Using Wearable Sensors to Measure and Predict Personal Circadian Lighting Exposure in Nursing Home Residents: Model...

Area of Science:

• Biomedical Engineering
• Chronobiology
• Health Informatics

Background:

• Circadian lighting significantly impacts individuals with dementia, affecting sleep, alertness, and behavior.
• Personalized light exposure measurement is vital for understanding health outcomes.
• Wearable sensors offer a practical solution for tracking individual light exposure.

Purpose of the Study:

• Develop and validate calibration and predictive models for wearable lighting sensors.
• Accurately assess individual circadian light exposure.
• Provide a reliable method for healthcare researchers to optimize lighting for dementia care.

Main Methods:

• Combined laboratory experiments and on-site data collection with spectrophotometer ground truth.
• Developed calibration models for photopic lux and correlated color temperature.
• Created predictive models for circadian stimulus using machine learning (random forest) and regression.

Main Results:

• Calibration models showed high accuracy (Adj. R² 0.858 for lux, 0.982 for CCT).
• Random forest model accurately predicted circadian stimulus (Adj. R² 0.915, CV R² 0.857).
• Significant individual variations in light exposure were observed, confirming the need for personalized assessment.

Conclusions:

• Wearable sensors with predictive modeling offer a scalable, cost-effective approach to circadian light assessment.
• This method replaces labor-intensive spectrometer measurements for continuous monitoring.
• Future work should focus on sensor refinement and expanding applications for vulnerable populations.