• E-mail
• Print
• Comment
• Font Size
• Digg
• del.icio.us
• Discuss article

Eligibility Rates of Traditionally Underserved Individuals With Disabilities Revisited: A Data Mining Approach

Posted on: Sunday, 16 October 2005, 03:00 CDT

By Chan, Fong; Wong, Daniel W; Rosenthal, David A; Kundu, Madan M; Dutta, Alo

Abstract - In this study we used a data mining approach to examine factors affecting eligibility decision-making in the vocational rehabilitation process. Severity of disability was found to be the most important variable differentiating individuals who were accepted from those who were rejected for vocational rehabilitation services. Individuals with severe disabilities had a significantly higher rate for acceptance (93%) than people without severe disabilities (55%). Race was found to be a slight differentiating factor in eligibility decisions. This study found that the prevalence and opportunity for racial bias is smaller when the criteria for eligibility are clearly defined as in the case of severe disability. Racial bias has a higher propensity to manifest itself when the criteria for eligibility are more ambiguous as in the case of justifying eligibility for vocational rehabilitation services of those without severe disabilities.

The study of the effect of racial bias on acceptance rates (also known as eligibility rates) in state vocational rehabilitation agencies has been an important topical area in rehabilitation counseling research. Concerns about the detrimental effect of racial bias had been highlighted in the Rehabilitation Act Amendments of 1992:

Patterns of inequitable treatment of minorities have been documented in all major junctures of the vocational rehabilitation process. As compared to European Americans, a larger percentage of African-American applicants to the vocational rehabilitation system is denied acceptance. Of applicants for service, a larger percentage of African-American cases is closed without being rehabilitated. Minorities are provided less training than their European American counterparts. Consistently, less money is spent on minorities than their European American counterparts (p. 4364).

Atkins and Wright (1980) first documented that the acceptance rate for African Americans was proportionally lower (about 5.5%) than European Americans in the federal-state vocational rehabilitation program. Herbert and Martinez (1992) conducted a similar study using the Rehabilitation Services Administration (RSA) 911 data. They found that both African Americans and Latino Americans were more likely to be found ineligible for vocational rehabilitation services than their European American counterparts. These two studies have garnered considerable attention among multicultural rehabilitation researchers and additional acceptance rate studies conducted by these researchers seemed to support the assertions of the Rehabilitation Act (e.g., Dziekan & Okocha, 1993; Feist-Price, 1995; Wilson, 2000).

However, it should also be noted that several researchers (e.g., Capella, 2002; Wheaton, 1995; Wheaton, Wilson, & Brown, 1996) have questioned the methodology of many racial bias studies used to analyze RSA-911 data. For example, since Atkins and Wright (1980) did not formally test for statistical differences, Bolton and Cooper (1980) questioned whether the results (a 5.5% difference in acceptance rates) were a true indication of differential acceptance rates for African Americans. Noting the limitations of the statistical test used by Dziekan and Okocha (1993), Wheaton (1995) re-analyzed the same state agency data previously used by Dziekan and Okocha with a more appropriate chi-square statistical procedure (chi-square test of homogeneity of proportions) and found that although European Americans had a 5.4% higher acceptance rate than African Americans, there was no statistical difference in the proportion of acceptance rates for rehabilitation services between the two groups. Capella (2002) also enumerated several limitations of using chi-square statistics in previous research:"...it [chi- square statistic] provides little information. It also does not allow the researcher to control for other relevant variables that may contribute to acceptance for and outcomes of vocational rehabilitation services" (p. 145). She further identified the focus of previous studies on only two racial groups (European Americans vs. African Americans) as another limitations.

