A key assumption guiding the community-based delivery of health and human services has been the need to integrate these services, typically through cooperation among provider agencies. Nonprofit agencies, especially those providing care to so-called "vulnerable populations"1 such as the poor, the elderly, the homeless, or individuals with serious mental illness, have traditionally tried to work together to integrate their diverse services. Such services include income assistance, rehabilitation, physical health care, mental health treatment, recreation, housing, counseling, substance abuse, legal, and more. The belief has been that clients benefit from integration of services by providing a complete continuum of care while agencies are able to use their limited resources more efficiently.2-4
Unfortunately, research on the subject has been equivocal regarding the link to client outcomes. For instance, a 20-year retrospective review of findings on services integration conducted by the U.S. Department of Health and Human Services (1991) concluded that there is "insufficient knowledge" about whether or not service integration is an effective system strategy. More recently, research by several scholars has indicated that services integration has had little impact on services and outcomes.5-7 Other researchers have found a positive link between integration and outcomes, but with the proviso that the relationship is conditional on such factors as stability and funding.8,9 Despite mixed research findings, interorganizational collaboration in the delivery of health and human services to vulnerable populations has become widely accepted in practice, and is often seen as a system-level indicator of how well services are being provided.
In recent years, the introduction of managed care to community-based health and human service delivery systems has introduced additional uncertainty about how services might best be provided across agencies.10 In particular, it is not clear whether agencies under managed care will be willing to cooperate with one another or to compete in the face of increased pressures to control costs. While there has been some recent research on fully capitated, at-risk, publicly funded systems for services to vulnerable populations like the homeless and individuals with serious mental illness,11,12 there have been almost no studies that have examined the impact of managed care on services integration and cooperation among nonprofit provider organizations (see Johnsen et al. for an exception).6
In community mental health, the general lack of research on fully capitated delivery systems has mainly been due to the relative absence of such systems nationwide until very recently. A key concern among public funders has been that mental health services, especially for the severely mentally ill, could not be financed using a fully capitated, risk-based format because of the large and unpredictable costs involved if adequate levels of service are to be maintained.10 For this reason, those states and cities that have now moved to some form of managed care and risk-based capitation for individuals with serious mental illness have mostly done so only with their Medicaid populations, as a way of testing the efficacy of such an approach. However, the movement to full risk managed care systems for all state-funded clients with serious mental illness is a trend that seems likely. In view of this trend, it is extremely important to develop a better understanding of how managed care might affect the extent to which nonprofit agencies are willing and able to work together effectively to provide needed patient services.
One factor that seems critical for explaining the impact of managed care on interorganizational collaboration is the role of the organization that manages and monitors the system. Despite the focus in recent years on managed care as a cause of major change in the health care system, it is not at all clear that it is managed care per se that causes these changes, especially for those changes that are not specifically financial in orientation, like collaboration. Rather, it may be the implementation of managed care, and the way in which that implementation is organized and managed, that determines whether or not managed care as a system of financing has a positive or a negative impact on the extent to which providers work together or compete.
Essentially, this can be viewed as a principal-agent problem.13,14 Although individual agencies may collaborate with others based on resource dependencies15 and pressures for legitimacy,16 agencies are often embedded in a larger system of care that is funded, managed, and monitored by a single entity. This entity is often the agent of the state, which provides Medicaid and other state funds. But it is also the principal for service providers at the community level—the local agents. Thus, the managerial dilemma for the principal is how to provide incentives and sanctions for local agents to control costs, while also encouraging them to provide high quality services. One way of doing this is to encourage agencies to work together, enabling them to utilize resources more efficiently, while still offering a broad range of coordinated services. The question is, can this be done when the principal operates under risk-based managed care, where cost cutting often produces an environment that encourages competition rather than collaboration?
Consistent with the discussion above, the research we report here will address three critical issues. First, are the pressures of financial risk under managed care beneficial or detrimental to cooperation among mostly nonprofit service providers? Second, to what extent is the provision of services affected by managed care? And third, what is the role of the core managed care organization for implementing and managing the system in a way that either facilitates or undermines collaboration among providers and the provision of services? We examine these issues empirically using longitudinal data collected from a single system of health and human service delivery funded through a managed care model.
