Medicaid Managed Care Payment Methods and Capitation Rates

John Holahan, Suresh Rangarajan, Matthew Schirmer

Assessing the New Federalism is a multiyear Urban Institute project designed to analyze the devolution of responsibility for social programs from the federal government to the states, focusing primarily on health care, income security, employment and training programs, and social services. Researchers monitor program changes and fiscal developments. In collaboration with Child trends, the project studies changes in family well-being. The project aims to provide timely, nonpartisan information to inform public debate and to help state and local decisionmakers carry out their new responsibilities more effectively.

Key components of the project include a household survey, studies of policies in 13 states, and a database with information on all states and the District of Columbia, available at the Urban Institute's web site. This paper is one in a series of occasional papers analyzing information from these and other sources.

The nonpartisan Urban Institute publishes studies, reports, and books on timely topics worthy of public consideration. The views expressed are those of the authors and should not be attributed to the Urban Institute, its trustees, or its funders.

Note: This report is available in its entirety in the PDF format, which many users find more convenient when printing.

---

---

Introduction

Discussions between states and managed care plans about capitation rates have become increasingly contentious. In numerous markets, health plans are reportedly responding to low payment rates by either withdrawing from or limiting participation in Medicaid managed care. In particular, an increasing number of mainstream health maintenance organizations (HMOs), those with a majority of commercial and Medicare enrollees, are reported to have ended or limited participation in Medicaid managed programs in Arizona, California, Connecticut, Florida, Georgia, Illinois, Missouri, New York, New Jersey, Ohio, and Pennsylvania (Rabinovitz 1998; Meyer 1997; Langreth 1998; Kilborn 1998).

Some of these plans may have made strategic decisions to focus on commercial or Medicare-risk business or on specific geographic markets. Nonetheless, health plans' negative view of Medicaid payment rates and administrative requirements of Medicaid contracts may constrain states' continued efforts to both "mainstream" Medicaid beneficiaries into the private health care system and achieve cost savings by moving larger numbers of Medicaid beneficiaries into managed care.

This paper provides new information from an Urban Institute survey of state Medicaid managed care payment methods and rates. In particular, the paper summarizes various payment methodologies used in 41 of the nation's 45 states (including the District of Columbia) with capitated Medicaid managed care programs, and it examines interstate and intrastate variation among payment rates. We begin the paper with a brief review of the history of federal and state attempts to use payment arrangements, including capitated managed care programs, to encourage the provision of "mainstream" health care to low-income populations. We then present results from a nationwide managed care payment survey of voluntary or mandatory capitated Temporary Assistance for Needy Families/Aid to Families with Dependent Children (TANF/AFDC) and TANF/AFDC-related managed care programs. This survey was conducted as part of the Urban Institute's Assessing the New Federalism project. Concluding the paper is a comparison of Medicaid capitation rates with Medicare adjusted average per capita cost rates.

Background

In 1965, the federal government enacted Medicaid, a joint federal- and state-financed health insurance program for the poor. One of the primary goals of the program was to provide the uninsured with access to private or "mainstream" health care systems. Over the next 25 years, both program enrollment and expenditures grew well beyond original expectations. Although expanding coverage to needy groups (e.g., children and pregnant women, the disabled, and the medically needy) was a dominant goal of the program, expenditure growth and the pressure it placed on state budgets made cost containment a central objective as well.

Medicaid cost containment has taken many forms, but many states initially found cutting or keeping physician fees low an attractive option. But while states were limiting fee growth, physician fees in the private market continued to rise. This divergence of Medicaid and private fee-for-service rates led to inadequate physician participation in Medicaid programs (Holahan and Held 1985). Consequently, a growing number of Medicaid beneficiaries were left with no choice but to receive primary care through either hospital emergency rooms or physician practices that often provided low-quality care to very high volumes of Medicaid patients.

Some states responded to declining beneficiary access to private-office-based providers (physicians who also see privately insured patients in noninstitutional settings) by offering enhanced physician fees on selective office-based primary care and preventive services (Mitchell 1991; Cohen 1989, 1993; Cohen and Cunningham 1995). But despite resulting increases in office-based visit rates, physician participation in Medicaid continued to lag the private market. Moreover, these efforts to bring the poor into mainstream care were at odds with efforts to contain costs.

Other states implemented Medicaid managed care alternatives, such as primary care case management (PCCM) programs and capitated HMO financial arrangements, to both improve beneficiary access and control costs (U.S. GAO 1993). Two types of PCCM programs arose. One version paid primary care physicians a fixed fee of $3 to $6 per member per month (pmpm) to manage primary care of individual beneficiaries; the other version rewarded primary care physicians with potential bonuses from inpatient cost savings. Unlike the PCCMs, capitated HMOs assumed financial risk for inpatient and outpatient services; each plan received a fixed dollar amount pmpm for providing a comprehensive set of services. Both PCCMs and HMOs had similar effects: Inappropriate emergency room use declined, access to office-based primary care improved, and expenditures fell 5 to 15 percent below fee-for-service levels (U.S. GAO 1993; Hurley, Freund, and Paul 1993; Hurley and Freund 1991).

Over time, states have migrated toward capitated HMO alternatives as the preferred strategy to not only improve access and accountability and reduce costs, but also achieve budget predictability. Furthermore, many states chose to build upon voluntary managed care programs by enrolling beneficiaries on a mandatory basis into capitated managed care programs under 1915(b) freedom of choice or Section 1115 of the Social Security Act managed care demonstration waivers. As of June 1997, Alaska and Wyoming were the only states without some form of managed care. Moreover, as of June 1998, 17 states had implemented Section 1115 waivers, which expand the role of Medicaid managed care.⁽¹⁾

Table 1 provides data on Medicaid managed care penetration by state. The table shows that 15 percent of Medicaid beneficiaries on June 30, 1997, were enrolled in PCCM programs and 30 percent were enrolled in full-risk HMOs. The table also shows large differences in managed care penetration among states. Some states, including Alabama, Mississippi, and Texas, have very little managed care of any kind. Other states, such as Georgia and North Carolina, have high levels of PCCM but very little enrollment in capitated HMOs. Finally, states such as Arizona, Hawaii, Oregon, and Tennessee have very high levels of enrollment in capitated plans.

States have recently found the transition to capitated managed care more difficult than expected. After a wave of interest by mainstream plans in the early 1990s, commercial plans have been hesitant to enter the Medicaid market, and a number of plans have either ended or limited participation in existing Medicaid managed care programs. Many plans leaving the market have cited increased frustration with Medicaid managed care, namely, inadequate capitation rates coupled with the administrative demands of state Medicaid programs.

In response, states contend that they are limited by the "upper payment limit," which restricts their ability to pay more than fee-for-service expenditures for comparable populations. States also argue that mainstream plans were interested in the Medicaid program during the early stages of voluntary programs because rates were attractive, but plans are now not willing to accept more "realistic" rates. Specifically, states perceive rates in the early voluntary programs to have been unnecessarily high. Rates were high because managed care plans initially attracted healthier beneficiaries, who incurred less than the average level of health care expenditures on which rates were based. As voluntary and mandatory programs have expanded, opportunities for risk selection have been diluted by additional enrollment of Medicaid beneficiaries who are less "healthy" into managed care.

On the other hand, mainstream plans see these more "realistic" rates as being (1) a result of state budgetary pressures, not the market, and (2) inadequate to both care for Medicaid beneficiaries and meet the state's administrative and quality reporting requirements. Moreover, the number of beneficiaries a plan expects to enroll has become intertwined with rate discussions. Plan enrollment has lagged behind expectations because of the decline in Medicaid rolls, which is caused by both the nation's economic recovery and the implementation of state and federal welfare reforms. Lower-than-expected enrollment has led plans to question the decision to enter or remain in the Medicaid market. Mainstream plans have also argued that state efforts to protect safety-net-based Medicaid-only plans (i.e., through preferences in autoassignment or higher rates) have created an unlevel playing field for Medicaid managed care.

With the limited amount of hard information available, it is difficult to sort out these competing arguments. Although a number of factors may influence mainstream plan participation, payment rates will probably continue to be a major decision factor for plans and a key policy lever for states during the development of Medicaid managed care programs.⁽²⁾ Without an increased understanding of alternative payment methodologies and payment rates, it is increasingly likely that capitation rates will reflect traditionally low Medicaid fee-for-service payments. Further, payment methodologies are not likely to effectively compensate plans for different levels of risk. As a result, mainstream plan participation will continue to wane, and a growing number of Medicaid beneficiaries will receive care from safety-net-based and Medicaid-only HMOs. This is not to say that safety-net-based and private Medicaid-only HMOs will not provide good quality care. Rather, they will participate not because of rate adequacy but because of their inability to leave Medicaid managed care markets. This paper provides information that will help states establish rates that will attract efficient mainstream plans as well as provide Medicaid-only plans with the resources necessary to care for a Medicaid population. At a minimum, it will help states understand how others are addressing the same issues.

Survey Design

To collect this information, we mailed surveys to Medicaid directors in all 50 states and the District of Columbia and asked only those with capitated HMO programs to respond.⁽³⁾ The survey was designed to elicit only information that was not available from secondary sources. It was also limited to capitated HMO programs for AFDC and AFDC-related eligibility groups (i.e., low-income unemployed parents, poverty-related children, and poverty-related pregnant women). Children's Health Insurance Program expansions, the elderly and disabled, and the medically needy were not included. The survey was mailed to states in March 1998. It asked questions about rate-setting methods, adjustments for disproportionate share hospital payments and graduate medical education, and services carved out from capitation. Each state was also asked to submit a rate sheet that listed, for the most recent contracting period, either fixed capitation rates or average accepted bids for state-specific eligibility categories across regions and pmpm equivalent values for carved-out services. We also asked questions related to add-on payments and adjustments for maternity care and deliveries, safety net providers, high-risk cases, and state administrative savings, as well as rate updating processes and methodologies. If written responses were unclear or additional information was needed, we telephoned the appropriate state contact for clarification.

Out of a possible 45 states, including the District of Columbia, with capitated Medicaid managed care, 41 responded.⁽⁴⁾ Delaware, Oregon, Montana, and Nebraska declined to participate. In aggregate, the participating states accounted for approximately 94 percent of the nation's Medicaid beneficiaries and 96 percent of Medicaid managed care beneficiaries enrolled in capitated managed care programs. As table 2 shows, 22 of the 41 states enrolled beneficiaries on a mandatory basis only; of these, 9 permit beneficiaries a choice between HMOs and PCCM programs. Fourteen states had mandatory programs operating in some areas of the state and voluntary programs in others. The remaining five states had only voluntary capitated programs. All states provided information on payment methods. Thirty-six states provided actual payment rates. (Alabama, Michigan, North Carolina, Pennsylvania, and Vermont provided information on methods but not rates.)

