Usefulness of a Tri-Axis Accelerometer for Sleep-Wake Scoring

Introduction A tri-axis accelerometer has been extensively used in recent years. We report our recent evaluation of the usefulness of a three-axis accelerometer for sleep-wake scoring of healthy individuals and patients with sleep-disordered breathing (SDB).

Methods

• Subjects
  • Seven healthy university students
    • 4 men, 3 women, mean age of 21.1±0.7
  • Seven SDB patients
    • 6 men, 1 women, mean age of 60.7±13.5
    • Apnea hypopnea index >20/h
  • Our hospital’s Ethics Committee approved the study, and informed consent was obtained from every subjects following full explanation of the study.
• Polysomnography (PSG)
  • Digital electroencephalograph
    • Neurofax EEG-1200 (Nihon Koden, Tokyo, Japan)
    • Polymate AP1532 (TEAC, Tokyo, Japan)
  • Analysis Software
    • NightOwl Professional (NoruPro Light Systems)
  • Analyses were based on the AASM Manual for the Scoring of Sleep and Associated Events, 2007.
• Tri-axis accelerometer
  • ActiSleep (ActiGraph, Florida, USA)
  • Analysis software
    • ActiLife version 5.6.1
  • Scoring of sleep/wake
    • Sadeh algorithm(SA)
    • Cole-Kripke algorithm(CKA)
• Uni-axis accelerometer
  • Actiwatch64(Mini Mitter)
  • Analysis software
    • Actiware5.0
  • Scoring of sleep/wake
    • The high, medium, and low thresholds are set at 80, 40, and 20 activity counts, respectively.
• In each subject, the accelerometer was attached to the wrist of the hand, opposite to the predominant hand, during PSG. The results obtained using the accelerometers were compared with the PSG-based visual scoring.

Results In healthy volunteers, 94.1 to 96.0% of all PSG epochs were correctly identified by all algorithms. All algorithms showed high sensitivity. However, they were not specific(Table a). In the SDB group, the accelerometer-based SA and CKA results exhibited greater variation as compared with PSG-based findings (agreement rate, 45.3 to 94.7%). In the SDB group, with each algorithm, 68.7 to 91.9% of the epochs determined as wakefulness by PSG were found to be sleep by accelerometer(Table b).

Bland-Altman pot showed the distribution of the differences against average of PSG and accelerometer sleep measurements. Differences are shows in the y-axis and averages in the x-axis. The mean bias of total sleep time was -57.2 (95% confidence interval: -106.9 – -7.5), and fixed bias was observed (Figure a). Proportional bias were observed in sleep efficiency, sleep latency and wake after sleep onset (Pearson’s correlation coefficient r=0.752, 0.839, 0.777).

Conclusions The use of accelerometer for sleep-wake scoring gave high agreement rates with PSG in healthy individuals. All algorithms showed high sensitivity. However, they were not specific, since they detected only a small proportion of waking epochs. The fixed and proportional biases of the accelerometer were reflected the low specificity.

Regarding the use of accelerometer in patients with SDB, the results of this study suggest that an accelerometer is a useful tool for scoring sleep-wake in cases where sleep efficiency has been improved, e.g., in patients after introduction of continuous positive airway pressure.

Conflicts of Interest This was not an industry supported study. The authors have indicated no financial conflicts of interest.