THE RESEARCH SETTING
In late 1995, the behavioral health system in Tucson/Pima County, Arizona changed from fee-for-service, where governance and accountability was only loosely tied to cost control, to a fully capitated model. All state and Medicaid mental health funds were to be locally managed and controlled by a newly established nonprofit Regional Behavioral Health Authority (RBHA) called Community Partnership of Southern Arizona (CPSA). CPSA would be responsible for the needs of Pima County and southern Arizona in the areas of serious mental illness (adults), general mental health, substance abuse, and children's behavioral health. Client services for adults with serious mental illness (SMI) was, and continues to be, the largest of the program areas, at more than 40 percent of the total CPSA budget.
SMI services in Pima County would be delivered by three primary, and one smaller, provider networks, each headed by a nonprofit mental health agency, referred to as an At Risk Provider, or ARP, which would be fully at-risk financially. The three major ARPs are Carondelet Behavioral Health Services, COPE Behavioral Services, and La Frontera Center, each of which were randomly allocated from 1,000 to 1,200 clients. The fourth ARP, CODAC Behavioral Health Services, was allocated 250 clients. In fiscal year 1996, all four ARPs were paid the same monthly capitation rate of $378 per month for AHCCCS (the state's Medicaid plan) clients and $326 for clients not eligible for AHCCCS. By fiscal year 1999, the capitation rate had increased to $501 for AHCCCS and $500 for non-AHCCCS clients. Approximately 48 percent of the CPSA system's adult SMI clients were enrolled in AHCCCS in 1999. The total allocation to each ARP could potentially be adjusted on a yearly basis if the number of SMI clients actually served by that ARP were to increase or decrease. In practice, new clients to the system have typically been absorbed by the ARPs with no additional funding.
A key aspect of the new plan was that each ARP was required to set up its own network of collaborating agencies, with which they could contract either on a fee-for-service, block purchase of service, or risk-sharing basis. These network agencies offered services that the ARP either did not provide or that the ARP provided only in limited quantities, such as inpatient beds and housing. A critical concern expressed by many in the system during its early years was that the size of the county's adult SMI population, about 3,500 enrolled individuals, was too small to sustain four separate networks of providers. Many expected that there would be significant competition across the networks, resulting in an eventual shake-out of the system, leaving only one or two large ARPs that could deliver services more efficiently. As of 2000, no consolidation of ARP networks had occurred, and the basic structure of the system that was established at founding remained in place. The formal funding relationship between the state, CPSA, the ARPs, and the other provider agencies is outlined in Figure 1.
Data on system/network integration were collected at two points in time. The first wave of data were collected in early 1996, shortly after the new system was formed, while the second wave began in November, 1999 and was completed by March, 2000. Network data were collected from each of the 42 providers in 1996 and from 44 providers in 1999/2000 using structured surveys. In both cases, the network of potential SMI providers was identified first through discussions with CPSA staff and then by asking the executive directors of the ARPs to review the list to add and/or delete agencies. The list was then refined and finalized, becoming the population of agencies we would study.
The primary focus of our data collection efforts was to obtain information about integration among the agencies that comprised the broad network of service delivery. Specifically, we were interested in determining the extent to which each agency was involved with every other agency in the system regarding formal funding/contractual ties and more informal client referral ties, both sent and received. For both waves of data, respondents were asked to indicate their links "over the past 6 months or so." In addition, for the 1999-2000 study, we also obtained data on the extent to which agencies were linked through shared information. For both studies, we used what has been defined previously as a "service implementation network" approach,8,17 where focus is solely on the network of relationships concerning provision of services to the adult SMI population. Thus, we were not interested in ties among agencies that might have nothing to do with our target population.
For the 1996 data collection, we obtained data only on network integration. For the 1999-2000 effort, our questionnaire was more extensive, allowing us to obtain data on additional issues such as relationship quality and attitudes concerning how the system has progressed since inception. We also conducted in-depth interviews (up to 1 ½ hours) with agency directors and/or their key staff involved with serving SMI adults. These interviews allowed us to develop a strong understanding of how the system had evolved, including its main strengths and shortcomings.
