## Abstract

A multivariate analysis was done to determine the relative importance of facility, resident, and community characteristics to expenditures. Facility factors associated with higher expenditures included ownership, facility size, facility services, and location. Individuals with a greater number of activity of daily living limitations, developmental disabilities, and more severe levels of mental retardation had higher expenses. Findings could improve our understanding of the costs of long-term residential care, assisting us to economically and effectively bring this population into the community. Data used are from the 1987 National Medical Expenditure Survey Institutional Population Component.

In the wake of large social, economic, and political changes over the past 40 years, the provision of residential services for persons with mental retardation and developmental disabilities has changed radically. This population, which was housed primarily in large institutional settings in the 1960s, has moved into community residential settings. In 1998, two thirds of the participants in Medicaid programs for persons with mental retardation were funded by Home and Community-Based Services (HCBS) Waiver programs and lived in some form of community setting (Lakin, Anderson, & Prouty, 1999). According to Anderson, Lakin, Mangan, and Prouty (1998), between 1987 and 1997, every state in the union reduced their population of persons with mental retardation and developmental disabilities (MR/DD) residing in state institutions. In fact, over half of residential service recipients now live in settings with 6 or fewer residents. In states such as Alaska, Michigan, New Hampshire, and Vermont, over 80% of individuals with MR/DD live in such facilities, whereas in at least 4 states, including Louisiana, Missouri, Kansas, and Virginia, the trend is much slower, with less than 20% living in households with 6 or fewer persons (Anderson, Polister, Prouty, & Lakin, 1997).

The huge success of the deinstitutionalization process since the late 1960s has masked several remaining problems related to residential services. First, a major problem hindering the continued movement of the persons remaining in institutions out of large congregate facilities is their age and level of impairment because of associated higher levels of care necessary to accommodate them in community settings. This highlights the underlying impact of cost, which influences the development of services for this population. Earlier researchers demonstrated cost savings for those moved out of institutions (Intagliata, Wilder, & Cooley, 1979; Jones & Jones, 1976; Mayeda & Wai, 1976; Murphy & Datel, 1976). However, results from more recent research reflect the increasing costs associated with the deinstitutionalization of the population of individuals with severe impairments, costs that more closely approximate those in institutional settings (Jones, Conroy, Feinstein, & Lemanowicz, 1984; Knobbe, Carey, Rhodes, & Horner, 1995; Nerney & Conley, 1992). These estimates are based either on small samples or on samples from specific areas of the country, which limit their generalizability to the national population using these services.

A second problem facing states and providers who want to supply community-based residential services is the large waiting lists for such services, estimated in 1997 to total between 83,000 and 90,000 persons—excluding those already in institutions (Anderson et al., 1998; Lakin, 1998; Prouty & Lakin, 1998). Combined with those remaining in institutions, approximately 57,000, a total of almost 150,000 persons are in immediate need of community residential facilities (Anderson et al., 1998). In addition to this population that needs services immediately, Ashbaugh (1999) has pointed to the demographics of the population of the United States that will intensify this problem over the next 20 years. He estimated that in 1996 there were almost 500,000 persons with developmental disabilities in the United States currently living with caregivers who were 60 years of age or above. Although it is likely that those on the waiting list to a large extent are a part of the population with aging caregivers, the list represents less than 20% of that potential tidal wave of need on the horizon. The 500,000 estimate also does not include the baby boomer cohort that has not yet reached the age 60 mark but will in the next 8 to 10 years (Ashbaugh, 1999). Based on these numbers, there is an urgent need for planning and financing to take care of the large population of persons with developmental disabilities who currently need community residential services as well as those who will require them in the upcoming years. Relative to the problem of those on waiting lists, Lakin (1998) predicted that “we are 9 years from chaos” (p. 157).

New and innovative solutions are called for to address these needs in a timely manner and at a reasonable cost. In 1998, Lakin estimated the need to increase the total national service capacity by 27%. In order to facilitate that planning, cost estimates should be based on the service needs of the individual rather than the organizational costs of service systems that are already in place, some of which are being phased out. Current trends toward person-centered community services, particularly those provided by the HCBS Waiver system, necessitate a more in-depth understanding of the specific needs of the population to be served and the resources available in the communities than has been previously considered.

Much of what we know about costs for services for the MR/DD population is based on analyses of data provided at the institutional or programmatic level. National research has demonstrated that as the institutional population decreases, the institutional cost per client increases, primarily because of the fixed costs of maintaining an institutional setting as well as government regulations that require increased staffing (Braddock, Fujiura, Hemp, Mitchell, & Bachelder, 1991; Braddock, Hemp, & Fujiura, 1987). In addition, the remaining population, more likely to have multiple problems, is generally a population that would generate higher expenditures no matter where they are located. Comparisons of institutional and community care expenses in this context are suspect because of the factors contributing to increasing institutional costs.