Payment Methods

Administrative Pricing, Negotiation, and Competitive Bidding

Methods for setting rates are summarized in table 3. About half (21) of the respondent states continue to use administered pricing—setting rates that are open to all plans willing to participate. This approach is analogous to the Medicaid fee schedules that states traditionally use to pay physicians who are willing to accept Medicaid-set rates (managed care plans, of course, must also meet a variety of standards in order to participate). These states typically use data from their fee-for-service experience to establish rates. Depending on state policy objectives, states apply some discount from fee-for-service, typically 5 to 15 percent, and then adjust rates on an annual basis. At least one state, Rhode Island, has found competitive bidding to be administratively burdensome and plans to return to administered pricing. On the other hand, a number of the administered rate-setting states plan to (1) implement rate negotiation or competitive bidding as a means of achieving immediate savings and (2) recalibrate state Medicaid managed care rates in the absence of comparable fee-for-service data as more beneficiaries enroll in managed care.

Of the remaining states, 6 negotiate rates with plans individually, while an increasing number of states (14) have moved toward either competitive bidding or competitive bidding with final rate negotiations to procure plans. In general, even the states with competitive bidding end up with a considerable amount of negotiation. Some states, such as New York and Washington, have found that their efforts to arrive at rates through competitive bidding were considerably altered through subsequent negotiations. States also vary in the length (years) of the contracts, the amount of information provided to plans before bidding, and the strictness with which target ranges are adhered to. Some examples of state approaches to negotiation and competitive bidding follow.

Massachusetts is a state that negotiates rates with each plan. Before negotiating rates, the state projects the enrollment distribution by geographic region of the state, indicating the enrollment that the plan would expect under Medicaid managed care during the next contract period. The state then negotiates a single rate for each plan. Plans have the option to include or exclude prescription drugs from the capitation rate. All plans that participate in the state employees' insurance program must submit a "good faith" proposal to contract for the state's Medicaid population.

Minnesota essentially sets rates but only after a round of negotiation. Minnesota first asks plans to submit rate and geographic coverage proposals. After examining all the proposals and discussing rates and geographic coverage with the plans, the state establishes a set of rates that apply to all plans within each managed care region. As a condition of HMO licensure, all Minnesota health plans must participate in Medicaid managed care.

States that use competitive bidding also differ in important ways. Arizona currently competitively bids and negotiates its rates under five-year contracts that include interim adjustments for inflation and programmatic and legislative changes. The state contracted with outside actuaries to develop a set of rate ranges as a frame of reference for the process. The state did not disclose these ranges to plans but compared them against submitted bids as a starting point for negotiations. Rate ranges were developed using (1) state expenditure and utilization data from the previous two and a half years and (2) various trending facts that include changes in utilization and service cost, and programmatic or legislated changes. Rates were also adjusted downward for expected cost of reinsurance and third-party recoveries and adjusted upward for administrative costs, expected profits, and other contingencies.

Hawaii competitively bids contracts every three years. The state does not automatically adjust rates every year; in fact, in the past rates have remained the same or have been lowered. During the last process, the state allowed plans that passed the technical review to visit the state's actuary and discuss assumptions used to prepare its bid. Plans were then given another opportunity to make final adjustments to their submitted bids. However, the state indicated that it will disclose rate ranges during the next bid process and will require plans to submit bids within the range. Individual meetings with plans will not be held, and bids outside the range will be considered nonresponsive.

The state of New York competitively bid rates across selected counties in 1996 with the assumption that it could achieve significant savings relative to fee-for-service expenditures. Before bidding, the state set a range of plus-or-minus 5 percent of a discounted fee-for-service midpoint for six age/gender groups for each of nine geographic regions (groupings of counties). Plans that bid below the range for a group were brought up to the lower payment limit, and plans that bid within the range were contracted at bids. For plans that bid above the range, the state set their rates at the lower payment limits as well. Nearly two-thirds of the plans had rates set at the lower payment limit. Soon after the 1996 process, plans claimed that rates were too low to cover eligible populations. In response, the state commissioned an independent study by Arthur Andersen, Inc., which concluded that rates were not adequate and recommended a 3 to 5 percent increase in capitation payments.

Since the 1996 study, the legislature has raised rates three times. The state raised rates 2 percent in New York City and up to 7 percent in upstate counties beginning in January 1997. Then the state approved two additional rate increases in August 1997. Rates to HMOs and prepaid health service plans for the period January 1, 1997, through March 31, 1998, were retroactively raised 5 to 11 percent. Approximately $20 million in rate increases was split between general increases to all plans ($10 million) and earmarked increases for those plans in need of additional managed care capacity. Using similar methodology, the state scheduled a raise in rates once again in April 1998 by 3 to 4 percent. In addition, each plan received a trend adjustment of approximately 3 percent to account for medical cost inflation during the January 1997 through March 1998 contracting period. Together, these rate adjustments could have potentially increased plans' rates from 13 to 25 percent since the 1996 bid process. Because of its unhappiness with this experience, New York intends to negotiate rates with individual plans in the future.

Finally, the state of Washington competitively bids contracts every three years, with interim adjustments for changes in service coverage, budgeted utilization trends, and inflation. During the last process, the majority of plans bid above the state's undisclosed upper payment limit. As a result, the state had to receive legislative approval for an increase in rates to ensure adequate participation.

Capitation Rate Adjustments

Most states establish different rates for separate demographic groupings. As shown in table 4, these vary from as few as 2 (Kentucky) to as many as 25 (Mississippi). States often make additional adjustments for selected health conditions, such as pregnancy or HIV/AIDS, which are discussed later. Most states also vary rates by region. In general, rate adjustments made solely to directly reflect differences in health status are not very sophisticated. However, five states—Colorado, Maryland, Minnesota, Michigan, and Washington—are moving ahead in designing diagnosis-based risk-adjustment payment systems for Medicaid populations (State Health Watch 1997). These systems are intended to (1) pay plans enough to ensure access for high-risk patients, including allowing them to contract with providers that specialize in high-risk Medicaid patients; (2) protect plans from adverse selection; and (3) make health plans compete on efficiency and quality and not just on the ability to select good risks.

Thus far, only Colorado and Maryland (of the responding states) have begun to implement more sophisticated risk-adjustment systems for their Medicaid populations. Both states perform risk assessment (measuring the predictors of health care costs of the individuals enrolled in a plan) and risk adjustment (compensating plans for differences in risk, as measured by risk assessment). Colorado relies on a disability payment system (DPS), which identifies and groups diagnoses that are chronic in nature and are associated with higher future costs. Approximately 2,400 chronic diagnoses were grouped into 43 major categories that correspond to body systems or specific types of illness or disability. Using two years of data, the state measured the annual cost of the average person in each of three eligibility categories of aid (disabled, AFDC adults, and AFDC children) during the second year. The state then measured the extent to which being in a particular DPS diagnosis group in the first year contributed to the person's health care expenditures in the second year. The correlation between diagnoses groups and second-year costs were then used to determine risk weights attached to each diagnosis group and to persons without any of the diagnoses. The relative risk weights and the frequency with which an HMO's (or fee-for-service's) enrollees fell into each diagnostic group determined an average case-mix factor for each HMO and for the fee-for-service population. This calculation was performed separately for each category of aid. Finally, the case-mix factors for each HMO were multiplied by a standardized baseline rate, which was lower for HMOs than for fee-for-service, to determine final rates for each HMO.⁵

Similarly, Maryland has also implemented a more sophisticated risk adjustment for its Medicaid populations, including AFDC and AFDC-related beneficiaries enrolled in managed care.⁶ The state relies on the Johns Hopkins Adjusted Clinical Groups' (ACGs, formerly Ambulatory Care Groups) diagnosis classification system to risk-adjust demographic rate cells. Under ACGs, enrollees are assigned to 1 of 52 unique morbidity groups based on their age, sex, and the inpatient and ambulatory ICD9-coded diagnoses observed from their previous medical claims. The state then uses aggregate ACG data to define a limited number of risk-adjustment categories (RACs) that reflect levels of relatively homogenous resource utilization and reduce the number of rate cells to a more manageable number for the state. Each individual is then assigned to one of nine RACs based on his or her original ACG. Individuals who do not have sufficient diagnostic data are assigned to one of eleven standard demographic rate cells—covering approximately 60 percent of AFDC enrollees.

Additional Adjustments and Supplemental Payments

Outside of limited risk adjustment, states typically base administrative prices, upper payment limits, or rate ranges for each demographic category on historical fee-for-service experience. Adjustments are made for plan benefit design, Federally Qualified Health Center (FQHC) adjustments, prescription drug rebates, expected changes in unit costs and utilization trends (managed care savings), inflation, and legislative changes.⁷ A number of states (Arizona, Georgia, New Jersey, New York, New Hampshire, Missouri, and South Carolina) also take into account the expected administrative savings of 1 to 4 percent to the state under managed care through administrative allowances by adjusting rates upward. In addition, some states make adjustments for expected favorable risk selection under voluntary enrollment. Georgia made a downward 5 percent adjustment to baseline capitation rates to reflect positive health status of voluntary enrollees. Likewise, South Carolina made a downward 3 percent adjustment to payment rates for AFDC and poverty-related pregnant women.

Seven states made provisions for FQHC or Indian Health Service (IHS) contracts. States either include 100 percent of FQHC costs into capitation calculations or paid plans with FQHC or IHS contracts an additional capitation payment. Five states (Alabama, Florida, New Mexico, New York, and Texas) reported that they pay 100 percent of FQHC costs (that is, no reduction is assessed for managed care savings) in setting capitation rates. Texas estimated that the impact of including 100 percent of FQHC payments in capitation payments results in a $.10 to $3.00 increase in pmpm payments across the different service areas. Connecticut pays a "bump" payment to plans that have contracts with FQHCs. Colorado pays two "bump" rates, one for hospital-based FQHCs and one to free-standing FQHCs. The state estimated that plans that contract with hospital-based FQHCs receive an additional $11.89 pmpm, whereas plans that contract with free-standing FQHCs receive an additional $19.20 pmpm. In Connecticut, these payments range from $1 to $15 pmpm depending on the level of FQHC utilization by plan. Likewise, California's two-plan model pays local initiatives that have contracts with public providers (including FQHCs) a higher capitation rate than it does to mainstream plans. Capitation rates for mainstream plans are approximately 4 to 5 percent lower than for the local initiative.

More than half of our respondent states make special payment arrangements for maternity-related expenses, rather than incorporating these costs into base capitation rates already paid to a plan for AFDC enrollees. The majority of these states make direct lump-sum payments to health plans for these expenses in addition to paying regular capitation rates. Lump-sum payments are typically used to reimburse plans for prenatal, delivery, and postpartum care costs. Colorado was one exception: The state pays plans a lump-sum payment to cover facility costs. Other states pay separate capitation rates for all pregnant women or only poverty-related pregnant women (PRW). California pays a separate PRW rate that varies for each of the counties in the two-plan system. Missouri combines both approaches: The state pays a lump-sum delivery payment for all deliveries and pays a separate PRW rate. Similarly, Texas currently pays a separate rate for PRWs, but plans to move to base rates for AFDC women and PRWs, plus a supplemental lump-sum payment that covers prenatal care, delivery, and postpartum services.

