This article examines expenditures for a random sample of 1,421 adult Home and Community Based Services (HCBS) and Intermediate Care Facility/Mental Retardation (ICF/MR) recipients in 4 states. The article documents variations in expenditures for individuals with different characteristics and service needs and, controlling for individual characteristics, by residential setting type, Medicaid program (ICF/MR or HCBS), and state. Annual average per-person Medicaid expenditures for HCBS recipients were less than those of ICF/MR residents ($61,770 and $128,275, respectively). HCBS recipients had less severe disability (intellectual, physical, health service needs) than ICF/MR residents. Controlling these differences, and for congregate settings, HCBS were less costly than ICFs/MR, but this distinction accounted for only 3.3% of variation in expenditures. Persons living with families receiving HCBS ($25,072) and in host families (including foster, companion, or shared living arrangements; $44,112) had the lowest Medicaid expenditures.
Two large Medicaid programs provide long-term services and supports to persons with intellectual and developmental disabilities. The Medicaid Intermediate Care Facility/Mental Retardation (ICF/MR) program was first authorized in 1971. It provided the first Medicaid long-term services and supports benefit specifically for persons with intellectual and developmental disabilities. In 1981 the Medicaid Home and Community Based Services (HCBS) program was authorized as an alternative to the institutional standards of the ICF/MR program. The HCBS program provided a community complement to the ICF/MR program at a time of accelerating commitments within states to develop of community and family supports. Reflecting those commitments, between June 1982 and June 2005, the number of people receiving paid support while living in community homes of 3 or fewer persons with intellectual and developmental disabilities increased from 15,700 to 184,000, and the number in settings with 4 to 6 residents with intellectual and developmental disabilities increased from 17,500 to 107,100 persons. In contrast, the number of persons with intellectual and developmental disabilities in public and private institutions of 16 or more residents decreased from about 180,100 to about 67,000 (Prouty, Smith, & Lakin, 2006).
Medicaid HCBS played a significant role in financing the substantial shifts from institutional to community supports. Between June 30, 1992, and June 30, 2005, the number of HCBS recipients with intellectual and developmental disabilities increased from about 62,500 in June 1992 to about 443,600. During the same period, ICF/MR residents decreased from 146,300 residents in June 1992 to 101,800 in June 2005 (Prouty et al., 2006).
Parallel changes were evident in HCBS and ICF/MR expenditures. Between Fiscal Years (FY) 1992 and 2005, annual federal and state expenditures for HCBS and ICF/MR services for people with intellectual and developmental disabilities increased from $10.5 billion to $29.3 billion. HCBS expenditures, which increased from $1.65 billion in 1992 to $17.2 billion in 2005, made up 82.5% of this growth (Prouty, Smith, & Lakin, 2005). Between 1993 and 2005, average annual per-person expenditure for HCBS increased from $25,176 to $39,656 (57.5%), and the average annual per-person expenditure for ICF/MR residents increased from $62,180 to $117,600 (89.1%), or by 18.4% and 47.2%, respectively, in inflation-adjusted dollars (Prouty et al., 2006).
A number of studies have documented differences in average expenditures for ICF/MR, HCBS, and other community service expenditures (Lakin, Hewitt, Larson, & Stancliffe, 2005; Nerney, Conley, & Nisbet, 1990). However, in these studies, “program type” was confounded by different characteristics among the ICF/MR and HCBS recipients. Other studies that have examined the association between individual characteristics and costs (Ashbaugh & Nerney, 1990; Campbell et al., 2005; Greenberg, Lakin, Hall, Bruininks, & Hauber, 1985) consistently have noted a relationship between greater degrees of intellectual disability and/ or adaptive behavior and service expenditures but have been confounded by the association between program type and expenditures. However, there is little research that has examined the association between costs and program models while controlling for degree and type of impairment.
In this article, we examine individual and program factors associated with Medicaid expenditures for 1,421 adults who were Medicaid HCBS and ICF/MR recipients within a larger random sample of 1,650 adults receiving state-administered intellectual and developmental disability services in four states. The data set and sample size permitted control of individual and program variables that have been confounding factors in most other expenditure studies.