In this paper we have attempted to demonstrate the need for and usefulness to the current planning process of data collected for individuals residing in facilities for people with mental retardation at the person-level rather than data collected at the organizational level (highly aggregated data). Using data collected at the organizational level to try to solve the kinds of problems facing states today limits the possible solutions to the forms of organizations already in place. Focusing instead on the types of persons for whom services need to be developed allows for more flexibility to ask questions outside the limits that organizational structures impose. National data, which reflect the variety of contexts within which persons with developmental disabilities find themselves, also allow for analysis that examines the interplay of individual characteristics with the type of facility and other contextual factors, such as regional differences in income, employment, and other factors that can contribute to increased expenses for one client versus another. Ideally, this national data would represent all persons with developmental disabilities rather than just those in residential settings or only those who receive Medicaid funding.

Understanding how both resident and facility characteristics are associated with expenditures (for purposes of these analyses, expenditures refers to payments received by the facility for care provided) is an important first step necessary for planning and expanding public policy for institutional and residential care. Ashbaugh and Nerney (1990) found, for example, that among residents living in Michigan and Nebraska, personal characteristics were not a factor and accounted for very little variation in residential expenses. At the same time, they did find that expenses varied largely as a function of the type of residential arrangements, particularly staffing levels, which generally are predicated on the level of client need. The work of Campbell and Heal (1995) is a good example of examining all the factors that contribute to the costs of providing these services in a manner that can enlighten the ongoing planning needed to accomplish a response to the problems on the horizon. Looking at data for one state (South Dakota), they found that local characteristics contributed about 11% of the variation in costs, client characteristics contributed about 62%, and provider characteristics about 12%. A drawback to both analyses is that because the samples are regional, they do not necessarily represent the national picture, nor do they allow for all of the variety of the residential approaches that exist. Although Campbell and Heal incorporated more useful variables, their results also depend on mean costs and do not allow for the variations that could be associated with actual individual data.

In the present analysis we sought to offer a way around those two drawbacks to analysis of cost issues. Our first objective was to provide nationally representative data about persons with mental retardation living in a variety of residential settings, using information about specific randomly chosen individual residents in order to develop data generalizable to the mentally retarded residential population as a whole. Even the expenditure data were captured at the individual level, allowing us to avoid the use of means or any other constructs. A second objective was to pinpoint which characteristics of the individual, the facility context, as well as the local economic situation contribute most to expenditures. We did not compare costs of different types of residential arrangements per se, nor did we limit the types of residential settings (other than the limitation in bed size inherent in the data) or the areas where the residences are located. The focus of our analysis was the individual and what it costs to provide a certain type of individual with residential services and ways in which the characteristics of the individual in certain types of facilities or certain locations combine to reduce or increase costs. Using regression analysis, we have sought to predict which characteristic or combination of characteristics of the individual, the facility, and the location combine to increase or decrease the per diem costs for providing the needed services. We used the results of this analysis to demonstrate how both national data and data focused on the client rather than the organization can be useful to seeking solutions to the residential and service gaps that are building to crisis proportions.

## Method

### Data Sources

Data for this study are from the Institutional Population Component of the 1987 National Medical Expenditure Survey (hereafter called the Expenditure Survey), which was sponsored by the Agency for Healthcare Research and Quality (AHCQ). (Though the age of the data may be questioned, these are the only data available at this point in time that provide a sample that, when weighted, provides nationally representative estimates of the residents with mental retardation in facilities with 3 or more beds.)

The purpose of the Expenditure Survey is to provide national estimates for health care use and expenditures for persons residing in or admitted to facilities for persons with mental retardation during 1987 (Edwards & Edwards, 1989). Although the data were collected before much of the population in Intermediate Care Facilities for the Mentally Retarded (ICF/MR) moved into community facilities, the data do include person-level data for a sample that resided in facilities of 3 or more beds. The only members of the population not included are those who were living in facilities with 1 or 2 beds or those who were living in their own homes or with their parents. All facilities of 1 or 2 beds were excluded because of undercoverage in the sampling frame. Approximately 80% of the facilities in this study were 3 to 15 beds, with 20% of the facilities selected from the larger institutions. Approximately 29% of the sample respondents were residents in the smaller facilities. Although this does not reflect the current arrangements where the majority of persons in residential settings are living in small community facilities, it does provide a large enough sample of such a population on which to do analysis. There is no national survey that includes persons in all forms of residential settings. Surveys of such residential settings today require a complete sampling frame that can be used to identify all known facilities that fulfill the definition. Such a sampling frame is almost impossible to construct because of the movement into and out of existence of many of the very small residential settings. This precludes the ability to draw a truly representative national sample of such residential settings at a reasonable cost and makes fielding such a survey both very expensive and probably inaccurate. In addition, more recent work does not provide data at the person-level, but, rather, data being aggregated at some higher level (e.g., by state).

