Data from the Center for Outcome Analysis were subjected to independent secondary analysis involving comparisons by ethnic group of eight indicators of service quality for users of adult developmental disability services in four states. Ethnic group membership had very limited or no association with the consumer outcomes and service inputs evaluated. Where significant ethnic group differences were found, they were small, and there was no consistent pattern of a particular ethnic group doing better. These findings may relate to the substantial influence of developmental disability services in the lives of service recipients and to the effect of disability itself, in that adaptive behavior was strongly related to most consumer outcomes.
Ethnic and racial status is related to social and economic disadvantage among the general population in the United States and individuals with developmental disabilities from ethnic minority groups who live with their family likely also experience disadvantage. Fujiura and Yamaki (1997) found that among United States households with a member who has developmental disabilities, there was lower income and higher dependence on welfare than for the United States population as a whole. Moreover, minority households with a member with developmental disabilities had lower income, larger household sizes, and greater dependence on welfare than did Caucasian households with a member who had developmental disabilities. That is, the impact of disability and ethnic group on family economic circumstances appeared to be additive.
In the current study we did not focus on the family but, instead, examined indicators of service quality (consumer outcomes) and service inputs for users of developmental disabilities services in several states, 97% of whom lived outside the family home. Our central research question was, Are differences in service quality (as assessed by consumer outcomes and service inputs) related to minority group status among recipients of developmental disability services?
Studies of the impact of minority status on educational attainment frequently examine educational outcomes (test scores, graduation rates, entry to college) and educational inputs (class sizes, amount of instruction, availability of computers) as measures of the quality of educational services. Likewise, in the present study of minority status and developmental disabilities services we used consumer outcomes and service inputs as indicators of the quality of these services. A finding of consistently poorer outcomes and/or inputs for particular ethnic group(s) would indicate a pattern of poorer quality developmental disabilities services to service recipients from that group.
The dependent variables in our analyses were consumer outcomes (weekly earnings, integrative activities, freedom from staff control, number of close friends, family contact, quality of life), and service inputs (number of doctor visits, day program hours). Consumer outcomes allowed us to compare minority groups in terms of equality of outcome, whereas service input variables relate to issues of equality of opportunity. Consumer outcomes embody the central goals and purposes of developmental disabilities services and so should be a primary focus of research into services and service quality. We chose to focus on these particular consumer outcomes and service inputs for several reasons that derive from developmental disabilities research generally, from examinations of ethnicity and disadvantage, and from the constraints of the available data.
First, consumer outcomes are widely seen as valid indicators of the quality of developmental disabilities services (Emerson et al., 2001; Gardner & Carran, 2005; Gardner, Nudler, & Chapman, 1997; Human Services Research Institute, 2003). Because these variables are assessed for each service user, it is possible to compare outcomes for service recipients with different characteristics (such as minority status). Second, we focused on outcomes that people with developmental disabilities themselves say that they value: variables such as access to community activities (referred to as integrative activities in the present study), contact with family and friends (McConkey, Sowney, Milligan, & Barr, 2004), and self-determination (referred to as freedom from staff control in the present study) (Kennedy, 1996). The variables discussed thus far all arise from developmental disabilities research that has been conducted with little specific focus on ethnicity. Third, we included some variables that have been the focus of studies of ethnicity and disadvantage (see Hatton, 2002), such as earnings and access to/intensity of services (e.g., number of doctor visits, day program hours). Finally, our choice of dependent variables was constrained by the variables available in the original data set, which contained few of the social indicators (educational attainment, neighborhood safety, standard of living, home ownership, poverty indices, welfare dependence, health outcomes) that have more traditionally been used when identifying disparities between ethnic groups in the general population.
Previous research on consumer outcomes has shown that service users with better-developed adaptive behavior often experience better outcomes (Stancliffe & Lakin, 1998). It is important to examine an array of consumer outcomes because, once differences in individual ability (e.g., adaptive behavior) are controlled for, there are generally weak relationships among outcome variables (Emerson et al., 2001). This finding indicates that evaluating a single outcome will not provide useful information about, or serve as a proxy for, other important consumer outcomes. Consumer outcomes are also related to residential service type, with better outcomes reported for community living as compared to institutions (Stancliffe & Lakin, 1998). Smaller, less restrictive, and more individualized community living arrangements are also often associated with better outcomes (Burchard, Hasazi, Gordon, & Yoe, 1991; Conroy, 1996; Emerson et al., 2001; Howe, Horner, & Newton, 1998; Stancliffe, Abery, & Smith, 2000; Stancliffe & Keane, 2000). However, the research literature offers little guidance about what to expect when comparing consumer outcomes by service users' ethnic group.
Research suggests that service inputs are often related to service type and/or funding program and less strongly, if at all, to service-user characteristics (Stancliffe & Lakin, 1998). For example, Stancliffe and Lakin (1999) found that day program hours averaged almost 30 hours per week for both institution and community residents in Minnesota, regardless of individual characteristics, essentially because all were enrolled in state long-term care programs that mandated 30 hours per week.
Factors such as adaptive behavior strongly influence consumer outcomes among adult recipients of developmental disabilities services (Stancliffe & Lakin, 1998), so possible differences in personal characteristics between ethnic-group samples need to be examined and, if necessary, controlled. We achieved such control by using covariance analysis with adaptive and challenging behavior as the covariates.
There is an established literature on developmental disabilities and ethnicity concerning (a) the prevalence by ethnic group (Fujiura & Yamaki, 1997; Hatton, 2002); (b) the circumstances, supports, and experiences of families with members with disability from different ethnic groups (Glidden, Rogers-Dulan, & Hill, 1999; Hatton, 2002); and (c) differential access to services (Hatton, 2002). Research is largely silent as to whether similar patterns of disadvantage are also evident in the lifestyle/consumer outcomes of recipients of adult developmental disabilities services. When an individual with developmental disabilities moves to permanent out-of-home accommodation, the impact on that individual's lifestyle, their family's economic circumstances and living situation presumably diminishes substantially. Indeed, the impact of ethnic group membership may be overshadowed by the economic effects of disability because regardless of ethnicity, adults with developmental disabilities are quite likely to have limited incomes.