For both studies, we used revised versions of methods and measures used previously by Provan and Milward.8 The questionnaires were sent out with a cover letter explaining the project and intensive followup procedures were used to ensure a high response rate. In fact, we obtained response rates of 93.3 percent in 1996 (42 of 45 agencies surveyed) and 90.9 percent in 1999-2000 (40 of 44 agencies surveyed). High response rates are critical in network research since missing data can have a significant effect on the validity of results. For the four nonrespondent agencies in the 1999-2000 wave, we relied on the links claimed by their partner agencies and confirmed these with the actual links reported by these four agencies in 1996, enabling us to include all 44 agencies. By the end of 1999, one of the agencies surveyed in 1996 was no longer in business and two no longer served the target client group. Thus, 39 of the 42 agencies surveyed in 1996, or 92.9 percent, were identical matches with agencies in the 1999-2000 network. The 1999 network also included two agencies that were completely new to the system and three others that had been nonrespondents in 1996.
To enhance the reliability of the network findings, we relied only on confirmed data.18 That is, contractual and shared information (1999 only) links were coded as a '1' only if both an agency and its partner reported that that particular link existed between them. A '0' was recorded if no link of that type was reported by both agencies or if only one agency, but not the other, reported the link (i.e., not confirmed). A referral link was only established if one agency indicated that referrals were sent to a partner agency and the partner agency also indicated that referrals were received from the first. A one-way referral link (either confirmed sent or confirmed received) was coded as a '1' and a two-way referral link (both sent and received) was coded as a '2.' Network data were analyzed using UCINET software.19
We also obtained data directly from CPSA. CPSA was established not only to act as a conduit for state mental health funds, but also as a mechanism for monitoring and evaluating the clinical and managerial activities of service providers. To do this, CPSA collected data from providers since inception of the system on service "encounters." These data are based on the value of the SMI service encounters actually provided by each ARP (and thus, for the system as a whole) in such areas as inpatient/hospitalization, rehabilitation and housing, medication, and case management. The value of these services is determined by the state's Department of Health Services. While encounter data do not necessarily reflect the actual amount paid to the ARPs in providing the services, they are a good indicator of the relative costs incurred. Since FY 1996 (adjusted to reflect a 12-month year), overall SMI encounter costs have steadily risen from $18.67 million, to $20.25 million, to $22.87 million, to $27.78 million in FY 1999.
No data were available on actual client-reported satisfaction with services, although in our 1999-2000 questionnaire we asked respondents to evaluate the system in terms of its impact on clients and client services. In general, our findings allow us to compare how the system has evolved under managed care, focusing on agency cooperation/service integration and on any change in relative service emphasis.
The findings from the first wave of data collection indicated that despite the fact that the new managed care system had been set up so that competition among ARP provider networks seemed likely, and even desirable from the perspective of CPSA as a way of controlling costs, there was evidence of cross-network cooperation. All of the networks shared ties with many of the more focused service providers (vocational rehabilitation, recreation, housing, etc.) and the ARPs themselves often maintained contracts and referral ties with each other. These findings, and a discussion of the birth of the new system, have been reported elsewhere.20
Despite this apparent cooperation, the early data represented only one point in time, and that was a point at the very beginning of the evolution of the new system. Thus, it was not clear if the level of integration in 1996 was an anomaly, reflecting early efforts by CPSA to encourage cooperation to get the system going by building on past cooperative relationships. Once well established and operational, pressures to control costs might then be applied with greater force, minimizing integration across the full network of providers. This possible outcome is supported by the work of Johnsen et al.,7 who found that managed care actually reduced the level of networkwide integration among agencies serving the homeless mentally ill. Their conclusion was that managed care often results in tighter, more formalized networks of designated providers, rather than the more open networks that might evolve informally in the absence of managed care contracts.
Our first research question is an attempt to address this issue directly. That is, does the introduction of managed care, and the maintenance of a risk-based funding mechanism, result in increased or decreased levels of cooperation through services integration among provider agencies? We examined this question in two ways. First, how has integration changed across the network as a whole? Second, how has the clique structure of the network changed under managed care?