The sampling frame was developed by updating the 1982 National Census of Residential Facilities (Cohen, Potter, & Flyer, 1993). Selection was restricted to facilities that served 3 or more clients and was based on a 3-stage probability design, with facilities being selected in the first two stages. Criteria for selection were as follows: The facility had to be (a) a place or unit certified as an ICF/MR by Medicaid or (b) a state-licensed place or unit, with 3 or more beds for clients who reside there, providing services to persons with mental retardation with 24 hours a day supervision, 7 days a week; but not a licensed hospital (unless a hospital for persons with mental retardation) and not a family providing services exclusively to family members. Samples of current residents (residents on January 1, 1987) and new admissions (admitted between January 1 and December 31, 1987) were selected during the third stage of sampling (Cohen et al., 1993). A total of 3,920 (225,000 weighted) individuals, from 665 (12,050 weighted) facilities, were selected for this study.

Data on individuals residing in facilities for persons with mental retardation were obtained from a baseline questionnaire administered to facility staff responsible for direct patient care to the sample person or other designated staff. Medical records were also examined. Data collected included physical and mental health status and functional limitations of sample persons (activities of daily living), presence of developmental disabilities, other impairments (such as hearing and vision loss), diagnoses for medical conditions, demographic characteristics, and vital status of the sample person's parents, siblings, and children (Edwards & Edwards, 1989).

Information on use and related expenditures was collected in an institutional use and expenditure questionnaire that was given several times during the year. Information collected included the amount of expenditures and sources of payment (Medicare, Medicaid, Supplemental Security Income, Veteran's Administration, own or family income, private health insurance, prepayment to a continuing or life care community, and other—state or local government and charitable organization). Facility characteristics (region, ownership type, facility size, basic and additional services provided) were obtained through the facility questionnaire from information provided by facility administrators and designated staff in sampled facilities (Edwards & Edwards, 1989).

### Procedure

Resident characteristics obtained (see Table 1) included age, gender, race, 1986 individual annual income, and parental vital status (used in multivariate analysis only). Residents were grouped into those with both parents living, those with at least one living parent, and those with neither parent living or the vital status of both parents unknown. Respondents also indicated whether residents had any difficulties in walking in addition to performing activities of daily living because of physical or mental problems. Activities included bathing or showering, dressing, using the toilet, getting in or out of bed or chair, and feeding oneself. Level of mental retardation was reported in one of four categories: borderline/mild, moderate, severe, or profound. In addition, it was necessary to create a fifth category for individuals whose level of mental retardation was not specified. These individuals often had mental illness problems, few if any limitations in activities of daily living, and were assumed to be functioning in at least a moderate level and, thus, were included in the moderate category in the analysis. In addition, respondents were asked whether residents had any developmental disabilities, including epilepsy, cerebral palsy, autism, and spina bifida. Respondents were also asked to identify problem behaviors, such as getting upset, behavior hurtful to self or others, dressing inappropriately, crying for long periods, hoarding, getting lost or wandering, stealing, exposing oneself, and inability to avoid danger (Edwards & Edwards, 1989).

Table 1

Independent Variables

Facilities characteristics (see Table 1) ascertained were location, as represented by regions of the United States (East, Midwest, South, and West), ownership type, and facility size as measured by number of beds. In addition, facility administrators or their designated staff members were asked to answer either yes or no to a series of questions pertaining to services included in the facility's basic charge. Services considered were medical, physical, occupational and speech/hearing therapy; mental and dental health; test or x-ray services; and other medical services (including respiratory and intravenous therapies).

Also, the number of additional services offered was determined. The facility administrator or designated staff member was asked to answer either yes or no to a series of questions pertaining to services routinely provided on site, but not included in the basic charge. Services included self-care; independent skills or academic training; and recreation, social, and case management services.

One county and one state level variable were also included in the analysis: per capita income (county level) and a measure of state generosity (state level) for the year 1987 (Braddock, Hemp, Fujiura, Bachelder, & Mitchell, 1990; Health Resources, 1989). State generosity was calculated by dividing the amount of state plus federal funding (Braddock et al., 1990) for programs pertaining to persons with mental retardation by the number of persons with mental retardation receiving residential services in that state (Braddock et al., 1990) to give a state funding level per person for this special population. Such funding consisted of both state and federal funds allocated to MR/DD services for both congregate sites (16+ beds) and community sites (those with 15 beds or less). Finally, an interaction term of level of mental retardation and size of facility was included to control for the fact that persons with severe or profound levels of mental retardation are more commonly found in large facilities.

