was primarily carried out through the work of establishing patient-centered medical homes. One manager, expressing reasoning we heard at two other ACOs as well, said that these efforts were intended to “increase efficiency within [PCP] offices, [which] I think will help them, too, because then they can do what they need to do. They can be the doctor. They can have that relationship.”
Our analyses also suggest that ACO leaders believed that patient choices were shaped in part by the fragmentations, inefficiencies, and confusions created by the complexities of health and illness, the health care system, and other social systems. In particular, attention was focused on the complexities and vulnerabilities that emerge in liminal spaces, such as between sites of care or between health care services and social systems. We identified two strategic goals and four boundary-spanning practices that the ACOs were using to reduce their dependencies on their environments by improving the abilities of both health systems and individual patients to function reliably and resiliently in these liminal spaces.
Improve coordination among system components
The first strategic goal for coping with interdependencies arising from system complexities was to improve coordination among system components. ACOs were pursuing this goal through three boundary-spanning practices. First, all four ACOs were managing transitions of care. Most typically, this involved standardizing processes and designating roles for overseeing transitions from hospitals and postacute settings (e.g., SNFs) to the home or between institutional settings (e.g., from hospital to SNF). All of the ACOs had one or more providers (e.g., discharge navigators) focused on managing care transitions by calling recently discharged patients to make sure they understood their discharge instructions, doing medication reconciliation and helping patients obtain necessary follow-up appointments with specialists and PCPs. For patients enrolled in care coordination programs—typically high-cost patients with complex care needs—care coordinators worked with discharge navigators to enact this practice. Managing transitions of care appeared primarily aimed at avoiding overuse: Informants asserted that the practice was crucial to ensure patients don’t “fall through the cracks” or “back into the system” and “have to be readmitted.” One executive said that postdischarge, some patients are
…confused. They’re sick. Some of them have good home support; some don’t. And we hold their hand to try to keep them from having to be unnecessarily readmitted because they didn’t change their medications, or they didn’t follow up with the ENT or the cardiologist…. (Executive)
In other words, health system factors (needing to follow up with multiple providers and manage multiple medications), patient characteristics (confusion), and social conditions (home support) were perceived to interact in complex ways, undermining recovery and fostering overuse.
Second, two of the ACOs attempted to combat leakage resulting from system complexities by engaging in the practice of carefully managing referrals. Both organizations worked to refer patients within their ACOs when possible; however, neither of these ACOs included all necessary services, as explained above. Informants reported that when referring patients to services not available within these ACOs, ACO providers only referred patients to non-ACO providers within preferred referral networks. Informants acknowledged, however, that because patients were free to seek care from any provider they wished, the success of managing referrals rested heavily on the PCP–patient relationship. A physician-executive summed up the crucial role of PCP relationships in boundary-spanning to manage referrals:
Most of our patients, over the years, have come to understand that we are their advocate both in the quality of care and the cost of care. So they know that if we recommend something, we’re going to choose the most efficient facility. If we send them to a specialist, it’s somebody that we would ourselves go to. (Executive)
In other words, the practice of managing referrals relies heavily on relational mechanisms. Furthermore, to stimulate relational mechanisms for managing referrals, both ACOs rewarded their providers financially for keeping care in the ACO or referring only to preferred non-ACO providers. The practice of managing referrals was not observed at the pediatric ACOs, each of which owned the sole pediatric hospital and all of the pediatric specialty services in its market. Leakage did not emerge as a significant concern for these ACOs.
Whereas the first two boundary-spanning practices aimed at compensating for system complexities were focused on standardizing health system structures and routines, the third boundary-spanning practice—used by all four ACOs—was focused on facilitating and problem solving the challenges incurred by individual patients. In particular, care coordinators engaged in this practice to remove barriers that compromised outcomes for some of the most complex, high-need patients. For example, one pediatric ACO manager told us about a young boy, on a feeding tube, who was experiencing recurring infections and repeated admissions.
