Life Adapt Inc. (formerly Adaptelligence LLC) was recently awarded a $2.3M NIH SBIR Phase II grant entitled “Automated Health Assessment through Mobile Sensing & Machine Learning of Daily Activities” to conduct research aimed at refining and validating a novel technology for assessing and supporting functional independence in older adults, leveraging wearable sensor data and machine learning. The project is in its Phase II stage and is predicated on the growing need to address challenges posed by an aging population and age-related chronic health conditions.
Specific Aims of the Project:
Extract Comprehensive Activity-Based Digital Behavior Markers
This involves modeling the complete routine of an individual to accurately assess their functional health. The approach aims to extract markers that reflect a person’s activity patterns, encompassing routine behaviors, trends, and abrupt changes. To validate this technology, 100 adult participants with subjective cognitive complaints or mild cognitive impairment will be involved, ensuring diversity in the sample to minimize bias. The outcome is a web-based dashboard that graphically summarizes the behavior markers, trends, and disruptions in behavior patterns.
Automate Health Assessment from Daily Behavior Markers
The automated health assessment includes the introduction of machine learning methods to predict clinical scores for cognition, memory, mood, mobility, and functional health. These predictions are based on data collected from clinicians, ecological momentary assessments (EMA) self-reports, and smartwatch-based tests, further refined by GAN-based data imputation and domain adaptation techniques. The validation of these health assessment technologies will use data from the same sample of 100 participants. The anticipated outcome is a dashboard that not only summarizes predicted clinical scores and trends but also generates a comprehensive “wellness report” to assist family members and health providers in creating more informed and personalized treatment plans.
Generate Activity-Aware Alerts and Suggestions for Timely Intervention
This feature will alert both individuals and care providers about detected behavior anomalies, trends, and clinical score predictions that might necessitate formal health assessments. Additionally, the technology will provide personalized health behavior suggestions, offering educational materials and advice for sustainable new behaviors that correlate with improved health scores. To ensure user-friendliness and effectiveness, 18 participants from key stakeholder groups, including older adults, care providers, and clinicians, will be involved in two rounds of an iterative participatory design process.
This project stands out for its innovative approach that combines activity learning, mobile sensing, and clinical evaluation to provide real-time, activity-aware functional assessment and intervention strategies. The expected outcomes are novel methods for comprehensively modeling daily behavior, technology for predicting health scores from behavior markers, and activity-aware intervention strategies to promote functional health. The commercialization plan is robust, aiming to bring the technology to the market and seeking insurance reimbursement for the services. The technology promises to offer better evaluation and individualized care for older adults, particularly those living independently or in remote and underserved areas. By facilitating early detection and more effective treatment, it aims to slow cognitive decline and improve the quality of life for individuals and their caregivers. Given the costs associated with nursing home care, the impact of family-based care, and the general preference for staying at home, the technology holds significant commercial value.