Capella then examined potential inequities in the federal-state vocational rehabilitation program with a more sophisticated methodology (than the chi-square test) using logistic regression analysis. She tested several logistic regression models using gender and race as predictors of acceptance rates and employment outcomes and age, education, and severity of disabilities as control variables. She found that (a) acceptance rates favored European over African Americans after controlling for age and education; (b) employment outcomes favored European Americans over both African Americans and Native Americans; and ) quality of closures favored men over women, with age acting as a modifier; as the age of women increased, the odds of their being placed in a high quality employment decreased when compared to men. However, when Wilson et al. (2002) applied logistic regression analysis to examine the effect of race, gender, education, work status at application and primary source of support at application to predict acceptance rates, they came to a totally different conclusion about the effect of race. Wilson et al. found that after controlling for gender, education, work status at application and primary source of support at application, African Americans were twice more likely to be accepted for VR services than European Americans. The contradictory results between Capella and Wilson et al. are somewhat confusing, as they both extracted their RSA-911 data from the same fiscal year (i.e., FY 1997). If nothing else, the contradiction reveals that there are problems associated with using the traditional hypothesis testing approach to predict acceptance rate and rehabilitation outcomes based on large archival databases such as the RSA-911 case service report data.

Most recently, Rosenthal, Ferrin, Wilson, and Frain (in press) conducted a meta-analysis investigating client race and acceptance for VR services representing an aggregate total of nine years of RSA- 911 data (between 1989 and 1998) and revealed statistically significant differences between VR acceptance rates for European Americans versus African Americans. European Americans were found to be 1.54 times more likely to be accepted for VR services than African Americans. While some of the past investigations have reVealed that ethnicity accounts for a negligible amount of variance in VR acceptance (Wheaton, 1995; Wilson, 1999), the Rosenthal et al. (in press) meta-analysis indicated a more robust effect size.

Rehabilitation researchers have recommended the use of a pattern recognition techniques (viz., automatic interaction detector [AID] for rehabilitation outcome studies). Specifically, Bolton (1972) identified the following advantages of AID for analyzing large databases: (a) AID is the most flexible and required the least assumptions of any of the multivariate prediction methods, (b) AID sorts rehabilitation consumers into homogeneous groups using an array of independent consumer characteristics and process variables, ) AID produces results which are displayed in a multivariate expectancy table; and (d) AID allows for the definition of consumer end groups which can be identified by following the sequence of analyses, thus revealing the number of consumers in the specified groups and the proportion who are successful on the criterion. It appears that the RSA-911 data is ideally suited for this kind of analysis. Unfortunately, the pattern recognition approach to analyzing huge rehabilitation data did not receive the attention of other rehabilitation researchers, probably due to the lack of computer power and computer software to conduct this kind of analysis in the 1970s.

Data Mining Techniques

Recently, with the arrival of powerful microprocessors and the advent of e-commerce, pattern recognition has become an important research tool for business and marketing researchers (Nong, 2003). Marketing researchers use an array of data mining techniques (the modern name for pattern recognition), including chi-squared automatic interaction detector (CHAID) to study the interaction between demographic characteristics of web users and their consumer behavior on the Internet. Data mining is defined as the extraction of hidden predictive information from large databases (Nong, 2003). Using a data mining technique such as CHAID, researchers can use the information to segment Internet users into homogeneous subgroups and direct advertisement and promotion activities toward unique consumer behavior patterns of each homogeneous subgroup (Nong, 2003). Some of the important applications of data mining techniques include credit scoring, wage prediction, and market segmentation (Nong, 2003; SPSS, 1998).

Because of the success of data mining techniques in solving business and marketing research questions, these techniques have also increasingly been profitably used to analyze pattern recognition problems in large healthcare and social service databases, with encouraging results. For example, Smith and Grawe (2003) applied CHAID to conduct an exploratory search for meaningful patterns of process and session-outcomevariables, based on 3,383 10- min sequences of 740 rated counseling sessions. They uncovered from the CHAID analysis that clients' responsiveness to intervention and motivation were most predictive of session productivity. Melchoir et al. (2001) also applied CHAID to a large data base containing prognoses and characteristics of HIV patients. The author found that 17 indicators reflecting service needs, vulnerabilities, and demographic characteristics can be used effectively to predict the unmet needs of groups of traditionally underserved individuals with HIV/AIDS.

