"Second-generation" Medicaid managed care: can it deliver?

This article offers insight into what we term "second-generation" Medicaid managed care. In case studies of seven States, we examined three critical questions: (1) Does managed care experience facilitate program operations? (2) Can Medicaid managed care deliver on important goals? and (3) Can States extend the program beyond low-income families and children to others? The answers are encouraging but also suggest caution. Medicaid managed care is not a solution to fundamental problems facing the Medicaid program. It may be a tool to encourage better delivery of care. This requires a long-term commitment and adequate financing to develop stable partnerships with all stakeholders.

Rationale and Objectives

Medicaid managed care has been mandatory on a large scale in many States only since the mid-1990s (Hurley, Freund, and Paul, 1993; McCall et al., 1985; Hurley, 1998). States typically have pursued Medicaid managed care to achieve budget predictability, control costs, and improve access to and coordination of care. We have described key features of current programs and of the early experiences of five States actively involved in implementing Medicaid managed care (Gold, Sparer, and Chu, 1996). Others have conducted similar analyses (Holahan et al., 1998; Ku et al., 1998). These studies both provide valuable lessons and highlight the gaps in empirical evidence and the important issues that only additional experience and analysis can address.

This article aims to fill some of these gaps by focusing on the performance of “second-generation” managed care programs. We use the term second generation because our study includes two “tiers” of States: those with several years of experience under the most recent round of Federal waivers authorizing these programs and those whose late start presumably allowed them to benefit from the experience of the States preceding them. The article is based on insight developed through two rounds of case studies of States that have been actively engaged in Medicaid managed care.

The studies spanned the 5-year period between late 1994 and early 1999, which coincides with the most intense transition in States to mandatory Medicaid managed care. Through these studies, we sought to answer three questions that we believe are among the most critical to the future of Medicaid managed care:

Does experience count? Do programs operate more smoothly after overcoming initial implementation hurdles? Does experience facilitate implementation by enabling States to learn from their own experience and from the experiences of other States? What does the role of experience tell us about the administrative performance of mature programs and about the inevitability of transition issues?

What can we learn now about the ability of Medicaid managed care to achieve important health care delivery goals? While there is a lot we do not know even now, current experience can provide tentative insight into at least three important questions. First, can States attract and retain managed care plans in the Medicaid market? Second, how do States make tradeoffs between cost savings and improvements in access and quality that increase costs? Third, what can we learn about the tradeoff between Medicaid managed care and broader public health goals, particularly those relating to care for the uninsured and protection for the safety net providers who care for this population?

Can Medicaid managed care models be extended beyond their initial target of low-income families and children? Can States use managed care to improve health care for elderly people, people with disabilities, and others with special health care needs? These subgroups contribute disproportionately to costs but have substantial care needs.

The answers to these large, outstanding questions can help policymakers identify the potential and the constraints of Medicaid managed care. Our analysis relies on an issue-specific approach focusing on State experiences that are most relevant to the three questions. We direct readers wishing a more comprehensive analysis to the individual case studies and to tables that compare features of the State initiatives.

Study Methods and States

This article is based on information collected in site visits to seven States in our second round of indepth case studies (California, Florida, Maryland, Minnesota, Oregon, Tennessee, and Texas) conducted between late 1997 and early 1999. We selected these States because they are both geographically diverse and active in pursuing Medicaid managed care under mandatory models that rely heavily on capitated managed care. The second round of visits, like the first, are part of a larger initiative—The Kaiser/Commonwealth Low-Income Coverage and Access Project. We interviewed State officials, health plan staff, providers, and consumer advocates in two or three markets in each State that were affected by the initiative. Given the size and complexity of California, we focused on Los Angeles and Orange counties, which use distinct managed care models that also apply elsewhere in the State and are larger than many State Medicaid programs.

We included all the States except Maryland in the first round of site visits between 1994 and 1996, using methods and protocols similar to the ones used between 1997 and 1999. We therefore have information on the States' recent experiences and on their experiences over time. We included Maryland in the second round because we thought it would give us an opportunity to examine the ability of one State to learn from the experiences of others, since it had 1 year of experience with a relatively comprehensive and recent program for Medicaid managed care. Table 1 highlights the key characteristics of the study States and their Medicaid managed care initiatives. All seven States were implementing broad-based Medicaid managed care initiatives that relied heavily on capitated, risk-based managed care statewide.

Table 1

Profile of States and Medicaid Managed Care

State Characteristic	California1	Florida	Maryland	Minnesota	Oregon	Tennessee	Texas
Total Population 1997 (in Millions)	31.9	14.6	5.1	4.7	3.2	5.4	19.4
Percent Uninsured Among Non-Elderly 1994-1995	19.7	19.2	14.4	9.2	13.7	7.2	23.9
Total Medicaid Eligibles FY 1997 (in 1,000s)2	6,400	2,100	448	590	552	1,400	2,800
Total Medicaid Managed Care Enrollment 12/97 (in 1,000s)2	2,364	910	308	NA	305	1,281	440
Federal Waiver	1915b	1915b	1115	1115	1115	1115	1915b
Start of Mandatory Managed Care	Los Angeles (1/98), Orange (10/95)	312/95	6/97	1985	2/94	1/93	41/93
Managed Care Options
Fully Captiated Plans	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Primary Care Case Management	No	Yes	No	No	No	No	5Yes
Populations in the Mandatory Model
AFDC and Related	Yes	Yes	Yes	Yes	Yes	Yes	Yes
SSI and Related	6Orange	6Yes	Yes	No	Yes	Yes	No
Dually Eligible	No	No	No	No	Yes	7Yes	No
Institutionalized	No	No	No	No	No	No	No
Benefits Excluded from Capitation
Behavioral Health	FFS	In	8Carveout	In	8Carveout	Carveout	9FFS
Long-Term Care	FFS	FFS	FFS	FFS	FFS	FFS	FFS
Eligibility Expansion (Other than SCHIP)	No	No	No	Yes	Yes	Yes	No

The particular model varies by county. Los Angeles uses a two-plan model of capitated plans, and Orange County uses a county-organized health system model, with most care delivered under capitated subcontract arrangements.

