Denise Spät Albino and Dr. Timothy Byron Smith, Counseling and Special Education

For the past several years, the psychology literature has investigated the possibility that mental health services are racially biased. Studies assessing treatment outcomes in multicultural settings have increased over the past few years, with now hundreds of articles comparing treatment outcomes of people of color and whites. Therefore, the purpose of this project was to conduct a meta-analytic examination of recent studies to inform the field of the results accumulating on ethnically sensitive interventions.
The first step of this project consisted of locating data to be examined. Thus, in order to identify published and unpublished articles investigating outcomes of multicultural counseling through 2004, complete searches of five CD-ROM databases were conducted: Medline(medicine), PsycInfo (psychology), Sociofile (sociology), CINAHL (nursing), and ERIC (education). Also, reference sections of retrieved studies were analyzed to identify additional studies, and three professionals who have published extensively in the field were consulted to identify unpublished papers and conference presentations.

After the studies were located, they were coded. Coding consisted of two students who transformed all information of a certain article to quantitative data. Variables such as therapists’ and clients’ gender, age, ethnicity, education, etc. were quantified, and an averaged effect size was computed for each article. Then, for purposes of inter-rater reliability, the article was re-coded by two other students, who performed exactly the same procedure as the first two students. After the second coding process was finished, a student from the first team got together with a student from the second team to verify the information coded the first couple of times. This verification process consisted of comparing the quantitative data from the first and second coding processes, and, at this time, discrepancies between these two sets of information were located. This third team was responsible for getting into a consensus in order to leave just one set of coded information for each article.

After finishing coding all the articles collected, an average across all the studies was obtained so that overall results could be looked at (i.e., overall, is there a significant bias in therapy towards people of color?). It was expected that an average extracted across all the studies would demonstrate that there is a significant difference in the treatment outcomes of people of color and white clients.

The study has not been entirely completed yet, but it is very close to be so. Only a few more articles need to be coded. Therefore, the results that have been obtained so far are basically the same as the results that will be obtained when everything is completed. Also, when the project is completed, an analysis will be performed with specific variables (e.g., social economic status, race, gender, age, etc.) to identify their individual impact on therapy outcomes (i.e., does the effect differ gender?). It is expected that the study will be entirely completed in March of 2004, and then it will be submitted for publishing.

So far, data from 108 studies have been coded. These studies include data from over 44,000 research participants (with an average of 439 participants per report). The overall results yielded an effect size of d = -.14, which was statistically significant (p < .001) but which was small in terms of magnitude. Effect sizes ranged from .96 to -2.11, with substantial variability across studies. (Some studies showed that therapists were negatively biased towards clients of color whereas other studies showed that clients of color actually received preferential treatment compared to European Americans.) Nevertheless, on average the literature demonstrated a very small negative bias against clients of color. Given that the mental health literature is replete with reports of negative racial bias among mental health interventions, the overall effect size was much smaller than had been predicted.

These results can be interpreted within the larger context of North American society. Although some racial bias still exists in society (and in mental health treatments, apparently), this bias is quite small and is likely decreasing even further over time. The publication of these results will inform the mental health profession of ways to potentially improve racial equity in access to mental health care and in the outcomes associated with that care when it is provided.

The largest problem that we had to deal with was coding all the necessary variables from the articles. Although we had to code several variables (e.g., gender, ethnicity, social-economic status, age, etc.) from both the clients and the therapists in each article, many authors did not report the information for all these variables. Most of the time, only part of the information we needed was reported by the author. Thus, at many times, we had to leave a blank for the coding of certain variables. Moreover, the verification part of the coding process was a difficult task because when the information from the first code did not match with the information from the second code, we had to get into a consensus and choose one of the pieces of information.