Some states preferred to compensate plans for maternity expenses by transferring a portion of these expenses into newborn rate cells or through risk-sharing arrangements. Mississippi and Texas pay a substantially higher rate for poverty-related newborns to protect plans from these pregnancy-related costs. Similarly, Georgia transferred 35 percent of AFDC newborn birth charges to AFDC males and females one to two months old. Two states reported the use of risk-pooling or risk-sharing mechanisms to protect plans from above-average pregnancy costs. Utah retained a portion of newborn capitation payments to better allocate reimbursement among plans. These capitation deductions are accumulated into a pool and distributed based on an HMO's delivery experience. Rhode Island addressed plans' concerns over maternity risk by offering plans a reinsurance arrangement on neonatal intensive care services in return for a small deduction in capitation. Table 5 summarizes maternity care payment policies among surveyed states.

A few states have also begun to make similar adjustments or create reinsurance mechanisms for HIV/AIDS cases. Maryland treats HIV/AIDS patients as a separate category when setting capitation rates. Massachusetts has a separate rating category for high-cost members with HIV/AIDS for managed care organizations that ask and secure approval for this rating category. Utah makes supplemental monthly payments to plans for each HIV/AIDS patient. New York defrayed some of the risk plans face for symptomatic AIDS patients through a regional stop-loss arrangement. The state estimated that on average approximately 0.1 percent of enrollees or 20 percent of HIV/AIDS enrollees in a region are AIDS symptomatic. Under the arrangement, if a plan's symptomatic AIDS membership is greater than 0.1 percent of its total Medicaid enrollment, the state will pay the plan a 1 to 5 percent capitation increase—depending on the level of AIDS enrollment—across all of the plan's Medicaid patients in that region.

Pennsylvania is using both an HIV/AIDS adjustment and a risk pool in different geographic areas. For the southeast area, plans were asked to submit six months' worth of utilization data and unit cost information by service category (e.g., hospital inpatient, outpatient) for enrollees with and without HIV/AIDS. If the plan's enrollment of HIV/AIDS members was 0.2 percent more than the regional prevalence rate for each eligibility category, the state reimbursed the plan for 90 percent of the cost difference between its HIV/AIDS patients and non-HIV/AIDS patients. For the southwest region, the state has created a quarterly HIV/AIDS risk pool that is financed through a withhold from baseline capitation rates. After each quarter, each plan is required to submit information on its HIV/AIDS enrollees. The state then disburses funds from the pool to each plan based on the plan's share of the HIV/AIDS patients that are enrolled in Medicaid managed care within the region.

Disproportionate Share Payments and Graduate Medical Education

Finally, most states exclude disproportionate share hospital (DSH) payments from Medicaid capitation rates. However, Illinois, Minnesota, West Virginia, and Wisconsin reported that DSH payments were included in the rates. Hospitals have claimed that HMOs do not pass through DSH in the form of higher payment rates. Moreover, the 1997 Balanced Budget Amendment includes provisions that require states to make Medicaid DSH payments directly to hospitals rather than to managed care entities, except for "payment arrangements in effect" on July 1, 1997.

States have also moved forward in excluding both direct medical education and indirect medical education components of graduate medical education (GME) payments.⁸ Only 4 of the 41 surveyed states included DSH payments in HMO capitation rates, and 10 states included GME (see table 6). In these cases, DSH or GME payments were incorporated into capitation rates because they were included in the fee-for-service baseline data used to set rates, upper payment limits, or acceptable bid ranges. However, states varied in their reasons for including these payments. The most common reason for including these payments was that states did not have or were developing the data capability to separate GME or DSH amounts from historical fee-for-service data. Even states with proven data capability chose not to exclude DSH or GME from capitation rates. Massachusetts and New Jersey did not exclude GME payments, which they estimated at approximately $5 and $6 pmpm, respectively. Illinois estimated its DSH inclusion to be approximately $4.59 pmpm. Minnesota estimated that DSH and GME payments, make up 6 percent of Hennepin and 3 percent of Ramsey County rates. It also reported that it made plan-specific adjustments to capitation rates based on plan utilization of DSH and GME hospitals. Washington included 50 percent of DSH and GME payments, with the remaining funds flowing directly to Harborview and the University of Washington Medical Centers.

Carve-Out Services

States carve out services from HMO capitation arrangements for numerous reasons. In addition to potential improvements in quality and access, states may pursue carve-outs to reduce expenditures, limit selection bias, or to protect public or local providers. Often plans do not have experience in managing these services. States may also achieve volume discounts on carve-out services through specialized contractors. States may use behavioral health carve-outs to reduce the incentive of plans to disenroll high-risk beneficiaries as well as to protect plans from the costs of adverse selection of high-risk beneficiaries (Frank et al. 1997). Carve-outs may protect public providers, including mental and public health departments that have traditionally served the Medicaid population, as well as reduce intergovernmental or interagency tensions over Medicaid budgets (Bachman et al. 1997). Likewise, states may use pharmacy and chiropractic carve-outs to protect local allied professionals. According to our survey, the most common carved-out services among states (not including long-term care, which all states carve out) were mental health, substance abuse, dental, and pharmacy (see table 7).

Behavioral health

Of the 41 states, 23 fully or partially carved out mental health services, and 20 states fully or partially carved out substance abuse services. However, states varied in their behavioral health carve-out designs. Some states carve out mental health services to other public agencies, whereas others carve out mental health services to private organizations. Some address these issues with partial carve-out arrangements that require some plans to bear some costs for behavioral health services but establish arrangements to limit plan risks.

In California, responsibility for mental health services for Medicaid beneficiaries covered by the state's two-plan model has been contracted to separate county organizations through a 1915(b) mental health waiver, rather than assigned to two-plan model HMOs. Under the mental health waiver, the Medicaid agency transfers a portion of general revenues to the State Department of Mental Health to cover costs for mental health care previously provided through the Medicaid agency. The State Department of Mental Health then allocates these funds to individual counties, following an allocation formula based on historical utilization of Medicaid mental health services, the number of Medicaid beneficiaries, and other relevant factors. Counties then pool these funds with a portion of proceeds from motor vehicle and sales taxes that are earmarked for mental health services and federal matching payments to provide services to Medicaid beneficiaries. The county organization is free to contract with independent providers on a fee-for-service or capitated basis.

Washington has transferred funding for mental health to the State Department of Mental Health, which pays regional support networks a capitation amount to provide mental health services to Medicaid beneficiaries. Each network may contract with independent providers on either a capitated or fee-for-service basis. In addition to mental health, substance abuse treatment is also carved out, to the Division of Alcohol and Substance Abuse. Kentucky carved out mental health services from managed care organizations and continued delivering services primarily through its network of community mental health centers, Compcare, on a fee-for-service basis. However, the state plans to implement a 1915(b) waiver program that will pay one managed behavioral health organization on a capitated basis in the two managed care regions currently under mandatory enrollment beginning January 1999.

Other states carve out behavioral health services to private organizations that specialize in these services. Tennessee and Pennsylvania carved out mental and substance abuse health services to separate behavioral health contractors. Maryland carved out all mental health services to Maryland Health Partners, an administrative service organization that manages care rendered by public and private mental health providers. Colorado contracts with mental health assessment and service agencies that employ different models in seven different geographic areas.

In some states, health plans are required or allowed to provide a portion of behavioral health services. Michigan requires plans to cover 20 outpatient visits per year. In Georgia, plans must cover mental health-related inpatient stays of up to 30 days. A number of states carve out only community-based mental health or substance abuse services. Florida, New Hampshire, Ohio, and Virginia carved out community-based mental health services to fee-for-service. Ohio and New Hampshire carved out community-based substance abuse services as well.

Finally, a number of states chose not to fully carve out behavior health services but rather provide plans with separate stop-loss arrangements for mental health and substance abuse coverage. New York limits mental health coverage under managed care to 20 outpatient visits or 30 inpatient days (inpatient days are combined with substance abuse) a year. The state also limits substance abuse coverage under capitation to 60 outpatient visits or 30 inpatient days (inpatient days are combined with mental health) per year. Beyond these limits, plans must provide care and bill the state on a fee-for-service basis. Eventually the state wants to develop special needs plans that would provide services on a capitated basis for the chronically mentally ill. Likewise, Rhode Island has separate stop-loss provisions for mental health and substance abuse. The state limits plan coverage under capitation of both mental health and substance abuse services to 30 inpatient days or 30 outpatient visits per year through two stop-loss arrangements. If a plan exceeds these limits for an enrollee, the plan will be reimbursed at 90 percent of fee-for-service or 90 percent of the actual cost to the plan, whichever is less, from the state.

Dental

Twenty-four states carved out dental care services from HMO capitation. Most pay for dental care on a fee-for-service basis. Florida, North Carolina, and Wisconsin gave plans the option to provide dental services for higher capitation rates. Similarly, in New York, local service districts that are responsible for contracting with plans may capitate or carve out dental care at their discretion. This flexibility in contracting may reflect a low level of participating dentists in managed care. Two states separated carve-outs for adults versus children. Mississippi carved out adult dental services, whereas West Virginia carved out only children's dental care.

Pharmacy

Twelve states partially or fully carved out pharmacy services. Massachusetts gives plans the option to provide pharmacy services. As a result, only three of the state's eight contracted HMOs are capitated for pharmacy services. Moreover, the extent of pharmacy carve-outs varies across states. Kansas carved out only psychiatric, family planning, and hemophiliac prescriptions. Washington directly pays on a fee-for-service basis for prescriptions from health departments, family planning clinics, regional support networks, community mental health clinics, and providers of voluntary pregnancy termination.

Organ transplants

Six states reported that they carve out selected organ transplants. Alabama includes renal and corneal transplants in HMO capitation but carves out bone marrow, liver, and heart transplants to fee-for-service. New Jersey carves out heart, heart/lung, cornea, and bone marrow transplants. California carves out heart, heart/lung, bone marrow, and liver transplants from capitation payments. The state also protects plans from the higher incidence of comorbidity costs by requiring plans to disenroll transplant patients from managed care to Medicaid fee-for-service. Utah does not carve out organ transplants but has a stop-loss arrangement with plans, which are responsible for the first $40,000 of all approved transplants and then 80 percent of any amount above $40,000.