Includes new expansion enrollees in Oregon (94,000), Tennessee (383,000), and Minnesota (104,000). The figure for California is for 2/99; the figure for Texas is for 4/99.

The default at the start was only to Primary Care Case Management model. Health maintenance organizations were inlcuded in the mandatory model in 1996.

Restricted to pilots in Travis County (Austin) and the Tri-County (Galveston) area. Mandatory enrollment was rolled out in the San Antonio region in 9/96; in the Fort Worth and the Lubbock region in 10/96; and in Harris County (Houston) in 12/97.

In most counties. Exceptions are Travis County (Austin) and Fort Worth.

SSI-eligible persons are included in mandatory programs only under the county-organized health system model being demonstrated in five locations across the State. In Florida, the default to health maintenance organizations was added to a primary care case management default in 1997. Children with special needs are excluded.

Only for benefits not included under the capitation rate (mostly pharmaceutical coverage).

In Maryland, carveout applies to mental health, not substance abuse services. Primary mental health is included in the basic capitation rate to managed care plans. In Oregon, mental health, not substance abuse, is carved out.

A behavioral health carve-out arrangement is being implemented as part of the Dallas County roll-out.

NOTES: FY is fiscal year. NA is not available. AFDC is Aid to Families with Dependent Children. SSI is Supplemental Security Income. FFS is fee for service. SCHIP is State children's health insurance program.

SOURCES: Authors' analysis of materials collected from the State and health plans onsite, 1999. Data on uninsured from Kaiser Commission on Medicaid and the uninsured, Kaiser State Facts: Health Needs and Medicaid Financing, 2/98. Data on Medicaid-eligible persons from the Health Care Financing Administration, HCFA 2082 Report for Federal FY 1997, Table 15.

The States varied substantially in population, managed care experience, sociodemographic characteristics, and region of the county in which they are located. They also had varying degrees of operational experience specific to Medicaid managed care. Oregon, Tennessee, and Florida launched their Medicaid managed care initiatives during the early to mid-1990s. Minnesota had done so even earlier. The initiatives in Maryland, Texas, and California were more recent or, in some cases, were just being rolled out in major population centers. Different groups of States, therefore, give us a different view of the role of experience in program development.

The scope of the State initiatives also varied. Although all seven mandated enrollment for children and low-income families, the comprehensive initiatives of Oregon, Tennessee, and Maryland included all or most of the population receiving supplemental security income (SSI). These three States had also established relatively complex structures to provide behavioral health care through a separate, specialized managed care program. Oregon, Tennessee, and Minnesota had expanded eligibility substantially under 1115 demonstration waivers.

Does Experience Count?

Because Medicaid managed care programs are typically very complex, they are inherently challenging to manage. In earlier work (Gold, Sparer, and Chu, 1996), we found that all States that have introduced Medicaid managed care have had to resolve administrative issues, such as establishing eligibility and enrollment criteria and processes, and performing basic oversight of managed care. Problems related to eligibility and enrollment have included overloaded telephone lines, difficulties with the content and distribution of enrollment materials, and high rates of involuntary assignment often followed by beneficiaries' requests for change and their confusion about which plans and providers to see. Problems related to oversight are manifest in publicized marketing abuses, weak financial performance that often is detected late, and in allegations that networks are inadequate, quality is uncertain, access is poor, and State agencies are not positioned to make corrections. In the first round of visits, we found these issues to be challenging for States, but we did not know whether these problems were transitional or whether they could be avoided with more experience.

The second round of visits indicates that administrative processes mature with time and experience, and that some early problems are therefore largely transitional. States can learn from their own experiences in other locations or with previous models, and they can learn from other States. However, the benefits of experience are limited in the following important ways: (1) unique State personalities or other underlying and hard-to-modify dynamics predispose States to enact certain types of policies in certain ways, which can mean repeating the same kinds of mistakes; (2) turnover in key personnel interferes with administrative continuity and learning; and (3) achievements, while desirable, can raise the bar of expectations, encouraging dedicated and talented staff to seek new challenges (limiting program stability) and to be overly optimistic about the ability to solve problems.

Programs Stabilize Over Time

Programs stabilize over time as administrative systems develop and gaps in oversight are addressed. Oregon, Minnesota, and Orange County devoted substantial time to developing and introducing their initiatives. All three encouraged administrative stability through careful planning and a deliberate, step-by-step process of confronting challenges. Oregon waited until its program was in its second year to add managed care for SSI beneficiaries and foster children, delaying full implementation of the mental health components even longer so that it could sponsor demonstrations. Minnesota, which implemented Medicaid managed care in 1985, also used small demonstrations to test managed care models both for SSI-eligible persons and that might be attractive in rural counties. CalOPTIMA, sponsored by Orange County, began operations in October 1995 but did not include the SSI population until February 1996. These types of planned expansions enabled all three States to reduce the severity and extent of initial administrative problems. Obviously, this type of approach is the best one for minimizing problems. But it is not consistent with the political style of all States.

States also can benefit from their own early experience, especially if, like Texas, they phase in implementation. Texas has a relatively undeveloped managed care infrastructure and a strong State medical association that prefers primary care case management (PCCM) to health maintenance organizations (HMOs). Using two early pilots (in Austin and Galveston) to test managed care, the State expanded these efforts to San Antonio, Lubbock, and Fort Worth, followed by a rollout in Houston. As our study ended, Texas was implementing managed care in Dallas and El Paso.