Several years have passed since the passage of the 1992 Rehabilitation Act Amendments identified racial bias as a problem in the federal-state vocational rehabilitation program. Rehabilitation counselors are now better trained and may be more sensitive to multicultural issues. Pattern recognition techniques appear particularly suited to re-examine the effect of demographic variables on acceptance rates, as well as the effects of demographic variables, process variables, and their interaction effects on rehabilitation outcomes, using a current RSA-911 dataset. The purpose of this current study is to examine factors influencing VR acceptance rates of traditionally underserved individuals with disabilities using a data mining approach.

Method

Participants

Data for this study was extracted from the Rehabilitation Services Administration Case Service Report (RSA-911) for fiscal year (FY) 2001. The original database contained 639,823 individuals. The number of individuals was reduced to 628,248 after eliminating cases with missing data related to the dependent and independent variables in this study. This reduced database included 342,830 men (55%) and 285,418 women (45%). Racial and ethnic backgrounds were diverse, including 66% European American, 23% African American, 9% Latino, 1% Native American, and 1% Asian American.

Native Americans comprised the smallest group in the data base (n=6,283). To maximize statistical power, we drew a stratified random sample of this size from each other racial group, resulting in a total of 32,933 individuals from all racial groups (N=32,933). Mean age of participants in the stratified sample was 36.5 years (SD = 12.8), and 56% were men. About 43% had completed high school, 30% had less than high school education, 19% had a least some college education, and 8% had special education. In terms of severity 74% were considered having a severe disability and 26% a non-severe disability or no disability.

Table 1

Acceptance Rates of VR Customers by Severity of Disability, Race, and Gender

Variables

Variables commonly found in acceptance-rate studies in the rehabilitation counseling literature were selected for this study. The Exhaustive CHAID (Chi-squared Automatic Interaction Detector), a data mining technique, was used to analyze the data for this study. The Exhaustive CHAID requires the use of categorical variables. For this reason continuous variables (e.g., age and education) were receded to categories to conform to this requirement.

Dependent Variable. The dependent variable in this study is acceptance rate. For data analysis individuals who are closed in Statuses 26, 28, or 30 were coded as "1" accepted for services. Conversely, individuals who are closed in Statuses 08 and 38 were coded as "O" not accepted for services.

Independent Variable. The predictor variables include gender (male or female), race and ethnicity (European American, African American, Hispanic/Latino, Native American, and Asian American), severity of disability (severe vs. not severe), age (16-34, 35-54, 55-64, 65 and older), and education (special education, less than high school, high school graduate, and at least some college).

Data Analysis

In the current study Exhaustive CHAID was used to build classification trees. CHAID analysis is a useful, exploratory technique to investigate possible interactions in categorical data, testing the predictor variables one at a time (Kosciulek, 2004). CHAID predicts membership of cases in the classification of a categorical criterion variable from the measurements on specified predictor variable(s). CHAID classification trees are used to predict or explain responses on a specified categorical criterion variable with the primary goal of obtaining the most accurate prediction possible. The hierarchical nature of the CHAID classification tree readily lends itself to an easily-read, graphic display of the variables, providing a visual depiction of criterion and predictor variable interactions that may not be otherwise observable or detected in traditional analytic procedures. The alpha level for all statistical tests was .05, corrected for the number of statistical tests within each predictor using a Bonferroni correction. The statistical software SPSS AnswerTree 2.0 was used to conduct the Exhaustive CHAID analyses (SPSS, 1998).

Results

Descriptive Statistics

Information related to the demographic characteristics of VR customers and their acceptance rates for VR services in the current study is presented in Table 1. As can be observed the acceptance rate was 80% and the rejection rate was 20% for the overall sample. Acceptance rates for men and women were quite similar. European and Asian Americans appear to have higher acceptance rates than the African American, Latino, and Native American groups. The difference between the European and African American groups was about six percent, similar to the five percent difference reported consistently in the acceptance-rate literature (Capella, 2002). As expected, and consistent with the mandate of the Rehabilitation Act, individuals with severe disabilities had a much higher acceptance rate than people with less severe disabilities.