Table 2 presents descriptive results for the entire population. In addition, daily expenses were compared by level of mental retardation in Table 3 and two broad facility size categories were compared in Table 4: those with 3 to 15 beds (small) and those with 16 or more beds (large). This particular distinction between small and large facilities is found in the original ICF/MR standards, in the Life Safety Code of the National Fire Protection Association, and in other federal standards; it is commonly used in the MR/DD field. Facilities with less than 16 beds are generally the most homelike and differ significantly from facilities with 16 or more beds (Bruininks & Lakin, 1985; Lakin & Hall, 1990; Rotegard, Hill, & Bruininks, 1983). Small facilities are assumed to be in the community or community-based. Large facilities are assumed to be institutional. This dichotomy remains one of the most useful ways of differentiating facilities (Lakin & Hall, 1990). The difference between mean values for daily expenditures was tested using student's t test; two-tailed tests were conducted at the .05 level.

Table 2

Resident and Facility Characteristics: United States, 1987

Table 3

Range and Mean Daily Expenditures in Facilities by Resident and Facility Characteristics: United States, 1987

Table 4

Mean Daily Expenditures in Facilities by Resident Characteristics and Facility Size: United States, 1987

In order to determine the relative importance of facility and resident characteristics and community resources, we conducted a multivariate analysis after completion of preliminary bivariate analyses; the dependent variable was daily expenditures (again, for purposes of this analysis expenditures refers to payments received by the facility for care provided). A multivariate approach provides a more stable base for hypothesis testing and theory construction (Spiker, 1982).

The model for multiple linear regression is given by the following equation:

px = β0 + β1x1 + β2x2 + … + βpxp

The coefficient β1 expresses how much the dependent variable, daily expense, would change for unit change in x1, on the assumption that all remaining independent variables are held constant (Branch, 1984; Pindyck & Rubinfeld, 1981; Zar, 1984). The recommended approach to model building involves seeking the most parsimonious model that still explains the data. By minimizing the number of variables in the model, the resultant model is more likely to be numerically stable and is more easily generalized (Hosmer & Lemeshow, 1980).

In the current research we examined the degree of association of 35 independent variables (and 5 interaction terms) found, based on previous research, to be associated with daily expenditures that were either available from questionnaire items in the Institutional Population Component of the Expenditure Survey or other secondary data sources as described. The decision to include or exclude particular independent variables was made in conjunction with subject matter knowledge. Model development was not based merely on the use of tests of statistical significance (Hosmer & Lemeshow, 1989). All independent variables were entered into the regression model simultaneously.

The independent variables selected for this analysis can be grouped into the following categories: sociodemographic variables (age, gender, race); financial variables (individual income and county level per capita income); social support (parental vital status); health status (level of mental retardation, functional and developmental disability, and behavioral problems); facility characteristics (region, facility size, ownership type, and number of services offered); and state funding (state generosity). For the purpose of the multivariate analysis, we kept facility size as a continuous variable.

The model derived using multiple regression was used to estimate the total association between the independent variables and the dependent variable, the coefficient of multiple correlation. When squared, it gives the coefficient of multiple determination, R2, which indicates the proportion of the total variation in the dependent variable that is explained by the collective set of independent variables considered (Lutz, 1983). However, an adjusted R2 was determined to take into account the number of independent variables used in the regression (Pindyck & Rubinfeld, 1981).

In multivariate model development, there is the concern for multicollinearity. The major undesirable consequence of multicollinearity is that the standard errors of partial regression coefficients of the collinear variables may be quite large, meaning that the β1s are imprecise estimates of the relationships in the population and results are unstable. In the presence of multicollinearity, the interpretation of the coefficients is difficult (Breslow & Day, 1980; Kennedy, 1979).

In practice, multicollinearity among independent variables is ignored if it is of low magnitude, but if the intercorrelation is high (about .8 or .9 in absolute value), conclusions regarding the significance of the correlated x1s may be in question and indicates corrective action is necessary (Breslow & Day, 1980; Kenney, 1979; Pindyck & Rubinfeld, 1981; Zar, 1984).

In conducting significance tests for means (bivariate analyses) and β coefficients (multivariate analyses), we considered the complex survey design of the Expenditure Survey Institutional Population Component using SUDAAN. The application of SUDAAN entailed using two additional pieces of information, the primary sampling strata and the primary sampling units. These additional pieces of information were used to compute estimates of the variance and covariance to more closely reflect sample design effects (Shah, Barnwell, & Bieler, 1995; Skinner, Holt, & Smith, 1989).