The two early pilots provided valuable experience, but their small scale limited their teaching potential. In particular, the pilots enabled Texas to defer revising its administrative structure (which relied heavily on an external, indemnity-oriented private firm for administrative support) to match the emerging managed care model [which required a structure focused on education, network management, oversight of care (especially in terms of under-use, and resolution of conflicts of interest between contractors and networks)]. Thus, despite the experience with pilots, Texas localities in the first wave of rollouts still experienced substantial problems, with considerable beneficiary confusion about plan enrollment and assignments. Learning from this first wave of experience, Texas employed an enrollment broker for the Houston rollout, where there was less confusion.

Of the States we studied, Tennessee experienced the most severe problems at startup in 1994. State staff, plan providers, and beneficiaries all generally agreed that program administration during the first year was chaotic. By late 1997, Tennessee had substantially strengthened its oversight systems by requiring all plans to be licensed as HMOs and by making oversight the joint responsibility of the insurance commissioner's office, which had responsibility for financial and administrative systems, and the TennCare Bureau, which oversaw quality in collaboration with an external contractor. Able to move past enrollment and network problems (which became less severe), the State could then begin to develop systems for ongoing program monitoring. According to State-sponsored surveys, beneficiaries' satisfaction with their insurance plan, which had decreased at the start of the TennCare program in 1994, had returned to its pre-TennCare level by 1996 (Fox and Lyons, 1998).

Constraints on Learning

The benefits of experience are, unfortunately, constrained by at least three factors. First, States have unique personalities or styles, which tend to lead them to repeat some mistakes or to fail to move forward. For example, Tennessee's emphasis on rapid program implementation was not changed by the disruptive early experience of TennCare. Indeed, the State's decision-making style led it to undertake an equally ambitious and even more controversial initiative 2 years later—TennCare Partners—a behavioral health carve-out initiative. Apparent in TennCare Partners were some of the same design and implementation mistakes of TennCare. TennCare Partners also diverted managerial attention from strengthening the basic TennCare program.

Like TennCare, Florida's program also matured, but the extent of growth and learning was constrained by the continuing challenges of the State's political environment. In 1996, Florida focused on responding to highly visible reports in the press on plan marketing abuses, quality problems, and the absence of a centralized beneficiary education process that might limit abuse and inform choice. Florida curbed the allegations of abuse and strengthened oversight staff and activities by 1998. Nevertheless, its litigious environment hampered the release of reports on the quality of care and delayed the award of a contract with an enrollment broker that was designed to support more comprehensive choice counseling.

Second, in addition to State decision-making styles, the ability of States to learn from experience also is limited by staff turnover. Tennessee has struggled with extensive turnover in key leadership (Gold and Aizer, 2000). Other States also had trouble retaining staff. At the time of our second-round site visits, for example, the administrative infrastructure and attention to administrative detail continued to be weaker in Tennessee than in Oregon, Minnesota, and Orange County. Nevertheless, Oregon still had difficulty retaining experienced staff, as its salaries were much lower than those offered in the private sector. Some also found the mechanics of operating an existing program were less interesting than the challenges of developing new ones and thus moved on.

Third, learning is limited by the fact that experience seems to “raise the bar” of State expectations, encouraging staff to seek new challenges and to be overly optimistic about their ability to solve problems that arose in earlier efforts. For example, Texas also used the Houston rollout to test managed care models that might allow it to to expand its program to the non-Medicare SSI population. Similarly, the rollout in Dallas tested an expansion to cover behavioral health, which also brought new challenges. In Orange County, the perception that the initiative had achieved its goals created staff and county demands for the program to deliver more, including, for example, taking on responsibility for long-term care (LTC) payment and sponsoring demonstrations for the medically indigent population. Clearly, because new challenges like these breed new issues and problems, administrative learning becomes an ongoing process.

The double-edged sword of experience as both teacher and constraint is perhaps best exemplified by Maryland. The State's long experience with an extensive, voluntary HMO program and with a mandatory PCCM program facilitated its implementation of Medicaid managed care. Moreover, its extensive planning process, which involved the public and private sectors and contracts with experienced analysts from the State university for program development support, enabled the State to identify problems in other States and some of the solutions that had been proposed. Maryland's academic knowledge base and collaborative process helped stakeholders in the State identify both policy-related problems (for example, risk segmentation, special needs of vulnerable subgroups, and potential adverse effects on safety net services) and possible solutions. However, the birth of many “good ideas,” the involvement of multiple stakeholders, and the extensive development period led to a relatively complex and highly ambitious initiative, complete with inherent implementation challenges. Maryland's initiative was complex because of both the specifics of the policies (e.g., risk adjustment, detailed access and quality standards monitored with encounter data) and the sheer number of them to be implemented simultaneously. Given the scope of Maryland's initiative and the nature of its development process, design rather than implementation secured a disproportionate amount of attention. Thus, although the State invested a relatively substantial sum in enrollee education and delayed startup for 6 months to allow for more systems development, it experienced many of the same, extensive startup problems as did other States as a result of expectations that may have been higher than what was reasonable.

In sum, programs mature with experience, but idiosyncratic political processes and staff turnover can block the internal and cross-State learning that would normally come with program maturation. And the fluidity and complexity of programs continue to create challenges, even with experience.

Ability to Deliver on Key Goals

Medicaid managed care goals generally reflect a varying balance among cost, quality, and access. Many States also seek to mitigate adverse effects that may spill over from the Medicaid managed care program onto other important State objectives, such as care for low-income uninsured individuals, protection for traditional safety-net providers, and targeting of Federal Medicaid funds to support State-sponsored health services of various types.