Ethnic group membership is routinely reported as a descriptive characteristic by disability researchers but rarely analyzed as an independent variable. We found a study conducted in Illinois, in which Wilson, O'Reilly, and Rusch (1991) examined earnings by supported employees with disabilities (predominantly intellectual disabilities) by minority status. They found that employees from minority groups (81% African American) were younger; had higher IQs; and, contrary to the typical situation in the general community, earned higher monthly wages than did their Caucasian counterparts. Such findings may suggest that ethnically related patterns of disadvantage that occur in the broader community may not necessarily be repeated among recipients of adult developmental disabilities services, but much more research is needed before any broad conclusions can be drawn. Wilson et al.'s findings should be interpreted cautiously because the authors did not control for ethnic differences in age and IQ in their analysis of wages.
There are several possible reasons why research has largely overlooked ethnicity when examining the outcomes of developmental disabilities services for adults. Researchers, like many service providers, may simply have ignored such issues (Traustadottir, Lutfiyya, & Shoultz, 1994). Sampling challenges may also have constrained opportunities for researchers to investigate race and ethnicity. In disability research it is often difficult to explore lower incidence issues, such as membership of minority ethnic groups, because of limited sample size. For example, our Minnesota longitudinal deinstitutionalization study (Stancliffe & Lakin, 1998) involved 187 participants, 95% of whom were Caucasians, with the 7 African American participants being the largest minority group.
The strategy we employed in the present study to enable us to gather a sufficiently large sample of minority group members was to undertake independent secondary analysis of large-scale evaluations of developmental disabilities services. The database we examined was assembled from separate state-specific evaluations of developmental disabilities services in four states conducted by the Center for Outcome Analysis. These evaluations all used Center for Outcome Analysis's established instrumentation package, containing major components that were largely the same in each of these states. The samples in each evaluation were not representative of service recipients in the state as a whole but, instead, reflected the specific purpose of each evaluation. For example, the Michigan sample was composed of participants in the Michigan Self-Determination Project, whereas the California sample consisted of class members of Coffelt v Department of Developmental Services (1994), all of whom were former residents of California state institutions. Having data from several different states enabled us to include a variety of minority groups in at least some of our analyses.
States differ considerably in the quantity and quality of their developmental disability services (Human Services Research Institute, 2003; Prouty, Smith, & Lakin, 2004). Differences in state developmental disabilities service systems and the outcomes experienced by service users in each state are legitimate subjects of inquiry, but such research requires representative samples from each state (see Human Services Research, 2003). Because participants did not constitute representative samples for their state, we had no intention of conducting state-by-state comparisons. Even so, state differences could not be ignored because state was confounded with minority group membership (some minority groups were overwhelmingly drawn from a single state). Therefore, state was included as a factor in our multi-state analyses to control for state-related differences among state samples so that we could make valid ethnic-group comparisons, but not as a factor for detailed analysis in itself or for comparisons between states.
The present study is exploratory, in that there is little previous literature on ethnic-group differences in consumer outcomes among adult recipients of developmental disabilities services. The key research question addressed was, are there systematic differences in developmental disabilities service quality (as measured by consumer outcomes and service inputs) related to race and ethnic group?
Professionals at the Center for Outcome Analysis were contracted to conduct large-scale evaluations of developmental disabilities service systems in the four selected states using an established instrumentation package, major components of which were the same for each state. They developed an integrated, multi-state data set on community living outcomes that was subjected to independent secondary analysis by researchers at the University of Minnesota. The Center for Outcome Analysis typically conducts longitudinal evaluations, but these data were cross-sectional, consisting of the most recent available data in each state. Data were drawn from the following evaluations (main data-collection period shown in parentheses): (a) California— Coffelt class members as part of the Coffelt tracking project; all were former residents of California state institutions (September 1999 to February 2000). (b) Indiana—individuals who left the New Castle and Northern Indiana State Developmental Centers (June 1998 to July 1999). (C) Michigan—participants in the Michigan Self-Determination Project; no necessary history of institutional living (May 1997 to March 1998). (d) Oklahoma—all Oklahomans receiving case management and/or more intensive services and supports in 1997–1998 and who were living in community settings (July 1997 to June 1998).
Because the primary data collection was conducted by staff of the Center for Outcome Analysis, our information about data collection was drawn from written copies of the interview schedules, instruments, and instructions to the data collectors. The instructions specified that data be obtained during a personal visit by the data collector, which in most cases was to the person's place of residence. Most of the information was to be obtained from “whoever knows the individual best on a day to day basis” (Center for Outcome Analysis, n.d., p. ii). Given that almost all of the participants lived in settings outside the family home, we presume that this informant would usually have been a residential staff member, but data about the informants were not included in the data set. Instructions to data collectors also specified that they must have “access to the person's records, including medical records” and to a “health care professional familiar with the person” (Center for Outcome Analysis, n.d., p. ii). However, individual records did not indicate to what extent these two sources of information were used when obtaining answers to various questions. Center for Outcome Analysis data-collection procedures also required that the data collector attempt to directly interview the consumer, but no data from the consumer interview are reported in the current study.
