Symposia
Dissemination & Implementation Science
Patty Kuo, M.Ed.
University of Utah
Seattle, Washington
Vivek Srikumar, Ph.D
Assistant Professor
University of Utah
Salt Lake City, UT
Michael Tanana, Ph.D
Research Assistant Professor
University of Utah
Salt Lake City, Utah
Karen Tao, Ph.D
Assistant Professor
University of Utah
Salt Lake City, UT
Jake Van Epps, Ph.D
Staff Psychologist
University of Utah
Salt lake City, Utah
Zac Imel, PhD
Associate Professor
University of Utah
Salt Lake City, UT
Due to systemic experiences of discrimination, Black, Indigenous, People of Color (BIPOC) and lesbian, gay, bisexual, transgender, and queer (LGBTQ+) individuals experience not only high levels of psychological distress, but also disparities in quality of mental health care. A primary focus of efforts to reduce therapist contributions to mental health care disparities has been examining cultural competency (CC), which involves a therapist’s ability to navigate the cultural aspects of clinical interactions. Client ratings of CC are generally associated with treatment outcomes and therapeutic processes. While client perceptions of therapist CC are important, a reliance on retrospective client ratings limits what we know about the language that constitutes culturally sensitive care. Recently, Natural Language Processing (NLP) models have been applied to psychotherapy conversations to automatically capture the use of evidence based treatments, topics of conversation, and emotional expression. Prior research demonstrating the feasibility of automatically identifying topics of conversation in psychotherapy suggest that NLP models could be trained to automatically identify when clients and therapists are talking about cultural identities, and inform training and provision of rapid feedback to therapists to help address disparities in care. The purpose of the current study was to develop NLP tools that capture the cultural content of client-therapist interactions among REM and LGBTQ clients. Utilizing 103,170 labeled talk turns from 188 psychotherapy sessions, we evaluated the performance of support vector machine models in recognizing the discussion of cultural topics in psychotherapy at the session and utterance level. Our models were unable to consistently identify cultural conversations at the session level due to the relatively small number of sessions evaluated. Our unigram model that excluded stop words, and took into account contextual information from other utterances in the session, was able to identify some patterns in dialogue associated cultural conversations (Spearman’s rho=0.45; p< 0.001). Our results have important implications for application of NLP tools in facilitating multicultural training and feedback to clinicians, and understanding multicultural processes in therapy.