This study merged two data sets, one providing data on characteristics, services, and experiences of sample members gathered with the National Core Indicators (NCI) Consumer Interview and the other providing Medicaid expenditures for the same individuals.
The National Core Indicators
The NCI program (Smith & Ashbaugh, 2001) is an outcomes assessment program developed by the National Association of State Directors of Developmental Disabilities Services (NASDDDS) and the Human Services Research Institute (HSRI). It is used in more than 20 states to evaluate systems of support for persons with intellectual and developmental disabilities. States pay a modest fee to participate in the NCI and for that receive access to psychometrically tested instrumentation, technical assistance with sampling and surveying methods, direct support with data coding and analysis, and a standardized training program for interviewers, including training manuals, presentation slides, training videos, scripts for scheduling interviews, lists of frequently asked questions, picture response formats, procedures for drawing samples, and other resource materials. The standardization of the NCI allows for cross-state comparisons and the merging of data sets from multiple states.
In the analyses described herein, we used NCI and expenditure data from four states: Alabama, Kentucky, Oklahoma, and Wyoming. These states were selected based on (a) recent NCI data collection (between November 2003 and December 2005), (b) samples randomly drawn from service recipients from the full-range of state service options, (c) geographic distribution, (d) interviewee agreements that permitted for merging identifier-stripped expenditure data, and (e) state capacity to assist in building the files. Six states were originally approached about participating, but only four felt able to meet the last two criteria.
These analyses used individual descriptive and service-use data from the NCI survey. NCI descriptive variables include items on demographics (age, ethnicity, gender, etc.), functioning (mobility, vision, hearing, etc.), diagnoses (level of intellectual disability, cerebral palsy, autism, etc.), and behavioral health (mental health–psychiatric diagnosis, medications for behavior problems, frequency of a variety of challenging behavior, etc.). Service use variables include items on services/service sites and are supplied by organizations providing long-term services and supports to the sample members. Service and lifestyle variables include measures of friendships, community participation, family involvement, respect, employment, and so forth. Satisfaction variables include individual reports of satisfaction/liking of home and job and support providers, feelings of safety at home and in the community, sense of loneliness, and opportunities to do new things.
The NCI Consumer Survey was developed with extensive involvement of a Program Advisory Committee (PAC) of state officials and other advisors to ensure the NCI instrument validly reflects data needed to evaluate states' successes in achieving the most important goals of services for persons with intellectual and developmental disabilities. In addition, the draft survey was reviewed by a focus group of individuals with intellectual and developmental disabilities to pretest the face validity of the questions. Focus group participants highlighted problematic questions, identified words that needed further definition, and suggested alternative ways of phrasing questions. These modifications were incorporated into the final survey.
The NCI Consumer Survey underwent three separate interrater reliability tests between 1997 and 1999. Each test yielded interrater agreement of 92%–93%. A lone test–retest reliability evaluation resulted in 80% item agreement (Smith & Ashbaugh, 2001). No additional tests of reliability were conducted for the interviews gathered in this study.
A number of strategies were used to establish and maintain integrity of the procedures and data management across the four states. “Train the trainer” sessions and a standard set of training materials were provided for consistency of training across states. The training program included instruction in basic skills for interviewing and question-by-question review of the survey tool. To reduce coding errors, all states received standard data entry materials, including codebooks that outline the required variable formats and response codes and databases with controlled data entry forms.
Medicaid Payment Files
After state payment data were submitted by the states to the Centers for Medicare and Medicaid Services (CMS) and passed through the CMS VALIDS process for basic editing and consistency checks, Medicaid payments were extracted for and matched to each individual sample member. This procedure was carried out by a CMS contractor, Medstat/Thomson. Payment records included the HCBS, ICF/MR and “other Medicaid” payments made on behalf of sample member in the 12 months prior to the midpoint of the 2–3 months in which the NCI interviews were conducted in each state.