## Results

Table 2 presents descriptive results for residents and facilities. The resident results are presented at the person-level; one record represents one sample person, whereas results for facilities are presented at the facility-level; one record represents one facility. The majority of residents were White males between the ages of 22 and 44, with an income below $5,000 who had 2 or more behavior problems. For facilities the majority were between 3 to 15 beds, with a fairly even distribution between profit and nonprofit ownership. In comparing those sample persons with higher levels of functioning (borderline/mild, moderate, and not specified) to those having severe or profound mental retardation, we found that the mean daily expense was significantly greater for those with severe or profound mental retardation. This is the case for all resident and facility characteristics (see Table 3, results are at the person-level). When comparing daily expenses for facilities with 16 beds or more versus facilities with 3 to 15 beds, we also found a consistent pattern. The mean daily expense, in almost all cases, was higher in the larger facility (see Table 4, results are at the person-level). For residents' characteristics (age, gender, race, and income), values were all statistically higher in facilities with 16 or more beds. With respect to health status measures (activities of daily living limitations, level of mental retardation, developmental disabilities, and problem behaviors), values were again all statistically higher in facilities with 16 or more beds. For region and facility ownership the pattern was the same, with the exception of for-profit facilities. The trend is clear and consistent: Care costs more in facilities with 16 beds or more as compared to facilities with 3 to 15 beds. In an effort to explain observed patterns, we took into consideration resident health care needs (e.g., level of mental retardation and number of activities of daily living limitations) and services included in the basic charge. With respect to health care needs, differences in the level of mental retardation by facility size is apparent (data not shown). The percentage of residents in larger facilities who had severe or profound mental retardation was 57% versus 32% in facilities with 3 to 15 beds. In contrast, individuals with borderline/mild, moderate, or mental retardation not specified made up over one half of the population of smaller facilities and less than one third of the population of facilities with 16 or more beds. In addition, a higher percentage of the residents of facilities with 16 or more beds had two or more developmental disabilities than did residents of smaller facilities (43% vs. 28%, respectively). Residents of larger facilities were also more likely to have a greater number of limitations in activities of daily living. Approximately two fifths of the residents of these types of facilities had difficulty with three or more activities of daily living as compared to less than one fifth of residents of smaller facilities. Slightly over one half of those residing in facilities with 3 to 15 beds reported no activities of daily living limitations at all; this compares to less than one third of residents residing in larger facilities. With respect to problem behaviors, estimates suggest that problem behaviors were present in about the same proportion in either size facility. A little over one half of the residents of both sizes of facilities were reported to have two or more problem behaviors. Despite this similarity, findings show that the severity of impairment and, thus, the need for greater levels of care is higher among residents of facilities with 16 beds or more. Although larger facilities had greater daily and annual expenses, they also tended to have residents who needed more care. Without exception, facilities with 16 or more beds included a greater number of services in their basic charge as compared to facilities with 3 to 15 beds (see Table 5, results are at the facility level). Daily expenditures for facilities with 16 or more beds was consistently higher than those for smaller facilities, but the larger facilities clearly bundled more health services into the basic charge. Table 5 Services Included in Basic Charge by Facility Size: Proportion Reporting Yes #### Multivariate analysis Person-level multivariate analyses show that both facility characteristics and resident characteristics are important arbiters of daily expenditures (see Table 6). In addition, some community resources, at least as conceptualized and quantified here, also have a role. Facility factors with significant β coefficients, associated with higher daily expenditures, included residence in nonprofit or government facilities, larger size facilities (defined as the number of beds, continuous), facilities with a higher number of services included in the basic charge, and a higher number of additional services routinely provided. The location of the facility was also associated with increased rates if the facilities were in the Northeast or Midwest as compared to the West. Higher per capita county income of residence was also associated with increased daily expenditures. Table 6 Regression Results, Determinants of Daily Expense (N = 225,000) Resident characteristics were associated with higher daily expenditures as well. Younger residents (less than 22) and those having greater needs for assistance (2 or more activities of daily living limitations) had greater daily expenses. Residents with ambulation problems, developmental disabilities, or behavior problems did not have significantly different expenses than did residents without those problems. Level of mental retardation, however, was a resident characteristic significantly associated with daily expenditures. Residents with severe or profound levels of mental retardation had significantly greater daily expenditures than did persons with borderline/mild levels of mental retardation. Residents with greater financial resources (income greater than$5,000) had lower daily expenditures. After controlling for facility characteristics, levels of mental retardation, age, and need for assistance or control, we found that non-White residents had higher daily expenses than did White residents.

The β coefficients of the interaction terms also indicate a significant effect of the combination of facility size and level of mental retardation above and beyond the effects of facility size and level of mental retardation alone. Compared to the daily expense of a resident with borderline/mild retardation living in a larger facility, expenditures were significantly less for persons with unspecified or profound levels of mental retardation and greater for persons with moderate mental retardation. The β coefficients indicate that for persons with profound levels of mental retardation, increased expenses would be minimal when moving individuals to smaller facilities, and savings (or at least no additional cost) could be realized for persons with borderline, mild, moderate, and severe levels of mental retardation. The adjusted R2 equaled .54, and there was no indication of multicollinearity (as discussed in the Method section) among the independent variables.