Children with special needs

The survey asked about carve-outs for services provided to children with special needs. Only five states indicated they carved out these services. On the basis of subsequent discussions with state officials and other outside experts, we concluded that children with special needs are treated in different ways by different states. In some, these children are covered under Title V programs. In others states, they are kept out of managed care plans. In still others they must enroll in managed care but can choose between an HMO and a PCCM plan. And in others, children with special needs may be included in managed care plans without a carve-out or a rate adjustment. Unfortunately, we did not obtain information on which states fall into the different categories. On the basis of comments we received from reviewers, California seems to be a special case that carves out a substantial amount of money for services provided by California Children's Services—local agencies that provide services to children with special needs.

Vision

Six states partially or fully carved out vision services. Vermont and Washington carved out only eyeglass prescriptions. Vermont provided eyewear through a sole-source contract. Washington carves out the costs of eyeglass frames, lenses, and fabrication services to the Department of Correction, which is reimbursed on a fee-for-service basis.

Stop-Loss Arrangements

Stop-loss, often called reinsurance, is a mechanism used by states to reimburse Medicaid managed care plans for certain benefits that exceed a specified expense limit. These payments will not be included in capitation rates but rather will be paid separately. The specific provisions of these arrangements or programs vary widely across states. Several surveyed states provided detailed descriptions of their stop-loss arrangements, which shed some light on these variations. Rhode Island's stop-loss arrangement, for example, applies to a wide range of benefits, including outpatient and inpatient mental health care, outpatient and inpatient substance abuse care, and long-term care. The state specifies the benefits for which plans are responsible under capitation payments. After the limit has been reached, Rhode Island pays 90 percent of the Medicaid rate or 90 percent of the actual cost to the plan, whichever is less, for all additional services used. The state estimates that this provision results in additional payments of $2.61 pmpm, an amount not included in the rate information provided below.

Arizona has essentially developed two separate stop-loss arrangements. The first applies to all inpatient services and covers 75 percent of expenses incurred after a specified deductible limit is met. This deductible is determined by plan size. For example, in a plan with 15,000 beneficiaries, the plan is responsible for 75 percent of costs after a deductible of $20,000 per enrollee is exceeded. If plan enrollment is greater than 50,000, this deductible more than doubles to $50,000 per enrollee. Arizona's second stop-loss arrangement, what they called "catastrophic reinsurance," applies only to beneficiaries with hemophilia and certain organ transplants. For organ transplants, the plan is responsible for 85 percent of the rate or 85 percent of the contractor-paid amount, whichever is less. For beneficiaries with hemophilia, the plan pays for all medically necessary services at 85 percent of service rates. In both cases, there is no deductible because the entire service or range of services is subject to the stop-loss arrangement. The state estimates that both of these arrangements result in additional payments of between $.21 to over $22.00 pmpm, depending on the age and region of the enrollee, as well as plan size.

Utah's reinsurance program applies to all covered services. In this case, plans are responsible for all expenses for the first $40,000 per enrollee. After this, the state becomes responsible for 80 percent of all additional expenses. The state estimates that this provision results in additional payments of an average $10.40 pmpm.

Several other states (California, Hawaii, Iowa, and Missouri) indicated that they have operational reinsurance programs. However, they did not provide us with any details of these provisions.

Rate Standardization Methodology

Thirty-six states provided us with data on capitation rates. Table 8 shows the variation among states in the rate data originally submitted to us. Unfortunately, these data do not lend themselves to ready comparisons across states. First, states establish very different age-sex categories and vary in the extent to which they use regional adjustments. For example, table 8 shows that Mississippi uses a very large number of rate cells, while Kentucky establishes two rates for each of eight regions. Colorado has two sets of rates—one for Denver and one for the rest of the state—while New Jersey has separate rates for 21 different regions. Second, the cost of maternity care is sometimes included in the capitation rates and sometimes is paid separately. For example, table 8 shows a relatively low rate in Colorado for any AFDC adult because there is a separate lump-sum payment for pregnancy expenses. Third, most states exclude DSH payments and the cost of GME from the rates, but some do not. Finally, the number of carved-out services varies among the states.

In order to compare collected rate information across and within states, survey data were standardized for variations in each of the areas mentioned above. We did not make any other adjustments, including those for changes to rates that occurred subsequent to the initiation of contracts, presence of or variation in stop-loss arrangements, state AFDC eligibility criteria (which can affect the composition of the population), unreported differences in carve-outs across states, and selection bias under capitated programs with voluntary enrollment.

Standardizing Rate Cells across All States

To compare rates across states, we chose to aggregate state-specific rate information for each year of age, sex, and region into the following set of standardized rate cells:

• Infants up to 1 year old.
• Males and females ages 1 through 13.
• Females ages 14 through 34.
• Males ages 14 through 44.
• Females ages 35 through 44.
• Males and females ages 44 through 65.

To convert a state's rate information into the above-mentioned standardized rate cells, we first used 1990 U.S. Census data on the population below 200 percent of the federal poverty level (FPL) to disaggregate each state's population by year of age, sex, and region (males less than 1 year old in region A, males less than 1 year old in region B, females less than 1 year old in region A, females less than 1 year old in region B, males 1 to 2 years old in region A, females 1 to 2 years old in region A, etc.). We then used each state's rate data to attribute a rate to each cell (e.g., the rate for children ages 1 to 5 was attributed to males age 1, and so forth). Eleven states paid separate rates for PRW or poverty-related children. For nine of the states, we used the state's own data to estimate the percentages of AFDC and PRW or PRC on Medicaid and created a blended rate for each year of age, sex, and region. For the two states that did not provide enrollment data, we used an average of the other states for women and Health Care Financing Administration 2082 data for children.

With these blended rates for each age/sex cell, we aggregated, using population counts as weights, across cells to calculate rates for each of our six groups for each of the states' regions. As an example, assume a state provided the following data. The state has three separate rates: $300 for infants and children up to age 1, $100 for children ages 2 through 5, and $75 for children ages 6 to 13. Because of our rate categories, we want to establish average rates for children up to 1 year old and for those ages 1 to 13. In this example, we would use $300 as the rate for infants. To establish an average rate for children 1 to 13, we use 1990 Census data to establish population weights. For the example in figure 1, we would calculate $92.74 to be the rate for children ages 1 to 13. In reality, the state is paying more for children age 1 and less for children ages 6 to 13, but on average it is paying $92.74 for children ages 1 to 13. To calculate a statewide average rate for a particular standardized category, we used 1990 Census information on the population of below 200 percent FPL to estimate the distribution of AFDC and related populations across a state's regions.⁹

Standardizing Treatment of Maternity Expenses across States

As mentioned earlier, more than half of the states relied on special treatment of maternity expenses for AFDC or PRW. Fifteen states paid separate lump-sum payments directly to plans to cover maternity expenses. Eleven states paid separate rates for PRW only. Two states paid a separate rate for all pregnant women, AFDC or PRW; four states transferred a portion of maternity costs of AFDC pregnant women or PRW into newborn rates. Thus, in order to compare newborns and females in childbearing years (ages 14 to 34 and 35 to 44) across states, we had to make adjustments to incorporate these special maternity payments into standardized rate cells.

Four types of adjustments were made. First, to incorporate lump-sum payments into Medicaid capitation rates for females ages 14 to 34 and females ages 35 to 45, we converted the lump-sum payment to a monthly basis and applied these payments to the percentage of pregnant women within each of the two age groups. We used a two-year average of 1994 and 1995 Current Population Survey (CPS 1995, 1996) data to estimate the percentage of pregnant women among females ages 14 to 34 and 35 to 45 enrolled in Medicaid. On the basis of national 1995/1996 CPS data (sample sizes were too small for CPS state data), we estimated that 24 percent of females ages 14 to 34 and 4 percent of females ages 35 to 44 enrolled in Medicaid across the nation were pregnant and delivered during the year. Accordingly, we then assumed that plans received lump-sum payments for 24 percent of female enrollees ages 14 to 34 (and 4 percent of those ages 35 to 44) during the year and then made the appropriate adjustments to monthly capitation rates.¹⁰

Second, to combine separate PRW rates with AFDC rates for females ages 14 to 34 and 35 to 44, we had to estimate the number of women in Medicaid who were enrolled through AFDC or poverty-related criteria. Nine of the eleven states with separate rates provided program enrollment information. In the two states that did not provide enrollment data, we used the average of the number of AFDC and PRW from the nine states that did provide this data. We then used 1995/1996 CPS state data to estimate the percentages of Medicaid women who were ages 14 to 34 and 35 to 44 as a proxy for their enrollment distribution in Medicaid managed care.

Third, in cases where a state paid a separate rate for all pregnant women (AFDC and PRW), we used the CPS data described above to estimate the percentage of women ages 14 to 34 and ages 35 to 44 on Medicaid who are pregnant in a given year and created blended rates.

Fourth, for states that incorporated pregnancy expenses in newborn rates, we had to first estimate the amount of excess capitation (i.e., that which resulted from pregnancies) in those rates and then spread those funds across the female ages 14 to 34 and 35 to 44 rate cells. In all three cases reported in the survey, the state provided separate AFDC and poverty-related newborn rates but incorporated expenses into only one of the two rates. As a result, we used the rate that did not include maternity expenses as the best estimate of the portion of capitation related solely to newborns. Once the excess capitation amount was estimated, we applied it to the female ages 14 to 34 and 35 to 44 rate cells based on national 1995/1996 CPS estimates of the percentage of pregnant women on Medicaid in each of these age groups. The impact of maternity adjustments is summarized in table 9.

Standardizing DSH and GME Exclusions

The majority of survey states excluded both DSH and GME payments from capitation rates. However, four states included DSH payments in capitation rates, and nine states included GME payments. If a state included DSH payments and did not provide a pmpm equivalent for DSH amounts incorporated in rates,¹¹ we calculated the ratio of 1995 Medicaid spending for adults and children to 1995 Medicaid spending for the disabled and multiplied this ratio by 1995 DSH spending to calculate the "share" of DSH spending attributable to adults and children (Urban Institute and Kaiser Commission 1997). This amount, less a 5 percent adjustment for managed care savings, was then deducted from standardized capitation rates.

If a state included GME payments and did not provide a pmpm equivalent for GME amounts incorporated in capitated rates,¹² we first estimated a state's GME spending based on survey data on GME payments as a percentage of inpatient-care spending (Plumb and Henderson 1995). If a state did not respond to the survey, we used the average of respondents. We then converted aggregate GME spending for a state into a capitated equivalent based on the percentage of inpatient Medicaid spending attributable to AFDC adults and children (Urban Institute and Kaiser Commission 1997). This amount was deducted from standardized capitation rates as well. Table 10 summarizes DSH and GME adjustments.