There was a total of 4,591 individuals with developmental disabilities from the four states (CA, IN, MI, OK) with data on adaptive and challenging behavior and on race/ethnicity (Caucasian, African American, Hispanic, Native American, and Asian). The number of sample members by state in each ethnic group is shown in Table 1. As can be seen in the table, ethnicity and state were heavily confounded, 2(12, N = 4,591) = 556.83, p < .001. This was particularly true with Hispanics, 94.7% of whom were from California, and Native Americans, 85.3% of whom were from Oklahoma.
Mean age and adaptive and challenging behavior scores by ethnic group and state are shown in Table 2. Ethnic groups differed significantly on adaptive behavior, F(4, 4586) = 23.70, p < .001, challenging behavior, F(4, 4586) = 15.58, p < .001, and age, F(4, 4554) = 13.11, p < .001. State samples differed significantly on adaptive behavior, F(3, 4587) = 216.28, p < .001, challenging behavior, F(3, 4587) = 488.05, p < .001, and age, F(3, 4555) = 66.17, p < .001.
Of the 4,591 sample members, 207 (4.5%) lived in their own home, 148 (3.2%) lived in their family home, 201 (4.4%) lived in a non-Intermediate Care Facilities/Mental Retardation (ICF/MR) group home, 12 (2.8%) lived in a foster home, 744 (16.2%) lived in supported living settings, 3,164 (68.9%) lived in other settings (including small and large ICFs/MR, and other large facilities).
Adaptive behavior was measured with a 50-item scale derived from the AAMR Adaptive Behavior Scale (ABS). These items were summed to create a cumulative adaptive behavior score. A 32-item version of this scale has been found to be highly reliable, with interrater and test– retest reliabilities of .95 and .96 (Devlin, 1989) and .99 and .99 (Fullerton, Douglass, & Dodder, 1999).
This type of behavior was measured with 16 items on behaviors such as self-injury, aggression, property destruction, depressive-like behavior, temper tantrums, and inappropriate undressing. These 16 items yield a sum score that has been shown to have a combined interrater and test–retest reliability of .85 (Conroy, 1995).
Total scores on both the adaptive and challenging behavior scales were computed so that they could range from 0 to 100 points, with the higher scores being more favorable in each case. Thus, for challenging behavior, higher scores indicated less challenging behavior.
The Integrative Activities During the Past Month Scale (Conroy, 1997c) includes 16 specified community activities, such as shopping, sporting, leisure, social, and religious activities. Respondents specify the number of times the sample member participated in each activity during the past month. The total score is the sum of the 16 activity frequencies. Fullerton et al. (1999) found a 7-item version of this scale to have test–retest reliability of .89.
Freedom from staff control
The Decision Control Inventory (Conroy, 1997b) is used to evaluate who makes choices and exercises control in the lives of people with developmental disabilities in relation to everyday life and services, such as the use of personal money; choice of foods, homes, and case managers; and/or whether to have pets. Respondents rate decision-making on an 11-point scale, with a rating of 0 for decisions made entirely by paid staff and 10 for decisions made entirely by the individual and/or unpaid loved ones. The sum of the item scores is recoded into a 0 to 100 scale. Higher scores indicate greater freedom from staff control because of control being exercised by the person and/or unpaid family or friends (Stancliffe & Lakin, 2005). The California sample used a 29-item version of the Decision Control Inventory, whereas all other states used a 35-item version. All scores were recoded so that total scores ranged from 0 to 100. Internal consistency for the Decision Control Inventory (Cronbach's alpha) was .95, test–retest reliability, .98; combined test–retest and interrater reliability, .86, and the correlation between Decision Control Inventory scores and adaptive behavior was .71 (Conroy, 1995).
Quality of life
The Quality of Life Changes Scale (Conroy, 1997d) contains 14 items about different aspects of life, such as health, making choices, seeing friends, happiness, safety, and overall quality of life. Quality of life for each item is rated using a 5-point scale from 1 (very bad) to 5 (very good) for quality of life “then” (e.g., a year ago) and “now.” We used only data from the quality of life now items. The Indiana sample used a 15-item version of the Quality of Life Changes Scale; the 14-item version was used in all other states. Total scores were recoded to range from 0 to 100 points, with higher scores indicating better quality of life. Head and Conroy (2005) reported that the interrater reliability of the 15-item version of the scale was .76.
No psychometric information was available for the five variables listed below, although in most instances the information involved was basic factual data that should be able to be provided reliably by a reasonably knowledgeable informant.
Respondents reported the number of doctor visits in the last year.
Respondents reported the number of contacts by all relatives in the past year via telephone, mail, visits, outings, individual planning meetings, or consent for medical care.
Number of close friends
The Close Friends Scale (Conroy, 1997a) was used to obtain information on the person's close friends (as defined by the person or the informant) and various characteristics of each friend. For the present study only the number of close friends was evaluated.
Earnings per week
To determine earning per week, we asked respondents, “how much money does this person earn in an average week?”
Day program hours per week
Respondents specified the number of hours per week the person spent in a range of paid and unpaid day activities, including employment, volunteer work, education, retirement programs, and traditional day programs. Day-program hours per week was the total across all activities.
Sample Size, Statistical Significance, and Effect Size
With very large sample sizes, very small effects can attain statistical significance. Unlike significance tests, effect size indices are independent of sample size and, therefore, provide a measure of the impact of a given variable that may be compared across studies and across samples. For ANOVA analyses, partial eta squared (ηp2) describes the proportion of total variability of the dependent variable attributable to an effect. For an effect with a ηp2 of .070, 7% of the variability is uniquely attributable to that effect. Based on Cohen's (1988) definition of effect sizes, the following values of ηp2 may be regarded as the minimum: for a small effect, .010; for a medium-sized effect, .059; and for a large effect, .138. Effects with ηp2 less than .010 are described as very small. In the following analyses sample size varies widely, so both statistical significance and effect size are reported for ethnic group; state; and the covariates, adaptive and challenging behavior.