## Discussion

The results of the multivariate analysis indicate, at a national level, what Campbell and Heal (1995) found in South Dakota. Facility characteristics, resident characteristics, and even community resources play a part influencing daily expenses for residents in facilities both large and small. Smaller facilities with for-profit ownership, fewer basic services, and additional services located in Western states appear to have the least expensive daily rates for resident care. Residents under age 22, with deficits in two or more activities of daily living skills, and profound levels of mental retardation are more expensive to care for on a daily basis. The results also show that for persons with borderline, mild, moderate, or severe levels of mental retardation, it is more expensive to provide care in larger facilities. For individuals with profound mental retardation, the size of the facility is not a factor in daily expenses once the increased expenses for the level of mental retardation are considered. The level of affluence of a community (as measured by county per capita income), but not the generosity of state programs, plays a part in the level of daily expenses for residents.

Other personal characteristics are associated with daily expenses for persons in facilities as well. A surprising finding is that expenditures for White individuals were lower than for non-White residents. This may be caused by interaction effects not taken into account, such as the possibility that a higher number of African Americans were residents in facilities in the Northeast (a higher cost region) than in other areas of the country, or it may indicate that African Americans were more likely to live in larger facilities, again more expensive. This issue warrants further examination.

A resident's level of income is an additional factor that affects expenses; higher levels of income are associated with reduced expenses. The exact nature of the effect of income is not clear. Having controlled for level of mental retardation and functional abilities, we found that the income does not necessarily represent higher levels of functioning. It may, however, represent the ability of the residents to provide for some of the sources of cost (such as medical care or personal spending) on their own, thus reducing the individual's cost to the facility.

It is also interesting to note the characteristics that might be considered to be associated with increased expenditures but that are not. The client's mobility was not a factor, even though care for nonmobile individuals is difficult and nonmobility can be detrimental to health. The number of developmental disabilities that an individual had also does not seem to relate to expenditures, although the health complications of multiple disabilities might be considered a source of greater expense.

## Conclusion

This analysis demonstrates a way to examine some of the factors that are involved in the increasing expenditures for residential care for persons with mental retardation. However, because of the age of the data, some of the changes in the delivery of residential services may not be accurately reflected. Nonetheless, it presents a model worth exploring and makes a case for seeking nationally representative data on this population. In addition, the analysis probably raises as many questions as it answers and points to some further data deficits that limit our ability to address the policy issues associated with residential care. The following are just a few of the important findings and possible ways expansion of this research could improve our understanding of the costs of long-term residential care for this population, allowing us to more economically and effectively bring this population into the community.

Clearly, government facilities are more expensive, but care must be taken to discriminate among the factors that can influence expenses because expenses may be associated with the costs of a larger sized facility rather than just government management. Government facilities have larger physical plants to maintain, higher salary and benefit levels, and mandated staff:client ratios to accommodate. This kind of information needs to be included in future comparisons to test the differences in expenditures between government ownership and other ownership types.

Geographic location was found to be another strong influence on daily expenditures for residents. Although the wealth of communities (as represented by county per capita income) was a factor, the generosity of the state funding for mental retardation services was not. What are the factors driving the geographic differences? Such factors could be employment rates, wage levels, cost of living, and other economic characteristics. What, if anything can be done about the large geographic discrepancies in expenditures noted in this analysis? What forms of facility organization or techniques of management are transferable from the West to the Northeast that could moderate these differences? What are the effects of community resources, untapped in this analysis, that may be a part of this geographic effect?

Increased service levels are associated with greater expenditures, as demonstrated here. However, in communities with established health care and rehabilitation systems, are separate systems for residential facilities necessary? Can integration into existing health organizations help with savings? Does such integration carry a cost for clients, particularly the more fragile population with severe impairments whose health care needs are not familiar to general medical providers? Will medical providers who need to update or improve their skills for this population be willing to do so and at whose cost?

The results of this analysis also demonstrate that several resident characteristics are associated with higher daily expenditures, including level of mental retardation and number of activities of daily living skills. Persons with similar levels of dependence had different daily expenses, related to their level of mental retardation and, thereby, the ability to cooperate and communicate with caregivers. Though other resident characteristics, such as age and race, are not primary factors associated with expenditures for residential care, they are part of the equation and need to be explored. The present results are focused on resident and facility characteristics, ignoring two important aspects of long-term care for persons with mental retardation, namely, quality of care and the effect of case-mix in a facility. We know little about how the level of heterogeneity of impairments within the variety of facilities impact on individual outcomes or expenses. Large government facilities tend to have mostly residents with severe impairments, whereas small community facilities also tend to have more homogeneous groups at higher or lower levels of impairment. Would a placement approach that promoted a more heterogeneous population reduce the labor intensive problems associated with caring for persons with high levels of impairment and dependency? Could higher functioning individuals be a source of stimulation and assistance for lower functioning individuals, and would this in any way moderate expenses?