With guidance from officials in each of the four participating states, procedure codes for each service in the state's current HCBS program were assigned to 1 of 12 common service categories. These categories included (a) residential services, (b) personal assistance, (c) respite care, (d) employment and day services, (e) nursing services, (f) therapy services, (g) environmental modifications, (h) supplies and equipment, (i) transportation, (j) training, (k) case management, and (l) other. Creating these common service categories permitted comparisons across states and an aggregation of expenditures for similar types of service. Payment files were also used to obtain Medicaid ICF/MR and “other Medicaid” expenditures not included in HCBS or ICF/MR payments. Other Medicaid expenditures were aggregated into 6 categories: (a) prescription medications, (b) medical expenses, (c) social services, (d) personal care, (e) home health care, and (f) therapies and total “other Medicaid” expenditures variable for each sample member.
Annualized payment variable
Within the four states, there were 36 sample members who spent part of the year in an ICF/MR and part of the year receiving HCBS services; in all of these cases, the members ended the year receiving HCBS-financed services. As noted, the NCI interviews took place at the end of the year for which payments were recorded. Therefore, for these 36 sample members, the NCI data gathered reflected their characteristics and experiences with their HCBS financed services, but their HCBS payment files reflected less than a year of HCBS expenditures. In 90 other cases, people entered the HCBS and ICF/MR program during the year in which payments were being aggregated so that they had fewer than 12 full months of payments. To support comparisons across programs, annual expenditures were summed from monthly HCBS and ICF/MR expenditures for sample members who received HCBS or ICF/MR services for the entire prior year. Annualized expenditures were computed for persons receiving Medicaid HCBS and ICF/MR for less than a full year by computing average monthly expenditures for the actual time enrolled in the HCBS or ICF/MR program at the time of the consumer interview and then multiplying the monthly average by 12.
The study data set included 1,421 adult HCBS and ICF/MR service recipients who were a subset of a larger randomly selected sample of all adults with intellectual and developmental disabilities who were receiving state-administered services and supports in Alabama, Kentucky, Oklahoma, and Wyoming. The 1,421 HCBS and ICF/MR recipients constituted 86.1% of the entire state samples of 1,650 adult recipients of state-administered intellectual and developmental disability services. The NCI and the Medicaid Payment File data sets were merged based on Social Security numbers. Once merged, Social Security numbers and other identifiers were stripped from the new data set.
This article includes both descriptive and predictive analyses. The descriptive analyses summarize HCBS and ICF/MR payments and provide various breakdowns according to state, sample member characteristics, and service approaches. These variables are also included in multiple regression analyses.
Multiple regression analyses
Multiple regression was used to determine the extent to which variables hypothesized to be associated with expenditures could in combination account for variability in expenditures.
The dependent variable in these analyses was the annualized Medicaid payment for the year prior to the individual consumer interview.
Variables used in descriptive and/or multiple regression analyses included (a) level of intellectual disability, (b) medical care needs, (c) mobility, (d) vision, (e) self-injury, (f) disruptive behavior, (g) mental illness/psychiatric diagnosis, (h) autism, (i) gender, (j) age, (k) home/ residence type, (l) ICF/MR (vs. HCBS) financing, and (m) state. Level of ID was categorized by the standard definitions of intellectual disability (1 = mild or none, 2 = moderate, 3 = severe, and 4 = profound). Medical care needs were categorized from reports of how often sample members required medical care (1 = less than once/month to 4 = daily or 24-hr immediate access). Mobility was categorized as a dichotomous variable based on whether sample members were able to move themselves around the environment with or without aids (0 = mobile, 1 = nonmobile). Vision was coded into three categories (1 = sees well with or without lenses; 2 = vision problems affecting activities; 3 = very limited vision or legally blind). Self-injury and disruptive behavior were categorized by the frequencies of such behaviors (0 = none to 5 = one or more times/hr). Mental illness/ psychiatric diagnosis was coded based on whether the individual's treatment record indicated a psychiatric diagnosis (0 = no, 1 = yes). Autism was coded as a dichotomous variable based on whether the sample member's treatment record indicated autism (0 = no, 1 = yes). Gender was coded 0 = male and 1 = female. Age was coded into four categories (1 = 18–39, 2 = 40–54, 3 = 55–64, 4 = 65+ years). Home/residential service type was dummy-coded using congregate as the reference group (0 = congregate [24-hr, agency-operated, specialized institution, group home, apartment, or nursing facility]) compared with host family, own home (individually owned or rented home or less than full-time supervision in an apartment program), or family or relative home. ICF/MR was coded based on whether respondents were receiving HCBS- or ICF/MR-financed services at the time of the NCI data collection (0 = HCBS, 1 = ICF/MR). State was dummy-coded, with State 1 as the reference category.