Standardizing Benefit Packages across States

In order to make comparisons across states, it is necessary to standardize for differences in benefit packages. States differ in the coverage of certain services (e.g., dental and vision services). Some states include services such as mental health and substance abuse in their capitation rates, but a large number of states exclude them (i.e., carve them out). Our objective was to make benefit packages as comparable as possible in order to facilitate useful comparisons. We relied on assistance from professional actuaries—the Actuarial Research Corporation (ARC)—to make estimates for six services: (1) mental health, (2) substance abuse, (3) dental, (4) vision, (5) prescription drugs, and (6) transplants. In general, the remaining services that states sometimes carve out are relatively minor in terms of expenditures.

ARC used data on mental health and substance abuse expenditures from proprietary databases from a private-employer-covered population that had access to a generous benefit package, with downward adjustments for estimates made for managed care use. These data were used to calculate the amount spent on mental health and substance abuse services as a percentage of total spending. Comparisons with submitted Medicaid managed care data showed that although the populations are very different, the proportion of charges attributable to mental health was comparable, whereas the proportion attributable to substance abuse was lower for the private population.¹³ Data on dental and prescription drug expenditures were obtained from the Medicaid 2082 data for AFDC child and adult populations. Spending on dental care and prescription drugs as a percentage of a total were calculated. States with no data and those whose information was regarded as unreliable were omitted. Finally, data on vision and transplant services were obtained from another ARC proprietary database from private HMO plans. Again, spending on these services as a percentage of the total was calculated. Comparisons with submitted Medicaid managed care data showed that the proportions of changes attributable to vision were slightly higher for private participants.

ARC used data on the capitation rates for each of the six rate cells described above after adjustments were made for maternity expenses, DSH payments, and GME. Different adjustments were made if the state provided the monthly amount for the carved-out services in the survey but did not provide a basis for distributing them across the six rate cells. Another type of adjustment was made if a state informed us that the service was carved out but provided no dollar estimate.

In the first case, when the state provides the dollar amount of the carved-out services, ARC used the appropriate donor database (Medicaid 2082 or private market data) to estimate the relative use of the service by the populations in each of the six cells. For example, children ages 1 through 13 were found to have expenditures for mental health services that were 83 percent of those of the average person in the database, whereas females ages 35 through 44 had expenditures 135 percent of the average. ARC then used these relatives to distribute mental health services to each of the six cells. For example, if the state estimated that the mental health carve-out was $10 per person, the amount attributed to children ages 1 through 13 would be $8.30, or 83 percent of $10. Similar information on spending from each of the donor databases was used to distribute the pmpm estimates of the carve-out for each of the five remaining services across each of the six rate cells.

If the state did not provide us with carve-out amounts, ARC first used the donor databases to estimate spending on the specific service relative to total spending for adults (older than 14 years of age) and children (0 to 13 years). ARC then applied the resulting percentage to the capitation rate. For example, if mental health services amounted to 5 percent of total spending for adults in the donor database, ARC estimated that the mental health carve-out for adults would be equal to 5 percent of the capitation rate. Finally, ARC used relative spending ratios among adult and children groups to distribute the carve-out amounts to each of the four adult and two child rate cells.

There were a number of other special cases, for example, where states carved out only part of the service, typically for services over $1,000 per member per year or services beyond a certain number of visits or inpatient days, and so forth. ARC used the donor databases in various ways to make special adjustments. These are discussed in the appendix.

Table 11 provides our estimates of the costs of carve-out services. The higher panel provides the rates for those states that submitted carve-out pmpm data, while the lower panel provides the same information for states that indicated they carved out the service but did not provide data.¹⁴ Mental health and pharmacy have the highest carve-out amounts. For most services, the carve-out amounts in states that provided data are close to those we estimated. Pharmacy costs are higher in the estimated states, which we believe reflects the composition of the states in the two groups. The differences, however, are not large and will not have a substantial effect on the results presented below.

Payment Rates

The results are provided in tables 12 and 13. Table 12 provides data for each of the six categories for those states that provided detailed data by age and sex. The results show that the highest rates were for newborns, followed by females ages 14 to 34. The lowest rates are for children ages 1 to 13. The greatest variation among states is among newborns. The table shows about a five-fold variation in the rates for newborns but about a three-fold variation for virtually every other group.

The grouping approach could affect the rates in each of the cells. For example, imagine two different states, one that has separate rate cells for females ages 14 to 34 and females ages 35 to 44, and a second state that establishes one rate for all females ages 14 to 44. Assume that in the aggregate the rates for the two states are the same ($230 per month). However, in the first state, payment rates for females ages 14 to 34 are $240 per month and for those ages 35 to 55 are $200. In the way that we have constructed rates for these six rate cells, we would establish two separate rates for the second state, but each would be $230. Thus, a cross-state comparison for the 14-to-34 cell would conclude that the first state pays $10 more for that group but $30 less for the 35-to-44 group. This difference is, in fact, only an artifact of our grouping strategy. The effect of the grouping strategy on the interstate comparisons within the six rate cells is small and not a major cause of differences in interstate variations.

The grouping strategy will also affect comparisons across rate cells within a state for the same reasons. Again, these effects are not large; the main reasons for intrastate variations are the rate-setting practices of the states.

Table 13 provides statewide average capitation rates. We compute the statewide averages in two ways. The first uses population weights based on the actual distribution by age and sex of each state's Medicaid population (AFDC and poverty-related groups), as shown on the CPS. The second uses national weights. The latter eliminates any differences in the statewide average that would be caused by the different composition of the Medicaid populations within the state. Table 13 shows that there is a slightly more than two-fold variation in the statewide Medicaid managed care rates. The five states with the highest rates are Connecticut, Hawaii, Kentucky, Massachusetts, and Texas. The five with the lowest are California, Florida, Georgia, Indiana, and Tennessee.

In general, states with a high proportion of children will have lower rates when state rather than national weights are used, and vice versa. The difference between the results with national and state weights can be shown in a state such as Tennessee, which has a low statewide average rate, even using the state's own population weights. When the state introduced the TENNCARE program, it added many more adults to the rolls. Because adults are more costly than children, the statewide average rate based on national weights reduces Tennessee's statewide average rate even further.

There are several reasons why capitation rates vary among states. First, capitation rates generally reflect previous fee-for-service expenditures, less some discount. The level of fee-for-service expenditures as well as the discounts depend on the objectives of the state. If states have a long history of attempting to control fee-for-service costs through narrow benefit packages or low provider payment rates, this is likely to result in lower capitation rates. If states view managed care as a way of improving access to a wide range of physicians and other providers, rates are likely to be higher. Some states can be constrained by the upper payment limit, which constrains capitation rates to be no greater than the state's existing fee-for-service payments for comparable populations. Thus some states may wish to pay more but are limited by their prior expenditure levels. If states view the primary goal as controlling expenditures, discounts are likely to be greater, and Medicaid capitation rates lower.

These factors may explain at least some of the patterns we see. California has historically had very low Medicaid spending on a per-enrollee basis. Massachusetts has been a relatively high-spending state. The state of Washington, as noted above, explicitly chose to increase rates in order to encourage a large number of Medicaid managed care plans. Tennessee, on the other hand, was relatively aggressive in attempting to achieve large discounts relative to fee-for-service. This experience appears to be reflected in the capitation rates of these four states.

Rates will also vary because of differences in eligibility rules and participation rates in Medicaid. If a state has relatively strict eligibility criteria or if enrollment is difficult, it could lead to adverse selection: Only those with a high need for services would enroll in Medicaid. As a result, utilization, and therefore expenditures, will be higher. The extent of adverse selection will affect the fee-for-service base on which capitation rates are established. The composition of the AFDC/TANF and related populations can differ for other reasons as well. For example, some states include foster children in managed care, but most do not; the former would be expected to have somewhat higher rates.¹⁵

There are a number of other reasonable hypotheses to explain these differences. For example, some states have relatively new managed care plans, as shown in table 13. States such as Connecticut, Mississippi, New Mexico, and Texas, where Medicaid managed care is relatively new, may be attempting to encourage a large number of plans to participate and therefore establish relatively high rates, at least initially. Some states with voluntary programs (such as Georgia and South Carolina) report reducing rates because of expectations of favorable selection in HMOs. Other states with a combination of HMOs and PCCM (such as Florida) may be experiencing favorable selection in HMOs and lowered payment rates to reflect their expectations of favorable selection.

Rates may also be high in some states if they have included data on the medically needy or the Supplemental Security Income population in the database used to estimate prior fee-for-service expenditures. Connecticut, for example, used data on disabled children in calculating rates for all children, which would increase the rates that plans are paid for AFDC/TANF and PRC. We have no evidence that others states used such "contaminated" data, but it could be reflected in some of the higher capitation rates that were reported.

The geographic distribution of the Medicaid managed care population would also affect the interpretation of the rates. States where Medicaid managed care enrollment is predominantly in urban areas will have higher rates than will states where enrollment is distributed throughout the state. For example, rates for Massachusetts reflect predominantly Boston-based plans. This situation may be the case in several other states as well.

Table 14 provides data on the interstate variation in rates for those states that establish rates for substate areas. We found that 23 of the 36 states had substate rates, varying from three states—Colorado, Maryland, and Kentucky—that have 2 geographic rate cells to New Jersey with 21 and Washington with 39. Some of the variations in rates across regions are great. There is slightly more than a twofold variation in rates within Mississippi, a 70 percent difference in rates in Florida, and a 40 percent difference in Texas. At the other extreme, within-state rates vary by 4 percent in Oklahoma and Nevada, 5 percent in Maine, and 6 percent in Hawaii.

Table 15 provides a comparison of Medicaid rates with an alternative that provides a measure of variation in state fee-for-service expenditures. Column 1 provides the statewide adjusted Medicaid managed care rate shown in table 13. Column 2 provides the same data in the form of index numbers calculated by taking the ratio of the statewide Medicaid managed care rate to the 50th percentile. Thus, California has a ratio of 0.66, indicating that its rates are 34 percent below the national median rate. On the other hand, Connecticut has a ratio of 1.46, indicating that Connecticut pays 46 percent more than the national median rate.

The important question is whether these rates are low relative to the alternative rates available to commercial Medicaid managed care plans. If one of the objectives is to increase the range of plans providing care to Medicaid beneficiaries, then the levels of Medicaid rates relative to the private market alternative is of significant interest. Unfortunately, data on the private-market-determined capitation rates for commercial plans for nonelderly populations are not available on a systematic and consistent basis across states (e.g., comparable populations with comparable benefits).

Ideally we would have data on Medicaid fee-for-services expenditures to compare with the capitation rates. Unfortunately, data are not available to make these comparisons. First, the latest Medicaid data are for 1996; the survey data are from 1997-98. Second, data are not available for comparable populations. We attempted to compare the capitation rates with Medicaid expenditures in two ways. One way is to examine Medicaid expenditures for acute care services for adults and children on a per-full-time-equivalent enrollee. Unfortunately, full-time-equivalent enrollees include the medically needy, who have very high expenditures. This would have the effect of making all the Medicaid capitation rates look quite low. We also explored using data on the cash assistance population. However, our capitation rates include the poverty-related groups, which include pregnant women and young children. Comparisons of the capitation rates with only cash assistance populations would make the rates look too high.

