Developing a reminiscence therapy platform—LifeBio MemoryTM—with a novel machine‐learning‐based application that transfers speech to text and generates life stories for nursing home residents with dementia

Abstract

AbstractBackgroundNonpharmacologic interventions positively influence cognition, mood, and other behavioral and psychological symptoms of dementia. One such intervention, LifeBio, creates conversations through a series of questions to elicit memories in persons living with dementia. Gathering and processing life story data is often done manually, leading to challenges with feasibility. As a result, novel software, LifeBio MemoryTM, is being developed to streamline the process with the use of machine learning and artificial intelligence. LifeBio Memory will minimize the need to hand‐write or type individualized life story data with the use of speech‐to‐text technology and automatic processing.MethodSeven focus groups (n=35) were conducted with persons living with dementia, family care partners, and residential care staff. Feasibility of use and other general reactions to interacting with the new LifeBio Memory prototypes were examined. Each focus group was audiotaped and transcribed, followed by a thematic analysis of the data that resulted in a list of recommended changes that were first prioritized and then discussed by the research team.ResultResults from the work groups strongly suggest that the use of the LifeBio Memory application is the preferred method for obtaining life story information, showing less support for writing or typing responses as in previous versions of LifeBio. Staff participants report the app would create a more efficient workflow, allowing them to serve a larger number of residents. While a consensus was reached on the value of the concept of LifeBio as well as the LifeBio Memory prototypes, potential barriers to implementing the new platform in residential care settings with persons living with dementia were also identified.ConclusionThe existing LifeBio program faces challenges to scalability due to the time‐and labor‐intensive nature of processing life story information. Results suggest the feasibility of using machine‐learning as an effective, efficient, and preferred method for collecting life story data in place of manualized techniques of the past. Findings from this focus group study will be used to develop a list of specific changes to existing LifeBio processes and materials as well as a plan for developing a full‐scale system to automatically record and process individuals' life stories.