Standardizing and Harmonizing Behavioral and Social Science Research Factors in Alzheimer's Disease through Ontology-Based Approaches

ABSTRACT Behavioral and social science research (BSSR) is instrumental in comprehending Alzheimer's Disease and related dementia (ADRD) and its far-reaching implications for individuals, families, and communities. BSSR investigates how behavioral and social determinants, encompassing factors like physical activity, cognitive

engagement, diet, and social interactions, influence the risk of developing AD. This exploration aims to uncover non-pharmacological interventions that can manage symptoms and enhance the well-being of individuals with AD/ADRD, including cognitive stimulation programs, behavioral therapies, social engagement initiatives, and

caregiver training. Moreover, BSSR delves into the social, cultural, and environmental elements contributing to health disparities, informing tailored interventions and policies for various populations, especially underserved and minority communities. Collectively, this research enriches our understanding of ADRD and guides the

development of interventions, support systems, and policies to enhance the lives of those affected by the disease. Yet, there are challenges impeding the integration of ADRD-related BSSR data. A critical issue is the absence of formal representations for BSSR data and limited tools to link comprehensive BSSR information from

diverse sources. This hampers the holistic consideration of BSSR factors in AD-related research, undermining evidence-based care and support. In response to PAR-23-182, we propose pioneering ontology-based approaches to formally represent ADRD-related BSSR factors in a standardized manner. We will develop natural

language processing (NLP) methods to extract and normalize BSSR data from Electronic Health Records (EHRs) and literature. Our project aims to integrate structured and unstructured data across various research silos, culminating in a comprehensive and normalized knowledge graph incorporating BSSR factors for ADRD

cohorts. More specifically, in Aim 1, we will develop the Behavioral Social Data and Knowledge Ontology for ADRD (BSO-AD) to standardize BSSR factors. We will also assess the BSO-AD for correctness and suitability, refining it based on evaluation scores. Aim 2 employs NLP technologies, including state-of-the-art large language

models, to extract and normalize BSSR-related information from clinical notes and literature. This NLP system will ensure semantic interoperability and consistency in entity recognition and normalization. In Aim 3, we will create a knowledge graph (KG) to integrate annotated BSSR factors from structured and unstructured sources,

supporting ADRD-related research and applications. We will evaluate the ontology and KG through demonstration studies and disseminate these resources to the research community, promoting collaborative research efforts. In summary, our project aims to bridge the gap in ADRD-related BSSR data integration by

standardizing representation, enabling efficient extraction, and fostering collaboration within the research community. This endeavor will advance our understanding of ADRD and contribute to evidence-based care and support for affected individuals and their communities.