Full Network Analysis
Table 1 reports findings for the full networks (1996 and 1999) of SMI provider agencies, comparing the funding contract networks, the referral networks, and both types combined. The network of relationships through shared information, collected in 1999 only, is also reported. For each type of network, three types of measures of network integration are reported—density, degree centrality, and betweenness centrality.21 Density refers to the extent to which all agencies in the network are directly connected to each other, expressed as a proportion of the total possible number of links. Thus, a density of .20 would mean that for a particular type of link (contracts, referrals, etc.), 20 percent of the total possible number of relationships (n(n - 1)) were actually in place. Centrality scores reflect the extent to which an agency is a central player in the network, either because it maintains more direct links of a particular type than others (degree centrality) or is central in the network-wide flow of relationships of that particular type, including indirect links (betweenness centrality). Higher scores reflect greater centrality.
The differences in density between 1996 and 1999 are striking and run completely counter to the findings of Johnsen et al.7 Specifically, across both types of linkages, contracts and referrals, the network became substantially more integrated. Integration through contracts increased from a network density score of .049 to .062 (+126.5 percent), while integration through referrals increased dramatically, from .086 to .133 (+154.7 percent). When combined ("all links"), density increased from .125 to .160 (+128 percent).
These findings reflect substantial levels of overall network integration, considering both the absolute level of network density in 1999, and especially the increase in density from 1996 to 1999. The increase in referral density is especially striking, with referral ties growing at a rate that was twice as fast as those based on contracts. Thus, under the CPSA-implemented system of managed care, provider agencies were apparently more willing than before to go beyond their contractual ties to develop relationships that were based more on trust and the needs of clients. This conclusion is reinforced by the 1999 survey finding for shared information. The density score of .257 for this measure of integration is the strongest of any of the results, and indicates that agencies talk to one another about issues related to SMI services to an extent that goes well beyond any contractual or referral relationships they may have.
In general, the findings on network density provide strong evidence that the introduction of managed care, as a mechanism for funding and cost control, need not result in a breakdown of the system of service provision. In fact, depending on how it is implemented and managed, managed care may actually result in enhanced levels of integration across the network of providers. The findings were strongest for those ties that were informal; namely, referrals and shared information, which could indicate a significant buildup of social capital22 that might be drawn on to help resolve any potential problems that might emerge as the system continues to evolve.
The centrality scores are somewhat more difficult to interpret, but they too suggest a pattern of increased integration across the full network as the system evolved under managed care. Basically, the higher centrality scores for 1999, versus 1996, indicate not only that there are more links (as with density), but that more agencies are either directly (degree centrality) or indirectly (betweenness centrality) connected to multiple other agencies, providing a richer pattern of integration and with a broader range of potential service options for clients. As with the density results, the increases in network centrality were strongest for the more informal referral links. The findings for shared information (collected only in 1999) were also quite strong. The mean degree centrality score of 11.045 indicates that on average, agencies throughout the SMI network directly share information on SMI clients and services with more than 11 other agencies—slightly more than one fourth of all the other agencies in the network.
Table 2 reports the findings for the clique analysis. Basically, a clique is a "maximally connected" subgroup within a larger network. That is, every organization in a clique must be connected to every other organization. Unlike the analysis reported in Table 1, which reported links between pairs, or dyads, of agencies, cliques focus on relations among groups of three or more organizations. Networks having many cliques are thus more intensively connected than networks only having many dyadic links. This is especially true when cliques overlap. As discussed by Provan and Sebastian,9 overlapping cliques occur when the same group of three or more agencies is fully connected in multiple ways—for instance, through both referrals and contracts.
As reported in Table 2, the number of cliques of three or more agencies that contracted with one another grew from 12 in 1996 to 16 in 1999, an increase of 33 percent. For referral cliques, the growth was even more impressive, from 47 in 1996 to 70 in 1999, an increase of 48 percent. The number of cliques can readily exceed the number of agencies in the system since any one agency can be a member of multiple cliques. For instance, agency A might be in a clique with agencies B and C, while also being in a clique with agencies D, E, and F. These are separate cliques because agencies B or C may have no ties with agencies D, E, or F.