The ethnic groups were multiply confounded with adaptive and challenging behavior and state. To minimize the impact of confounding, we used adaptive and challenging behavior as covariates in all analyses. We included state as a factor in the multi-state analyses or held it constant when completing separate analyses of data from individual states. Even so, interpretation of state differences remained problematic because they could be attributed to differences in state developmental disabilities service systems or to differences in the characteristics and circumstances of the samples from each state. Consequently, no specific state-by-state comparisons were planned. For the sake of completeness, significant differences between states are reported, but no emphasis is placed on state differences in the discussion because of these interpretative problems.
We analyzed the following six consumer outcomes: weekly earnings, integrative activities in the last month, freedom from staff control, number of close friends, quality of life, and family contact. In addition, two service inputs were analyzed: number of doctor visits in the past year and day program hours. As Table 3 shows, there were several dependent variables for which data were not available from Oklahoma (freedom from staff control, number of close friends, family contact, quality of life, number of doctor visits, and day program hours). Findings are reported separately for variables with data from all four states and variables with data from only three states (i.e., excluding Oklahoma).
Using the multi-state sample, we undertook a univariate, general linear model (GLM) two-way ANCOVA for each dependent variable, with ethnic group and state as the independent variables. In this way, all the major variables were included in each analysis, but only Caucasian and African Americans were present in sufficient numbers across all states to be included in these analyses (N = 4,079).
Second, for each dependent variable, we completed separate univariate, GLM, one-way ANCOVA analyses by ethnic group for each state sample, which allowed us to include Hispanic, Asian, and Native American groups in some state analyses. Ethnic groups were included in these analyses only if the group had 30 or more sample members in that state. As a result, each analysis only involved between two and four of the five ethnic groups (see Table 1). Where a significant difference between ethnic groups was found involving three or more groups, we conducted follow-up comparisons of specific ethnic groups using simple contrasts. Adaptive and challenging behavior were used as covariates in all ANCOVA analyses.
Variables With Data From All States
Weekly earnings: Multi-state sample
Means adjusted for the effects of adaptive behavior and challenging behavior are shown in Table 4. Table 5 shows the overall R2 for each analysis, effect sizes for main effects, the State × Ethnic Group interaction, and covariates as well as the direction of the association between covariates and the dependent variables. African Americans earned significantly more per week than Caucasians, F(1, 4,069) = 5.37, p < .05, but the effect size for ethnic group was very small (accounting for 0.1% of variance). There was a significant difference in earnings between states, F(3, 4,069) = 9.07, p < .001, with a very small effect size (0.7% of variance). The interaction of Ethnic Group × State was not significant. Both adaptive behavior, F(1, 4,069) = 520.71, p < .001, and challenging behavior, F(1, 4,069) = 28.82, p < .001, were significant covariates, but adaptive behavior had a much larger effect size than did challenging behavior (11.3% and 0.7% of variance, respectively). Higher earnings were associated with better-developed adaptive behavior and better (less challenging) challenging behavior (the direction of these associations is denoted by a plus sign [+] or minus sign [–] in the relevant cells of Table 5). Although ethnic group had a statistically significant association with earnings, its ability to explain variability in weekly earnings was very modest, with adaptive behavior being by far the most important variable in the analysis.
Weekly earnings: Individual states
Means adjusted for the effects of the two covariates are shown in Table 6. The overall R2 for each analysis, effect sizes, and the direction of the association between covariates and the dependent variables appear in Table 7. The California sample had a statistically significant difference in adjusted average earnings between ethnic groups, F(3, 1958) = 6.15, p < .001, with African Americans earning significantly more per week than did Caucasians, difference = −7.27, SE = 1.72, p < .001; Hispanics, difference = −6.02, SE = 2.04, p < .01; and Asian Americans, difference = −8.36, SE = 3.58, p < .05. Better adaptive, F(1, 1958) = 283.75, p < .001, and challenging, F(1, 1958) = 47.79, p < .001, behavior were both significantly associated with higher earnings, with the direction of the associations is shown by plus sign or minus sign in Table 7.
In Indiana, African Americans had higher adjusted average weekly earnings, F(1, 253) = 3.92, p < .05. Adaptive behavior was significantly related to higher earnings, F(1, 253) = 10.06, p < .01. In the Michigan and Oklahoma analyses, there were no significant differences in earnings by ethnic group. Adaptive behavior was significantly associated with higher earnings in both Michigan, F(1, 526) = 62.82, p < .001, and Oklahoma, F(1, 1791) = 255.80, p < .001.
African Americans earned significantly more than did Caucasians in the multi-state analysis and in the state-level analyses for California and Indiana, but not in Michigan and Oklahoma. The effect size for ethnic group was consistently very small to small (accounting for 0.0% to 1.5% of variability in earnings). Adaptive behavior was a significant covariate in all analyses and had a much larger effect size (3.8% to 12.7% of variability). Challenging behavior was a significant covariate in only some analyses, and its effect size was very small to small (0.0% to 2.4% of variance). Overall, these findings show that ethnic group and challenging behavior had a (very) small association with earnings, but adaptive behavior had a consistently positive and mostly medium-sized effect.
There were no significant differences by ethnic group in number of integrative activities in the past month. There was a significant difference between states, F(3, 4060) = 4.93, p < .002. The interaction of Ethnic Group × State was significant, F(3, 4060) = 12.65, p < .001, with African Americans participating in more integrative activities in three states (CA, IN, OK), but with Caucasians having a higher mean in Michigan. Both adaptive and challenging behavior were significant covariates, Fs(1, 4060) = 552.88 and 15.98, ps < .001, respectively. More integrative activities were associated with higher adaptive behavior and more (worse) challenging behavior. Although challenging behavior and the State × Ethnic Group interaction were statistically significant, they explained little of the variability in integrative activities (both 0.4% of variance). Adaptive behavior had a much larger effect size (12.0% of variance).
