Predicting State Medicaid Home and Community Based Waiver Participants and Expenditures, 1992–1997

Purpose: The study examined trends and predictors of state Medicaid home and community based waiver participants and expenditures from 1992 to 1997 to identify factors of in- terest to policy makers and clinicians. Design and Methods: HCFA Form 372 data were collected from state officials for each waiver for each year. Two separate regression analyses were conducted to examine the effects of sociodemographic, economic, political, policy, and health services on state waiver participants and expenditures. Results: State waiver participants were positively associated with those aged 85 and over, personal income, residential care beds, and inpatient users and negatively with home health regulation and nursing home beds. State waiver expenditures were positively associated with democratic governors, personal income, home health re- imbursement methods, Medicaid eligibility, home health agencies, and Medicare home health users. Implications: The factors policy makers might consider changing include increasing the number of residential care beds and home health agencies, removing certificate of need for home health care, using Medicare home health reimbursement methods for Medicaid, and raising the Medicaid eligibility criteria. In some states with low nursing home occupancy rates, reducing the supply of nursing home beds may also be considered. All of these approaches would be controversial and should be based on additional cost-effectiveness analysis.

The Medicaid home and community based services (HCBS) waiver program was established by Congress in 1981under Section 1915 of the Social Security Act to shift services to the community and away from institutional settings (Miller, 1992;Miller, Ramsland, & Harrington, 1999a). By 1997, all states, except Arizona and the District of Columbia, had one or more 1915(c) HCBS waiver programs for longterm-care services (see Appendix, Note 1). By 1997, 228 different waiver applications had been submitted to the Health Care Financing Administration (HCFA; Miller et al., 1999a;Salo, 1998).
The importance of the home and community based services increased with the passage of the Americans With Disabilities Act (1990), because it outlawed certain practices of private and public entities that unreasonably restrained the participation of individuals with disabilities in society. More recently, the Supreme Court ruled in Olmstead v. L. C. (1999) that states may not discriminate against persons with disabilities by refusing to provide community services when these are available and appropriate.
The Medicaid HCBS waiver program has grown from $3.8 million for 6 waivers in 1982 (Miller, 1992;Greenberg, Schmitz, & Lakin, 1983) to $8.1 billion in 1997 (Burwell, 1999;Miller et al., 1999a). The program expenditures increased by 45% between 199645% between and 199745% between and by 12% between 199745% between and 199845% between (Burwell, 1999Miller et al., 1999a). Even with its growth, total HCBS waiver expenditures were only 15% of the $59 billion reported in total long-term-care Medicaid dollars in 1998, most of which was spent on institutional care (Burwell, 1999). State Medicaid programs may offer home health and personal care services in addition to the HCBS waiver program. When Medicaid HCBS waivers, home health, and personal care services are combined, they represent 25% of total long-term-care service expenditures in the United States (Burwell, 1999). The HCBS waiver program expenditures are less than institutional expenditures, in part, because of statutory and regulatory requirements imposed on the program to control costs (this point is discussed later). The HCBS waiver program also does not cover room and board expenses, although such expenses are paid in institutional settings. HCBS are optional services requiring state program funds. The spending patterns have varied dramatically by state depending on a wide number of factors (Miller, Ramsland, & Harrington, 1999b).
The purpose of this article is to present first-time trend data on the state HCBS waiver participants collected by the investigators from the states (on HCFA Form 372). The data include the number and types of 1915(c) HCBS waiver programs actually operated by the states and the number of participants and expenditures from 1992 to 1997. In addition, the study examines an array of state-level factors associated with the number of state participants and expenditures in the 1992 to 1997 period. The analytical model tests the effects of sociodemographic, economic, political, public policy, and health service factors on waiver participants and expenditures. The identification of statelevel variables associated with the development of waiver programs is important to the development of effective waiver policies. The information should inform policy makers and clinicians about which factors could be changed to expand the number of state waiver participants and/or expenditures.

Background
The HCBS waiver program was established under Section 1915(c) of the Social Security Act established by Public Law 97-35 and has been broadened by statutes and regulations since that time (see Appendix, Note 2). States are required to submit an application, be reviewed, and then approved by the Secretary for Health and Human Services before services become eligible for federal matching payments. The regulations require states to demonstrate that spending will not exceed the amount that would otherwise be spent on institutional services (42 U.S.C. 1396 [n][c][1]; Salo, 1998).