The long time horizon required to assess performance, the nature of the case study method, and the complexity of Medicaid managed care in the State context limit the insight we (and others) can derive on the States' performance in these areas (Gold, 1999; Starke, 1995; Norton and Lipson, 1998). Even now, 4-5 years after some programs were established, we know very little about their ultimate success. However, our study provides evidence on three dimensions of program performance that seem essential to achieving the goals of Medicaid managed care: (1) attracting and retaining plans and providers; (2) choosing the best tradeoff between a State's commitment to increasing access and providing quality care, and its cost-savings objectives; and (3) limiting adverse effects on the safety net.

Attracting and Retaining Plans and Providers

Low payment rates, extensive requirements, and complex beneficiary needs, among other challenges, have historically made it harder to attract providers to Medicaid than to the commercial market. States continue to grapple with these issues under Medicaid managed care (Felt-Lisk and Yang, 1997; Felt-Lisk et al., 1999).

The level of rates is a critical influence on participation, though the way that this relationship manifests itself is complex. Recent research shows that payment rates vary greatly across States, even after standardizing for key State features like benefits (Holahan, Rangarian, and Schimmer, 1999). Among the States we studied, rates are lowest in California, Tennessee, and Florida. They are highest in Texas and Minnesota. Low rates make it more difficult for a State to attract and retain plans. And while low rates may not prevent States from securing plan participation (as we found), they do affect the mix of participating plans and will discourage participation or continued program commitment from commercial plans that can do without the business.

Rates were more an issue in some States than others. Plans in Tennessee, California, Texas (the Dallas rollout), and Florida all expressed strong concerns about the overall adequacy of rates. In other States, the focus was less on rates overall than on adjustments for patient mix (Maryland and Oregon) and on cost factors stemming from adverse selection in safety net plans (Oregon, Tennessee, and the Texas earlier rollouts). In addition, plans universally believed they were not adequately reimbursed for the costs of administrative compliance with State-imposed requirements. Finally, commercial plans cited substantial differences between State requirements and commercial practice as reasons for not participating. These plans considered systems' redesign an unwarranted burden from a business perspective. These tensions appear to have grown over time with program experience.

A lack of consensus on the minimum number of plans needed to provide a real choice for beneficiaries and the failure of policymakers to agree on the kinds of choices that need to be offered make it difficult to assess whether “enough” plans and providers have been recruited and whether they can be retained (Felt-Lisk, Frazer, Gold, 1994). Further, the adequacy of choice becomes an issue if all options for beneficiaries essentially involve the same providers, which may or may not include the mainstream providers. Unfortunately, weakness in the data available for measuring Medicaid provider participation and care patterns that existed before Medicaid managed care also complicate analysis of change in provider participation as a result of Medicaid managed care.

A simple test of minimal adequacy can be based on the market and political process: whether enough plans and providers operate in a State to enable it to proceed with managed care, what tradeoffs were necessary to accomplish this, and whether participation can be sustained over time. Almost by design, in terms of their selection, all the study States “succeeded” initially because they were operational, HCFA-approved, and able to sustain their programs. For the most part, they provided at least two formal choices in all areas of the State and more choices in most urban areas. However, most States took longer than expected to meet this objective, so implementation fell behind schedule.

States had to make tradeoffs to achieve the goal of sufficient choice, including minimal initial requirements for participation and less focus on mainstreaming (Table 2). Florida and Tennessee had serious concerns about their ability to attract enough plans and therefore deliberately set requirements low, tightening them only after their programs were established. The experience in both States suggests that this strategy requires at least a short-term tradeoff between participation and instances of abuse, confusion, and limited access. In contrast, Minnesota's HMO licensure requirement emphasized network adequacy and fiscal solvency, among other features, even though this limited statewide implementation in rural areas.

Table 2

Selected Measures of Medicaid Managed Care Operations, by State

Plan Participation	California1		Florida	Maryland	Minnesota	Oregon	Tennessee	Texas
	Los Angeles	Orange
Number of MCOs
Initial Year	211	238	22	9	6	20	12	3
Most Recent Year	11	17	15	N/A	7	14	9	3
Percent Enrollment in Two Largest Plans
Initial Year	28	30	56	49	60	47	73	—
Most Recent Year	30	28	48	—	65	43	70	77
Percent Fully Capitated Plans (of managed care enrollees) 19973	92	—	40	78	100	100	100	55
HMO Licensure Requirements
Initial Year	Yes	4No	No	4No	Yes	4No	No	Yes
Most Recent Year	Yes	4No	Yes	No	Yes	4No	Yes	Yes
Competitive Bidding Used at Any Time in 1997-1998	No	No	No	No	No	No	No	Yes
Urban Institute Rate Analysis3,5
Statewide Adjusted MMC Rate	82.75	—	99.61	121.05	139.64	—	99.07	153.75
						—
Statewide MMC Relative to 50th Percentile	0.066	—	0.80	0.97	1.12	—	0.79	1.23
State AAPCC Rate to 50th Percentile	1.3	—	1.24	1.08	0.88	—	1.07	1.18
Medicaid MMC Relative to Medicare	0.51	—	0.64	0.89	1.27	—	0.74	1.04

The number of MCOs and percent in two largest plans are for San Antonio. The rest of the statistics are for all of Texas.

Prepaid Medical Assistance Program and General Assistance Medical Care Program.

(Holahan, Rangarian, and Schimmer, 1999.)

In Orange County, HMO licenses are required except for designated provider entities that are subject to alternative, though generally comparable, requirements. Maryland has provisions that give provider-sponsored plans in Medicaid some additional flexibility though this has not tended to be needed. Oregon does not license HMOs and Medicaid has always done its own oversight.