Data are available, however, for Medicare. Medicare expenditures are clearly higher than for a Medicaid or private nonelderly population but should be reflective of the relative variations among states in fee-for-service expenditures. If the variation in Medicare fee-for-service is highly correlated to the variation in private-sector fee-for-service expenditures, it will be indicative of the opportunity cost facing plans. We use the weighted monthly adjusted average per capita cost (AAPCC) for Medicare.¹⁶ We used rates for those counties for which states provided us Medicaid rates. Thus, in California, we used data on 11 major urban areas for which the state had provided Medicaid capitation rates. The AAPCC is shown in column 3; it will reflect differences in the demographic composition of the state population as well as in expenditures for the fee-for-service population.

The Medicare AAPCCs in all cases are, not surprisingly, above the Medicaid capitation rates. Column 4 provides a set of index numbers, again comparing the state AAPCC with the 50th percentile of the state AAPCC rates. Column 4 shows that California is 30 percent above the national average, whereas Iowa, North Dakota, and Utah are 85 percent of the national average.

We then computed ratios of the two index numbers. These ratios indicate whether a state's Medicaid capitation rates are high relative to the fee-for-service (Medicare) alternative. For example, if a state has a relatively low ratio for Medicaid but a high ratio for the AAPCC, the Medicaid/Medicare ratio would be relatively low and Medicaid rates would appear low. On the other hand, if the Medicaid/Medicare ratio would be relatively high, Medicaid rates would appear quite adequate.

The results show a very low correlation between Medicaid rates and the Medicare AAPCC. As a result, most of the states that were high when Medicaid rates are compared with one another remain high when the alternative Medicaid/Medicare ratios are employed. For example, Connecticut has a ratio of 1.50 for Medicaid and 1.11 for Medicare. This suggests that Connecticut is very high relative to other states for Medicaid but still above average for Medicare; thus its Medicaid rates are high relative to Medicare. Utah appears to be quite high for Medicaid relative to other states but below average for Medicare relative to other states. Thus in Utah, Medicaid capitation rates also seem relatively high relative to Medicare fee-for-service. Other states where Medicaid rates seem high relative to their own Medicare rates include Hawaii, Iowa, Kentucky, Minnesota, New Hampshire, New Mexico, and North Dakota.

On the other hand, California is below average for Medicaid, while Medicare fee-for-service is very high. Thus Medicaid managed care rates appear particularly low relative to Medicare. Other states that appear particularly low relative to Medicare fee-for-service alternatives include Florida, Georgia, Illinois, and Tennessee.

Because the California average rate seems so low by these measures, we looked for private data as another basis of comparison. Information on weighted average individual premiums per month for 1996 are available from Enthoven and Singer (1996) for several California private HMOs. Figure 2 shows the 1996 weighted average individual monthly premiums (for adult workers).

California did not provide separate capitation rates for adults and children. The $82.75 reported in table 14 is an average of children and presumably more expensive adults. In 1996, 3.3 million children and 1.9 million adults were enrolled in Medicaid. If the differences in costs of adults and children was 2.5:1, roughly the average for other states, then the implicit rates for adults and children would be about $133 and $53, respectively (1997-98). This is above The HIPC rate for adults (which has a fairly narrow benefit package) but below the others. Thus, by this measure California's capitation rates are low but not as low as when compared with other states or Medicare expenditures. It should be noted that the Medicaid population is poorer and most likely in worse health and more costly than the population in private plans. In addition, the Medicaid benefit package is typically richer, and there is little or no cost-sharing.

There are two explanations for the low California rates. First, California provides a substantial amount of money to hospitals through DSH payments based on the number of Medicaid patients served; thus, the state may believe capitation rates can be lower and still be adequate. Second, a large amount of money provided by Medicaid to California Children's Services is carved out from the capitation rates. Data provided us by PriceWaterhouseCoopers suggested that if the amount carved out for California Children's Services had been included in the 1998 capitation rates, those rates would have increased by 6.2 percent (Hunt 1999). Whether this amount should be included in the rates for comparison with other states is not clear. The answer depends on whether these children are included in the rates in other states. The survey responses we received as well as subsequent inquiries to states suggest that these children are sometimes enrolled in mandatory managed care plans, sometimes given a choice between HMOs and PCCMs, and sometimes excluded from managed care entirely.

Conclusion

On the basis of results from this nationwide survey of 41 of the nation's 45 states with capitated managed care programs, several conclusions emerge. First, there is considerable variation among states in how rates are set. About half the states simply establish rates that are available to any plan meeting state standards and willing to participate. These rates are based on fee-for-service data with some reductions to reflect states' cost-containment and access objectives. Six other states relied on negotiations. Another 14 states used competitive bidding, but even these states had a considerable amount of negotiation before final rates were set.

States vary in the extent to which they adjust for differences in risk. Some states have relatively few adjustments for age and sex, but others make many distinctions. Only two states, Colorado and Maryland, have implemented health-based risk-adjustment systems for this component of their Medicaid population. Most states adjust for maternity expenses, perhaps the major source of risk to a health plan for this population. A variety of approaches—lump-sum payments, special adjustments for poverty-related pregnant women, and higher newborn children's rates—are employed. Another way that states adjust for differences in risk is by carving out certain services from the benefit package for which the health plan is responsible. Mental health and substance abuse services are the most frequently carved-out services. These are usually carved out to other public entities, private organizations, or some combination of the two. Other states use different approaches to protect health plans from the potentially high costs of mental health and substance abuses services. Still other services that are frequently carved out are pharmacy and dental care, organ transplants, and vision services. Finally, several states also protect plans from high-risk cases through stop-loss arrangements.

States also vary in the extent to which they adjust for differences within states. We found as much as a two-fold variation within states, though this degree of within-state variation is unusual.

These differences in approaches to rate setting make rate comparisons among states difficult. We adjusted for differences in demographic groupings, treatment of maternity expenses, treatment of DSH payments and GME expenses, and the use of carve-outs. We found that capitation rates vary by a factor of more than two-to-one among states. We found more variation in rates for newborns than for other groups. States with the highest rates included Connecticut, Hawaii, Kentucky, Massachusetts, and Texas. States with the lowest rates were California, Florida, Georgia, Indiana, and Tennessee. We computed an index of Medicaid rates to an index of Medicare AAPCCs. We found that states that had high capitation rates in Medicaid were also high relative to Medicare and vice versa.

There are a number of reasons for variations in capitation rates among states. First, states that had high fee-for-service expenditures tend to have high capitation rates, and vice versa. This is because states use their prior fee-for-service expenditures as the basis for setting rates. Even if low-paying states wanted to increase payment rates, they are constrained by the upper payment limit, a federal regulation that constrains capitation rates to be no greater than the state's existing fee-for-service payments for comparable populations. The variation in capitation rates will differ from variation in prior expenditures, depending on the degree of savings that were achieved. States such as Tennessee that attempted to obtain large savings end up with lower capitation rates than states such as Washington, which acted less aggressively out of a desire to develop a competitive Medicaid managed care sector.

Rates will also be higher if states have relatively new programs. Several states that are just beginning Medicaid managed care set rates at relatively high levels to encourage the participation of many plans. States with more mature Medicaid managed care industries have often reduced or restrained the rate of growth in capitation rates. Rates will also tend to be lower in states that have voluntary plans. States such as Georgia and South Carolina have recognized that HMOs in a voluntary program will benefit from favorable selection. States in which Medicaid managed care is in predominantly urban areas will also have higher rates than in a state where Medicaid managed care is statewide. States that have a combination of HMOs and PCCM also seem to have lower HMO rates, perhaps assuming that HMOs are benefiting from favorable selection.

There is a great deal of interest in the question of the adequacy of Medicaid managed care rates. These data provide information that would allow analysts to make progress in assessing this issue, but the survey by itself does not answer the question of rate adequacy. The reasons for the entry and exit of plans are complex. It is clearly likely that the level of rates, as well as the way states adjust for risk, will affect the willingness of managed care plans to serve the Medicaid clientele.

It also seems clear that a number of other factors matter as well. One is the history of managed care in the state. If a state has many plans and a high level of competition, it is likely that more plans would be willing to participate in Medicaid at a given rate, all else being equal. On the other hand, if managed care is relatively new or there are few competing plans, rates will probably have to be higher to attract entry of a sufficient number of plans.

Second, the ability and willingness of plans to adapt to the product being purchased by the state will also vary among states. The costs of health care are not simply exogenous. Managed care plans have some flexibility in adapting the product (e.g., numbers of providers in the networks, utilization controls, and provider payment rates). The willingness to adapt the product may be limited by concerns about how a plan's commercial business might be affected.

Third, rate adequacy will be affected by the amount of excess provider capacity in the states. Plans that can take advantage of an excess supply of hospitals and doctors to limit rates will be more likely to accept lower capitation rates. In states with little excess capacity, plans will not be able to lower provider payment rates, and rates will have to be higher to attract plans.

Fourth, the degree of managed care regulation will affect the adequacy of rates. The more regulation, the higher the cost of managed care. Rates will have to be higher to attract a sufficient number of plans.

Fifth, other state policies can affect how plans will view the rates. For example, a state that makes a high level of disproportionate share payments to hospitals will probably be able to pay lower capitation rates. Hospitals will be able to accept lower rates from plans because they are receiving Medicaid revenues from other sources.

For all these reasons, analysts and policymakers should use these data carefully when making judgments that rates are too high or too low. We also remind the reader that we had to make several estimates—for example, maternity costs, GME and DSH, and carve-outs—that will affect any rate comparisons. We could not adjust for differences in the use of stop-loss arrangements and the impact of carve-outs other than those we discussed. Moreover, while we have attempted to define a fairly homogeneous population (AFDC/TANF and poverty-related groups), there can be differences in the composition of this population for which we cannot control.

Despite these caveats, the results show more than a twofold variation in rates among states for reasonably homogeneous populations. Further adjustments would not much alter that reality. The key issue is the implications of these rate variations for plan participation and for beneficiary access. If states with low rates have plenty of plans participating and satisfactory levels of access, perhaps rates in other states are too high. But most likely states are getting what they pay for and this is reflected in both the number and nature of plans participating and in levels of beneficiary access. States with low rates are probably not buying the same "product" as states with higher rates. The implications of these "product" differences for beneficiary access, quality of care, and health are empirical questions beyond the scope of this study. Hopefully these data will help further policy discussions and research on these issues.