The HCBS waiver program may be either statewide or confined to specific geographic areas (Section 1902 [a][1]; Gurny, Hirsch, & Gondek, 1992). A recent survey of state HCBS waiver programs found that 12 states out of 51 (including Washington, DC) had some waivers that were not statewide in 1999 ; see also Lipson & Laudicina, 1991). States have the option of applying the financial eligibility criteria that they use for institutions (hospitals, nursing facilities [NFs], or intermediate care facilities for the mentally retarded [ICF-MRs]) for up to 300% of the Supplemental Security Income program (SSI) for individuals living in the community who are in the HCBS waiver program (Hovarth, 1997). Fourteen states have income standards for institutional eligibility set at levels below 300% of SSI . The HCBS programs may also use the spousal improverishment standards for institutional services. In 1999, LeBlanc andcolleagues (2000) found that 7 out of 49 states (excluding Arizona and Washington, DC) did not use the same special income criteria for HCBS as for institutional services.
Under the statute and regulations, the HCBS program is limited to those who would otherwise require the level of care provided in institutions including hospitals, NFs or ICF-MRs. Although HCBS waiver participants must be eligible for institutional care, the states have wide flexibility to establish need criteria for the waivers that are the same criteria as the state's criteria for institutional care (Harrington, LaPlante, et al., 2000;O'Keeffe, 1996). Consequently, the need criteria vary widely for the waivers within and across states.
Under the HCBS waivers, states may target the 1915(c) waiver program to particular groups specified under each state's waiver plan; they are not required to offer services to all categorically or medically needy groups. (This is called a waiver of comparability;Section 1915[c][4][A]; 42 U.S.C. 1396n). States may provide waiver services up to the maximum number of HCFA-approved waiver slots or openings. When the state waiver slots are full, states may then establish waiting lists for program services or states may request additional waiver slots from HCFA as long as the state has funding for the waiver program. HCFA's legal counsel has stated that states are allowed to give priority to Medicaid participants within target groups as long as the criteria are not arbitrary and are clear and specified in the waiver application (O'Keeffe, 1996).
The shortage of HCBS waiver programs for the Medicaid population is a problem found across the states. Kassner and Williams (1997) reported that 33 states had waiting lists in 1996 for individuals who were Medicaid eligible and wanted to be in the Medicaid HCBS waiver program but the program services were limited. O'Keeffe (1996) also reported states had waiting lists because of inadequate funding or too few waiver slots. Massachusetts made a major shift from institutional to community care for both the mentally retarded and developmentally disabled (MR/ DD) and the mentally ill groups between 1990 and 1996, but the state had almost 1,400 people residing in nursing homes and on waiting lists for communitybased residential and day care (Holahan, Bovbjerg, Evans, Wiener, & Flanagan, 1997). A study in Texas also reported 30,000 people on the combined waiting lists for community-based MR/DD and mental health services (Wiener, Evans, Kuntz, & Sulvetta, 1997). Another survey of state Medicaid officials in 1998-1999 found that 27 states had waiting lists for HCBS waiver services, although some states could only estimate the numbers, and 42 states reported inadequate numbers of HCBS waiver slots (Harrington, LeBlanc, Wood, Satten, & Tonner, 2000). The different procedures used by states to collect data and screen individuals can affect the size of the waiting lists. According to state officials, the waiting lists generally reflected a lack of state funding for the HCBS waiver program to match federal Medicaid dollars . Other analyses have suggested that waiver availability is related to the overall resources of states (Miller et al., 1999b).

Factors Associated With State Waivers
Waiver growth is a dynamic process affected by public policy decisions and multiple market factors. On the basis of previous studies of state long-term care programs, five major types of independent variables were conceptualized to be related to the number of waiver participants and expenditures: (a) sociodemographic factors, (b) economic factors, (c) political factors, (d) state policies, and (e) health care services available in the market place. Sociodemographic factors can influence the overall demand for waiver programs. The aging of the state population, particularly the growth in the oldest old (Mendelson & Schwartz, 1993) would be expected to have a positive influence on the demand for services by waiver participants and on expenditures. The age of the population is a strong predictor of the use of formal home care services (Mauser & Miller, 1994;Wallace, Campbell, & Lew-Ting, 1994;Logan & Spitze, 1994;Houde, 1998;Kemper, 1992).