Components of Expenditures
HCBS and ICF/MR recipients are made up of “other Medicaid” and HCBS or ICF/MR components. The other Medicaid services include prescribed drugs, medical, social services, personal care, home health, and non-HCBS therapies. Expenditures in these categories were summed for HCBS, ICF/MR, and all sample members. Note that these analyses include only Medicaid expenditures. The contributions to the support of individual sample members from Social Security cash assistance programs and other non-Medicaid community programs were not available for sample members. Social Security cash assistance programs would likely have contributed approximately $6,000 per year to the support of the adult HCBS recipients.
As shown in Table 1, HCBS recipients had on average substantially greater “other” Medicaid expenditures than ICF/MR recipients (an average of $7,363 vs. $2,097, respectively). This difference reflects the more comprehensive (institutional) programs of ICF/MR settings that tend to bundle more services, including health services, under a single institution payment than do providers of HCBS. HCBS recipients are more likely to have such service paid for by the Medicaid state plan program. Table 1 also provides the annualized expenditure average for all HCBS recipients according to HCBS service categories (for ICF/MR residents, ICF/MR is the lone service category). Residential supports had by far the largest average expenditure for HCBS recipients (an average of $26,997 per person), followed by personal assistance services (an average of $11,971 per person), and day–vocational services (an average of $9,961). An average of less than $110 per person per year was spent on transportation, environmental modifications, supplies and equipment, and training. We believe it is noteworthy that 71.6% of HCBS expenditures went to services categorized as residential or personal care (63.1% of all Medicaid expenditures).
Because substantially larger amounts of the total Medicaid expenditures for HCBS recipients were paid by “other” Medicaid than was the case for ICF/ MR recipients ($7,363 or 11.9% vs. $2,097 or 1.6%, respectively), the individual expenditures used in the subsequent cost analyses include “other” Medicaid expenditures as well as the HCBS and ICF/ MR expenditures. In general, the expenditures for sample members were comparable with the average expenditures for all HCBS and ICF/MR recipients in the four states. In FY 2004, the four states reported average per-person ICF/MR expenditures of $123,824 compared with an average of $126,178 for sample members in ICFs/MR. In 2004 the four states reported average per-person HCBS expenditures of $47,382 compared with $54,407 for sample members (Prouty et al., 2005). It seems likely that much of this difference would be accounted for by the exclusion of children in the current sample. Children tend to have substantially lower HCBS costs because the majority of them live with family members and attend public education programs during the day.
Diagnostic Health, Physical, and Sensory Indicators
Table 2 presents a summary of annualized Medicaid expenditures for HCBS and ICF/MR recipients by (a) level of intellectual disability, (b) reported frequency of needed health care from a nurse or physician, (c) mobility limitations, and (d) visual limitations. The table shows not only the consistently lower Medicaid expenditures for persons receiving HCBS but a much more consistent association between intellectual, health, physical, and sensory impairment and Medicaid expenditures for services to HCBS recipients than to ICF/MR residents. For example, Medicaid billings on behalf of HCBS recipients by level of intellectual disability increased progressively as level of intellectual disability became more severe. Although this tendency was also evident for ICF/MR recipients, it was much less regular than for HCBS recipients.
In other characteristics examined in Table 2 (frequency of medical care needed, mobility, and vision), HCBS recipients had Medicaid expenditures that were more directly related to indicators of impairment than did ICF/MR recipients. In only one area did Medicaid expenditures for HCBS recipients exceed those for ICF/MR recipients: for persons requiring the medical care of a nurse or physician on a daily or more frequent basis. Such a finding may reflect some economies in delivering such services in larger ICF/MR settings but may also reflect the leveling effect of facility-based billings in institutional settings. Because ICF/MR rates are typically set for the facility, not for the individual residents, in a manner similar to managed care, true health service costs for persons who need daily health care are likely to be underestimated in institutional billings, just as they are likely to be overestimated for persons with relatively few health care needs. If it were possible to determine the actual proportions of staff and other resources allocated to each resident within the ICFs/MR (and other congregate-care settings), it seems likely that higher expenditures would be more directly associated with medical, mobility, and sensory impairments.