Although cliques reflect a more densely connected network than one consisting mostly of simple dyadic/pairwise ties, cliques can also indicate a fragmented network if cliques of one type (i.e., referrals) do not overlap with cliques of other types (i.e., contracts). Thus, an analysis of overlapping cliques is helpful. Consistent with prior work by Provan and Sebastian,9 we examined the extent to which agencies involved in cliques of contracting agencies were identical with cliques of agencies referring clients to one another. In 1996, there were no agencies involved in overlapping cliques. By 1999, however, 12 of the 44 agencies in the network, including each of the four ARPs, were in at least one overlapping clique. In addition, three of the ARPs were in overlapping cliques with 10 of the agencies while the fourth, La Frontera Center, was in overlapping cliques with all of the other 11 agencies (links to itself not counted).
Overall, the network moved from a system with a modest and highly fragmented clique structure when managed care was first introduced, to one where many more cliques were in evidence and where 27 percent of the agencies were involved in cliques of three or more agencies that shared both contracts and referrals. Thus, the clique analysis adds further support to the findings from the overall network results that under the managed care model implemented by CPSA, significantly greater coordination and collaboration was achieved in the delivery of client services. Unlike Granovetter's23 premise that weak ties are beneficial, our findings clearly indicate that agencies providing mental health services strive to strengthen their relationships with other agencies, and that managed care need not be detrimental to this process.
Services and Performance
The study did not directly address or measure client outcomes. However, we did generate some general findings on the quality of services provided by the system and how spending patterns might have shifted. In our 1999-2000 survey, we asked all respondents to evaluate the impact of managed care under the CPSA system on their agency's adult SMI services. Respondents indicated their responses on a five point Likert-type scale, ranging from 1 ("at-risk system has had a significant negative impact") to 5 ("at-risk system has had a significant positive impact"). A response of 3 indicated that "the impact has been pretty much neutral." Respondents were asked to apply these ratings to each of 6 aspects of client services and operations (mean response in parentheses): number of SMI services offered (2.6), number of SMI clients served (2.81), quality of SMI services (3.03), controlling cost of SMI services (3.09), coordinating SMI care with other agencies (2.94), and stabilizing the system of SMI care (3.14). When combined into a single scale, the mean for all six items was 2.94 (Cronbach's alpha = .87), which indicates a slightly negative, but generally neutral overall view of how the system of care for SMI clients has evolved under managed care. At the same time, these same respondents gave CPSA reasonably high marks for "the overall management and governance of the adult SMI delivery system" since becoming the RBHA in late 1995, with a mean score of 3.63 (scale ranged from 1 = "system management is much worse than before," to 5 = "system management is much better than before").
Respondents were also asked to rate every agency with which they had a link, based on the "overall quality" of the relationship. A scale ranging from 1 (poor relationship) to 4 (excellent relationship) was used. Not surprisingly, scores were uniformly high, since it is unlikely that agencies would maintain a relationship over time with another if the quality of that relationship were poor, unless of course, they had no choice. Nonetheless, differences in the quality of agency relationships with the four ARPs and CPSA were evident. The median score for all 44 agencies in the system was 3.14. Overall relationship quality scores were somewhat below the median for COPE (3.0), and nearly at the median for CODAC (3.13). Scores were above the median for La Frontera (3.29) and Carondelet (3.38). Consistent with the generally favorable attitude agency respondents had toward the job CPSA had done, agencies also indicated that their relations with CPSA were good, with a mean score of 3.52, the second highest score for any of the 44 agencies in the system.
Finally, we examined CPSA encounter data, which reports the value of all SMI service encounters in each of several major service categories. These data are reported in Table 3 and represent the assigned value (determined by the state) of the services actually provided to clients (i.e., service encounters). The encounter data provide a reasonably accurate picture of how each ARP and the system as a whole is allocating its resources for SMI service provision and even how much is actually being spent, even though the ARPs do not necessarily get paid enough through their capitation rate to cover these expenses. The figures are especially useful in demonstrating shifts in service emphasis from 1996 to 1999.
