Harnessing Data Science to Assess Racial, Ethnic, and Linguistic Diversity in the Clinical Social Worker Workforce

Background and Purpose. Persistent racial and ethnic inequities in mental health service utilization and barriers to creating a more diverse workforce remain significant challenges in the social work field. These disparities are evident in the passing rates on the Association of Social Work Boards exam among racially/ethnically minoritized groups and non-native English speakers. Embedding antiracist principles (Goings et al., 2023), this study aims to highlight existing structures, policies, and procedures that perpetuate racial inequities by employing innovative computational methods to assess the diversity of the social work workforce at the national level. This study is to demonstrate how a data science approach can be applied to quantify the diversity of the social work profession and offer insights into mechanisms perpetuating these inequities.

Methods. This study employs a computational approach as follows: (1) big data preprocessing with Python, (2) web scraping to create a comprehensive dataset of private practice social workers, (3) machine learning for predicting race/ethnicity, and diasporic origin of social workers from their name and zip-code, and (4) statistical analyses to examine social worker distribution and accessibility. Through the use of these computational methods, we aim to mitigate the limitations associated with traditional survey methods used by professional associations like NASW and CSWE, which frequently include sampling and response bias. Using multilevel multinomial and logistic regressions, the role of social work licensure on racial/ethnic and linguistic diversity was assessed. Further, multilevel modeling allows us to control for state-level variance with population-level differences.

Results. In the largest private practice therapist directory, which included 244,333 mental health professionals, 32% identified as social workers based on their titles. Multinomial logistic regressions with White therapists as the reference, showed that social worker licensure was associated with a 22% decreased probability of identifying as Asian (RRR = 0.78, p < .001) and an 8% decreased probability of identifying as Black (RRR = 0.92, p < .001) therapists. However, no significant association was found between social worker licensure and identifying as Hispanic or Latino (p > .05). Furthermore, social worker licensure was associated with a 14% increased odds of providing Spanish language services (OR = 1.14, p < .001), but no significant associations were found for providing Chinese, Korean, Vietnamese, or Tagalog language services (p > .05).

Discussion. This study illustrates how social work researchers can apply computational methods to assess the diversity of the social work workforce at the national level more precisely through the innovative application of computational analysis of web data collection. We will discuss how these and other results offer insights into regional gaps in the licensing and training system that should be addressed to foster a more diverse and culturally competent mental health workforce. The findings could provide evidence to address racial and ethnic disparities in therapist workforce diversity, which could be associated with the disparities in mental health service utilization considering the role of racial/ethnic and language concordance.