Table 3 presents differences in Medicaid billings for HCBS and ICF/MR by a number of behavior characteristics. These include self-injurious and disruptive behavior, a mental health or psychiatric diagnosis, or autism.
The absence of self injurious behavior was associated with lower expenditures among HCBS recipients (p < .001). There was a modest but inconsistent association between self-injurious behavior and expenditures among ICF/MR recipients. For ICF/MR residents, there was a paradoxical association between higher frequencies of reported disruptive behavior and lower average expenditures. The presence of a mental health–psychiatric diagnosis was not associated with HCBS expenditures and was modestly and inversely associated with ICF/MR expenditures (i.e., expenditures were lower for persons with mental health–psychiatric diagnoses). A diagnosis of autism among HCBS recipients was associated with higher expenditures than for individuals without autism ($76,868 vs. $62,351, respectively; p < .001), but no difference was noted among ICF/MR residents. Again, it seems likely that the greater use of facility-level rate setting among ICFs/MR would affect the degree of association.
Type of residence
There was a strong association between Medicaid expenditures and the residential circumstances of HCBS and ICF/MR recipients. Four residential arrangements were identified: (a) congregate–agency-operated housing, (b) host family (also called foster, shared living, or companion living), (c) own home (owned or rented by an individual or an apartment program with part-time staffing), and (d) family home (living with parents or other relatives). Table 4 presents a breakdown of Medicaid expenditures by type of arrangement. All ICF/MR residents resided in congregate–agency-operated settings.
Table 4 shows that Medicaid expenditures were substantially lower for adults living with parents or other relatives (40.6% of the average for all HCBS recipients vs. 35.5% of all sample members). Individuals living in host-family situations also had expenditures substantially below average for HCBS recipients (71.4% of the average HCBS expenditures vs. 62.4% of average expenditures for all). On average, the most costly of HCBS arrangements was own home (133.5% of the average HCBS expenditure but only 64.0% of the average ICF/MR expenditure). Own-home arrangements with one or two residents are vulnerable to higher costs when 24-hr staffing is required, so that a typical staff ratio cannot be held constant as the number of people in the home decreases (Felce & Emerson, 2005). Congregate–agency-operated settings housed half (50.6%) of the HCBS recipients with slightly higher than average HCBS expenditures (102.5% of the average). Average expenditures for HCBS recipients in congregate–agency-operated settings were about half (49.3%) of expenditures for ICF/MR residents of congregate–agency-operated settings.
Multivariate Analyses of Expenditures
Although expenditures for HCBS recipients were on average substantially less than ICF/MR residents, it was also clear that ICF/MR recipients tended to have on average more severe disability. For example, 20.4% of HCBS recipients had profound intellectual disability compared with 60.8% of ICF/MR recipients; 39.2% of HCBS recipients had mild intellectual disability compared with 9.5% of ICF/MR recipients. Similarly, HCBS recipients in the sample, when compared with ICF/MR recipients, were less likely to need medical services of a nurse or physician on a monthly or more frequent basis (24.5% and 84.7%, respectively), more likely to be mobile (88.0% and 66.1%, respectively), to see well (88.7% and 78.3%, respectively), and more likely to not have self-injurious behavior (79.8% and 61.5%, respectively). Because of such differences in the characteristics of individuals supported by HCBS and ICF/MR programs, and because such differences were consistently associated with expenditures in the total sample, we used multiple regression to examine the contributions of various individual and program characteristics to variations in expenditures for persons receiving HCBS or ICF/ MR services.
