Development and Pilot of ImpleMentor: An LLM-based interactive tool to support implementation of a technology based, non-trauma focused therapy.

Stanford University, Stanford CA

Investigators

Abstract

Project Summary. Agency administrators and clinic leaders have shown significant interest in making evidence-based psychotherapies (EBPs) for PTSD available in their programs, but have little access to support to address the considerable barriers to implementation. Implementation support strategies can be effective but are hard to scale due to the labor involved, the limited supply of trained facilitators, and lack of dedicated funding for implementation facilitation. This leaves low-resource programs serving underrepresented individuals with little implementation guidance and support beyond non-interactive online repositories of resources that do not offer tailored guidance. Lack of implementation support is a problem in community mental health settings as well as in Department of Veterans Affairs (VA) settings. Thus, there is a substantial need to develop scalable ways to support agencies that have the motivation but not the knowledge or support to implement effective practices. Large language models (LLMs) offer a potential solution to making EBP implementation support more accessible and equitable. LLMs can be built to support individuals or teams in developing implementation plans and troubleshooting problems as they attempt to execute those plans. This may enable fully self-directed implementation, or at least substantially reduce the level of human support required for implementation success. We have developed a prototype for an LLM-based virtual interactive facilitation resource called ImpleMentor. ImpleMentor has great potential to fill a pressing implementation need, as it can be used on-demand to develop an implementation plan, track progress, and troubleshoot challenges. In this exploratory project, we will refine ImpleMentor and pilot test it in 8 community and VA-based settings that serve underrepresented clinical populations as they implement webSTAIR, a technology-supported treatment for PTSD. We will examine its ability to activate mechanisms that support implementation. We will also evaluate outcomes including EBP reach and equitable use of webSTAIR, clinic leaders’ completion of implementation plans, and ImpleMentor’s costs and readiness for deployment. This project complements the other CREATE projects, which focus on therapist EBP training and patient support, by providing a scalable tool for supporting implementation in settings that lack resources for cost-and time-intensive human implementation support. It will lay groundwork for a fully-powered hybrid implementation-effectiveness trial or research to optimize the efficiency of implementation support.