The encounter data reveal that significant cost shifting across the system occurred as a result of managed care. Specifically, of the four largest cost categories (accounting for 91 percent of total SMI costs in 1999), inpatient/hospitalization costs declined from 5.4 percent in 1996 to 3.3 percent of total costs in 1999, while medication costs jumped from 16.8 percent to 23.7 percent. Case management costs declined from 37.4 percent in 1996 to 32.2 percent in 1999 while rehabilitation (including residential) increased from 27.4 percent to 31.8 percent of total encounter costs. Overall, what appears to be happening is that the cost control emphasis of managed care has led CPSA and its provider agencies to reallocate spending so that clients now spend less time in expensive inpatient settings, with less active case management once they are released. The focus has turned instead to greater use of medication with a focus on rehabilitation and outpatient residential care. Whether this shift in focus is appropriate for all adult SMI clients is not clear, although in our in-depth interviews of agency heads and clinical directors, many suggested it was not.
Further insight into the impact of managed care on client services was obtained by examining data from an ongoing study conducted by CPSA on the different costs of treatment for different categories of clients.24 Specifically, longitudinal data were collected from a small sample of SMI clients (the same clients tracked across four time periods) in three cost categories. High-cost clients (n = 27) were those whose yearly treatment costs in the year immediately preceding the shift to managed care under CPSA (1994-1995) ranged from $30,000 to over $100,000. Low-cost clients (n = 13) were those whose pre-CPSA costs ranged from $1,200 to $3,000. Middle cost clients (n = 13) ranged from $4,000 to $20,000 per year in treatment costs. The treatment costs of each group were tracked across four time periods. A summary of results is reported in Table 4.
Although this study was only based on a small sample of the total adult SMI population served, the results suggest that there has been a definite shift in resources under managed care. The encounter data reported in Table 3 indicated a shift in resources across service categories toward less resource-intensive treatment modes (i.e., medication versus inpatient treatment). The results in Table 4 indicate that this resource shift occurred across client groups as well. The most severely ill clients under the old fee-for-service system received services that were more than twice as resource intensive as the services these same clients received in the 1998-1999 period. In contrast, the low-cost utilizers increased their costs significantly compared to the pre-CPSA system, nearly doubling by 1997-1998, before falling back to an increase of slightly greater than 50 percent in 1998-1999. The mid-cost utilizers first experienced a big drop but their costs quickly returned to pre-CPSA levels.
Using 1999 data for the system as a whole, rather than focusing only on samples of clients, the findings on cost shifting were confirmed. Specifically, under the pre-CPSA fee-for-service system, the highest cost clients (top 10 percent) utilized slightly more than 70 percent of total adult SMI resources. Under CPSA's managed care model, the highest costing 10 percent of clients utilized only approximately 45 percent of total resources.
Taken collectively, the findings on service costs (Tables 3 and 4) provide strong evidence that risk-based managed care has had a major impact on how services are provided. While some clients may get fewer services than before, far more clients whose needs are not as severe can be provided enhanced levels of service under the new system. This reflects a conscious shift in service priorities that can have both clinical and political implications. One major service delivery implication of such a shift is the need for inter-agency collaboration. When not hospitalized, SMI patients are likely to need a broad range of services that are more likely to actually be offered when there is more money available through cost shifting. Thus, it is not surprising that the system of service delivery in Pima County has become substantially more integrated at the same time that pressures to control costs have resulted in shifts in spending toward patients that require more extensive, rather than intensive services.
Overall, the findings reported here provide strong evidence that the nonprofit managed care system in place in Pima County, Arizona had a major impact on how services were delivered to adult SMI clients. As the system evolved, it became considerably more integrated, with greater levels of cooperation and coordination across providers. At the same time, there were significant shifts in how clients were served, both in terms of allocation of resources across types of services (more medication and outpatient rehabilitation) and across clients, with lower-cost utilizers receiving more services than before, and the most severely ill, fewer. While the shift in resources could be viewed negatively, especially by the advocates of clients with the most severe problems, the subjective assessment of overall service quality we obtained indicated that while service quality did not necessarily improve, it was maintained at an acceptable level.
More generally, this study has been an attempt to shed further light on the impact of managed care on the structure and outcomes of health and human service delivery systems. Nationally, there has been endless debate about the impact of managed care on health care, but little of this discussion has focused on community-based, nonprofit settings. There has also been almost no attention paid to how the collaboration and cooperation that has traditionally characterized the activities of nonprofit health and human service agencies might be affected by the introduction of managed care.