Correlations between expenditures and independent variables
Table 5 presents a correlation matrix of variables used in these analyses. It shows the high correlations between annualized expenditures and living in an ICF/MR (r = .50), more frequent medical care needs (r = .48), level of intellectual disability (r = .44), and living with parents/relatives (r = −.37). There was a particularly high correlation between ICF/MR residence and reported medical care needs (r = .58), which may have reflected in part the presence of onsite medical personnel and regulatory requirements in the ICFs/MR. ICF/MR residence was also associated with level of intellectual disability (r = .35) and had weaker but positive correlations with mobility (r = .23) and vision impairment (r = .11). Such correlations demonstrated the need for multivariate controls.
Multiple regression analyses
To examine the relationships between independent variables and annualized expenditures, we used the ordinary least-squares (OLS) regression model. Independent variables were ordered in blocks, first reflecting personal characteristics and health support needed by individuals, followed by home/residential service type and HCBS or ICF/MR program participation. The specific block of variables in order included (a) level of intellectual disability, (b) health and sensory limitations, (c) challenging behavior, (d) mental illness–psychiatric conditions, (e) autism, (f) gender and age, (g) home/residential service type, (h) HCBS or ICF/MR program participant, and (i) state. Table 6 presents the results of the OLS regression according to the blocks of variables identified above. In Table 6, standardized coefficients (β) as well as unstandardized coefficients (b) are shown. Standardized coefficients are reported because unstandardized coefficients cannot be compared directly when estimated for arbitrary scales (e.g., in Table 6, level of intellectual disability and frequency of health care needed). By rescaling both independent and dependent variables in terms of their standard deviations, the standardized coefficients indicate the relative importance of predictor variables in accounting for variability in the dependent variable.
The results of the OLS regression showed that level of intellectual disability had a strong association with annualized expenditures, accounting for about 19% of a total variance (R2 = .193, p < .001), with more severe intellectual disability associated with greater expenditure. Within the second block of variables, medical care needs was a particularly strong predictor of annual expenditures, with higher expenditure associated with more frequent medical care. This entire block of independent variables increased the explained variance considerably (R2 change = .170). Together, the level of intellectual disability and health and sensory limitations accounted for more than one third of the variance in annualized expenditures (R2 = .363), with all component factors except visual limitations contributing at statistically significant levels (p < .001). Challenging behavior as a block contributed little to the variance accounted by the equation (R2 change = .003). Mental illness–psychiatric conditions contributed an additional .008 to the overall explained variance (p < .001), with the presence of this diagnosis linked to higher expenditure. Diagnosis of autism had a small but statistically significant effect on explained variance (R2 change = .001, p < .05) and was associated with higher spending. Gender and age had no statistically significant association with variance in annualized expenditures (although in combination, they added .002 to R2). Following the blocks of individual variables, the block of “home/residential service type” variables was added to the equation. Together they made a substantial contribution to explained variance (R2 change = .135, p < .001, total R2 = .513), with own home associated with higher expenditure relative to the reference category of congregate care (which included HCBS group homes) and family home and host family being linked to lower expenditure than this reference category.
In addition to the variance accounted for by the individual and residential type variables, ICF/ MR versus HCBS participation also accounted for a statistically significant amount of variation in expenditures. Adding ICF/MR as a predictive variable accounted for statistically significant (p < .001) additional .033 in explained variation in expenditures, with the total R2 being .546. That is, controlling for all the other factors examined in the preceding blocks, including home/residential service type, there remained a significant difference between HCBS and ICFMR expenditures, with ICF/MR being significantly more costly. Last, state, as a series of dummy variables, contributed to a statistically significant (p < .001) degree in accounted-for variability in expenditures (R2 change = .021). Controlling for all the other factors noted, states provided different amounts of funding for services to adults with intellectual and developmental disabilities. Overall, the variables included in the regression analysis accounted for 56.8% of the variation in the annualized expenditures of the 1,240 individuals in the regression sample.
