Accelerometer Cut-Points for Physical Activity in Older Adults
Author Information
Author(s): Skjødt Mathias, Brønd Jan Christian, Tully Mark A., Tsai Li-Tang, Koster Annemarie, Visser Marjolein, Caserotti Paolo
Primary Institution: University of Southern Denmark
Hypothesis
Can machine learning models improve the classification of physical activity intensity in very old adults compared to traditional methods?
Conclusion
The study developed machine learning models that provide better classification of physical activity intensity in very old adults than traditional ROC cut-points.
Supporting Evidence
- The study established cut-points for physical activity intensity using both ROC analysis and machine learning.
- Machine learning models showed higher accuracy in classifying physical activity intensity compared to traditional methods.
- The study included a diverse range of activities to better reflect real-life scenarios for older adults.
Takeaway
This study helps us understand how active older people are by using special devices to measure their movements, which can help keep them healthy.
Methodology
The study used accelerometers on various body parts to measure physical activity while participants performed daily activities and walking tests, comparing results from ROC analysis and machine learning models.
Limitations
The study's findings may not generalize to frailer older adults, and different accelerometer brands were used across wear sites.
Participant Demographics
Participants were community-dwelling Danish older adults aged 75 years and older, with a mean age of 80.2 years and 43% female.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website