Hebrew SeniorLife's Hinda and Arthur Marcus Institute for Aging
Research, Harvard Medical School
Developing an AI-based Measure for Behavioral and Psychological Symptoms of Dementia Using Nursing Home Electronic Health Records Data
Dr. Park is a geriatrician and an assistant scientist at the Hinda and Arthur Marcus Institute for Aging Research
and an instructor in medicine at Harvard Medical School. Her research focuses on improving medication safety
and care quality for people living with dementia using large-scale real-world data, including Medicare claims
and nursing home electronic health records. She has developed and validated novel dementia severity
measures and leads pharmacoepidemiologic studies evaluating medication use and outcomes in frail older
adults. Through this Career Development Award, she seeks to advance expertise in artificial intelligence (AI)-
enabled outcome measurement, implementation science, and embedded pragmatic clinical trials to design
scalable interventions that improve care in long-term care settings.
People living with dementia (PLWD) in nursing homes frequently experience behavioral and psychological
symptoms (BPSD), which drive antipsychotic use, care burden, and adverse outcomes, yet these symptoms
are poorly captured in structured data. Current measurement approaches lack scalability and fail to leverage
rich information embedded in unstructured clinical notes in electronic health records (EHR). This project
addresses this gap by developing and validating an artificial intelligence (AI)-enabled, natural language
processing (NLP)-based measure of BPSD using nursing home EHR data. This Career Development Award will
provide training in clinical NLP, AI-based outcome measure development, implementation science, and
embedded pragmatic clinical trial (ePCT) design to support the development and real-world integration of
scalable dementia outcome measures. This training will support the following Specific Aims: (1) To develop an
AI-enabled, NLP-based measure of BPSD using unstructured nursing home EHR data, and (2) To externally
validate this measure across diverse long-term care settings for use in pragmatic clinical trials. This work will
establish a scalable, EHR-based outcome measure that enables efficient evaluation of interventions targeting
BPSD in real-world settings. It will lay the foundation for future ePCTs to improve medication safety and care
quality for PLWD in nursing homes.