Dementia Care Research and Psychosocial Factors

Area of Science:

• Gerontology
• Neuroscience
• Computer Science

Background:

• Driving is crucial for independence in older adults.
• Cognitive changes with aging increase driving risks.
• Driving behavior may signal early cognitive decline.

Purpose of the Study:

• To determine if machine learning can detect cognitive status from driving performance.
• To identify driving metrics indicative of cognitive impairment.

Main Methods:

• Older adults (cognitively impaired and unimpaired) drove a simulator.
• Driving metrics (steering, lane deviation, acceleration, braking) were analyzed.
• A Random Forest model was used to classify cognitive status.

Main Results:

• The Random Forest model achieved 66.67% accuracy using all features.
• Key predictors included acceleration, steering angle, and lane gap.
• A refined model with top features improved accuracy to 70.83%.

Conclusions:

• Machine learning can classify cognitive status using driving data.
• Driving metrics show potential for early cognitive decline detection.
• Future work includes detecting pre-clinical decline and assessing AI reliability.