(Holahan, Rangarian, and Schimmer, 1999.) Oregon declined to participate. California data are for the two-plan model and hence, do not apply to Orange County. Rates are adjusted to allow comparisons of rates using standardized rate cells, maternity expense policies, disproportionate share hospital, and graduate medical education exclusion policies and benefit package.

NOTES: MCO is managed care organization. NA is not applicable. HMO is health maintenance organization. MMC is Medicaid managed care. AAPCC is adjusted average per capita cost.

SOURCE: Authors' analysis of materials collected onsite from the State and health plans, 1999.

In Maryland and California, the key tradeoff was between the mainstreaming objective, which these States viewed as less critical, and other more important objectives. Maryland designed a program to encourage traditional providers to form plans or to join existing networks. California developed a two-plan model in 12 counties to allocate a share of enrollment to plans affiliated with public and other safety-net providers. In Los Angeles, however, adequate public capacity to support the model was lacking, and commercial plans played a larger role than originally envisioned. As a result, 10 plans, through complex subcontracting arrangements, actually shared risk for care (Draper, Gold, and Hudman, 1999).

The study States—with the exception of Texas and Los Angeles—accepted all qualified plans, at least initially. Some, like Orange County, which had an option for provider capitation, initially set enrollment thresholds low to encourage participation but subsequently raised them. This action encouraged consolidation, thus reducing administrative burden and increasing economies of scale. Only Texas used competitive bidding. Plans there were concerned about obtaining enough enrollment to achieve economies of scale and welcomed a limit on participants. Despite concerns over rates and other issues (in Texas as in other States), all the Texas county rollouts had more plan applicants than slots until recently, when several plans withdrew from the Dallas rollout and left slots unfilled.

In most States, consolidation and withdrawals reduced the number of participating plans after startup (Table 2). Typically, however, two or three plans dominated over time, accounting for a large share of statewide enrollment. Commercial plan participation in Medicaid managed care is an issue (Felt-Lisk and Yang, 1997). The experience from our study suggests that the dynamics of State involvement vary substantially across States, and that commercial plans are more likely to remain committed if they have participated since the inception of managed care (as in Maryland and Oregon) and are strongly locally based. Full participation by all commercial plans is probably contingent on three factors: State requirements or strong incentives (as in Minnesota), a strong focus on actuarially sound rates (as in Oregon), and a pre-existing managed care infrastructure that favors an integrated model (as in Minnesota and Oregon, which have extensive managed care histories and which provide organized information groups and systems). Even with a commitment, however, the number of participants will diminish if the industry is consolidating (as it is now). High fixed costs for participation (to meet Medicaid requirements) also encourage economies of scale and consolidation. This pattern of results suggests that States must work seriously to foster adequate plan participation, the absence of which, as others have found, can be a serious barrier to the continuation of Medicaid managed care. Experience further suggests that the issue of commercial versus Medicaid-dominant plans is an important one, but that it plays out differently in different markets.

Our findings also show why plan participation and provider participation are distinct concepts that should be monitored separately. We found that patterns of care and provider participation under Medicaid managed care are shaped by the patterns that existed before managed care. Thus, the States and communities in which office-based physicians had been extensively involved in Medicaid (Minnesota, Oregon, Orange County, and San Antonio) retained that participation under Medicaid managed care. Medicaid managed care helped States increase the number of office-based participants (in Florida, Houston, and western Tennessee) by expanding alternatives to the public hospital and county clinic system. Shifts from expanded provider choice also are more likely in communities in which traditional providers have a poor public image and unattractive facilities. Florida's experience shows, for example, that States can also accomplish this under a fee-for-service (FFS) system if Medicaid raises rates (as Florida did, for example, for primary care). Such strategies are especially likely to be effective when increased pressure to contain costs on the commercial side makes Medicaid relatively more attractive.

Thus, our research suggests that the study States have been able to attract enough plans to make a managed care strategy initially feasible, but participation varies across States, and some States have had to make more tradeoffs than others to operationalize this strategy. Most critically, all have experienced some contraction in participation over time. In securing plan participation, States do not necessarily achieve or ensure access to mainstream care (typically through office-based physicians). Barriers to mainstream care appear particularly difficult to overcome in communities and States in which this care was unable to thrive under traditional FFS Medicaid. Both rate levels and a State's attitude as a business partner are important factors in retaining participants. If States consider these goals to be important, they must work in partnership with plans, be committed to long-run goals, develop better databases to monitor provider participation and enrollee care patterns, and shape the policies necessary to change counter-productive patterns.

Attaining Cost Savings Versus Other Objectives

All the States initially emphasized cost savings. Most of them used administered pricing and included a specified percentage reduction in estimated FFS spending to calculate capitation rates. In Tennessee, the initial TennCare rates were set to less than three-quarters of estimated costs, based on estimated savings that included offsets for assumed savings in provider charity care, with the difference in State spending applied to expand eligibility. Besides Tennessee, the savings assumed in calculating capitation rates were greatest in Maryland (10 percent) and in Florida (8 percent from 1997). However, not all States achieved their expected savings rates because of errors in rate calculation (Maryland) or because of the failure to adjust for selection into HMOs (in States offering a PCCM option). In general, capitation, compared with FFS, seems to make it easier for States to cut payments, as in Florida where capitation payments were reduced from 95 to 92 percent of FFS in 1997. It is possible that these reductions are easier because the effects on providers of payment cuts in capitation are less direct, lessening opposition from providers. But capitation can also increase pressure to selectively raise rates, as when Minnesota's legislature directed that rates be increased to improve access in rural areas.