Figure 1. CHAID Diagram of the Splits of Eligibility Decisions for Individuals with Severe Disabilities

Figure 2. CHAID Diagram of the Splits of Eligibility Decisions for Individuals without Severe Disabilities

Data Mining Results

For the Exhaustive CHAID analysis with race, gender, severity of disability, age, and education as predictor variables and eligibility as the criterion variable, a solution was found with a risk of false classification of 16% and a risk of 17% for cross- classification. In general, the predictors are better at predicting acceptance (88%) than rejection (66%). The overall correct classification of 84% is only a slight improvement over the base rate of 80%. In order to fit the CHAID diagram to the page, the tree diagram is depicted in two figures. Figure 1 shows the right split of the decision tree predicting the acceptance rates of people with severe disabilities and Figure 2 shows the left split of the tree predicting the acceptance rates of people without severe disabilities.

The most significant predictor of acceptance rate is severity of disability. The effect size (odds-ratio) for severity is computed to be 16.4 and is considered a large effect. People with severe disabilities (93%) are more likely to be accepted for services than people without severe disabilities (45%). The CHAID analysis split the sample to 24 end groups (terminal nodes).Seven groups were found to have significantly higher rates of acceptance than the acceptance rate (80%) of the overall sample. These seven groups represented all individuals with severe disabilities in the sample. The following is a brief description of these high acceptance rate groups:

Node 14. Node 14 represents 153 European American with severe disabilities who were 65 years or older, with 152 people accepted for services, or a 99% acceptance rate. These 153 elderly individuals represent 0.49% (node percent) of VR consumers in the overall sample. The 152 individuals accepted for services represent 0.60% (accept percent) of all people accepted for VR services in the overall sample. An index score of the ratio of these two percentages shows how the proportion of people accepted for services in this group compares to the proportion of people accepted for services in the overall sample. For node 14 the index score is about 124% (0.60/ 0.49), meaning the proportion of people accepted for services in this group is about 124% the acceptance rate for the overall sample.

Node 21. Node 21 represents 585 African Americans with severe disabilities who had received special education. The acceptance rate for this group is 97% and the index score is about 120%.

Node 10. Node 10 represents 4711 Asian Americans with severe disabilities. The acceptance rate for this group is 96% and the index score is about 120%.

Node 13. Node 13 represents 4751 European Americans with severe disabilities between 16 and 64 years of age. The acceptance rate for this group is 93% and the index score is about 116%.

Node 9. Node 9 represents 4,520 Native Americans with severe disabilities. The acceptance rate for this group is 93% and the index score is about 116%.

Node 22. Node 22 represents 402 Latino Americans with severe disabilities who had received special education. The acceptance rate for this group is 91 % and the index score is about 113%.

Node 16. Node 16 represents 8039 African Americans or Latino Americans with severe disabilities who did not receive special education (i.e., less than high school education, high school graduates, or college education). The acceptance rate for this group is 91% and the index score is about 113%.

The following eight groups demonstrated significantly higher rates of rejection than the rejection rate (i.e., 20%) for the overall sample. The following is a brief description of each of these high rejection-rate groups:

Node 4. Node 4 represents 1631 African Americans without severe disabilities, with 1021 people rejected for services, or a 63% rejection rate. These 1631 individuals represent 5.19% (node percent) of VR consumers in the overall sample. The 1021 individuals represent 16.53% (reject percent) of all people rejected for \VR services in the overall sample. An index score of the ratio of these two percentages shows how the proportion of people rejected for services in this group compares to the overall proportion of people rejected for services. For node 4, the index score is about 318% (16.53/5.19), meaning the proportion of people rejected for services in this group is about three times the rejection rate for the overall sample.

Node 5. Node 5 represents 1,763 Native Americans without severe disabilities. The rejection rate for this group is 60% and the index score is about 303%.

Node 17. Node 17 represents 1,272 Asian or Latino American men without severe disabilities who had completed either a special education, a high school education, or some college education. The rejection rate for this group is 55% and the index score is about 280%.

Node 3. Node 3 represents 1,379 European Americans without severe disabilities. The rejection rate for this group is 55% and the index score is about 277%.

Node 24. Node 24 represents 380 Latino American men without severe disabilities who did not complete high school. The rejection rate for this group is 51% and the index score is about 260%.

