# Use of topic modeling and stakeholderengagement to map determinants of implementation disparities in addiction and pain research

> **NIH NIH U2C** · STANFORD UNIVERSITY · 2024 · $190,697

## Abstract

ABSTRACT
Despite progress in the development of population-specific innovations and culturally-tailored interventions,
disparities in access, receipt, use, and quality of health care delivery persist for populations that have
historically been underserved, underrepresented, and marginalized in dissemination and implementation (D&I)
research. This is partially due to a need for consistent language, methods, and data elements for
characterizing determinants of health inequities and disparities in D&I research, which often vary based on
population or issue studied. In addition, we lack systematic tools for identifying and mapping population-
specific and culturally-relevant determinants onto existing implementation constructs (e.g., inner setting,
innovation characteristics). The parent grant (U2CDA057717), the Research Adoption Support Center for the
Helping to End Addiction Long-term® (HEAL) Data2Action (HD2A) Program, aims to increase the D&I
capability of opioid use disorder and pain management treatment research. This supplement seeks to develop
and validate a health equity taxonomy using mixed methods (i.e., stakeholder engagement, machine learning).
The proposal breaks new ground in the D&I field by leveraging domain expertise (i.e., D&I and health
equity/health disparities) and machine learning (ML) to develop and validate a taxonomy that will serve to
prioritize the needs of those impacted by health disparities and inequities, articulate culturally-relevant and
population-specific determinants known to influence implementation, and heighten researchers’ ability to
assess which health equity issues are being addressed in D&I research and efforts. Specific aims include: 1)
Develop and validate a taxonomy that serves to assess equity-focused D&I efforts; 2) Build, fine-tune, and
evaluate topic models (i.e., unsupervised machine learning models) using off-the-shelf topic modeling
algorithm tools to identify, extract, and describe the most important topics in a sample of HEAL research
abstracts that address underserved populations; 3) Integrate taxonomy with topic modeling results via rapid
feedback with a Board of Domain Experts and the Mentorship Team. The third aim involves collaborating with
the Board of Domain Experts, including stakeholders, and the candidate’s mentorship team to 1) assess
taxonomy completeness by comparing and contrasting taxonomy categories with themes uncovered via topic
modeling, and 2) evaluate consistency between the taxonomy and uncovered topics in terms of co-occurrence
and distribution of topics across categories. The candidate will refine the taxonomy categories based on
feedback from the Board of Domain Experts and return to members for a final review. Results from this project
will inform the development of a centralized public use product that can be utilized by future HD2A studies to
assess key determinants of D&I related to health equity.

## Key facts

- **NIH application ID:** 10987451
- **Project number:** 3U2CDA057717-03S1
- **Recipient organization:** STANFORD UNIVERSITY
- **Principal Investigator:** WILLIAM C BECKER
- **Activity code:** U2C (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2024
- **Award amount:** $190,697
- **Award type:** 3
- **Project period:** 2022-09-30 → 2027-08-31

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/10987451

## Citation

> US National Institutes of Health, RePORTER application 10987451, Use of topic modeling and stakeholderengagement to map determinants of implementation disparities in addiction and pain research (3U2CDA057717-03S1). Retrieved via AI Analytics 2026-05-23 from https://api.ai-analytics.org/grant/nih/10987451. Licensed CC0.

---

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