And come to find out, once the care coordinator became involved, they recognized that…it was actually just a matter of having uniform supplies and making sure that [the family] were always using the same tools. Because when they would switch tools, it would take the family time to relearn how to use those tools. So once they were able to be connected with a consistent service provider, and always had the same equipment, the infection rate went down. And it was just a matter of somebody saying, “Okay, let's look at why this is happening and fix this.” (Manager)
Care managers facilitated and problem-solved by arranging transportation to appointments for patients who needed it and frequently attending appointments with patients to help them interpret information from providers. At the two pediatric ACOs, we were told stories of staff intervening in complicated home situations, including where drug-addicted parents were neglecting to administer appropriate medications to their children, in order to improve pediatric patient outcomes. More so than most other practices we observed, facilitating and problem solving entailed customized actions, adapted to specific situations, highlighting the fact that this practice appeared to be particularly important for handling contingencies that threatened specific, vulnerable, high-cost patient cases. The practice demonstrates that informants perceived that the complexities of patients’ lives, including particular manifestations of the social determinants of health (e.g., poverty), as well as the complexities of the health care system (e.g., finding consistent suppliers of feeding tubes) contributed to both over- and underuse and, therefore, were important factors influencing ACO dependence.
Improve patient ability to self-manage
The second strategic goal for coping with interdependencies arising from system complexities was to improve the abilities of patients to manage their own care and cope more effectively in the midst of complexities. All four ACOs were pursuing this strategy through the practice of educating and equipping patients. This practice included directly educating patients about effective care and the appropriate use of health services and disseminating formal educational materials and tools for managing chronic conditions. For example, informants at all four ACOs said staff were instructing patients who had been identified as overutilizing services about alternatives to high-cost services (e.g., urgent care rather than the ED). They were also educating patients who had self-referred about the benefits of staying in the ACO: “if you want your care coordinated, it’s better to keep it all within the [ACO], because we share our medical record. We’re able to do better handoffs” (Executive). An example of formal educational materials was one ACO’s online decision tool that instructed patients on the comparative benefits and risks of treatment options. In terms of tools, another ACO was piloting “care kits,” including one for congestive heart failure that provided, “the scale, the blood pressure pump, health management tools, education materials” (Manager). In addition, at least two ACOs were experimenting with motivational interviewing as a method to engage patients in their care.
Care Provided to ACO Patients by Non-ACO Providers
Our analyses further suggest that ACO leaders believed that, because ACO performance was impacted by the costs and quality of any care provided to ACO patients by non-ACO providers, care also shaped interdependencies. Although informants said they lacked access to data on and had no direct influence over non-ACO providers, there were important exceptions in the cases of the two ACOs that had formed arrangements with select non-ACO providers (described above). As a strategic goal, both ACOs attempted to improve their control over the actions of these non-ACO providers by intervening directly in care decisions as those decisions were being made. That is, these ACOs leveraged the connections they had built with preferred non-ACO providers to exert direct influence over those providers.
The ACOs pursued this strategic goal through a single boundary-spanning practice: They used their own staff for overseeing care provided by preferred non-ACO facilities (i.e., SNFs, hospitals). When ACO patients presented at the EDs or were admitted to the hospitals or SNFs of preferred non-ACO providers, key ACO staff (e.g., hospitalists, advanced practice nurses, “SNFists” specializing in SNF care) would engage in care decisions on behalf of ACO PCPs and help coordinate care. Critically, these key ACO staff members were often embedded within these non-ACO facilities, which allowed them to respond quickly to influence decisions.
We have our team that looks at our patients in the hospitals, and helps get them transferred to SNFs if they need to be…you know, “how many hospital days have they had in there? Are they ready to move out, or are they not ready?” (Manager)
In addition, these key ACO staff members also had access to the ACOs’ EHRs, “so they have the last…history, you know, the prescriptions that they’re on, the last diagnostics, what the primary care is thinking, the problem list” (Executive). Informants reported that this practice addressed overuse by reducing lengths of stay or avoiding unnecessary admissions and services.
Uncertainties Related to the Environment
Finally, our analyses suggest that ACO leaders believed that interdependencies related to patient choice were shaped in part by uncertainties about the environment, including about patient and non-ACO provider actions and about the effectiveness of ACO efforts to influence those actions. We identified two complementary strategic goals and three boundary-spanning practices that the ACOs were using to address uncertainties.
Improve ACO knowledge of environmental constraints
One strategic goal aimed to improve ACO knowledge of environmental constraints, the stable features or patterned actions of environmental actors. We found two boundary-spanning practices under this strategic goal. First, all four ACOs were monitoring population health and utilization patterns, which reportedly required extensive investments in information infrastructures, including EHRs, disease registries, data warehouses, and data analysts. The ACOs ran reports at regular intervals (e.g., daily, monthly, quarterly) to monitor underuse, overuse, and leakage patterns and enable other boundary-spanning practices. For example, monitoring utilization informed the practice of communicating and reminding (i.e., “outreach”).