Notes

1. The 17 states are Alabama, Arizona, Arkansas, Delaware, Hawaii, Kentucky, Massachusetts, Minnesota, Maryland, New Jersey, New York, Ohio, Oklahoma, Oregon, Rhode Island, Tennessee, and Vermont ("Medicaid Letter" 1998).
2. For evidence that Medicare payment rates (average adjusted per capita cost, or AAPCC) affect participation in Medicare, see Porell and Wallack, and Pai and Clement.
3. The survey included the District of Columbia. In cases where states had both voluntary and mandatory programs, we asked the state to provide information on the mandatory program only.
4. States with only fee-for-service or primary care case management programs were not asked to respond.
5. Adjustments to baseline fee-for-service data were made to reflect differences in the populations entering managed care plans vis-?-vis the fee-for-service population as well as savings assumptions.
6. Controversy developed over the Maryland risk-adjustment system because of overpayments attributed to the use of two years of data on expenditures to calculate rates, resulting in an overpayment of $8 million. This seems to be a serious calculation error and not a flaw in the underlying system.
7. Except in some demonstration projects, the final capitation rates must comply with federal Medicaid regulations, which require that capitation rates for Medicaid prepaid health plans not exceed fee-for-service costs for an actuarial equivalent population.
8. Direct medical education costs of providing graduate medical education (GME) are those costs related to salaries and fringe benefits for residents and teaching physicians, conference and classroom space, and equipment and supplies used for instructional purposes. Indirect medical education costs are those related to higher patient care costs that teaching hospitals are thought to incur because of such factors as increased diagnostic testing, increased number of procedures, higher staff ratios, and increased recordkeeping (U.S. GAO 1997).
9. In many cases, numerous counties were grouped together by Federal Information Processing Standard code to calculate population distributions in the state-selected regions. To the extent the Medicaid population differs by age and region from the population below 200 percent of the federal poverty level (FPL), the aggregation could be affected. Disaggregated data at the age, sex, and regional level for the Medicaid population are not as reliable in most states as the population below 200 percent of the FPL.
10. As a check on the reasonableness of this assumption, we called a number of states to see if they could provide estimates of the proportion of female Medicaid enrollees who are pregnant in a given year. Rhode Island responded that 17.4 percent of their female group ages 15 to 44 were pregnant in a given year. Washington estimated that 10.1 percent of women ages 6 to 64 and 17.2 percent of females ages 19 to 34 were pregnant in a given year. Maryland provided data that suggest approximately 14.3 percent of women ages 15 to 44 were pregnant during fiscal year 1998.
11. Illinois included disproportionate share hospital (DSH) payments in rates and provided a pmpm equivalent of $4.59. Minnesota included DSH and GME in capitation rates, but reported that DSH and GME totaled 6 percent and 3 percent of rates in Hennepin and Ramsey Counties.
12. Massachusetts included GME payments, but provided a per-member per-month (pmpm) equivalent of $5.00. Likewise, New Jersey included GME payments in rates, but provided a pmpm equivalent of $6.00.
13. The substance abuse estimates were lower by a factor of roughly three, but the resulting differences in absolute dollars are not large. An alternative would have been to use information from the states that did provide data. We concluded that these data varied too much to rely on them for making estimates for other states.
14. Because we used the donor databases to compute the amount in the carve-out service as a percentage of the total, the estimates in the two panels can vary because of differences in the underlying capitation rates.
15. To examine the impact of including foster children in Medicaid expenditures, we analyzed Medicaid claims data from California, Georgia, and Tennessee. The results showed that rates for children would increase by 7.2 percent in California, 1.0 percent in Georgia, and 11.5 percent in Tennessee. Including foster children would increase the statewide average rates in table 1, which include adults, by 1.7 percent in California and 5.2 percent in Tennessee, but would reduce rates by 0.2 percent in Georgia.
16. The AAPCC reflects average Medicare fee-for-service expenditures in each county, adjusted for differences in age, sex, Medicaid enrollment, institutionalization, and work status. It therefore reflects variations in health care costs across states but is also affected by Medicare policies on these outlays. For example, the Medicare AAPCC tends to be low in states such as Minnesota and Oregon, where health care costs have been low because of managed care.

---

Estimates were made of the monthly cost of specific Medicaid carved-out services for managed care plans by the Actuarial Research Corporation (ARC). Estimates were limited to six service classes because of constraints in available data. The services estimated were mental health, substance abuse, dental, vision, prescription drugs, and transplants. We did not have sufficient data on the number of special needs children or HIV patients in each state to make reliable estimates.

The starting point for estimating the cost of carved-out services is a matrix provided by the Urban Institute (UI) that shows which states carve out which services. State-level tables provided by UI display starting capitation rates and whatever carve-out information was provided by the state in question. Starting capitation rates are given for six age-sex cells: age 0, ages 1 to 13, females ages 14 to 34, females ages 35 to 44, males ages 14 to 44, and all persons ages 45 to 64. For many states, these rates were given by region and were aggregated across all regions in the state by UI in order to obtain state-level rates.

UI adjusted rates to account for (1) disproportionate share hospital (DSH) and graduate medical education (GME) and (2) maternity expenses. DSH and GME were removed from the rates for those states that included them. Finally, we used counts of Medicaid cash (Aid to Families with Dependent Children, or AFDC) recipients from the March 1997 Current Population Survey (CPS) to calculate composite rates across age-sex cells.

Composite rates were calculated from these starting rates in order to make them comparable to the data we have available on specific covered services. In general, data were available either as an overall adjustment or an adjustment made for adults and children separately. We did have some detailed data on the distribution of mental health and substance abuse services by age. We defined "adult" to be made up of the age-sex cells that contained persons 14 years and older. We defined "children" to be made up of the age-sex cells containing persons 13 years of age and younger. Composite rates were calculated by multiplying the population weights (i.e., Medicaid AFDC counts) by the capitation payments for the age groups in question, and then dividing by the total population in question. As mentioned above, Medicaid AFDC counts were used as a proxy for the health maintenance organization (HMO) counts so as to not give undue weight to the male cells.

Data for Carve-Out Cost Estimates

Data on mental health and substance abuse were obtained from ARC proprietary data for a private employer population, with estimates made for managed care use. Data are person level and can be aggregated up from inpatient and outpatient services. The information used is the amount spent on the service as a percentage of total spending. Comparing these percentages with those calculated from states with carve-out amounts for mental health or substance abuse or both led to the conclusion that although the populations are very different (private health versus Medicaid), the proportion of charges attributable to mental health or substance abuse appears to be of a reasonable magnitude.

Data on dental and prescription drugs were obtained from Medicaid 2082s for AFDC child and adult populations. Data have been adjusted to be the conditional average for states with "valid" data for services in question, thus omitting those states with either no information or whose information were outliers.

Data on vision and transplant services are from ARC proprietary data for private HMO plans. No differentiation was made for adult-versus-child costs, beyond that implied by looking at the amounts as a percentage of the total.

Classifying States by Amount Known

States were classified by the amount that was known about the carve-out for the particular service. Three categories were used for the classification: (1) full data on service known, (2) partial data on service known, and (3) no data on service known. The simplest cases were those where either full data or no data were given by the state for a service. Partial data involved determining the relationship between what was given and what was actually carved out.

States Providing Full Data

Full data were defined to be either age breaks for a particular service (best case) or a weighted average capitation amount for a particular service (that was later adjusted to be age specific if the state used age-rating). States that provided data in this manner are listed by service by state.

Full data on mental health costs were provided by California, Colorado, Kansas, New Jersey, Ohio, and Utah. Full data on substance abuse costs were provided by Kansas, New Jersey, Ohio, and Utah. Full data on prescription drug costs were provided by Kansas and Utah. Full data on vision costs were provided by California and Massachusetts. Full data on dental costs were provided by California, Hawaii, Kansas, Massachusetts, and Utah.

States Providing Partial Data

Partial data encompass different types of information, including states that have restrictions on what portion of a service is carved out, as well as states that know the cost only for a subgroup of the covered population.

For mental health or substance abuse or both, the following states provided either limited coverage or limited cost data on the coverage provided: Colorado, Florida, Georgia, Kentucky, New Hampshire, New York, Ohio, Rhode Island, South Carolina, and Virginia. Tennessee provided data on mental health plus substance abuse combined. Arizona also provided these data combined, but UI deemed they possibly contained disabled costs and so they were not used. For dental, the following states provided either limited coverage, or limited cost data on what coverage was provided: Mississippi and West Virginia. For prescription drug coverage, Washington provided only a limited carve-out. For vision, the following states provided limited coverage in a carve-out: Vermont and Washington. More detail on the carve-out calculations is provided in the section on special cases.

States Providing No Information on Carve-Out Costs

Several states provided no cost information on carved-out services. We estimated costs on a "percentage of total" basis to account for the different levels of base costs in the states in these cases. States in this category are listed by service. No cost data on mental health were provided by Iowa, Kentucky, Maine, Maryland, Michigan, Mississippi, North Carolina, Pennsylvania, Washington, or West Virginia. No cost data on substance abuse costs were provided by Iowa, Indiana, Maine, Michigan, Mississippi, Missouri, North Carolina, Pennsylvania, Virginia, Washington, or West Virginia. Costs for both mental health and substance abuse were calculated for Arizona despite the state's concerns about the comparability of the included population. Cost data on dental carve-outs were not present for Georgia, Iowa, Illinois, Maine, Michigan, Mississippi, Missouri, Nevada, North Dakota, New Hampshire, Rhode Island, South Carolina, Vermont, Washington, or West Virginia. Prescription drug carve-out cost data were not provided by Iowa, Maine, Missouri, North Carolina, North Dakota, New Hampshire, Texas, or West Virginia. Vision carve-out cost data were not provided by North Dakota or South Carolina, and transplant carve-out cost data were not provided by California, the District of Columbia, Maine, Missouri, or New Jersey.

Calculation of the Costs of Carve-Outs

States that had fully carved-out services and no information on the service amounts were assigned percentage amounts on an adult-versus-child basis. For states that had carved-out services and provided data, the amounts as provided were used. Special-case states (partial data or partial coverage or both) were estimated on a service-by-service basis (see below).

Once estimates of the cost of services on a percentage basis were obtained, the next step was to convert these percentages to capitation amounts. This was done by looking at what the basis was for the particular service and determining the appropriate algebraic formulas for each state involved. Cost bases were as follows. For mental health and substance abuse, "percentage of total" refers to the total costs including mental health, substance abuse, and prescription drugs, but not including dental, vision, or transplant services. For dental and prescription drugs, "percentage of total" refers to the total costs of all services involved. For vision and transplant services, "percentage of total" refers to the total costs, including these services as well as prescription drugs, but not including dental.

The next step was to convert the dollar values (either overall or adult versus child) to age-sex-specific rates for the six rate cells of interest. This calculation was done only for states that did not provide age data and did age-rate their capitation payments.

Data on the cell-based ratios were obtained from our proprietary data source for mental health and substance abuse and were used to calculate initial age-rates. These rates were then normalized using the Medicaid/AFDC population counts from the March 1997 CPS in order to hit the starting costs.