The quality of care is another area that has not been addressed in assessing the expenses of care and is one that should be of concern, particularly in today's good job market where many jobs are available with better pay, better working conditions, and more benefits. As noted by Duker, Seys, Van Leeuwe, and Prins (1991), employee turnover rates in residential settings point to a possible problem associated with working conditions and rewards that could have an impact on quality of care for residents.

As we face the pressing need to respond to persons on waiting lists, the continued need to move residents from institutions to the community, and preparation for the large group of persons with developmental disabilities who will be leaving aging parents over the next 10 to 20 years, planning is essential. Guiding that planning process, we should use all the data that are relevant and available and develop new data sources at the person-level. Though the Expenditure Survey data are from 1987, they can be used in a variety of ways to assist in the upcoming planning process. Expenditures information can be adjusted to provide current dollar equivalents. Analysis can be limited to the sample of community residents. Other area information, such as employment levels and number of physicians and hospitals, can be appended to the Expenditure Survey data using the Area Resource File. The Expenditure Survey can also be used as a model of the kind of data, with individual information and expenditures, to be collected in other studies, either local or national.

As the need for community living arrangements increases, it is important to understand how organizational type, resident characteristics, number and types of services, and location come together to influence expenditures in order to develop the necessary resources for proposed health care delivery plans. Examining expenses from the individual rather than the organizational perspective allowed us to examine this complicated puzzle in a different way. It provided us with some new insights into how both organizational and individual characteristics contribute to expenditures as well as generate new, important questions about the impact of location and area resources on solutions to expanding services and controlling costs.