Another source that we use is just data queries. So for example…asthma. We look at, you know, how many kids who have had more than three ED visits for an asthma related diagnosis in the past six months. And so we’ll generate a list of those patients and outreach to [them]. (Manager)
ACOs were also monitoring population health and utilization patterns to assess the effectiveness of their interventions, which informants said helped them decide where to focus their efforts.
Second, the two ACOs with preferred non-ACO provider arrangements were monitoring non-ACO provider performance. They analyzed data, particularly for evidence of overuse, to differentiate among non-ACO providers and inform the practices of managing referrals and developing and maintaining relationships with these providers.
There are certain hospitals in the…area that we know are—you know, you drop people in, and it's like putting them in a black hole. You can't get them out…and also the flip side of that is, you know, what are the most efficient facilities that we should be steering people to? (Executive)
This monitoring involved not only data analyses but also a “subjective gestalt about both the quality and cost efficiency” (Executive) of non-ACO providers, developed over many years.
Improve ACO awareness of environmental contingencies
The other strategic goal pertaining to environmental uncertainties aimed to improve ACO awareness of environmental contingencies. The dynamic, emergent nature of contingencies means that they are largely unpredictable; however, we found that ACOs were attempting to improve their abilities to identify these contingencies early enough to intervene by engaging in the boundary-spanning practice of monitoring individual high-risk patients (those with the highest costs, highest utilization, or most complex care needs). This monitoring was often accomplished by care coordinators who would follow panels of high-risk patients by checking up on them during regular phone calls or in-person meetings and by reading provider documentation in the EHR. A manager at one ACO gave example questions that care coordinators might try to answer—initially and ongoing—regarding severe asthma patients: “Are they maintaining their prescription medication? Are they avoiding triggers in the home?”
We found additional instances of monitoring individual high-risk patients at the two ACOs that embedded their own providers in non-ACO facilities (e.g., hospitals, SNFs). The ACO providers stationed in those facilities would monitor the conditions of and care provided to ACO patients during admissions at those facilities. In these instances, the monitoring directly informed their own practices related to overseeing care provided by non-ACO providers (described above). But the practice of monitoring individual high-risk patients involved more than these embedded providers. For example, at one ACO this monitoring involved a nurse coordinator, with access to local hospital information systems from which she would develop a census of any hospitalized ACO patients. She would then notify PCPs and embedded providers.
As the classic work of open-system theorists (e.g., Fennell & Alexander, 1987; Pfeffer & Salancik, 2003; Thompson, 1967) would predict, we found that ACOs were engaging in boundary-spanning in order to understand, cope with, or alter interdependencies related to patient choice. Our findings not only confirm these earlier conceptualizations but also elaborate and reframe them in ways that are more useful to researchers attempting to study ACOs and administrators attempting to design and manage them. Specifically, our findings provide a process theory of ACO boundary-spanning. Our theory suggests that each individual boundary-spanning practice contributes to a broader strategic goal, through which it may impact a particular aspect of interdependence and thereby reduce underuse, overuse, or leakage.
Many of the boundary-spanning practices we found relate to evidence-based care. For example, the boundary-spanning practices of communicating and reminding and educating and equipping represent attempts to stimulate involvement of patients in their own care, a central aim of patient activation and engagement (PAE) programs (Greene et al., 2015; Hibbard & Greene, 2013). Although ACOs are only beginning to embrace PAE programs and leaders have mixed perceptions of the usefulness of such programs (Shortell, Sehgal, et al., 2015), we found evidence of PAE efforts at all four ACOs in our diverse sample. Related, some boundary-spanning practices (e.g., facilitating and problem solving) were enacted as part of care coordination programs for patients with complex illnesses. Evidence suggests that these programs are important for reducing overuse and improving health outcomes (McCarthy, Ryan, & Klein, 2015). Similarly, some boundary-spanning practices (e.g., improving the efficiencies and effectiveness of PCP processes) were instantiated as part of patient-centered medical homes, important mechanisms for addressing both under- and overuse (Hoff, Weller, & DePuccio, 2012). Although previous studies have individually identified these mechanisms, we find their simultaneous use in the context of ACOs coping with environmental interdependencies. This finding demonstrates both that ACO models may indeed be encouraging the uptake of evidence-based care and that the uptake of evidence-based care may improve organizational performance by helping ACOs cope with interdependence.