Node 19. Node 19 represents 683 Asian American women without severe disabilities. The rejection rate for this group is 50% and the index score is about 255%.

Node 20. Node 20 represents 857 Latino American women without severe disabilities. The rejection rate for this group is 44% and the index score is about 226%.

Node 23. Node 23 represents 289 Asian Americans without severe disabilities who did not graduate from high school. The rejection rate for this group is 40% and the index score is about 200%.

Table 2

Reasons for Rejecting People for VR Services by Race

Reasons for Refusal

Overall, refusal of services, failure to cooperate, and the category "all other reasons" appear to be the most common reasons for closure. Specifically, a higher percentage of Asian Americans (24%) was closed as "unable to locate or contact" than other racial groups. European Americans had a higher percentage (26%) of closure for refusal of services than did other racial groups, whereas Latino Americans had the lowest percentage (15%). African Americans had the highest percentage (25%) of closure for failure to cooperate among racial groups, and Asian Americans had the lowest percentage (12%). Latino Americans had the highest percentage (13%) of closure for no disabling conditions among racial groups (see table 2).

Discussion

The most important variable that differentiates people who were accepted for VR services from those who were rejected for services is severity of disability. Individuals with severe disabilities had a significant higher rate for acceptance (93%) than people without severe disabilities (55%). The importance of severity of disability as a factor in eligibility decision is consistent with the Rehabilitation Act, which mandates services to individuals with the most severe disabilities as the highest priority. The contribution of severity of disability in eligibility decision-making is also significantly overlooked in the previous acceptance rate literature. Race was the second most important variable in explaining eligibility decisions. In the severe disability group, Asian American had the highest acceptance rate (96%) and African American and Latino had the lowest rate (91%). The acceptance rate for European American was 93%. In the non-severe group, the differences were even more acute. Asian and Latino American had the highest acceptance rate of 50%, while African American had the lowest rate of 37%. The difference between European American (45%) and African American was about 8%. Interestingly, Asian Americans also had the lowest rate (12%) of non-acceptance for the reason of failure to cooperate and the highest rate (24%) of non-acceptance for the reason of unable to locate or contact. African American had the highest rate of closure due to failure to cooperate (25%) and European American had the highest rate of closure due to refusal of services (26%).

The results of this study and the Rosenthal et al. (in press) meta-analysis stand in contrast to several earlier studies that found ethnicity and VR acceptance to be independent (e.g., Wheaton, 1995; Wilson, 1999). The studies conducted by Wheaton, and Wilson reported variances explained by ethnicity were only 3% and less than 1 %, respectively. Although the present study and a majority of others have demonstrated statistically significant race effects in regards to VR acceptance rates, it is important to note that, in this study, and in general, the amount of explained variance has been small.

Although there are many variables that may influence the findings that VR consumers from underrepresented groups are less likely to be accepted for VR services than are European Americans, one potential influence that must be considered is the possibility of racial discrimination against particular subgroups (Herbert & Martinez, 1992; Wilson, 2000; Wilson, Harley, McCormick). Many studies within the social sciences, particularly within social cognition research, have found that racial stereotypes tend to be negative and enduring and that they are often implicit (out of the realm of awareness) and thus, resistant to change (Devine, Plant, & Buswell 2000). Boski (1988) reported that when African Americans present themselves in ways that are consistent with negative stereotypes or schemas held by European Americans, they may trigger or exacerbate negative evaluations that may not be within the evaluator's awareness. As postulated by Dziekan and Okocha (1993), a VR counselor's negative perception of a potential customer's capacity for success (or failure) may produce an inaccurate determination of the customer's ability to benefit from VR services, resulting in disproportionate numbers of underrepresented customers found ineligible for VR services.

In the current study, VR consumers can be further partitioned into 24 homogeneous groups that were statistically different from each other based on combinations of the gender, race, severity, education, and age variables. For example, for some unknown reason, European Americans who were 65 or older with a severe disability had the highest acceptance rate (99%) than any other groups. African Americans with severe disabilities who had completed special education also had a high acceptance rate of 97%. The data mining approach also provides more detailed information and insights about interaction between demographic variables and acceptance rates through the segmentation of the sample into 24 mutually exclusive groups.

