Awardee Organization(s): University of Pennsylvania
Principal Investigator(s): Nancy A. Hodgson, PhD, RN, FAAN
Official Project Title: Using AI to predict depression & burden AD/ADRD caregiving conversations
AITC Partner: PennAITech
Website(s): https://www.nursing.upenn.edu/live/profiles/7116-nancy-hodgson
Clear communication between clinicians and caregivers of people living with dementia (PLWD) is essential to delivering quality dementia care. Objective, empirical assessment of these clinical communications can support the timely evaluation and management of care needs for PLWD and their caregivers but is currently too time-consuming and prone to clinician bias.
This is a secondary analysis of data collected during an implementation study evaluating the translation of an evidence-based dementia program. Conversational speech data from 125 hour-long (on average) sessions between clinicians and dementia caregivers along with repeat assessments of caregiver depression and burden will be leveraged to predict clinically meaningful treatment outcomes, (caregiver burden, depression and PLWD healthcare utilization) via a machine learning (ML) model.
The study aims to: 1) use an ML model to identify patterns in clinical conversations linked to dementia caregiver depression and burden, 2) detect patterns predicting PLWD healthcare utilization (e.g., 911 calls, hospitalizations) and 3) analyze ML outputs to enable early, targeted interventions. This third aim will be guided by an advisory group of healthcare providers and dementia caregivers. The results will demonstrate the potential of ML and data science to improve health outcomes for over 11 million U.S. dementia caregivers. The long-term goal is to develop a scalable technology-based intervention to address caregiver depression and burden, reduce costly care, and enhance quality of life.