Poster #177 - Examining Caregiver Perceived Need for and Utilization of Adolescent Counseling Services through Machine Learning

• In Event: 2-055 - Poster Session 06
  In Poster Session: PS 06 Section - Race, Ethnicity, Culture, Context

Integrative Statement

Current literature indicates low service utilization among adolescents in need of mental healthcare (Merikangas et al., 2011). Moreover, research highlights lower rates of service use among Black youth when compared to White youth, even in the face of significant impairment. While investigators have identified logistic barriers (e.g., transportation, cost) that caregivers face as a potential predictor of disparities (Scheppers et al., 2006; Garland, Landsverk, & Lau, 2003), existing evidence also highlights differences in perceived need for care between cultural groups (Cauce et al., 2002). Further examination of factors that impact service utilization differently between demographic groups may contribute to the development of more effective, culturally robust ways to increase access to appropriate care and reduce mental health disparities. The current study employs a random forest algorithm to assess predictors of perceived need and utilization of adolescent services, and evaluate differences between Black and White caregivers.

The present study uses data from 574 caregivers (68.3% Black, 31.7% Non-Hispanic White) who completed the age 16 wave of the Longitudinal Studies of Child Abuse and Neglect (LONGSCAN). Caregivers reported on family demographic information (e.g., gender, race/ethnicity, income), adolescent internalizing and externalizing problems (i.e., Child Behavior Checklist; Achenbach, 1991), prior Child Protective Services (CPS) reports, common caregiver stressors (i.e., transportation, income, child behavior), and perceived need for and utilization of adolescent counseling services in the prior year.

Random forest models included 32 variables drawn from these responses (see Figure 1). Data were grouped by caregiver race/ethnicity (i.e., White, Black), and 500 trees were grown for each outcome – perceived need, and service utilization – to determine relative importance. Inspection of the relative importance plot allowed for identification of variables of high importance (i.e., variables above the curve demonstrating significant increase in predictive discrimination; see Figure 1).

Model accuracies ranged from 80-90% classification. Findings suggest differences in the predictors of perceived need for and utilization of services for adolescents. Among Black caregivers, youth internalizing problems and caregiver stress regarding behavior problems were important predictors of both outcome variables. Similarly, White caregivers’ stress regarding their adolescents’ behavior problems was also an important predictor of their perceived need for services, and adolescents’ internalizing problems also contributed significantly to White caregivers’ likelihood of accessing care. However, among White caregivers, youth externalizing symptoms contributed meaningfully to perceived need and use of services. Notably, chi-square tests suggested report of adolescent internalizing, externalizing, and behavioral stress differed by caregiver race (p < .005). White caregivers reported higher externalizing and internalizing symptom severity, and Black caregivers reporting higher rates of stress relating to youth behavior. Thus, group differences in the importance of variables as predictors of these outcomes should be interpreted with these results in mind. Nonetheless, findings highlight potential differences in the variables that drive perceived need for and access of counseling services between caregivers of different demographic groups, to be tested prospectively in future work. Implications for clinical practice, dissemination, and program implementation will be discussed.

Authors

• Alejandro L. Vazquez, Utah State University
  Presenting Author

• Tommy Chou, Florida International University
  Non-Presenting Author

• Cynthia Navarro, Utah State University
  Non-Presenting Author

• Tyson S. Barrett, Utah State University
  Non-Presenting Author

• Melanie M. Domenech Rodríguez, Utah State University
  Non-Presenting Author