Accelerating Platforms for Evaluating and eXecuting AI Lifecycles in ADRD (APEX-ADRD) - PROJECT SUMMARY/ABSTRACT | OVERALL STRATEGY Despite the 7M Americans living with Alzheimer’s disease and related dementias (ADRD) at an annual cost exceeding $781B, and nearly half of dementia cases being potentially preventable, artificial intelligence (AI) has yet to deliver its promised transformation in aging and ADRD care. Few AI-based clinical decision support tools for ADRD exist and were typically cleared through low-evidence pathways. Markedly, current validation pro- cesses often require years, which is too slow for AI, a technology that is quickly evolving. As a result, clinicians and patients are without trustworthy guidance on which tools help or harm. The gap is growing between innova- tion and evidence, particularly for vulnerable older adults with ADRD. Objectives and impact. The UCLA Accelerating Platforms for Evaluating and Executing AI Lifecycles in ADRD (APEX-ADRD) Collaboratory will transform how AI is developed, tested and deployed for aging and ADRD pop- ulations. Our goal is to stimulate transformative research across the NIA Stage Model (Stages 0-V) and compress timelines from years to months while preserving rigor, safety, and fairness. APEX-ADRD will serve as a national resource to accelerate responsible AI by: 1) building governance and infrastructure for rapid, reproducible eval- uation; 2) catalyzing innovation via design studios that embed co-design and ethics early in the process, along with access to unique data resources; and 3) establishing best practices for ethical and fair AI including contin- uous post-deployment fairness monitoring. UCLA brings unmatched multidisciplinary capacity to lead this effort with demonstrated track records in both the development of novel AI methods deployed into real- world healthcare settings as well as dynamic and comprehensive evaluation strategies. Notably, through our Collaboratory, pilot projects will have access to the University of California (UC) Health Data Warehouse (8.7M patients; 152K ADRD) for in silico validation with electronic health record (EHR) data and imaging; the ATLAS biobank (150K UCLA genotyped patients with EHR data); and other national datasets. This infrastructure is coupled with active guidance from leaders in geriatrics, neurology, biomedical informatics, computer sci- ence/AI, biostatistics, bioethics, health economics, implementation science, and the human-AI interface. Im- portantly, APEX-ADRD brings a strong industry and stakeholder team and will work closely with the AITC Coor- dinating Center and the trans-AITC Stakeholder Engagement Team to ensure all pilots integrate end-user feed- back, algorithmic fairness metrics, and common evaluation tools. Collectively, these facets are integrated throughout the Collaboratory’s services to ensure the success of AI pilot projects in aging and ADRD. By creating scalable models for safe and fair AI evaluation, UCLA’s APEX-ADRD will accelerate national progress toward improving health, independence, and quality of life for older adults and persons living with ADRD.
