Table 2

Comparison of automated wear-time estimation algorithms* in classifying time as sedentary or non-wear against activity log

Algorithm 1Algorithm 2Algorithm 3
Sensitivity (%)
 Overall (95% CI)95.61 (95.16 to 96.05)95.43 (94.96 to 95.89)95.10 (94.64 to 95.56)
 Worst day43.1600
 Best day100100100
Specificity (%)
 Overall (95% CI)95.61 (95.14 to 96.09)97.40 (96.99 to 97.81)97.67 (97.32 to 98.03)
 Worst day19.9320.1119.93
 Best day100100100
Sedentary time classed as non-wear
 Days with misclassification (%)31.919.418
 Minutes, median (min, max)78 (6, 620)75 (26, 600)72 (1, 600)
Non-wear time classed as sedentary
 Days with misclassification (%)42.843.751.3
 Minutes, median (min, max)49 (1, 329)50 (1, 338)44 (1, 329)
 Mean difference (95% limits of agreement) estimated – criterion wear time−11 (−172, 149)5 (−134, 143)9 (−124, 142)
  • * The algorithms define non-wear periods as non-movement (ie, 0 cpm and any allowed interruptions) of 60-min duration: Algorithm 1 permits ≤2 consecutive active counts of maximum intensity 99 cpm; Algorithm 2 permits ≤2 active counts of maximum intensity 49 cpm and Algorithm 3 allows no interruptions (requires all counts to be 0 cpm).

  • Median (minimum, maximum) minutes misclassified (against criterion) on days where misclassification occurred.