Importantly, the problem of racial bias against African American and to a lesser degree Latino American and Native American is less in the severe disability group than in the non-severe group. The results suggested that the prevalence and opportunity for racial bias is smaller when the criteria for eligibility are clearly defined, as in the case of severe disability. Racial bias has a higher propensity to manifest when the criteria for eligibility are more ambiguous, as in the case of making a justification to determine an individual without severe disabilities as eligible for VR services. These results are supported by social cognition research demonstrating that counselors' susceptibility to clinical biases are heightened when they are faced with ambiguous data regarding clients (Lopez, 1989).

The results of the current study are consistent with Strohmer and Leierer's (2000) review of the counselor bias literature. Counselors have been found to be prone to being susceptible to systematic biases associated with specific client variables such as gender, age, sexual preference, social class, and disability type. Phenomena such as diagnostic overshadowing may lead counselors to give undue weight on one salient variable, while disregarding or missing other important information (Spengler, Strohmer, & Prout, 1990). Once counselors formulate negative hypotheses regarding clients, they may demonstrate confirmatory bias, seeking confirmatory information while paying less attention to disconfirmatory information, even in the face of contradictory evidence (Strohmer & Shivy, 1994; Strohmer, Shivy, & Chiodo, 1990). McGinn, Flowers, and Rubin (1994) further suggested that cultural biases were, at least in part, responsible for the inequitable patterns of rehabilitation counseling acceptance rates and service delivery for African American consumers. It is important to note that such biases may not be overt, intentional or even within consciousness of the practitioner. Research regarding the implicit nature of stereotypes indicates that even when stereotypes are not explicitly recognized or noted, implicit stereotypes can have significant influence on perceptions (Fazio & Olson, 2003), thus, influencing clinical decisions (Rosenthal, 2004; Rosenthal & Berven, 1999).

Solution focused research to identify ways to reduce the automatic activation of stereotypes would seem important. Garb (1998) stated "by learning about the conditions when biases and errors are likely to occur, clinicians may become more adept at making judgments and they may become better at deciding how likely it is that their judgments are correct" (p. 4). Garb also reviews "debiasing" strategies that have been recommended in the research literature, with the most frequent being the need to consider multiple alternatives when making judgments, such as diagnostic determinations, explanations or attributions for observed behavior (including possible influences of the surrounding context), and treatment and service plans. Specifically, rehabilitation counselors can study how initial biases may affect the counselor's determinations regarding acceptance and choice of rehabilitation services.

This is the first study to present data that illustrates the propensity for reha\bilitation counselors to have varying susceptibility to racial bias depending on the certainty (acceptance for consumers with severe disabilities) or the uncertainty (acceptance for consumers without severe disabilities) of the data regarding the consumer. Thus, given the results of the present study, eligibility determinations may be influenced by implicit negative, racial stereotypes and attitudes that tend to exist in more ambiguous situations. There are several implications for minimizing racial bias in the eligibility decision-making process. First, counselors must be made aware of their potential for implicit racial bias as uncovered in archival analysis of the RSA-911 data and other experimental studies of racial bias. Second, rehabilitation administrators and policy makers can minimize the influence of racial bias in the eligibility decision-making process. Eligibility criteria and decision-making steps should be operationalized as clearly as possible to minimize subjectivity. Lastly, rehabilitation counselors and rehabilitation counselor educators work to minimize the deleterious affects of racial bias by gaining awareness of the potential for bias to influence clinical decisions, particularly in the face of ambiguous data. Thus, subsequent sensitivity training can be developed to address the contextual nature rehabilitation counselor clinical decisions based on the certainty or ambiguity of the available consumer data. Lastly, we must emphasize that although race was the second most important factor in determining consumer acceptance or non- acceptance for VR services, given the small percentages of difference across race indicates that, although they are statistically significant differences, they may be of limited practical significance.