In the California sample, there were significant differences by ethnic group in integrative activities, F(3, 1953) = 3.53, p < .05. Simple contrasts, with Caucasians as the reference group, showed that Hispanics participated in significantly more integrative activities than did Caucasians, difference = 4.19, SE = 1.39, p < .01, whereas African Americans and Asian Americans did not differ significantly from Caucasians. Both adaptive, F(1, 1953) = 538.60, p < .001, and challenging, F(1, 1953) = 6.78, p < .01, behavior were significant covariates, with more integrative activities associated with higher adaptive behavior scores and more (worse) challenging behavior.
There was no significant difference by ethnic group in integrative activities among Indiana sample members. Better adaptive behavior was significantly associated with more integrative activities, F(1, 252) = 53.85, p < .001. In the Michigan sample there was a significant difference by ethnic group, F(1, 523) = 18.48, p < .001, with Caucasians engaging in more integrative activities than did African Americans. More integrative activities were significantly related to higher adaptive behavior scores, F(1, 523) = 38.23, p < .001.
In the Oklahoma sample there was a significant difference by ethnic group, F(2, 1791) = 4.35, p < .05. Both African Americans, difference = 2.55, SE = 1.14, p < .05, and Native Americans, difference = 2.91, SE = 1.40, p < .05, participated in significantly more integrative activities than did Caucasians. More integrative activities were significantly associated with better adaptive behavior, F(1, 1791) = 148.32, p < .001, and worse challenging behavior, F(1, 1791) = 5.02, p < .05.
Consistent with the significant ethnic Group × State interaction found for the multi-state sample, state-by-state ethnic comparisons differed. Significant ethnic differences in integrative activities favoring non-Caucasians were found in two of the four states examined, with differences in Michigan that favored Caucasians. Effect size analysis revealed a consistent picture of a very small to small effect size for ethnic group, 0.5% to 3.4% of variance, whereas adaptive behavior was a significant, positive covariate in all cases and had a much larger effect size, 6.8% to 21.6% of variance. Challenging behavior was associated with integrative activities in some states but not others, and effect size was very small in all cases, 0.0% to 0.4% of variance. Overall, these findings show that ethnic group and challenging behavior were, at best, minor influences on integrative activities, whereas adaptive behavior was a medium to large influence.
Variables With Data From Three States
Data on several variables were not collected for the Oklahoma sample, so data in this section are only from California, Indiana, and Michigan. As before, adjusted means for the multi-state sample are shown in Table 4, and R2 and effect sizes in Table 5. For the individual state analyses, adjusted means are again shown in Table 6, and contrasts, R2, and effect sizes in Table 7.
Freedom from staff control: Multi-state sample
There were no significant differences by ethnic group in freedom from staff control, but substantial differences by state, F(2, 2361) = 44.33, p < .001. The ethnic group by state interaction was also significant, F(2, 2361) = 6.01, p < .01, with differences favoring Caucasians in Michigan, but African Americans in Indiana. Greater freedom from staff control was associated with better developed adaptive behavior, F(1, 2361) = 1265.69, p < .001, and more (worse) challenging behavior, F(1, 2361) = 8.36, p < .01. Adaptive behavior had a very large effect size (34.9% of variance), whereas state, ethnic group, State × Ethnic Group interaction, and challenging behavior had very small to small effect sizes (0.1% to 3.6% of variance).
Freedom from staff control: Individual states
For the California sample there were no significant differences by ethnic group. Higher adaptive behavior scores, and worse challenging behavior were significantly associated with greater freedom from staff control, Fs(1,1948) = 1207.03 and 19.20, respectively, ps < .001.
In Indiana there were no significant differences between Caucasians and African Americans. Better developed adaptive behavior was significantly associated with greater freedom from staff control, F(1, 251) = 100.31, p < .001.
For Michigan, there was a significant difference by ethnic group, F(1, 524) = 13.16, p < .001, with Caucasians having higher scores than did African Americans. Greater freedom from staff control was associated with better adaptive behavior and less challenging behavior, Fs(1, 524) = 163.78 and 16.16, respectively, ps < .001.
Consistent with the significant State × Ethnic Group interaction, the state-by-state analyses revealed that Caucasians had greater freedom from staff control in Michigan, but there were no ethnic group differences in the other two states. There were very small to small effect sizes for ethnic group, 0.2% to 2.4% of variance, and challenging behavior, 0.1% to 3.0% of variance, but large effect sizes for adaptive behavior, 23.8% to 38.3% of variance.
Number of Close Friends
There were no significant differences by ethnic group in the number of close friends, but substantial differences by state, F(2, 2235) = 137.20, p < .001. The Ethnic Group × State interaction was not significant. Individuals with higher adaptive behavior scores had more close friends, F(1, 2235) = 4.63, p < .05.
State had the largest effect size (10.9% of variance), whereas ethnic group, State × Ethnic Group interaction, adaptive and challenging behavior had very small effect sizes (0.0% to 0.2% of variance). These findings may have resulted from the different prior histories of the various state samples and the effects of these histories on sustaining relationships.
In the California sample there was a significant difference by ethnic group in the number of close friends, F(3, 1913) = 4.37, p < .01. Simple contrasts showed that the Hispanic group, difference = 0.53, SE = 0.20, p < .01, had more close friends than did Caucasians, who in turn had more close friends than did African Americans, difference = −0.52, SE = 0.25, p < .05. Caucasians and Asian Americans did not differ significantly. Those with better developed adaptive behavior, F(1, 1913) = 8.62, p < .01, and worse challenging behavior, F(1, 1913) = 5.67, p < .05, had more close friends.