Managed care has generally been viewed as a boon to cost saving but detrimental for client services. However, this simplistic view is starting to be challenged. In particular, there is evidence that nonprofit systems of managed care, when compared with for-profit HMOs, can both save money and deliver high quality services.25 Our research provides preliminary evidence that nonprofit agencies can learn to operate reasonably well within a risk-based managed care system, developing increased levels of cooperation and system integration. Although we did find shifts in how services and service spending were allocated, our findings on performance did not indicate any significant detrimental effects on the adult SMI client population as a whole.
A key factor for explaining our findings is the role of CPSA in managing the system. As noted earlier, CPSA can be viewed as the principal in a principal-agent relationship with its service providers. Unlike a traditional principal-agent relationship, however, in which the principal funds and monitors while its agents merely deliver services, CPSA and its primary service providers, the ARPs, were closely aligned. In fact, three of the four ARPs were equity partners in CPSA, having provided the financial stake to allow CPSA to bid for the original contract with the state (a $10 million performance bond was required). While the agencies in our study were not necessarily in favor of managed care, once it was imposed on the RBHA by the state, many of our respondents told us that they were committed to ensuring that the CPSA model worked. As its name implies, CPSA was indeed a "community partnership," and agency heads were willing to work together to balance costs and services, so that the state would not be tempted to select a for-profit firm in the next round of bidding for the RBHA contract.
What this study demonstrates is that the introduction of managed care to a system of nonprofit providers can be successful, but that system change is likely. Some of these changes, like increased system integration, are very positive and desirable, especially from the perspective of those clients, like individuals with serious mental illness, whose services need coordinating. Other changes, like shifts in funding priorities, may be viewed positively by some, but negatively by others. These changes are likely to be accepted by provider agencies when the funding, monitoring, and management body; in this case, CPSA, works closely with providers to ensure that the changes brought on by managed care are understood and accepted. If managed care is simply imposed on providers by a management entity that acts not as a partner, but as a traditional agent of the state, then costs may be contained but enhanced collaboration among provider agencies is not likely to occur. As a result, client services are likely to suffer.
We believe that this close working relationship is most likely under a nonprofit managed care model, where the goals of both principal (CPSA) and agents (the provider agencies) are closely aligned. If the system had been established and run by a state-imposed agency or an out-of-town for-profit organization, the results might have been quite different. In particular, it is reasonable to expect that a for-profit managed care organization might not shift cost savings from one client group to another, but instead, divert these savings to higher salaries and profits. In addition, a for-profit managed care organization might also encourage cost-cutting competition among key provider agencies, creating a climate that would not be conducive to enhanced levels of collaboration and integration. Further research on alternative models of risk-based systems would certainly be helpful in determining the generalizability of the results reported here.
Despite our focus on the evolution of a single mental health system, the conclusions generated from this study should be useful more broadly, enabling the management of both service providers and funding agents in other cities and in other health and human service settings to make better decisions about how their systems might be structured and managed. In general, our findings should enable nonprofit managers to understand better how collaboration might be affected under conditions of managed care and financial risk, how such systems evolve, and how these systems can be managed to minimize disruption to the agencies that comprise the system and their clients.
1. Aday, L.A. At Risk in America.
San Francisco: Jossey-Bass, 1993.
2. Dill, A.E.P., and Rochefort, D.A. "Coordination, Continuity, and Centralized Control: A Policy Perspective on Service Strategies For the Chronic Mentally Ill." Journal of Social Issues
45, no. 3 (1989): 145-59.
3. Alter, C., and Hage, J. Organizations Working Together.
Newbury Park, CA: Sage, 1993.
4. Provan, K.G., and Milward, H.B. "Do Networks Really Work? A Framework For Evaluating Public Sector Organizational Networks." Public Administration Review
61 (2001): 414-23.
5. Bickman, L. "Implications of a Children's Mental Health Managed Care
Demonstration Evaluation." Journal of Mental Health Administration
23 (1996): 107-17.