References

Alston, R. J., & Mngadi, S. (1992). The interaction between disability status and the African-American experience: Implications for rehabilitation counseling. Journal of Applied Rehabilitation Counseling, 23 (2), 12-16.

Atkins, B. J., & Wright, G. N. (1980). Three views: Vocational rehabilitation of Blacks: The statement. Journal of Rehabilitation, 46(2), 40, 42-46.

Beynon, M., Curry, B., & Morgan, P. (2001). Knowledge discovery in marketing: An approach through Rough Set Theory. European Journal of Marketing, 35, 915-935.

Bolton, B. (1972). Predicting client outcome from intake data. Paper presented at the meeting of the Arkansas Rehabilitation Association, hot Springs, Arkansas.

Bolton, B., & Cooper, P. G. (1980). Three views: Vocational rehabilitation of Blacks: The comment. Journal of Rehabilitation, 46, 41-49.

Boski, P. (1988). Cross-cultural studies of person-perception: Effects of ingroup/outgroup membership and ethnic schemata. Journal of Cross-Cultural Psychology, 19, 287-328.

Brann, M., Anderson, J., Brann, M. (2002). E-medicine and health care consumers: Recognizing current problems and possible resolutions for a safer environment. Health Care Analysis, 10(4), 403-415.

Capella, M. E. (2002). Inequities in the VR system: Do they still exist? Rehabilitation Counseling Bulletin, 45, 143-153.

Devine P. G., Plant, E. A., Buswell B. N. (2000.) Breaking the prejudice habit: progress and obstacles. In: S Oskamp (ed.) Reducing Prejudice and Discrimination. New Jersey: The Clairement Symposium on Applied Social Psychology, 185-210.

Dudley, W. N., Dilorio, C., & Soet, J. (2000). Detecting and explicating interactions incategorical data. Nursing Research, 49(1), 53-56.

Dzieken, K.. I., & Okocha, A. G., (1993). Accessibility of rehabilitation services; comparison by racial-ethnic status. Rehabilitation Counseling Bulletin, 36, 183-189.

Epstein, I., Blumenfield, S. (2001). Clinical data-mining in practice-based research: Social work in hospital settings. New York: Haworth Social Work Practice Press.

Feist-Price, S. (1995). African Americans with disabilities and equity in vocational rehabilitation services: One's state's review. Rehabilitation Counseling Bulletin, 39, 119-129.

Garb, H. N. (1998). Studying the clinician: Judgment research and psychological assessment. Washington, DC: American Psychological Association.

Goldberg, L.R. (1959). The effectiveness of clinicians'judgment. Journal of Consulting Psychology, 23, 25-33.

Herbert, J., & Martinez, M. Y. (1992). Client ethnicity and vocational rehabilitation case service outcome. Journal of Job Placement, 8(1), 10-16.

Hiler, E. W., & Nesvig, D. (1965). An evaluation of criteria used by clinicians to infer pathology from figure drawings. Journal of Consulting Psychology, 29, 520-529.

Houtenville, A. J. (2004). Disability statistics in the United States. Ithaca, NY: Cornell University Rehabilitation Research and Training Center, www.disabilitystatistics.org. Posted May 15, 2003. Accessed August 02, 2004.

Johnston, R., & McNeal, B.F. (1967). Statistical versus clinical prediction. Journal of Abnormal Psychology, 72, 335-340.

Kosciulek, J. F., (2004). On Classification Trees. Rehabilitation Education, 18, 139-145.

Lopez, S.R. (1989). Patient variable biases in clinical judgment: Conceptual overview and methodological considerations. Psychological Bulletin, 106, 184-203.

McGinn, F., Flowers, C.R., & Rubin, S. E. (1994). In quest of an explicit multicultural emphasis in ethical standards for rehabilitation counselors. Rehabilitation Education, 7, 261-268.

Melchoir, L. A., Huba, G. J., Gallagher, T., Jean-Louis, E., McDonald, S. S., Smereck, G. A. D., German, V. F., Brown, V. B., & Panter, A. T. (2001). Unmet needs in groups of traditionally underserved individuals with HIV/AIDS: Empirical models. Home Health Care Services Quarterly, 19(1), 29-51.