This study computed and compared annualized expenditures for HCBS and ICF/MR services for a sample of 1,421 adults (18 years and older) with intellectual and developmental disabilities in four states. HCBS and ICF/MR expenditures were computed from payment records for individuals in the 12 months prior to a comprehensive interview with individuals and support providers, which focused on individual characteristics, program participation, quality of life and services, and other data. These 12-month periods varied by state but fell between January 2003 and December 2005. Because there were higher average expenditures for “other” Medicaid expenditures among HCBS recipients ($7,363) than among ICF/MR recipients ($2,097), “other Medicaid expenditures” were added to HCBS and ICF/MR program expenditures in each individual record. To be able to make comparisons that included individuals who spent less than a full year in an HCBS or ICF/MR program, an annualized expenditure was computed. Annualized expenditures and a comprehensive data set on individual, program, and quality of life and service variables were merged into a single record for each sample member.
The findings of the analyses indicated that HCBS services, including other Medicaid services, were substantially less costly than ICF/MR services. The differences were not only evident in overall average expenditures ($61,770 for HCBS recipients vs. $128,275 for ICF/MR recipients) but also were evident in virtually all comparisons for individuals with similar characteristics. For example, the average annualized expenditures for persons indicated to have profound intellectual disability were $80,626 for HCBS recipients compared with $134,918 for ICF/MR recipients. A major contributing factor to lower expenditures for adults receiving HCBS were the substantially lower total expenditures for adults who were living with parents or other relatives (average annualized expenditures of $25,072) and adults who were living in host-family (companion, foster, shared living) arrangements (average annualized expenditures of $44,112). By definition all ICFs/MR were congregate–agency-operated settings. Half of the HCBS sample members (50.6%) in the states with expenditure data also lived in settings defined as congregate–agency operated. Expenditures for adult HCBS recipients in congregate–agency-operated settings were slightly above average for HCBS recipients ($63,286) but still much less (50.7% less) than the average expenditures for persons living in ICFs/MR. It is important to note, however, that room and board costs in ICFs/MR are included in the service expenditures, whereas, for HCBS recipients, housing costs are not paid by Medicaid but by the individual, almost always from cash benefits of Social Security disability programs (Supplemental Security Income or Social Security Disability Insurance), which on average would be nearly $6,000 per person.
There was much stronger differentiation of expenditures by personal characteristics for HCBS than for ICF/MR recipients. This suggests that HCBS expenditures are more clearly based on the individual service-user characteristics related to support needs. The relative lack of differentiation in expenditures among ICF/MR recipients suggests that individuals with milder intellectual disability, few medical care needs, and/or few mobility or visual impairments may often be in environments that provide many services they do not need, or at least not at the intensity with which they are available. Because there was only one factor—people needing daily care from physicians and nurses—on which ICF/MR residents made up a majority of the total sample and on which ICF/MR service costs were less than HCBS, it would be hard to argue from the data gathered in this study that ICF/MR was a cost-effective program for more than a small proportion of the combined ICF/MR and HCBS sample.
In comparing expenditures for HCBS and ICF/ MR services, it was clear that there were many confounding factors. For example, the average expenditure for adults with profound intellectual disability within the total sample ($97,921) was 38.7% more than the average expenditure for all sample members ($70,601) and 85.1% above the average expenditure for persons with mild or no intellectual disability ($52,888). In sum, 60.8% of ICF/MR residents, but only 20.4% of HCBS recipients, had profound intellectual disability; only 9.5% of ICF/ MR residents, but 39.2% of HCBS recipients, had mild or no intellectual disability. These confounding relationships were evident in a number of correlations between ICF/MR residence and factors associated with expenditures: for example, level of intellectual disability (.35), frequency of medical care needed from nurses of physicians (.58), and being nonmobile (.24).
To examine the extent to which such factors in combination could account for the variation in individual expenditures, and whether, having done so, the ICF/MR program model would remain statistically associated with higher expenditures, we used OLS regression. The resulting equation accounted for 57% of the variation in individual expenditures. Controlling for level of intellectual disability; health, physical, and sensory limitations; behavioral, psychiatric, and autism diagnoses; gender and age; and type of residence, ICF/MR was still a statistically significant predictor of higher expenditures (p < .001). ICF/MR did, however, after controlling for the many other variables related to cost, predict only an additional 3.3% of variation in expenditures. It is notable that the home/residential service type block, entered into the regression before ICF/MR status, accounted for 13.5% of variance. Residence type is an inherent characteristic of ICF/MR, and much of the variance in expenditure attributed to residence type could be attributed to ICF/MR status.