6. Lehman, A.F., Postrado, L.T., Roth, D., McNary, S.W., and Goldman, H.H. "Continuity of Care and Client Outcomes in The Robert Wood Johnson Foundation Program on Chronic Mental Illness." Milbank Memorial Quarterly
72 (1994): 105-22.
7. Johnsen, M.C., Morrissey, J.P., Landow, W.J., Starrett, B.E., Calloway, M.O., and Ullman, M. "The Impact of Managed Care
On Service Systems For Persons Who Are Homeless and Mentally Ill." In Research in Community Mental Health, Volume 10
, edited by J.P. Morrissey. Greenwich, CT: JAI Press, 1998, pp. 115-37.
8. Provan, K.G., and Milward, H.B. "A Preliminary Theory of Interorganizational Network Effectiveness: A Comparative Study of Four Community Mental Health
Systems." Administrative Science Quarterly
40 (1995): 1-33.
9. Provan, K.G., and Sebastian, J.G. "Networks within Networks: Service Link Overlap, Organizational Cliques, and Network Effectiveness." Academy of Management Journal
41 (1998): 453-63.
10. Arons, B.S., Frank, R.G., Goldman, H.H., McGuire, T.G., and Stephens, S. 1994. "Mental Health
and Substance Abuse Coverage under Health Reform." Health Affairs
13 (1994): 192-205.
11. Callahan, J.J., Shepard, D.S., Beinecke, R.H., Larson, M.J., and Cavanaugh, D. "Mental Health
/Substance Abuse Treatment in Managed Care
: The Massachusetts Medicaid Experience. Health Affairs
14, no. 3 (1995): 173-84.
12. Fried, B.J., Topping, S., Morrissey, J.P. et al. "Comparing Provider Perceptions of Access and Utilization Management in Full-Risk and No-Risk Medicaid Programs For Adults With Serious Mental Illness." The Journal of Behavioral Health Services & Research
27 (2000): 29-46.
13. Milward, H.B., and Provan, K.G. "Principals for Controlling Agents: The Political Economy of Network Structure." Journal of Public Administration Research and Theory
8 (1998): 203-21.
14. Miller, G. "Above Politics: Credible Commitment and Efficiency in the Design of Public Agencies." Journal of Public Administration Research and Theory
10 (2000): 289-328.
15. Pfeffer, J., and Salancik, G. The External Control of Organizations.
New York: Harper and Row, 1978.
16. Suchman, M.C. "Managing Legitimacy: Strategic and Institutional Approaches." Academy of Management Review
20 (1995): 571-610.
17. Provan, K.G., and Milward, H.B. "Institutional-Level Norms and Organizational Involvement in a Service-Implementation Network." Journal of Public Administration Research and Theory
1 (1991): 391-417.
18. Marsden, P. "Network Data and Measurement." In Annual Review of Sociology, Volume 16
, edited by W. Richard Scott. Palo Alto, CA: Annual Reviews, 1990, pp. 435-63.
19. Borgatti, S., Everett, M., and Freeman, L. UCINET V Version 1.0.
Natick, MA: Analytic Technologies, 1999.
20. Provan, K.G., Milward, H.B., and Roussin, K. "Network Evolution to a System of Managed Care
For Adults With Serious Mental Illness: A Case Study of the Tucson Experiment." In Research in Community Mental Health, Volume 10
, edited by J.P. Morrissey. Greenwich, CT: JAI Press, 1998, pp. 89-113.
21. Scott, J. Social Network Analysis.
London: Sage, 1991.
22. Putnam, R. Making Democracy Work: Civic Traditions in Modern Italy.
Princeton, NJ: Princeton University Press, 1993.
23. Granovetter, M.S. "The Strength of Weak Ties." American Journal of Sociology
78 (1973): 1360-80.
24. Berren, M., Duran, R., Saravia, G., and Eck, R. "Two Approaches to Managed Care
and Case Management: Impact on Cost of Services." Outcomes, Innovations and Best Practices
2, no. 1 (2000): 1-4.
25. Himmelstein, D.U., Woolhandler, S., Hellander, I., and Wolfe, S.M. "Quality of Care in Investor-Owned vs. Not-For-Profit HMOs." Journal of the American Medical Association
282, no. 2 (1999): 159-63.