Table 4 Performance of non-wear detection methods in visual quality control.

From: A standardized workflow for long-term longitudinal actigraphy data processing using one year of continuous actigraphy from the CAN-BIND Wellness Monitoring Study

 

Wear Sensor Mean (SD)

Choi Algorithm Mean (SD)

Troiano Algorithm Mean (SD)

Van Hees Algorithm Mean (SD)

4-method Majority Algorithm: Wear Sensor, Choi, Troiano, Van Hees Mean (SD)

3-method Majority Algorithm: Wear Sensor, Choi, van Hees Mean (SD)

Total Rows

6,702,388

6,702,388

6,702,388

6,624,026

6,624,026

6,624,026

Total Days

4,654.44

4,654.44

4,654.44

4,600.02

4,600.02

4,600.02

Missing

1,823,252

1,823,252

1,823,252

1,901,614

1,901,614

1,901,614

Accuracy

0.8839 (0.2722)

0.9816 (0.0564)

0.9609 (0.0683)

0.9866 (0.0474)

0.9884 (0.0526)

0.9887 (0.0517)

Positive Predictive Value

0.6197 (0.4703)

0.9101 (0.2762)

0.6723 (0.4528)

0.9515 (0.2063)

0.9665 (0.1711)

0.9641 (0.1767)

Sensitivity

0.9444 (0.2225)

0.9617 (0.1707)

0.9823 (0.1067)

0.9289 (0.2013)

0.9608 (0.1609)

0.9592 (0.1620)

Specificity

0.9154 (0.2326)

0.9885 (0.0491)

0.9632 (0.0717)

0.9967 (0.0370)

0.9982 (0.0125)

0.9972 (0.0280)

  1. Mean and SD of the algorithm performance statistics were calculated at the day level.
  2. SD = standard deviation.