Higher percentages of women in the labor force (who would be unavailable to care for elderly and disabled family members) should increase the demand for long-term-care services (Silverman, 1990;Kemper, 1992;Houde, 1998). Nyman, Sen, Chan, and Commins (1991) and Kenny (1993) found that urban residents were more likely to use home health services than rural residents. Thus, we expected that states with a higher percentage of the population living in metropolitan areas should increase demand and be positively associated with waiver participants and expenditures.
Although African Americans are more likely to require assistance for limitations in daily living (Harpine, McNeil, & Lamas, 1990), Kemper (1992) found that African Americans and Hispanics were less likely to receive formal home health services. Although the findings are mixed, other studies have found less access to long-term-care services in minority populations (Cagney & Agree, 1999;Wallace, Levy-Storms, Kington, & Andersen, 1998;Houde, 1998;Murtaugh, Kemper, & Sillman, 1990). States with large non-White populations may have fewer waiver participants and expenditures.
Economic factors can of course affect the demand for waiver services as well as the input prices and labor availability of waiver providers. States with higher personal income per capita are expected to have a higher demand for long-term care, but these states may also be more generous in their funding of Medicaid waiver programs (Buchanan, Cappellini, & Ohsfeldt, 1991;Schneider, 1993;Kane, Kane, Ladd, & Nielson, 1998;Miller et al., 1999b). States with high poverty may have increased demand for services if individuals are unable to pay for long-term care. On the other hand, high poverty rates may lower the demand for waiver services because more individuals may be unemployed and thus available to provide informal care services to family and friends.
Political factors should have some direct effects on the amount of waiver participants and expenditures. Those state politicians with conservative voting records generally have been considered to be less likely to support Medicaid programs, whereas liberals traditionally have supported funding for public programs. (See the approach of Barrilleaux & Miller, 1988;Lanning, Morrisey, & Ohsfeldt, 1991). The role of the governor is important in shaping state policies (Schneider, 1993;Schneider & Jacoby, 1996). Democratic governors may be more politically liberal and more likely to support Medicaid home and community based waiver programs. Finally, the aging lobby is expected to vary across states in terms of its political power. The percentage of membership in the American Association of Retired Persons (AARP) is used as a proxy measure for the political power of the aging population in a state. These factors were developed by Lanning and colleagues (1991) in his study of hospital regulation.
Waiver participants and expenditures may be positively related to the use of certificate of need (CON) and/or moratorium regulation of nursing homes. Where nursing home beds are controlled, there may be fewer nursing home participants and expenditures, and consequently more state funds may be available for HCBS waiver services (Harrington, Swan, Nyman, & Carrillo, 1997). Where states use CON and/or moratorium for home health, the supply of waiver services may be diminished. State reimbursement rate policies may also influence state long-term-care programs (Swan, Harrington, Grant, Luehrs, & Preston, 1993). Where state Medicaid reimbursement rates for nursing homes are high, states may have less funds available to pay for waiver participants and expenditures. On the other hand, where states have more generous reimbursement rates for home health care services (i.e., use Medicare methodologies), these states may have a higher supply of waiver providers and more waiver participants and expenditures.
State Medicaid eligibility policies should also have a direct effect on Medicaid waiver participants and expenditures. Those states that use more restrictive eligibility policies, such as the special 209(b) eligibility rules that limit the number of aged, blind, and disabled who are eligible for Medicaid, may reduce access to HCBS waiver participants and consequently reduce expenditures (Miller et al., 1999b). Those states with more generous eligibility in terms of the dollar threshold for the medically needy program, compared with those that have no medically needy spend down program or have low threshold levels, may consequently have higher waiver participants and expenditures.
The supply of health care services in a state can also directly influence waiver participants and expenditures. Greater numbers of nursing home beds per population should reduce the available funds for state waiver participants and expenditures (Miller et al., 1999b). Larger numbers of residential care beds and certified home health care agencies should have a positive influence on the number of waiver participants and expenditures. Increased rates of Medicare home health users per 1,000 Medicare beneficiaries should also increase the demand for all long-term care, including the HCBS waiver participants and expenditures (Harrington et al., 1997;Swan & Harrington, 1990;Kenny, Rajan, & Soscia, 1996). More home health services in an area may lead to the identification of more individuals in need of HCBS services or it may be a proxy for higher disabilities rates in an area. Thus, these factors will be taken into account in the analytical model for the present study.