In general, it appears that the States could most easily obtain support for trading-off cost savings against other objectives when they first launched their initiatives. For example, the Harris County Hospital District (Houston) forced legislation that guaranteed it and other hospital district plans an automatic slot in Texas' program. Similarly, Tennessee's need to secure advocate support (to encourage HCFA to approve the waiver) led to provisions on coverage expansion. Once a program is adopted, achieving far-reaching change seems much more difficult. Minnesota, Oregon, and Tennessee, which included large eligibility expansions in their initiatives, had difficulty garnering support for additional expansion after the first year; most growth in enrollment from new eligible individuals occurred early in the program, and enrollment levels remained relatively flat thereafter. Thus, both spending and tradeoffs tend to involve marginal actions in response to issues that various stakeholders raise. For example, savings may be reduced as some benefits, such as certain pharmaceuticals, are removed from capitation and provided on an FFS basis; special funds may be appropriated to offset potential adverse effects on safety net providers; or additional funds may be added to a State's budget to enable State officials to respond to highly publicized problems.

The change in political climate over the study period appears to have favored cost savings over other objectives. The welfare rolls in all the study States decreased between January 1993 and June 1998, ranging from 64 percent in Florida to 16 percent in California (United States Department of Health and Human Services, 1999). Comparable data on reductions in Medicaid eligibility levels were not available, but most of the States indicated that welfare-related eligibility for Medicaid had fallen substantially. This heightened their concerns over the rising number of uninsured individuals and the potentially reduced ability of safety-net providers to respond to rising demand (Ellwood and Ku, 1998).

The State Children's Health Insurance Program (SCHIP) has provided a counter-vailing set of pressures that partly offset both the decline in Medicaid enrollment associated with welfare reform and the destabilizing effects of the growth in uninsured populations on the safety net. In some cases, however, the States simply used SCHIP to shift priorities for expansion, for example, from below-poverty families overall to previously uninsured low-income children (Oregon). SCHIP's provisions for a private-sector alternative to Medicaid also meant that some States shifted administrative resources to develop coverage options outside the traditional Medicaid program. Florida, for example, expanded its small Healthy Kids program (a joint public-private effort offered through schools) to cover new SCHIP-eligible children of school age. In California, SCHIP has been implemented partly through the Healthy Families Program, which is administered mostly separately from Medi-Cal.

In sum, the States' efforts to achieve cost savings account for the competing concerns about access and quality, but these concerns appear more likely to lead to a reduction in expected savings rather than to an increase in overall spending. Also over the study period, it appears that spending commitments to expanded coverage, access, and quality are more likely to be made at the start of a program, when political support from stakeholders is essential to program progress. Thereafter, the States may make tradeoffs, but they are likely to succumb to pressure to generate savings rather than take on new obligations. Any increases in spending, typically at the margin, are responses to the actions of organized interest groups or to highly publicized problems. Finally, erosion in support for public programs over time has encouraged some States to explore models that involve more partnering between the public and private sectors.

Protecting the Safety Net

As States move to Medicaid managed care, support for the safety net is likely to erode even though State policy may benefit newly covered individuals through expanded coverage, and serious expansions may help the providers previously serving the uninsured. Our research suggests that the effects on the safety net and the tradeoffs vary with the strength of both the independent funding stream available to safety net providers and the management infrastructure they can use to help them compete. Thus, hospitals in Maryland have benefited from a rate-setting scheme that includes all payers and compensates hospitals for reasonable costs for the uninsured. Under managed care, the system generally continues paying hospitals the same rates they received under FFS. The exception is approved arrangements when the hospital assumes risk. In Florida and in some Texas hospital districts like Dallas (Parkland Hospital), tax-supported local financing and strong teaching affiliations that generate a diverse patient mix have enhanced the capacity of the public hospital to compete. Conversely, some hospital districts in Texas (Houston and Fort Worth) appear to have been affected adversely by Medicaid managed care despite some independent funding because they still depend on Medicaid funds and have weaker management.

The role of safety net providers in a community before the introduction of Medicaid managed care is also important. Medicaid managed care appears to have the greatest adverse impact on safety net providers that have long shouldered most of the burden of care for both the uninsured and Medicaid patients. Consequently, Medicaid and indigent care funds comingle in a way that makes the potential loss of Medicaid patients more acute. Thus, California implemented its two-plan model in order to maintain the flow of revenue to large public systems, such as those in Los Angeles. But such adverse effects were less a concern in Orange County because of the absence of a public hospital.

Smaller and non-hospital-based safety net providers appear to be the most vulnerable to reductions in revenue resulting from the introduction of Medicaid managed care. Typically, these community clinics have few sources of revenue other than grants and Medicaid. To support operations, many have depended on cost-based reimbursement under Medicaid's federally qualified health center (FQHC) provisions. Under Medicaid managed care waivers, many clinics lost that form of payment. Their volume of Medicaid patients also fell. Part of this situation reflects the expanded choice of providers allowed by Medicaid managed care and the preferences of patients when faced with these choices. But patient flow to safety net providers also seems to be adversely affected by the startup confusion and involuntary assignments. Further, safety-net providers, even more than others, are challenged by the demands associated with managing risk and capitation. For example, risk-based safety-net providers need adequate systems to monitor incurred-but-not-reported claims and to match service with eligibility and benefits. But they often do not have systems that are up to meeting this challenge or the resources needed to build such systems.

Community clinics seem to be especially harmed by transition-related problems. For example, they may lose out if auto-assignment is dominant in the program, as their patients seem to be more likely to be affected. Or the design of enrollment materials may put them at a disadvantage, especially if these materials are organized by physician name when their patients identify more with the clinic. Because clinics depend more on Medicaid revenue than do other providers, they are more sensitive to initial payment delays, which occurred in many States with the introduction of Medicaid managed care.