There was no significant difference by ethnic group in the Indiana sample. Better adaptive behavior and better challenging behavior were significant covariates, Fs(1, 231) = 13.37 and 6.32, ps < .001 and .05, respectively. In the Michigan sample, there was no significant difference by ethnic group in number of close friends nor were adaptive behavior or challenging behavior significant covariates.
There were no significant differences by ethnic group except for the California sample, where there was a very small effect size, 0.7% of variance. Better developed adaptive behavior was significantly associated with having more close friends in all analyses except Michigan, but effect sizes were very small or small, 0.2% to 5.5% of variance. Challenging behavior was a significant covariate in two analyses (CA and IN) but the direction of this relation differed. None of these variables was consistently and strongly associated with number of close friends.
Twenty-five sample members were excluded from the analyses of number of family contacts in the past year because of extremely high values, 369 to 5037.
Caucasians had more family contact than did African Americans, F(1, 1783) = 3.96, p < .05. There were significant differences by state, F(2, 1783) = 5.24, p < .01. The Ethnic Group × State interaction was not significant. Higher levels of family contact were associated with better developed adaptive behavior, F(1, 1783) = 107.97, p < .001. Adaptive behavior had the largest effect size, 5.7% of variance, whereas state, ethnic group, State × Ethnic Group interaction, and challenging behavior had very small effect sizes, 0.2% to 0.6% of variance.
In the California and Indiana samples, there was no significant difference in family contact by ethnic group. Better adaptive behavior was significantly associated with more family contact in both California, F(1, 1407) = 145.20, p < .001, and Indiana, F(1, 231) = 13.37, p < .001.
In the Michigan sample, there was a significant difference by ethnic group, F(1, 432) = 4.28, p < .05, with Caucasians having more family contact than African Americans. Greater family contact was significantly related to higher adaptive behavior scores, F(1, 432) = 8.64, p < .01.
Ethnic group membership was significantly related to the amount of family contact for the multi-state analysis and for the Michigan sample, but effect sizes were very small or small, 0.2% to 1.0% of variance. Higher levels of adaptive behavior had a small to moderate association with more family contact in all analyses, 2.0% to 9.4% of variance. Challenging behavior was not significantly related to family contact in any of the analyses.
Quality of Life
Caucasians had higher quality of life scores than did African Americans, F(1, 2315) = 6.76, p < .01. There were significant differences by state, F(2, 2315) = 116.06, p < .001. The Ethnic Group × State interaction was not significant. Neither adaptive behavior nor challenging behavior were significant covariates.
In the California sample, there was a significant difference by ethnic group, F(3, 1955) = 4.68, p < .01. Simple contrasts showed that the Caucasian group had higher quality of life scores than did African Americans, difference = −2.07, SE = 0.75, p < .01. Caucasians did not differ significantly from Hispanics or Asian American Californians. Higher quality of life was significantly associated with lower adaptive behavior scores, F(1, 1933) = 14.62, p < .001, and worse challenging behavior, F(1, 1933) = 10.04, p < .01.
In the Indiana and Michigan samples, there were no significant differences by ethnic group. Challenging behavior was a significant covariate in both Indiana, F(1, 244) = 10.51, p < .001, and Michigan, F(1, 493) = 25.05, p < .001, and in both samples better quality of life was related to less (better) challenging behavior.
Mean quality of life scores were higher for Caucasians than for African Americans, in the multi-state comparison and for California, and in all cases the effect size was very small, 0.1% to 0.7% of variance. Challenging behavior was a significant covariate for each state comparison, but not for the multi-state sample and had a very small or small effect size, 0.5% to 4.8% of variance. The direction of these associations was inconsistent. There was a negative relation between challenging behavior and quality of life (persons with worse challenging behavior had higher quality of life) in California but a positive relation in Indiana and Michigan.
Health Services–Doctor Visits
There were no significant differences by ethnic group in number of doctor visits in the last year. There were significant differences by state, F(2, 2273) = 6.47, p < .002. The Ethnic Group × State interaction was not significant. Attending more doctor visits was significantly associated with higher adaptive behavior scores, F(1, 2273) = 5.84, p < .05, and worse challenging behavior, F(1, 2273) = 11.80, p < .001.
In the California and Michigan samples, there were no significant differences by ethnic group. In Indiana, African Americans had significantly more doctor visits, F(1, 225) = 6.38, p < .05, than did Caucasians.
Adaptive behavior was not a significant covariate in any of the states. Challenging behavior was not significantly related to the number of doctor visits in Indiana or Michigan, but in California worse challenging behavior was significantly related to more doctor visits, F(1, 1905) = 11.24, p < .001.
Ethnic group differences in the number of doctor visits in the last year were not significant (multi-state sample, California, and Michigan) or statistically significant but small (Indiana, 2.8% of variance). The relation to adaptive behavior was very small to small, 0.1% to 1.2% of variance, and mostly nonsignificant. Likewise, the association with challenging behavior was very small, 0.0% to 0.6% of variance, although it attained statistical significance for the multi-state and California samples.
Day Program Hours
Twenty-five sample members had very high to implausibly high values for day program hours (from 51 to 280 hours per week) and were excluded from the analyses.
There were significant differences by ethnic group in weekly day program hours, F(1, 2340) = 4.59, p < .05, with African Americans receiving more hours than Caucasians. There were significant differences by state, F(2, 2340) = 81.43, p < .001. The Ethnic Group × State interaction was not significant. Individuals with worse challenging behavior had more day program hours, F(1, 2340) = 12.33, p < .001.
In the California sample, there were no significant differences by ethnic group, F(3, 1951) = 0.83, p = .48. Adaptive behavior was not a significant covariate, but worse challenging behavior was significantly related to more day program hours, F(1, 1951) = 15.45, p < .001.