Analytical Model
Independent variables for the period of 1991-1996 were used to predict the number of waiver participants and expenditures in the 1992-1997 period, where the state was the unit of analysis. This 6-year period was examined because it was the only one in which a complete data set was available for the waivers. The independent variables were used to predict the dependent variables in the following year, because it was expected that their effects would require a year to have an impact. This approach also reduced the likelihood of any problems of endogeneity.
The waiver participants and expenditures were grouped together for the dependent variables across different types of waivers. Although the waivers involve many different target groups, some of the waivers had very few participants, were not operational for the entire study period, and were not available in all states. Thus, it would not be feasible to conduct an analysis of each waiver type separately. Moreover, the factors that affect the development of waivers for one target group within and across states are conceptually likely to affect the waivers for other target groups. For example, all target groups use nursing homes, residential care, and home care. Moreover, the sociodemographic, political, and economic factors were expected to have similar effects on waivers for different target populations. Although the waivers for the developmentally disabled and mentally retarded might be expected to be somewhat different from the aged and disabled, when we tested this group separately, the regression model results were very similar to the results for the total waiver groups as a whole.
The study used an ordinary least squares (OLS) regression model with a random effects panel analysis. This allows a test of the effects of the independent variables across states, rather than one using only a fixed effects model that would test differences within states. We were interested in which factors across states contribute to the number of waiver participants and expenditures. A Hausman test of the parameter estimates of the two models showed that the random effects model was consistent and efficient and that use of the random effects model was appropriate (see Hausman, 1978;Greene, 1990).
The waiver participants and expenditures were standardized by 1,000 population rather than by total Medicaid recipients to allow for comparisons across state populations. This allowed us to test for the effect of state Medicaid financial eligibility criteria as an independent variable.
The independent variables were tested for multicollinearity by completing an SAS correlation matrix. None of the variables was found to be highly correlated (above a .65 correlation). In addition, tolerance tests in the regression analyses did not show multicollinearity to be a problem. The LIMDEP program Version 7.0 was used to conduct the regression analyses, using the adjustment for autocorrelation (Greene, 1991).
Arizona and Washington, DC, were eliminated from the data set because they had no Medicaid waiver program in place during the 1992-1997 period. Thus, 49 states were included with 294 observations over a 6-year period. The significance test for the model was a 2-tailed test. The following equations were examined: where E it ϭ random error terms.
The waiver participants showed a normal distribution, but the waiver expenditures were skewed, so the log of waiver expenditures was used as the dependent variable, adjusted for annual increases in the consumer price index. The analytical models are reduced-form equations that include factors influencing both the demand and the supply of HCBS waiver services that impact both the participants and the expenditures in the states. We examined the plots of the residuals from the two regression equations and found that the pattern of error terms was almost evenly distributed within the band of plus or minus 2.5, showing that the estimates were not biased.

Data Sources for the Independent Variables
Sociodemographic Variables.-Secondary data on the percentage of persons aged 85 and over and the total population for each state were collected from the U.S. Bureau of the Census (USBOC;1991-1996a. The percentage of women in the labor force came from U.S. Department of Labor statistics (1991)(1992)(1993)(1994)(1995)(1996). Data on the percentage non-White and percentage living in metropolitan areas came from the U.S. Bureau of the Census (1991-1996b ; 1991-1996). The poverty rates came from the U.S. Bureau of the Census (1991)(1992)(1993)(1994)(1995)(1996)(1997).
State Policies.-The state reimbursement rates and methodologies for nursing facilities and for home health care were obtained from telephone surveys of state officials (Harrington, Swan, Wellin, Clemena, & Carrillo, 2000b). State policies on certificate of need and moratorium for nursing homes and for home health were also obtained from telephone surveys of state officials (Harrington, Swan, et al., 2000b). The state eligibility policies and the eligibility threshold for the medically needy were collected from the U.S. Health Care Financing Administration (HCFA;1991-1996a. Health Care Services.-The numbers of nursing home and residential care beds and the number of certified home health agencies were obtained from primary data collected from state officials (Harrington, Swan, Wellin, Clemena, & Carrillo, 2000a). The number of Medicare home health users were obtained from secondary HCFA Medicare data (U.S. Health Care Financing Administration, 1991-1996b). These data were standardized for each 1,000 Medicare beneficiaries in a state.