Medicaid agencies can implement policies to mitigate these adverse effects. For example, they can use an appropriate risk adjustment technique to equitably pay plans and providers who may treat more severely ill patients. But good techniques for risk adjustment are limited both in Medicaid and in the commercial sectors, and few plans employ even those techniques that are available. Among the study States, only Maryland was experimenting with a major risk-adjustment technique (based on adjusted clinical groups, (ACGs); others enacted more limited rate adjustments (e.g., maternity “kick” payments at time of newborn delivery) or made discretionary provider payments on the basis of adverse selection. Although Maryland's approach garnered strong support, its application requires the availability of encounter data. Though plans are required to submit such data, their quality has been poor, thus limiting Maryland's adjustment to a subset of beneficiaries for whom prior FFS data are available.

Policies that protect the safety net often raise other issues, as they generally require a tradeoff between managed care and broader health objectives. The tradeoffs are most obvious in Maryland's initiative, which focused extensively on linking Medicaid managed care and these broader health objectives. For example, Maryland's policy of encouraging self-referral to promote continuity of prenatal care as well as unfettered access to publicly funded immunizations and school-based services limited the ability of managed care plans to coordinate that care but held them responsible for its costs. Maryland also implemented a controversial policy of requiring plans and providers to report on whether welfare recipients received treatment for substance abuse. While the intent was to encourage treatment and keep people eligible for welfare coverage despite not being in the workforce, the policy was burdensome and raised ethical concerns for providers.

Participation of safety net providers that shoulder most of the burden for the uninsured can encourage continuity of care for Medicaid-eligible persons who enter and exit eligibility rolls. In Texas, for example, several hospital districts compete by guaranteeing eligibility for care even if independents lose their Medicaid coverage. To be effective, however, these provider systems must have the resources and management infrastructures to mount competitive managed care plans.

Some policies are basically problematic. The requirement that managed care plans contract with designated safety net providers was beneficial in States in which the requirement targeted highly specialized providers. But the same requirement was relatively ineffectual in other States because the issue for traditional providers was less a matter of a contract than one of ensuring a flow of patients that maintained a revenue stream. By requiring HMOs to pay cost-based reimbursement to FQHCs, Florida made it difficult for FQHCs to contract with managed care organizations.

Our analysis suggests that the move to Medicaid managed care will inevitably draw funds away from uncovered low-income individuals unless States deliberately consider the tradeoffs and structure their policies to minimize the chance that this will happen. Protecting the safety net seems more important to some communities than to others, depending upon the role of the safety net before managed care and on the extent to which program funding streams in that community and State are intertwined.

States that wish to limit adverse effects on the safety net have three main, and only partly satisfactory, options. First, they can expand eligibility and coverage. However, as Oregon, Minnesota, and Tennessee discovered, while significant coverage expansion is possible, universal coverage seems an unrealistic goal in the present political climate, regardless of whether that coverage would come through Medicaid or through insurance reform in the private sector (as Oregon attempted to use). Second, States can try to protect safety-net providers by structuring Medicaid policy to include risk adjustment and roles for public systems. However, some safety net providers are too weak to benefit from this option, and patients may not always view public systems as attractive. Third, States can directly pay for care for the uninsured by explicitly supporting safety-net providers through grants or programs that are unaffected by the shift to managed care. However, it is not clear that the political climate would allow new programs like this to go forward, as they could require new taxes (in counties or hospital districts) or State regulation (all-payer ratesetting).

Extending Medicaid Managed Care

Medicaid is the primary means of covering certain especially needy populations. Two-thirds of program expenditures are targeted to individuals who are eligible because of disability or age (65 or over), but these groups together make up only about one-quarter of the total enrollment (Kaiser Commission, 1998). LTC accounts for more than one-third of spending for intermediate care facilities for people with mental retardation, and for nursing home, home health, and mental health benefits. Excluding these components from Medicaid managed care would limit the States' ability to generate savings and to use managed care to deliver care to the neediest people in a more organized way. Consequently, more States are expanding Medicaid managed care to SSI beneficiaries, although LTC for this group is not likely to be included for some time. Changes in Federal policy have also encouraged States to expand their Medicaid coverage by making it easier for them to do so (Schneider, 1997). And unlike low-income families, for whom we have data on commercial managed care experience, many kinds of subgroups in the Medicaid SSI population have no commercial parallel.

About one in four non-elderly persons with disabilities in Medicaid were enrolled in managed care in 1998, two-thirds of whom were in capitated plans (Regenstein and Schroer, 1998). Our study included three of the five States that had at least 75 percent of their Medicaid under age 65 population with disabilities in capitated managed care: Tennessee (100 percent), Oregon (78 percent), and Maryland (75 percent). Moreover, two-thirds of Florida's Medicaid population with disabilities were in managed care programs that offered a choice of capitation or PCCM; two-thirds of this enrollee group had chosen PCCM. Orange County covered SSI beneficiaries, as did several other States, through small pilot programs (Table 1).

Yet, even as States proceed with expanding Medicaid managed care eligibility, there are still significant gaps in coverage. Medicaid managed care in the study States focused only on acute care; coverage of the institutionalized was typically excluded, and LTC was carved out. Dually eligible persons are another issue. Only Oregon and Tennessee included individuals dually eligible for Medicaid and Medicare, and each made special provisions about jointly covered services (crossover benefits, which only Oregon covered). When Medicare and Medicaid both covered a service, Medicare is the primary payer and overrides State Medicaid policy. States have no authority to waive Medicare policy, and ratesetting is extremely challenging because each program maintains its own claim records. In expanding to include SSI-eligible persons, Florida and Tennessee established models that had relatively uniform requirements for all eligible beneficiaries, including SSI beneficiaries. However, both States have had to resolve problems arising because of unique characteristics of the SSI program, such as the involvement of the Social Security Administration in eligibility determination, which complicated the way that standard enrollee education and plan selection processes were structured. Maryland's emphasis on providing coverage to groups with special care needs meant that special program features for many SSI beneficiaries were mandated even though these were not targeted at SSI explicitly. All the States established special arrangements to resolve problems related to including seriously and persistently mentally ill individuals in Medicaid managed care.

