
Investigators:
Jennifer Rae Myers, Musical Health Technologies
Chelsea S. Brown, Musical Health Technologies
MassAITC Cohort: Year 3 (AD/ADRD)
Project Accomplishments: This pilot project focused on developing an equitable machine learning (ML)-based music intervention for older adults at risk for Alzheimer’s disease. The study progressed through two phases, beginning with IRB approvals and participant recruitment in mid-2024. Data collection included online surveys and music preference assessments from rural-residing participants, with 76% reporting at least one dementia-related risk factor. Analysis revealed that participants preferred songs with positive sentiment, which were also perceived as more beneficial for mood improvement. These findings informed the development of a hybrid music recommendation model that integrates user preferences and song characteristics to enhance emotional well-being.
The ML recommender system was built using knowledge graph embeddings (KGE), achieving a normalized discounted cumulative gain (NDCG) of 0.009—more than double the random baseline—demonstrating early promise despite limited user feedback. User satisfaction data from Phase II is currently being analyzed, with a goal of reaching 85% satisfaction. The project has been presented at multiple national conferences and is under review for publication. These efforts lay the groundwork for future refinement, user testing, and potential commercialization of a personalized music intervention aimed at supporting cognitive and emotional health in older adults.
Initial Proposal Abstract: Older adults in rural communities face unique health challenges that place them at a greater risk for Alzheimer’s disease (AD). Social determinants of health such as healthcare access and social engagement coupled with psychosocial barriers (e.g., stigma, medical mistrust) often exacerbate their risk. Although the adoption of digital health interventions has shown promise in reducing cognitive-related health disparities in rural communities, there remains a lack of consideration regarding rural culture in the design and implementation of such interventions. By creating a more comprehensive, individualized platform through collaborative filtering models, SingFit can serve as a more inclusive and culturally relevant tool for at-risk older adults in rural communities who have been traditionally underrepresented in the development of music-based digital health interventions.
The specific aims are to 1) Develop culturally inclusive user personas: The unique needs, preferences, and backgrounds of at-risk rural-residing older adults will be identified to learn more about their goals, challenges, and perspectives regarding music-based digital health interventions. 2) Design an Intelligent Recommendation System: ML models will be developed using collaborative-based filtering that utilizes user data to identify optimal therapeutic music components for enhancing engagement and emotional resonance among rural-residing older adults at-risk for AD. Finally 3) Assess acceptability of personalized therapeutic music sessions: An online survey will be conducted to evaluate user feedback of the song/guided prompts generated by the recommendation system among a diverse group of rural-residing older adults at-risk for AD.
Surveys will collect perceived benefit, sentiment, relevance, and preference for SingFit algorithms and subsequent collaborative filtering algorithms. Additional data regarding demographic characteristics and musical preferences will also be collected. The data will be used to develop culturally inclusive user personas (e.g., segment analysis) and preprocessed in order to design and evaluate an Intelligent Recommendation System.