Data Sources and Data Collection of the Dependent Variables
Previous studies have examined the HCBS waiver programs in states primarily by using data from HCFA Form 64, the quarterly financial management data that states submit to the federal Medicaid program to obtain federal matching funds. These data do not include information on HCBS waiver participants (Burwell, 1999;Miller et al., 1999a). States are required to report on their waiver programs on an annual basis by filing HCFA Form 372. These forms contain information on waiver participants and expenditures for each waiver. Separate applications must be submitted to HCFA for approval of each waiver, which can be given for an initial period of 3 years and then renewed or extended for up to 5 years. Initial 372 reports are due 6 months after the end of the first year of a waiver. One year later (18 months after the end of each waiver year), states are required to submit "lag (i.e., final) reports" which include all revisions, adjustments, refunds, recoupments, cost settlements, disallowances, and other changes. Overall, the data from Form 372 were considered to be relatively accurate, even though there were discrepancies with the HCFA Form 64 data from the Medicaid HCBS expenditure claims. Because HCFA Form 372 is the only data available with participant data for the 1915(c) HCBS waiver program, there is no other comparison source.
Lists of state HCBS waivers were obtained from the HCFA waiver application lists (Salo, 1998). This study built on our initial effort in 1994 to collect Form 372 from the states for the year 1992. Between 1997 and 1999, we called all state Medicaid programs by tele-phone and sent faxes to collect HCBS waiver data for the study period. There was an average of 4 waivers per state for each year, so the data collection to reconstruct historical files from the states required a great effort. Overall, between 3-5 calls were made by study researchers to each state each year to collect the data for 1992-1997 period. The investigators were reasonably confident that all waivers had been identified and data either collected or estimated. Thus, the HCBS waiver data presented here represent the best available reports for the waiver information as well as the most recent complete data set of actual participants and expenditures.
Once the HCFA Form 372 reports were obtained, the data were coded and entered into a SAS database. States were asked to estimate data when the Form 372s were unavailable. The project collected data on a total of 155 waivers in 1992, which increased to 211 waivers in 1997. Data were obtained for a cumulative total of 1,111 waivers over the 6-year period. Of this total, 23% of total waivers only had initial reports available. Six percent of the total were estimated by states. Where states did not provide estimates, missing data for participants and expenditures (12% of the total) were estimated by the investigators. Because the data set was made up of cross-sectional time series data, linear interpolation was used to develop the estimates based on the other values provided by the states for the 6-year period. Table 1 shows that there were 211 total HCBS waivers in 49 states in 1997 (excluding Arizona and Washington, DC). Of the total, 75 waivers (36%) were targeted toward individuals with mental retardation or developmental disability, including both adults and/or children with MR/DD. Sixty waivers (28%) were targeted to the aged and/or disabled and 27 waivers (13%) were targeted to the disabled and physically disabled population. This group included not only those with functional or physical impairments but also those with functional impairments that were related to cognitive or mental impairment. There were 20 waivers (10%) targeted to children, primarily for children categorized as medically fragile or with special problems. Fifteen waivers (7%) were for AIDS/ARC, 12 waivers (6%) were for traumatic brain injury or head injury, and only 2 (1%) were targeted to those with mental health problems. All states had at least one waiver for MR/DD groups and some had 3-5 waivers for these groups and at least one or more waivers for the aged, aged/disabled, or disabled. The most uncommon waivers were for mental health, primarily because institutions for mental disease for the 22-64 year old population are not paid for by the Medicaid program. It was therefore more difficult for states to meet the cost neutrality requirements of the waiver program for the mentally ill than for other target populations that received institutional services. The total number of waivers varied from 2 (in Arkansas, Massachusetts, Montana, Oregon, and West Virginia) to 9 in New Jersey and New York and 10 in Colorado.