In the Indiana and Michigan samples, no significant ethnic differences were found. Adaptive behavior was a significant covariate in Indiana, F(1, 234) = 9.76, p < .01, and Michigan, F(1, 517) = 6.14, p < .05, but the direction of these associations was opposite, with better developed adaptive behavior associated with more day program hours in Indiana but fewer hours in Michigan.
Differences in day program hours by ethnic group were weak (multi-state sample, 0.2% of variance) or nonsignificant (all three individual states). Day program hours are a service input not a consumer outcome. Unlike many of the consumer outcomes examined, day program hours had a weak or nonsignificant relation to individual characteristics, such as adaptive behavior (0.0% to 4.0% of variance) and challenging behavior (0.1% to 0.8% of variance).
Because the main focus was on ethnic group differences, and the analyses were rather complex, Table 8 presents a summary of all the ethnic group comparisons both for the multi-state sample and for each state sample in relation to the eight dependent variables. This table shows whether the comparison favored Caucasians (C), minority groups (M), or whether there was no significant ethnic group difference (0).
In the present study we asked whether there are any systematic differences in service quality (as measured by consumer outcomes and service inputs) related to race and ethnic group among recipients of developmental disabilities services in several states. We found that minority status had limited, inconsistent, or no association with consumer outcomes or service inputs. That is, there was no compelling evidence for systematic inequality of outcome or inequality of opportunity related to race or ethnic group. Rather, consumer outcomes were most strongly and consistently related to personal characteristics associated with disability, such as adaptive behavior.
Our analyses involved comparisons by ethnic group of six consumer outcomes (earnings, integrative activities, freedom from staff control, number of close friends, family contact, and quality of life) and two service inputs (doctor visits and day program hours). In these analyses adaptive behavior and challenging behavior were used as covariates to control for the confounding effects of differences in personal characteristics among the various state and ethnic group samples. State was included in the multi-state analyses as a control variable not an analytic one. The differences between samples from different states are reported for the sake of clarity and completeness, but their implications will not be discussed because we were not able to conduct valid state-by-state comparisons due to the differing and unrepresentative samples of service users from each state.
In the multi-state comparisons of Caucasians and African Americans, four of the eight analyses showed significant ethnic group differences, with two comparisons (weekly earnings and day program hours) favoring African Americans and two (family contact and quality of life) favoring Caucasians. Even these comparisons had very small effect sizes for ethnic group that accounted for between 0.1% and 0.3% of the variance. That is, these multi-state analyses revealed no consistent ethnic group differences and none of practical significance.
Ten of 26 individual state ethnic-group comparisons were statistically significant. One comparison revealed mixed findings regarding Caucasians and minority groups (number of close friends in CA), four comparisons favored Caucasians, but another five revealed better outcomes for minority groups (mostly African Americans) relative to Caucasians. Overall, the individual state comparisons did not reveal a consistent pattern of better consumer outcomes or service inputs for any particular ethnic group. Moreover, these comparisons mostly had very small or small effect sizes that only explained 0.5% to 1.5% of the variance. Such differences are of limited or no practical significance.
In the Michigan sample, there were three comparisons showing that Caucasians had better outcomes than African Americans (see Table 8), with two of these having nontrivial effect sizes: integrative activities, 3.4% of variance, and freedom from staff control, 2.4% of variance. In Indiana, African Americans had more doctor visits than did Caucasians, with an effect size of 2.8% of variance. These findings raise the possibility that differences in consumer outcomes related to ethnicity may be more prominent in some locations than in others. Inconsistent findings between states regarding the relation between consumer outcomes and minority group status may relate to the reality that the effects of specific minority group membership might differ from state to state. For example, being Hispanic in a state such as California with a large Hispanic population may be substantially different from having the same ethnic background and living in Michigan.
This possibility is further supported by the multi-state finding of significant Ethnic Group × State interactions (albeit with very small effect sizes) for integrative activities and for freedom from staff control. The interactions indicate that the magnitude and/or direction of ethnic group differences varied from state to state. This is most clearly illustrated for integrative activities, with significantly higher scores for non-Caucasian groups in California and Oklahoma, but the opposite in Michigan (see Tables 7 and 8). It was notable that sample members from Michigan had a much higher proportion of people living in their own homes (31.8%) or in their family's home (16.5%). It is possible that the impact of ethnically related disadvantage was more prominent for individuals living in their own home/family home. Empirical examination of these complex issues was beyond the scope of the present study.
Overall, ethnic group membership had very limited or no association with the consumer outcomes and service inputs evaluated in the current study. Where significant ethnic group differences were found, they were small, and there was no consistent pattern of a particular ethnic group doing better. Some outcomes favored Caucasians but others favored minority groups. Of course, this finding does not preclude ethnic differences being identified among other important outcomes that were not examined here, outcomes such as social indicators (educational attainment, neighborhood safety, standard of living, housing quality, home ownership, poverty indices, welfare dependence, and health outcomes).
We found no consistent pattern of minority group disadvantage for consumer outcomes among individuals who already were recipients of developmental disabilities services, but this finding does not compensate for disadvantage that may be experienced by other minority group members in accessing the service system. There is some evidence of underrepresentation of minority ethnic communities in residential services for people with developmental disabilities (Hatton, 2002). For example, Hewitt, Larson, and Lakin (2000) found underrepresentation of minority groups among recipients of Home and Community-Based Services (HCBS) for people with developmental disabilities in Minnesota, with 5.2% of all HCBS recipients being from minority groups, whereas 9.2% of Minnesota's total population consisted of minority group members.
