AI-Enhanced and Culturally Adapted STAR-C for Latino Caregivers: A Stage I Pilot Study

University Of Washington, Seattle WA

Investigators

Abstract

PROJECT SUMMARY/ABSTRACT Latino adults face a higher risk of Alzheimer’s disease and related dementias (ADRD) than non-Latino White adults. The number of cases is expected to rise sharply in the coming decades, with Latinos facing the largest increase among all racial and ethnic groups. This growth will intensify caregiving demands on Latino families, who already face challenges managing behavioral and psychological symptoms of dementia (BPSD) and often lack culturally appropriate professional support. STAR-C is an evidence-based behavioral intervention that equips caregivers with strategies for managing BPSD. It was originally delivered in-person by a coach but was later adapted to a virtual format with e-learning modules and phone check-ins with a coach. However, STAR-C was not developed with consideration of Latino cultural values, beliefs, and practices. As a result, the content lacks cultural and linguistic relevance for Latino caregivers. Latino caregivers’ engagement with the STAR-C virtual intervention also remains underexplored since prior testing included mostly non-Latino White participants. To enhance access for Latino caregivers, it is crucial to address cultural and linguistic relevance, time constraints, and the need for personalized support. Artificial intelligence (AI) offers a promising solution by providing culturally and linguistically tailored on-demand support. We propose to integrate an AI-driven virtual assistant with chatbot functionality into STAR-C. This virtual assistant will provide real-time support between coach check-ins and reminders to engage with e-learning modules that have been culturally adapted. This Stage I study has three specific aims. Aim 1 is to develop and integrate an AI-driven virtual assistant in the STAR-C intervention. We will follow our recent approach for creating a Large Language Model (LLM)-based chatbot. We will refine the model with input from STAR-C coaches, selecting the best-performing version. We will conduct usability testing with Latino caregivers to further optimize the virtual assistant based on their feedback. Aim 2 is to evaluate fidelity and caregiver acceptance of the AI-enhanced STAR-C intervention. Fifty Latino caregivers will participate in a 6-month intervention involving e-learning modules, phone check-ins with a human STAR-C coach, and on-demand support from the virtual assistant. The study will assess fidelity of intervention delivery and caregivers’ perceived usefulness, ease of use, behavioral intention, and actual use of the intervention. Aim 3 is to assess Latino caregivers’ attitudes toward using the AI-enhanced STAR-C intervention. Qualitative interviews with the 50 participants will assess their perceived benefits, challenges, and emotional responses to the AI-enhanced STAR-C intervention. These qualitative findings will be integrated with quantitative data from Aim 2 to provide a comprehensive understanding of Latino caregiver acceptability of the AI-enhanced and culturally adapted STAR-C intervention. The findings of this Stage I study will lay the groundwork for advancing along the NIH Stage Model to develop a scalable and effective intervention that improves health outcomes for Latino caregivers and their family members with ADRD.