1992-1997 and Participants Per Capita
The total number of HCBS participants increased from 235,668 in 1992 to 561,510 in 1997, or by 138% (see Table 2). Some states like Alaska were slow to begin their programs (Alaska's started in 1994), whereas many states began waivers early in the 1980s after the federal legislation was passed. Thus, states with the highest growth rate during the 1992 to 1997 period tended to be those states that were catching up with other states that had more established programs. For example, Iowa had the highest growth (2,957% over the 6-year period) and reached the U.S. average for participants per 1,000 by 1997. Louisiana and Mississippi also had fairly high growth rates between 1992 and 1997, but were among the five lowest states in the nation in participants per capita.  Injury  Total   Alaska  2  1  0  1  0  0  0  0  4  Alabama  1  0  1  1  0  0  0  0  3  Arkansas  1  1  0  0  0  0  0  0  2  California  1  0  2  2  0  1  0  0  6  Colorado  5  Notes : MR/DD ϭ mentally retarded or developmentally disabled; AIDS ϭ acquired immunodeficiency syndrome; ARC ϭ AIDSrelated complex; TBI ϭ traumatic brain injury. Vol. 40, No. 6, 2000 A few states had high growth rates but were also among those states with high participants per capita. For example, Kansas had a growth rate of 395% and had the second highest number of participants per capita in the U.S. (5.92 per 1,000 population). Wyoming had a growth rate of 493% over the period and was eighth in the nation in participants per capita (3.63 per 1,000 population).
The top five states in HCBS waiver participants per capita were Oregon (7.91 per 1,000 population), Kansas (5.92 per 1,000 population), Rhode Island (5.79 per 1,000 population), Missouri (4.41 per 1,000), and Vermont (3.84 per 1,000 population; see Table 2). The lowest five states in HCBS waiver participants per capita in the nation were Indiana, Louisiana, Tennessee, Maryland, and Mississippi. Table 3 shows that total 1915(c) waiver expenditures increased from $2.17 billion in 1992 to $7.87   Table 4 for the independent variables used in the analysis. Separate regression models are shown for the two dependent variables: waiver participants per 1,000 population and waiver expenditures per 1,000 population (log). Table 5 shows the regression model for factors associated with state waiver participants per 1,000 population over the 1992-1997 time period. In terms of sociodemographic factors, the percentage of persons aged 85 and over in a state was a positive predictor of waiver participants. For every 10% increase in the percentage of the aged, the number of waiver participants increased by 17 participants per 1,000. None of the political factors were associated with waiver participation rates.

Growth in Expenditures by State in 1992-1997 and Expenditures Per Capita
States with higher personal incomes per 1,000 population had more waiver participants per 1,000 population. A $1,000 increase in per capita income resulted in an increase of 281 waiver participants per 1,000 population. States that used a certificate of need or moratorium on home health care agencies had lower numbers of waiver participants per 1,000 population.
The number of nursing home beds per 1,000 population in states was a strong negative predictor of the waiver participants. Increasing the bed supply by 100 beds per 1,000 population decreased the number of waiver participants by 22 participants. In contrast, the number of residential care beds per 1,000 population was a positive predictor. An increase of 100 residential care beds per 1,000 population increased the number of waiver participants by 16. The number of Medicare home health users per 1,000 Medicare beneficiaries in states was also a positive predictor of the waiver participants. For every additional 1,000 Medicare home health users in a state, the number of waiver participants increased by 12. Overall, the state effects alone predicted 81% of the variance, and the independent variables predicted 30% of the variance. The combined state and independent variables in the model predicted 90% of the variance. Table 5 shows the factors associated with 1915(c) waivers in the state in the 1992 to 1997 time period.

Factors Associated With State Waiver Expenditures Per Population
None of the sociodemographic factors were predictors of the amount of waiver expenditures. Of the political factors, having a democratic governor was a positive predictor of the amount of waiver expenditures. Personal income per 1,000 population was a positive predictor of the amount of expenditures. An increase of $100 in personal income per 1,000 population increased waiver expenditures by 14.5%.
In terms of public policies, using the Medicare home health reimbursement methodology increased the waiver expenditure rate. The dollar level of the eligibility thresholds for the medically needy was a positive predictor of the amount of state expenditures for waivers. For every $100 increase in the eligibility threshold, the waiver expenditures per 1,000 population in a state increased by 0.2%. Nursing home beds per 1,000 population were negative factors on waiver expenditures as well as on waiver participants. An increase in 100 beds per 1,000 population decreased waiver expenditures by 16%. An increase in certified home health care agencies per population was also a strong predictor of the amount of waiver expenditures. An increase in one home health agency per 1,000 population increased waiver expenditures per 1,000 population by 14%. The Medicare home health users per 1,000 Medicare beneficiaries was a positive predictor of the amount of state waiver expenditures. An increase in Medicare home health users by 100 increased the HCBS expenditures by almost 1%.