The number and variety of boundary-spanning practices that we identified highlight the breadth and complexity of the challenges ACOs face and a considerable range of ways these interdependencies can be conceptualized and addressed. In fact, although the unique interdependencies that ACOs face arise from patients’ freedom of choice, we did not find all boundary-spanning efforts aimed directly at patients. Most significantly, two ACOs engaged in multiple boundary-spanning practices aimed directly at monitoring and influencing non-ACO providers. Few ACOs include all services across the care continuum (Shortell et al., 2014). Thus, developing arrangements to ensure patients have access to missing services may be crucial for ACO performance (Shortell, Colla, et al., 2015). The arrangements developed by the two ACOs in our study may seem unusual; however, the fact that two significantly different organizations—one a small primary care group practice and the other a large integrated health system—both relied on these complicated arrangements suggests that such arrangements may be practical for a wide range of other ACOs. However, if an ACO’s network of independent community-based physician practices is potentially vulnerable to market dynamics, such as other health systems purchasing those practices (Mick & Shay, 2016), then these complicated arrangements with non-ACO providers might be equally, if not more, vulnerable to such dynamics. Furthermore, the contracted providers may eventually demand a share of the savings or other rewards, suggesting that these arrangements may evolve. These possibilities suggest that further research is warranted.
In proposing a theory of ACO boundary-spanning, our study contributes to efforts to evaluate ACO models (Fisher, Shortell, Kreindler, Van Citters, & Larson, 2012). Specifically, our theory identifies five categories of factors—that is, important aspects of—interdependencies related to patient choice. If indeed these factors shape interdependencies and thus ACO performance, then efforts to categorize and evaluate ACOs (Shortell et al., 2014) may need to be expanded to capture dimensions related to these factors. For instance, measures of the availability of services or of the frequency, efficiency, and quality of interactions with patients might be useful for distinguishing among ACOs and predicting performance.
Our theory also proposes a number of relationships that subsequent research should test (Table 2). For example, throughout we point to general problems (i.e., underuse, overuse, leakage) that each boundary-spanning practice was intended to address. Subsequent research can test these relationships. This would include examining how variations in practices are more or less effective for reducing these problems, because likely there are many possible variations of the practices we identified. Similarly, our theory proposes that each boundary-spanning practice was aimed at improving ACO performance in light of some specific aspect of interdependency (e.g., system complexities). Although our theory is grounded in ACO leaders’ beliefs and perceptions and has face validity, further research is needed to know whether or not these practices—or specific variations of them—actually alter interdependencies or otherwise enable ACOs to cope more effectively. The contribution of our study is to elicit these beliefs and perceptions and articulate them so that they may be tested.
Our study has a few limitations. First, we cannot evaluate the effectiveness of any boundary-spanning practice or the extent to which it may be used. These are important questions for subsequent research for which our theory provides a useful conceptual framework. Second, we do not report patient perspectives on boundary-spanning practices, which might yield different insights. This, too, is an important opportunity for further research. Our study of patients’ awareness of these ACOs more generally is forthcoming (McAlearney, Song, & Hilligoss, 2016). Third, although some physicians were included as informants and a number of managers and executives that we interviewed were also physicians and nurses, we do not claim to have broad representation of clinician views. Our focus was not on clinical practice, but rather on strategic perspectives related to the environmental interdependencies of ACOs.
Despite limitations, our theory provides managers with a framework for developing and refining their own strategies for managing interdependencies related to patient choice. For example, in identifying five categories of factors that shape interdependencies, our findings may be useful for evaluating how well one’s organization is prepared to address each category or for setting goals aimed at improving the organization’s position relative to that category. Similarly, our theory identifies 15 boundary-spanning practices managers might implement, adapting them as necessary to local conditions or to leverage existing mechanisms to achieve those goals.
The authors are extremely grateful to the organizations and informants who participated in this study and to the members of our Project Advisory Team. We also thank our research team members who assisted at various stages of this project: Meredith Rosenthal, Julie Robbins, Jennifer Hefner, Daniel Gaines, Lindsey Sova, Kelsey Murray, Pamela Thompson, Pamela Beavers, Alexandra Moss, Megan Sinclair, Kendall Haas, Mary Frances Gardner, and Jessica Stewart. The Robert Wood Johnson Foundation funded this research; however, the study sponsor had no involvement in the collection, analysis, or interpretation of data; in the writing of this manuscript; or in the decision to submit the manuscript for publication.