Table 2

Continued

Table 3

Continued

Table 4

Continued

Table 6

Continued

## References

References
Anderson
,
L.
,
K. C.
Lakin
,
T.
Mangan
, and
R.
Prouty
.
1998
.
State institutions: Thirty years of depopulation and closure.
Mental Retardation
36
:
431
443
.
Anderson
,
L.
,
B.
Polister
,
R.
Prouty
, and
K. C.
Lakin
.
1997
.
Services provided by state and nonstate agencies in 1996.
In R. Prouty & K. C. Lakin (Eds.), Residential services for persons with developmental disabilities: Status and trends through 1996. Minneapolis: University of Minnesota, Institute on Community Integration
.
Ashbaugh
,
J.
and
T.
Nerney
.
1990
.
Costs of providing residential and related support services to individuals with mental retardation.
Mental Retardation
28
:
269
273
.
,
D.
1999
.
Aging and developmental disabilities: Demographic and policy issues affecting American families.
Mental Retardation
37
:
155
161
.
,
D.
,
G.
Fujiura
,
R.
Hemp
,
D.
Mitchell
, and
L.
Bachelder
.
1991
.
Current and future tends in state-operated mental retardation institutions in the United States.
American Journal on Mental Retardation
95
:
451
462
.
,
D.
,
R.
Hemp
, and
G.
Fujiura
.
1987
.
National study of public spending for mental retardation and developmental disabilities.
American Journal of Mental Deficiency
92
:
121
133
.
,
D.
,
R.
Hemp
,
G.
Fujiura
,
L.
Bachelder
, and
D.
Mitchell
.
1990
.
The state of the states in developmental disabilities.
Washington, DC: American Association on Mental Retardation
.
Branch
,
L.
1984
.
Relative risk rates of nonmedical predictors of institutional care among elderly persons.
Comprehensive Therapy
10
:
33
40
.
Breslow
,
N.
and
N.
Day
.
1980
.
Statistical methods in cancer research, Volume 1, The analysis of case-control studies (IARC Scientific Publications No. 32.).
Switzerland: World Health Organization, International Agency for Research on Cancer
.
Bruininks
,
R. H.
and
K. C.
Lakin
.
1985
.
Living and learning in the least restrictive environment.
Baltimore: Brookes
.
Campbell
,
E.
and
L. W.
Heal
.
1995
.
Prediction of costs, rates and staffing by provider and client characteristics.
American Journal on Mental Retardation
100
:
17
35
.
Cohen
,
S.
,
D.
Potter
, and
P.
Flyer
.
1993
.
Sample design of the National Medical Expenditure Survey: Institutional Population Component (AHCPR Publication No. 94–0017, National Medical Expenditure Survey, Methods 6).
Rockville, MD. Agency for Health Care Policy and Research, Public Health Service
.
Duker
,
P.
,
D.
Seys
,
J.
Van Leeuwe
, and
L.
Prins
.
1991
.
Occupational conditions of ward staff and quality of residential care for individuals with mental retardation.
American Journal on Mental Retardation
95
:
388
396
.
Edwards
,
W.
and
B.
Edwards
.
1989
.
National Medical Expenditure Survey: Questionnaires and data collection methods for the institutional population component, Methods 1 (DHHS Publication No. PHS 89–3440).
Washington, DC: Department of Health and Human Services, Agency for Health Care Policy and Research, Public Health Service
.
1989
.
Area Resource File (ARF) System.
Information for health resources planning and research (ODAM Rep. No. 7–89). Washington, DC: U.S. Department of Health and Human Services, Health Resources and Services Administration, Public Health Service, Bureau of Health Professions, Office of Data Analysis
.
Hosmer
,
D.
and
S.
Lemeshow
.
1980
.
Goodness of fit tests for the multiple logistic regression model.
Communications in Statistics, A
9
:
1043
1069
.
Hosmer
,
D.
and
S.
Lemeshow
.
1989
.
Applied logistic regression.
New York: Wiley
.
Intagliata
,
J.
,
B.
Wilder
, and
F.
Cooley
.
1979
.
Cost comparison of institutional and community-based alternatives for mentally retarded persons.
Mental Retardation
17
:
154
156
.
Jones
,
P.
,
J.
Conroy
,
C.
Feinstein
, and
J.
Lemanowicz
.
1984
.
A matched comparison study of cost-effectiveness: Institutionalized and deinstitutionalized people.
Journal of the Association for the Severely Handicapped
9
:
304
313
.
Jones
,
P.
and
K.
Jones
.
1976
.
The measurement of community placement services and its associated costs (Interim Rep. No. 1–5).
Waltham, MA: Brandeis University
.
Kenney
,
P.
1979
.
A guide to econometrics. Cambridge: MIT Press.
Knobbe
,
C.
,
S.
Carey
,
L.
Rhodes
, and
R.
Horner
.
1995
.
Benefit-cost analysis of community residential services versus institutional services for adults with severe mental retardation and challenging behaviors.
American Journal on Mental Retardation
99
:
533
541
.
Lakin
,
K. C.
1998
.
On the outside looking in: Attending to waiting lists in systems of services for people with developmental disabilities.
Mental Retardation
36
:
157
162
.
Lakin
,
K. C.
,
L.
Anderson
, and
R.
Prouty
.
1999
.
Medicaid HCBS “Waiver” recipients are now twice the number of Medicaid ICF/MR residents.
Mental Retardation
37
:
341
343
.
Lakin
,
K. C.
and
M. J.
Hall
.
1990, December
.
Medicaid financed residential care for persons with mental retardation.
Health Care Financing Review, 1990 Annual Supplement, 149–160
.
Lutz
,
G.
1983
.
Understanding social statistics.
New York: MacMillan
.
Mayeda
,
T.
and
F.
Wai
.
1976
.
The cost of long-term developmental disability care,.
Washington, DC: Department of Health, Education, and Welfare
.
Murphy
,
J.
and
W. A.
Datel
.
1976
.
Cost-benefit analysis of community versus institutional living.
Hospital & Community Psychiatry
27
:
165
170
.
Nerney
,
T.
and
R.
Conley
.
1992
.
A policy analysis of community costs for persons with severe disabilities.
Journal of Disability Policy Studies
3(2)
:
32
51
.
Pindyck
,
R.
and
D.
Rubinfeld
.
1981
.
Econometric models and economic forecasts (2nd ed.).
New York: McGraw-Hill
.
Prouty
,
R.
and
K. C.
Lakin
.
(Eds.).
1998
.
Residential services for persons with developmental disabilities: Status and trends through 1997.
Minneapolis: University of Minnesota, Research and Training Center on Community Living, Institute on Community Integration
.
Rotegard
,
L.
,
B.
Hill
, and
R.
Bruininks
.
1983
.
Environmental characteristics of residential facilities for mentally retarded persons in the United States.
American Journal of Mental Deficiency
88
:
9
56
.
Shah
,
B.
,
B.
Barnwell
, and
B.
Bieler
.
1995
.
SUDAAN user's manual: Software for the statistical analysis of correlated data.
Research Triangle Park, NC: Research Triangle Institute
.
Skinner
,
C.
,
D.
Holt
, and
T.
Smith
.
1989
.
Analysis of complex surveys.
New York: Wiley
.
Spiker
,
V.
1982
.
An analysis of factors antecedent to the institutionalization of the elderly (Working Paper No. 95).
Chicago: Northwestern University, Center for Health Services and Policy Research
.
Zar
,
J.
1984
.
Biostatistical analysis (2nd ed.).
Englewood Cliffs, NJ: Prentice-Hall
.

## Author notes

Authors:Jeffrey A. Rhoades, PhD, Survey Statistician, Agency for Healthcare Research and Quality, 2101 E. Jefferson St., Suite 500, Rockville, MD 20852 ( jrhoades@ahrq.gov). Barbara M. Altman, PhD, Special Assistant for Disability Statistics, National Center for Health Statistics, 6525 Belcrest Rd., Room 730, Hyattsville, MD 20782.