Nong, Y. (2003). Handbook of data mining. Mahwah, NJ: Lawrence Erlbaum Associates.

Rehabilitation Act Amendments of 1992, 29, U.S.C. 701, et seq.

Rosenthal, D. A. (2004). Effects of client race on clinical judgment of practicing European American vocational rehabilitation counselors. Rehabilitation Counseling Bulletin, 47, 131-141.

Rosenthal, D. A., & Berven, N. L. (1999). Effects of client race on clinical judgment. Rehabilitation Counseling Bulletin, 42, 243- 265.

Rosenthal, D., Ferrin, M., & Wilson, K., & Frain, M. (in press). Client race and vocational rehabilitation acceptance: A meta- analysis. Journal of Rehabilitation..

Smith, E. C., & Grawe, K. (2003). What makes psychotherapy session productive? A new approach to bridging the gap between process research and practice. Clinical Psychology and Psychotherapy, 10, 275-285.

Sonquist, J. A. (1970). Multivariate model building: The validation of a search strategy. Ann Arbor, MI: Institute for Social Research.

Spengler, P. M., Strohmer, D. C., & Prout, H. T., (1990). Testing the robustness of the diagnostic overshadowing bias. American Journal on Mental Retardation, 95, 204-214.

SPSS (1998). AnswerTree 2.0 user's guide. Chicago: Author.

Strohmer, D. C., & Leierer, S.J., (2000). Modeling rehabilitation counselor clinical judgment. Rehabilitation Counseling Bulletin, 44(1), 3-11.

Strohmer, D. C., & Shivy, V. A.,(1994). Bias in counselor hypothesis testing: Testing the robustness of counselor confirmatory bias. Journal of Counseling & Development, 73(2), 191-198.

Strohmer, D. C., Shivy, V. A., & Chiodo, A. L. (1990). Information processing strategies in counselor hypothesis testing: The role of selective memory and expectancy. Journal of Counseling Psychology, 37, 465-472.

Sue, D. W., Arrendondo, P., & McDavis, R. J. (1992). Multicultural counseling competencies and standards: A call to the profession. Journal of Counseling and Development, 70, 477-486.

Wheaton, J. E. (1995). Vocational rehabilitation acceptance rates for European Americans and African Americans: Another look. Rehabilitation Counseling Bulletin, 38, 224-231.

Wheaton, J. E., Wilson, K. B., & Brown, S. M. (1996). The relationship between vocational rehabilitation services and the consumer's sex, race, and closure status. Rehabilitation Counseling Bulletin, 40, 116-133.

Wilson, K. B. (2000). Predicting vocational rehabilitation acceptance based on race, education, work status, and source of support at application. Rehabilitation Counseling Bulletin, 43(2), 97-105.

Wilson, K. B., Alston, R. J., Harley, D. A., & Mitchell, N. A. (2002). Predicting VR acceptance based on race, gender, education, work status at application, and primary source of support at application. Rehabilitation Counseling Bulletin, 45, 132-142.

Wilson, K. B., Harley, D. A., McCormick, K., Jolivette, K., & Jackson, R. L. (2001). A literature review of vocational rehabilitation acceptance and explaining bias in the rehabilitation process. Journal of Applied Rehabilitation Counseling, 32(1), 24- 35.

Fong Chan is a professor and David A. Rosenthal is an assistant professor, Department of Rehabilitation Psychology and Special Education, University of Wisconsin-Madison. Daniel W. Wong is a professor and coordinator of the doctoral program, Department of Rehabilitation Counseling, East Carolina University. Madan M. Kundu is professor and chair and Alo Dutta is an assistant professor, Department of Rehabilitation and Disability Studies, Southern University.

Acknowledgment Preparation of this report was supported in part by a Disability and Rehabilitation Research Project ("Rehabilitation Research Institute for Underrepresented Populations"), which was funded by Grant#H133A031705 from the National Institute on Disability and Rehabilitation Research to Southern University.

Copyright National Rehabilitation Counseling Association Fall 2005

Source: Journal of Applied Rehabilitation Counseling

More News in this Category