Dental rates were assigned as is on an adult-versus-child basis except for those states that provided age-specific values. Children age zero were not assigned a dental cost, so that the entire cost of the children's dental care was assigned to the 1 to 13-year age grouping. Prescription drug rates were assigned to the age-sex cell based on the percentage of the total amount represented. Vision and transplant costs were assigned uniformly across age-sex cells, unless different information was provided by a particular state.

Our estimates for the cost of carved-out service as a percentage of total costs are depicted below. Note that for mental health and substance abuse, "total" represents total inclusive of these services and prescription drugs. For dental and prescription drugs, "total" represents total overall services; for vision and transplant services, "total" represents total inclusive of all services except for dental.

Below is a table that shows, for Iowa, starting capitation rates (obtained from UI) and counts of persons (from Medicaid/AFDC recipients on the March 1997 CPS—used for distribution only), as well as calculated rates for adult versus child.

Iowa carves out the following services: mental health, substance abuse, dental, and prescription drugs. Estimates of the services (as a percentage of total) for adults and for children are listed below. For adults, the estimates are as follows:

• Mental health = 5.4 percent of total, where total includes capitation rate plus mental health plus substance abuse plus prescription drugs.
• Substance abuse = 0.83 percent of total, where total includes capitation rate plus mental health plus substance abuse plus prescription drugs.
• Dental = 2.74 percent of total, where total includes capitation rate plus mental health plus substance abuse plus dental plus prescription drugs.
• Prescription drugs = 7.05 percent of total, where total includes capitation rate plus mental health plus substance abuse plus dental plus prescription drugs. For children, the estimates are as follows:
• Mental health = 7.08 percent of total, where total includes capitation rate plus mental health plus substance abuse plus prescription drugs.
• Substance abuse = 0.03 percent of total, where total includes capitation rate plus mental health plus substance abuse plus prescription drugs.
• Dental = 3.42 percent of total, where total includes capitation rate plus mental health plus substance abuse plus dental plus prescription drugs.
• Prescription drugs = 4.85 percent of total, where total includes capitation rate plus mental health plus substance abuse plus dental plus prescription drugs.

These relationships can be expressed algebraically as follows: Variables are defined as

C_a = capitation rate, adults
M_a = mental health rate, adults
S_a = substance abuse rate, adults
D_a = dental rate, adults
P_a = prescription drug rate, adults
C_c = capitation rate, children
M_c = mental health rate, children
S_c = substance abuse rate, children
D_c = dental rate, children
P_c = prescription drug rate, children

And the equations are as follows:

M_a = 0.054 * ( C_a + M_a + S_a + P_a )
S_a = 0.0083 * ( C_a + M_a + S_a + P_a )
D_a = 0.0274 * ( C_a + M_a + S_a + D_a + P_a )
P_a = 0.0705 * ( C_a + M_a + S_a + D_a + P_a )
M_c = 0.054 * ( C_c + M_c + S_c + P_c )
S_c = 0.0083 * ( C_c + M_c + S_c + P_c )
D^c = 0.0274 * ( C^c + M^c + S_c + D^c + P^c )
P^c = 0.0705 * ( C^c + M^c + S^c + D^c + P^c )

Each system of equations is solved for each of the unknowns (e.g., M_a, S_a, D_a, and P_a). Once the carve-out values are solved, they are then adjusted for age-sex variation among the six groups and normalized in order to reach the calculated values.

Mental health and substance abuse were adjusted by age using relationships from our underlying data on the level of payments across age groups. The relationships are defined below.

For adults, if M_a = mental health rate, then the rates for the four adult cells are initially calculated as follows:

Females 14-34 = 1.1884 * M_a
Females 35-44 = 1.2670 * M_a
Males 14-44 = 0.9324 * M_a
All Persons 45-64 = 0.7808 * M_a

For adults, if S_a = substance abuse rate, then the rates for the four adult cells are initially calculated as follows:

Females 14-34 = 0.6160 * S_a
Females 35-44 = 0.4234 * S_a
Males 14-44 = 1.9602 * S_a
All Persons 45-64 = 0.5928 * S_a

For children, if M_c = mental health rate, then the rates for the two children cells are initially calculated as follows:

Age 0 = 0.1147 * M_c
Ages 1-13 = 1.1014 * M_c

For children, if S_c = substance abuse rate, then the rates for the two children cells are initially calculated as follows:

Age 0 = 0.00 * S_c
Ages 1-13 = 1.1146 * S_c

The calculations for dental and prescription drugs are simpler. For dental coverage, the calculation is much simpler because the child rate is initially applied only to the 1 to 13 age group, and the adult rate is applied uniformly across the four adult cells. For prescription drug coverage, the child rate as a percentage of total is applied to each of the child groupings, and the adult rate as a percentage of total is applied to each of the adult groupings.

Special Case Carve-Out States and Services

Mental Health Services: Colorado

Sixty-five percent of the population are in plans where mental health is carved out, and the remainder are in plans that receive carve-ins. Colorado provided cost information (by adult versus child) for the carve-out and carve-in. This information was used along with the proportions of the population to calculate the weighted average effect of the mental health cost.

Mental Health Services: Florida

Rates for Florida mental health carve-outs were not given in the same manner as coverage offered. Rather, rates were given by six age groups (that do not correlate with the six groups of interest) for total mental health care for one region [Hillsborough, Hardee, Highlands, Manatee, and Polk Counties] containing approximately 12 percent of the population of interest. Adjustments were made to convert down to community mental health services (10 percent of outpatient) and to convert from region-specific rates to overall average rates for state minus region. The full rates and the community mental health rates were then averaged together. Finally, the age groups were converted to the six groups of interest by assuming level rates within group.

Mental Health and Substance Abuse Services: Georgia

Georgia carves out mental health and substance abuse except for inpatient hospital stays of up to 30 days. We estimated the proportion of inpatient costs in excess of the 30-day limit and added this to the cost of outpatient in order to come up with mental health and substance abuse costs for Georgia.

Substance Abuse: Kentucky

Kentucky does not cover substance abuse, except for inpatient detoxification (detox). Assuming that detox represents approximately 50 percent of inpatient substance abuse costs (the rest being mainly rehabilitation), we were able to remove this from the estimate of what a full substance abuse program would cost if Kentucky were to cover it.

Mental Health and Substance Abuse: New Hampshire

New Hampshire carves out community mental health and substance abuse. An adjustment factor (10 percent) was used to reduce the cost of outpatient mental health and substance abuse services to reflect only those offered through a community health center.

Mental Health and Substance Abuse: New York

Plans are responsible for 20 outpatient visits or 30 inpatient days of mental health treatment annually and 60 outpatient visits or 30 inpatient days of substance abuse treatment. Plans are paid fee-for-service for amounts beyond this, and this is the amount we estimated as the "carve-out" because it is a separate reimbursement. The amount was estimated in the same manner as the Georgia adjustment: by looking at the proportion of charges in excess of the day or visit limits and adjusting the total amounts downward in order to calculate only this excess cost.

Mental Health: Ohio

Community mental health and substance abuse are carved out in Ohio, and providers are paid fee-for-service. Ohio has provided the per-member per-month equivalents for community mental health as $5.41 and substance abuse as $1.44.

Mental Health and Substance Abuse: Rhode Island

Rhode Island has stop-loss for mental health after 30 inpatient days or 30 visits and for substance abuse after 30 visits. These amounts were obtained by determining costs for days/visits in excess of these thresholds and applying these to our starting-point mental health and substance abuse numbers.

Mental Health and Substance Abuse: South Carolina

South Carolina carves out mental health and substance abuse after $1,000 per member per year. We compared the average rates from South Carolina with our overall average rates from the data set used to determine mental health and substance abuse costs, adjusted the $1,000 threshold to be comparable, and looked at what proportion of charges were in excess of $1,000.

Mental Health: Virginia

Virginia carves out substance abuse and community mental health services. Substance abuse was estimated separately as a full carve-out, and an adjustment factor (10 percent) was used to reduce the cost of outpatient mental health services to reflect only those provided through a community health center.

Dental Services: Mississippi

Dental is carved out only for adults in Mississippi, so no children's cost was calculated.

Dental Services: West Virginia

Dental is carved out only for children in West Virginia, so no adult cost was calculated.

Vision: Illinois

Vision care was not covered in Illinois, so we estimated the amount that would have to be added in if it were to be covered (1.5 percent), in order to maintain a comparable benefit plan.

Vision: Washington State

Vision care is carved out for glasses only, so a smaller amount (0.5 percent instead of 1.5 percent) was used to calculate the carve-out amount.

Vision: Vermont

As with Washington, vision is carved out for glasses only, so the carve-out amount was calculated at 0.5 percent instead of the full 1.5 percent.

Prescription Drug Services: Washington

Washington State has a limited carve-out for prescription drugs: prescriptions from health departments, family planning clinics, regional support networks, community mental health clinics, and dentists, as well as protease inhibitors. While protease inhibitors are beyond the scope here, the other limitations we assume make up about 10 percent of normal prescription drug cost.

---

References

---

About the Authors

John Holahan is the director of the Urban Institute's Health Policy Center. He has authored several publications on the Medicaid program. He has also published research on the effects of expanding Medicaid on the number of uninsured and the cost to federal and state governments. Other research interests include expanding health insurance for children, health system reform, changes in health insurance coverage, physician payment, and hospital cost containment.

Suresh Rangarajan is currently completing his prerequisites for medical school in Columbia University's Post-Baccalaureate Premedical Program. Prior to attending Columbia, Suresh was a research associate at the Urban Institute, where he focused on Medicaid managed care programs nationwide and the evaluation of Section 1115 Medicaid waivers. Before coming to Urban, Suresh worked as both a health care consultant and an investment banker. He has a master's degree in health policy management from the Harvard School of Public Health.

Matthew Schirmer was a research assistant at the Urban Institute when this publication was written. His work focused on both fee-for-service and managed care Medicaid reimbursement policy, as well as various issues regarding access to care and insurance coverage for low-income individuals. Matthew is currently working as a health care consultant. He has a bachelor's degree in sociology from Vanderbilt University.

---

Usage and reprints: Most publications may be downloaded free of charge from the web site and may be used and copies made for research, academic, policy or other non-commercial purposes. Proper attribution is required. Posting UI research papers on other websites is permitted subject to prior approval from the Urban Institute—contact [email protected].

If you are unable to access or print the PDF document please contact us or call the Publications Office at (202) 261-5687.

Disclaimer: The nonpartisan Urban Institute publishes studies, reports, and books on timely topics worthy of public consideration. The views expressed are those of the authors and should not be attributed to the Urban Institute, its trustees, or its funders. Copyright of the written materials contained within the Urban Institute website is owned or controlled by the Urban Institute.