# Development of a natural language processing (NLP) based therapist tool for culturally responsive care

> **NIH NIH K23** · UNIVERSITY OF PENNSYLVANIA · 2024 · $194,184

## Abstract

PROJECT SUMMARY
 Institutional discrimination has limited access to important resources for Black, Indigenous, People of
Color (BIPOC) and Lesbian, Gay, Bisexual, Transgender, and Queer (LGBTQ+) communities, and have
impacted the quality of mental healthcare that these communities receive. Researchers have focused on
promoting cultural competency (CC) among therapists to address mental healthcare disparities for BIPOC and
LGBTQ+ communities. Therapist CC is broadly associated with improved clinical outcomes and therapeutic
alliance. However, current methods of assessing therapist CC are infeasible to implement in clinical settings.
Therapist-client power differentials limit clients from providing specific feedback, and behavioral coding is
infeasible to implement in healthcare systems given difficulties with scaling up session evaluations. Natural
language processing (NLP) provides a promising alternative to current methods of assessing and providing
feedback on therapist CC, and recent applications of NLP based tools to psychotherapy settings indicate that
they are viable methods of improving culturally responsive care. The purpose of the current study is to develop
a prototype of a NLP-based therapist CC feedback tool (HEAL). We will first create an annotated dataset by
coding 300 therapy sessions with a behavioral coding system assessing culturally responsive care (Aim 1). We
will then create a prototype of HEAL (Aim 2) by (a) developing a feedback visualizer based on input from
community advisory boards (CABs) comprised of supervisors and therapists, and BIPOC LGBTQ+ clients, (b)
selecting a speech recognition software, and (c) developing and validating NLP models based on the annotated
dataset from Aim 1. We will use rapid cycle prototyping and testing to iteratively revise HEAL, meeting monthly
with CAB members. We will assess acceptability, appropriateness, feasibility, and usability through interviews
with CAB members, and quantitative measures. Finally, we will pilot the prototype of HEAL with 15 therapists
using standardized patients, and assess acceptability, feasibility, appropriateness, and usability through
individual feedback interviews and quantitative measures (Aim 3). The results of this study will culminate in the
development of a novel therapist CC support tool, and ideally position me to pursue R01 funding to evaluate its
effectiveness in a randomized clinical trial. The proposed aims are in line with NIMH's strategic plan to develop
innovative service delivery models to dramatically improve the outcomes of mental health services received in
diverse communities and populations. The current study and training plan will allow me to develop expertise in
community partnership in user centered design, and understand how cultural contexts and biases impact NLP
models, and strongly position me to be a leader in multidisciplinary community-based research developing
innovative methods of addressing mental healthcare inequities.

## Key facts

- **NIH application ID:** 10949091
- **Project number:** 1K23MH137394-01
- **Recipient organization:** UNIVERSITY OF PENNSYLVANIA
- **Principal Investigator:** Patty Beyrong Kuo
- **Activity code:** K23 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2024
- **Award amount:** $194,184
- **Award type:** 1
- **Project period:** 2024-08-07 → 2029-07-31

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/10949091

## Citation

> US National Institutes of Health, RePORTER application 10949091, Development of a natural language processing (NLP) based therapist tool for culturally responsive care (1K23MH137394-01). Retrieved via AI Analytics 2026-06-12 from https://api.ai-analytics.org/grant/nih/10949091. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*