One finding worthy of closer examination concerned weekly earnings. As noted, Wilson et al. (1991) found that minority group members in Illinois earned higher monthly wages than did Caucasians. Results of the current study confirm and extend Wilson et al.'s findings by showing that across four additional states, with adaptive and challenging behavior controlled using covariance, African Americans earned more than Caucasians. When the samples from each state were analyzed separately, African Americans earned significantly more in two states and earnings did not differ significantly by ethnic group in the remaining two states. Although our findings were consistent with Wilson et al.'s earlier work, it is important to note that the effect size was small. At most, ethnicity accounted for 1.5% of the variability in wages, whereas adaptive behavior accounted for up to 12.7%.
In contrast with the findings on ethnicity, adaptive behavior was consistently related to consumer outcomes, with individuals who had more adaptive skills typically enjoying better outcomes. Effect size analyses showed that adaptive behavior frequently had a moderate to strong relation with consumer outcomes. That is, a defining feature of intellectual disability, such as adaptive behavior, was much more strongly and consistently related to consumer outcomes than was ethnicity. Previous research has also shown a clear relation between adaptive behavior and many consumer outcomes (e.g., Stancliffe & Lakin, 1998).
The relation between adaptive behavior and service inputs was weak and inconsistent, with small or very small effect sizes. Such findings suggest that adaptive behavior was not strongly related to the service inputs examined. This result is consistent with previous research showing a weak or nonexistent association between service inputs and personal characteristics, for example, adaptive behavior (Stancliffe & Lakin, 1998). However, a variable such as day program hours does not account for differences in the type of employment or day programs made available to individuals with differing levels of disability. Other research has shown that better remunerated and more socially integrated employment options, such as competitive and supported employment, are mostly available to persons with milder disabilities (e.g., Stancliffe & Lakin, 1999).
Challenging behavior was a significant covariate for some consumer outcomes and service inputs, but findings were inconsistent across states and/or in direction, and effect sizes were small or very small. Overall, the relation between challenging behavior and consumer outcomes and service inputs was weak or nonexistent.
Disability services have been characterized as being unresponsive to the specific needs of ethnic minorities and as ignoring issues such as race, ethnic group, and socioeconomic status (Traustadottir et al., 1994). Few sample members lived in their own home (4.5%) or their family home (3.2%), with the vast majority (92.3%) living in residential service settings where their lifestyle likely was heavily influenced by service providers. Our finding of very small and nonsignificant differences by minority group status may relate to the substantial influence of developmental disability services on the lives of service recipients. The supports and services they receive, and the outcomes they experience, may be determined mostly by their status as service recipients, and little by their ethnic minority status. For example, many small group home residents, unlike other nondisabled individuals with poverty-level incomes, live in a middle-class house in a middle-class neighborhood.
In future research, it is important that investigators evaluate whether the present findings hold true for persons who live in their own home or their family home, where presumably familial and cultural influences are more prominent and service providers are less influential. Likewise, it will be necessary to broaden the range of dependent variables examined to include more social indicators (e.g., educational attainment, neighborhood safety, standard of living, housing quality, home ownership, poverty indices, welfare dependence, health outcomes) that have frequently been overlooked in the developmental disability field but are often the subject of disparities between ethnic groups in the broader community.
Consistent with much previous research, adaptive behavior had a strong relation with outcomes and a weak relation with service inputs (Stancliffe & Lakin, 1998). The impact of this defining characteristic of intellectual disability on consumer outcomes may have been so great as to overwhelm any differences related to minority group membership. In this context, it is useful to consider the nature of the consumer outcomes commonly used in developmental disability research (and in this study). Undoubtedly, variables such as community participation (integrative activities in the present study) and self-determination (freedom from staff control in the present study) are the focus of research attention, in part, because they relate in expected ways to adaptive behavior and to the quality of developmental disabilities services. The association between consumer outcomes and ethnic group has remained largely unexamined in developmental disability research. Variables drawn from other research traditions, such as social indicators, may yield a different picture regarding ethnicity from the findings reported here on consumer outcomes.
Significant differences by state were found for all outcomes and service inputs, but given the differences in samples among states, it was impossible to determine whether these variations were attributable to differences between state service systems or to sampling factors. We reiterate that no state-by-state comparisons should be attempted using the data reported here. Readers interested in such comparisons should consult studies where the research design and sampling allow for valid comparisons to be made between states (see Human Services Research Institute, 2003; Prouty et al., 2004).
Other limitations of the current study need to be considered. We had no information about participants' location within states (e.g., urban/suburban/rural) and, therefore, were unable to control for such factors. There can be important variations in service availability and service quality within states that we did not consider in the present
In this study we examined differences (or lack of difference) in consumer outcomes and service inputs among ethnic groups, without attempting to explain the causes of any such differences. Ethnic group membership, in itself, is not posited as a causal agent. Rather, ethnic group is an indicator of social and economic factors, such as greater poverty, discrimination, or lack of access to services, which are considered to be likely causal variables. If future researchers identify important differences in outcomes or services by ethnic group, it will be vital to go further and identify the causal factors underlying such differences.
This paper was prepared with financial support from the National Institute on Disability and Rehabilitation Research (NIDRR), U.S. Department of Education (Grant H133B031116). The data analyzed in this article were gathered by the Center for Outcome Analysis. Through an agreement with the Center for Outcome Analysis, independent analyses of these data were conducted at the University of Minnesota, Research and Training Center on Community Living. This article represents the opinions of the authors and does not necessarily reflect official positions of NIDRR, the Center for Outcome Analysis, or any of its funding agencies.
Authors: Roger J. Stancliffe, PhD (firstname.lastname@example.org), Consultant Research Associate, 780 Werombi Rd., Theresa Park, NSW 2570, Australia. K. Charlie Lakin, PhD, Director, Research and Training Center on Community Living, Institute on Community Integration, University of Minnesota, 214 Pattee Hall, 150 Pillsbury Dr. SE, Minneapolis, MN 55455