Participating states were entered into the regression equation as individual dummy variables. The results indicated that, in addition to variation accounted for by all the different individual and program variables, state accounted for an additional 2.1% of the variability in expenditures. Beyond differences attributable to individuals' support needs and program models, some states may simply allocate more funding for services to persons with intellectual and developmental disabilities.
Together, the variables in the regression analysis accounted for 57% of the variation in annualized expenditures, but it seems likely that higher predictability might have been attained had gradations within the independent variables been more precise (e.g., there were only four levels of intellectual disability, two levels of mobility, no severity indicator for challenging behaviors, simple diagnosis/no diagnosis indicators for psychiatric disability, etc.). Still, this study suggests that in addition to whatever social benefits may accrue to persons with intellectual and developmental disabilities as a result of the notable shifts from ICF/MR to community services financed by HCBS (Lakin et al., in press; Stancliffe & Lakin, 2005), substantial financial benefits also accrue to state and federal governments. Continued efforts within federal and state governments to encourage continued HCBS development seem well founded.
A number of limitations of this study should be noted. First, as already noted, the available expenditure data were limited to Medicaid expenditures. Other contributions to the support of sample members, most notably Social Security cash payments to HCBS recipients, were not included in the data set. Second, as noted, a number of the predictor variables were somewhat imprecise in scaling factors of hypothetical importance to expenditures (e.g., psychiatric and behavioral disorders, mobility limitations). Third, the variable available to measure medical care need may have been confounded with program type; specifically, defining medical care need by, “How often does this persons require medical care [performed or delegated by a nurse or physician],” may have confounded ICF/MR regulatory requirements with actual health care needs (the correlation between greater levels of medical care need and ICF/MR was .58). Fourth, although the NCI has established reliability and a standardized interviewer training program to promote common, consistent data collection practices across states, there was no specific assessment of variability and reliability of intrastate or interstate data collection conducted as part of this study.
Last, despite its limitations, this study demonstrates the benefits of establishing and maintaining policy-relevant data sets with sufficiently large, representative, and multistate samples to address key topics, such as expenditures, with the capacity to attend to important demographic and diagnostic subpopulations, service options within larger programs, and other research considerations that may arise. This multistate study of HCBS and ICF/MR recipients, outcomes, and expenditures was based entirely on extant data sets that were merged for the purposes of creating an integrated data set of unique capacity at low cost. Whether through extant or original data collection, ongoing research and evaluation of Medicaid programs for persons with intellectual and developmental disabilities seem essential, because these programs have surpassed $30 billion dollars per year in expenditures for about 600,000 individual service recipients.
This research was funded through a contract (500-98-0005-T03) between the Centers for Medicare and Medicaid Services (CMS) and the Lewin Group and a cooperative agreement between the National Institute on Disability and Rehabilitation Research and the University of Minnesota (Grant H133B031116). We appreciate the careful attention of Susan Radke, Bill Clark, Paul Boben, and Kathryn Anderson of CMS and Lisa Alecxih of the Lewin Group to the data collection and analysis. We are grateful to Kate Sredl of MEDSTAT/ Thomson for her essential assistance in preparing the Medicaid payment files used in this study. The National Association of State Directors of Developmental Disabilities Services contributed to the planning, instrumentation, and implementation of this study. We are, of course, deeply indebted to the four states and 1,421 sample members who agreed to participate in the study.
Authors: K. Charlie Lakin, PhD (email@example.com), Director; Robert Doljanac, PhD, Research Associate; Soo-Yong Byun, MA, Research Assistant, Research and Training Center on Community Living, University of Minnesota, 214 Pattee Hall, 150 Pillsbury Dr., SE, Minneapolis, MN 55455. Roger J. Stancliffe, PhD, Associate Professor, Faculty of Health Sciences, University of Sydney, P.O. Box 170, Lidcombe NSW 1825, Australia. Sarah Taub, MMHS, Senior Policy Specialist, and Giuseppina Chiri, MA, Senior Research Analyst, Human Services Research Institute, 2336 Massachusetts Ave., Cambridge, MA 02140