The independent variables predicted 37% of the variation across states. The state effects alone predicted 60% of the variance, with a combined effect of 72% of the variance explained by the random effects panel model.

Summary and Discussion
Overall, the states have expanded both the number of HCBS waiver participants (by 138%) and the expenditures (by 263%) substantially over the 6-year period of 1992-1997. In spite of the growth, the findings from this study confirm those from other studies that the growth in waiver expenditures has been very uneven across states (Miller, 1992(Miller, , 1999a(Miller, , 1999bLadd, Kane, & Kane, 1996;Kane et al., 1998). The states with the largest HCBS waiver programs were Oregon (7.9 participants per capita) and Kansas (5.9 participants). Fifteen states had more than 3 participants per capita, whereas 10 states had about 1 or fewer participants per capita. Expenditures across states also varied dramatically from $96 per capita in Vermont to $4 per capita in Mississippi. Individuals living in states with more waiver participants per capita were more likely to be able to remain at home or in the community and to avoid institutionalization. States with more expenditures per capita may receive more services or more expensive services. Thus, substantial inequities exist in access to HCBS services across the states.
Although previous studies have examined factors related to spending on home and community based services, this is the first study that has examined the factors associated with waiver participants in states. State variation in waiver participants and expenditures is related to a number of factors. As expected, the percentage of state population of persons aged 85 and over was a positive predictor of the number of waiver participants, responding to the great demand for services by the oldest old. Other sociodemographics and political factors did not predict waiver participation in states, but states with democratic governors were more likely to have higher waiver expenditures per population.
As expected, states with higher personal incomes had more waiver participants. This is probably because these states have more resources to pay for long-term-care services. This suggests that perhaps one policy approach to increase HCBS services is to increase the federal financial participation (FFP) rates (over the current levels) to states with low incomes as a means of encouraging these states to increase state waiver expenditures. Such a policy would require a statutory change in the Medicaid FFP rates. Another approach would be to offer special grants to low income states to help expand the number of participants and/or to build their HCBS waiver programs.
The findings suggest that states that used a certificate of need or moratorium on home health care agencies had lower numbers of waiver participants, probably because such policies restrict the supply of home care services. States that used Medicare home health reimbursement methods for the Medicaid program were associated with higher waiver expenditure levels. Medicare reimbursement methods are generally more generous that Medicaid rates (Buchanan et al., 1991). Medicare home health payment methods, however, did not translate into higher waiver participation rates.
Perhaps, the medically needy income criteria did not increase the number of waiver participants when controlling for other factors because most states had limits on their waiver slots and waiting lists for services . Increasing the Medicaid eligibility criteria for the medically needy did increase Medicaid waiver expenditures, because more individuals are allowed to spend down and receive waiver services. If state medically needy financial criteria were made more generous, more individuals would also be able to spend down to become eligible for institutional services. If given a choice, however, more individuals may choose the waiver services over institutional services if the waiver services were readily available.
As expected, health care service availability did have an impact on waiver participation and expenditures. The number of nursing home beds per 1,000 population in states was a strong negative predictor of both waiver participants and expenditure levels. High ratios of nursing home beds per population increases access to institutional services and consequently increases Medicaid costs (Harrington et al., 1997). If states want to expand access to waiver programs, one approach would be to reduce the ratio of nursing home beds per population. Many states have fairly low nursing home occupancy rates (84% average occupancy in 1998) and these have declined steadily from 88% in 1992 (Harrington, Carrillo, Thollaug, Summers, & Wellin, 2000). The number of nursing home beds per 1,000 aged population has also declined over the period. Thus, state reductions in nursing home beds would not appear to compromise access and may allow state policy makers to direct more funds to the waiver program. On the other hand, lowering the number of nursing home beds may be a difficult task to accomplish because of the political influence of the nursing home industry who would support increased Medicaid funds for institutional care rather than for HCBS services.
Interestingly, this study did not find that having a CON/moratorium for nursing homes had an impact on waiver participants or expenditures, when the model controlled for nursing home beds per population. This lack of association may be due to the delayed influence of CON/moratorium legislation, because CON/moratorium policies for nursing homes can only prevent future growth (Harrington et al., 1997). Perhaps CON/moratorium controls are best combined with other policies such as increasing the supply of residential care and home care services.