Of the study States, Oregon conducted the most planning for the expansion. It did not include SSI beneficiaries until the second program year, and it sought participation from a cross-section of stakeholders, including advocates, plans, providers, and State agencies. All participating plans were required to enroll SSI beneficiaries, and Oregon added three mechanisms to the program to facilitate the transition: (1) communication reinforcing the fact that comorbidities under the State's priority-based benefit package were covered even if the condition would not be treated otherwise because it was “below the line” and thus excluded from the benefit package; (2) a requirement that each plan hire and train at least one exceptional needs care coordinator (ENCC) to assist SSI beneficiaries in navigating the system, and (3) an ombudsman's office to serve as a client advocate.

Oregon's careful planning contributed to a relatively smooth transition, but the State still had to resolve several issues. For example, ENCCs are viewed as valuable by some advocates, but their perceived effectiveness varies in part because State requirements are vague, so plans have used these coordinators in different ways. Some have used them as both patient advocate and high-cost case manager, which can cause conflicts. Because knowledge of the ENCC role varies across plans, providers, and beneficiaries, the coordinators are not used as fully as they might be. Moreover, coordination among ENCCs, the ombudsman, agencies, and other actors is less than optimal.

Monitoring the inclusion of SSI beneficiaries has been challenging. All the study States had problems developing data to monitor performance specific to SSI-eligible individuals, but these data are needed to ensure that individuals without an active voice do not become invisible by virtue of a focus on the average experience. Oregon has evaluated its ENCC program and reviews such conditions as depression and diabetes in its quality oversight. However, it has not structured systems to assess how subgroups of SSI beneficiaries are faring. Thus, although the State's experience suggests that with careful planning, these subgroups can be integrated into managed care, it is not yet possible to measure outcomes for them or to confirm that the relative absence of highly visible complaints signifies good performance.

It is substantially more difficult to move dually eligible individuals than it is to move non-Medicare SSI-eligible persons into Medicaid managed care. Administrative coordination with HCFA has been a challenge for States like Oregon. A key issue is the disparity in lag time for enrollment across the two programs, which complicates care management. Both Oregon and Tennessee covered dually eligible persons, but Tennessee covered only non-Medicare benefits (mainly pharmaceuticals). Medicare coverage of physician services enabled beneficiaries to go to any physician in Tennessee, but under TennCare, prescriptions could only be filled at some pharmacies, and beneficiaries were often unaware of this. This structure built confusion about and barriers to filling prescriptions while making it difficult for the State to control the costs of pharmaceuticals, since there were no controls on physicians. (TennCare's dually eligible policy has since been changed.)

Oregon was more successful in integrating the programs, largely because Medicare managed care penetration in the State is high, and four of the six Medicare HMOs participate in both programs. Overall, one-half of dually eligible individuals in Oregon are in fully capitated plans, and the State designed its policies to prevent beneficiaries from enrolling in more than one health plan. But few States are like Oregon, where jointly participating plans are predominant and Medicare managed care penetration is high.

Conclusions

We suggest that the findings of this study show that managed care is not a direct solution to the fundamental issues that Medicaid must confront. Managed care does, however, provide a tool that, appropriately used, may encourage better care. In a society in which commercial coverage focuses on acute care, universal coverage is absent, and support for spending by the public sector is limited, government has expanded Medicaid to serve a variety of needs but has often failed to provide enough funds to achieve its goals. Clearly, moving to Medicaid managed care does not eliminate these tensions. In fact, it may accentuate them.

States can look to Medicaid managed care as a tool for encouraging the development of more accountable and coordinated systems. Nevertheless, States should know that such development requires a substantial time commitment, with early years devoted to creating basic administrative systems rather than to enhancing the delivery of care at the local level. States should not expect substantial savings, at least not if they want to encourage broad-based plan participation. Because States can resolve problems over time, they can focus on encouraging plans to better manage clinical care. However, a State's ability to learn from its experience and from the experience of other is diminished for four main reasons: (1) because States vary, what works in one State may not work in another, (2) idiosyncratic decisionmaking styles can lead States to repeatedly make some of the same mistakes, (3) State staffing is not stable, and (4) States do not always achieve stable participation from plans or providers. The ability of the programs themselves to stabilize is limited by the complexity inherent in many Medicaid managed care models and by the States' desire to confront new challenges, perhaps before they are prepared to do so. Our findings also show why expanding Medicaid managed care to subgroups of particularly needy individuals, such as SSI beneficiaries, is especially challenging even though this is where the potential for savings is greatest because more care is used.

In sum, the ability of Medicaid managed care to deliver is limited by the same features that limit the traditional Medicaid program. Medicaid has the potential to enhance the delivery of care for beneficiaries, but realizing this potential means that States must make a long-term investment of fiscal and administrative resources. Some States have made this investment, but others have not, and few have invested as much as they optimally could to support the most effective programs. When support for resources is limited, Medicaid managed care requires State policymakers to make difficult tradeoffs between competing goals: improving Medicaid access, providing care for the uninsured, serving those with special needs who depend on State-funded programs, and apportioning State funds among these and other competing interests. Making such tradeoffs inevitably leads to compromise that will limit the gains from Medicaid managed care.