The number of residential care beds per 1,000 population was a positive predictor of waiver participation in states but not of waiver expenditures. This suggests that if states want to expand their waiver programs, one approach is to expand residential care beds per population as substitutes for nursing home beds. This may lower Medicaid costs, because residential care programs generally are less expensive than nursing home programs.
The expansion of home health agencies did not increase the number of waiver participants by a significant level but did increase the amount of waiver expenditures. It may be that states using independent home health care providers, rather than agency providers, are able to expand participation without increasing expenditures. This would be an important hypothesis to test in the future.
As expected, the number of Medicare home health users per 1,000 Medicare beneficiaries in states was also a positive predictor of the number of state waiver participants and the amount of expenditures. Perhaps, higher home health utilization indicates that the population has higher disability rates. Or perhaps high Medicare home health utilization is associated with more individuals needing long-term HCBS care or results in the identification of more individuals who need long-term-care services beyond those offered by Medicare. State policy makers probably have little influence over Medicare home health policies.
The 1915(c) Medicaid waiver program has proved its importance in providing long-term-care to individuals with severe disabilities and chronic illness since its inception in 1981. The program is a particularly popular and sought-after Medicaid program among those with disabilities, because it is a way to prevent institutionalization and to offer choice of long-termcare setting. Disability advocates have lobbied for the passage of legislation, such as the Medicaid Community Attendant Services and Supports Act of 1999, that would provide personal care in the home as an alternative to institutional care. Moreover, the Supreme Court decision in the Olmstead v. L.C. (1999) suggests that states must begin to address how to ensure that individuals have the option to remain in the community rather than in institutional care.
The data from this study show that states have been able to keep the average Medicaid waiver costs ($14,016) well below the average Medicaid institutional costs for long-term care. The average institu-tional cost per recipient for SNF and ICF-MR services combined were $23,225 in 1997 (Harrington, Swan, et al., 2000a). Institutional costs, of course, cover room and board expenses, but HCBS waiver services are not allowed to pay for such expenses. This issue is important because states must demonstrate that each waiver is no more costly than institutional care (cost neutral), and the states must ensure that every waiver participant meets the need criteria for institutional care. Thus, waivers are required by Medicaid statute to be direct substitutes for institutional care.
The lower waiver costs per participant than institutional costs are consistent with reports that have suggested that the HCBS waiver program has the potential for being cost effective. A 1996 study of Washington, Oregon, and Colorado concluded that the expansion of home and community-based services was cost effective in these states (Alecxih, Lutzky, & Corea, 1996). A 1994 study had similar conclusions about HCBS in Washington, Oregon, and Wisconsin when coupled with decreased institutional capacity (U.S. General Accounting Office, 1994aOffice, , 1994bWiener & Stevenson, 1997). More research is needed to determine whether the waivers are truly cost effective in the sense of adding better value for the money spent. This would involve comparisons of waivers with institutional services that take into account functional and mental status, quality of care, quality of life, and other measures.
The major problem with increasing the HCBS program is the potential cost implications for the Medicaid program, unless waivers are designed to substitute directly for institutional care (Snow, 1996;Wiener, 1996;U.S. General Accounting Office, 1999). Although some states already have extensive waiver programs, other states would have to expand their HCBS waiver programs, and that could result in greater costs (Wiener, 1996). Policy makers are concerned about the potential for new applicants for HCBS services who refuse institutional care. On the other hand, the general growth in the aged and disabled population will continue to increase the demand for long-termcare services over time. Increasing HCBS services may require less capital and other investment to meet the future demand than would increasing institutional care. All of these considerations must be taken into account by state and federal policy makers in trying to shape Medicaid long-term-care services.
In summary, the growth of the aging and disabled populations is difficult to address, but new policies related to HCBS resources can be developed. A new focus on expanding federal and state resources for HCBS services, especially for states with low personal incomes, could encourage these states to expand their programs. At the same time, removing the regulatory barriers to the growth of home care services and increasing reimbursement rates for home care may encourage the growth of home care providers. Policies that control the growth of nursing homes and expand residential care and home care, along with policies that increase the Medicaid medically needy eligibility criteria appear to be the most likely