Premium
Accuracy of Activity Monitors in Assessing Sleep
Author(s) -
Mitrzyk John,
Swalve Natashia,
Harfmann Brianna,
Montoye Alexander HK.
Publication year - 2018
Publication title -
the faseb journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.709
H-Index - 277
eISSN - 1530-6860
pISSN - 0892-6638
DOI - 10.1096/fasebj.2018.32.1_supplement.588.5
Subject(s) - sleep (system call) , electroencephalography , audiology , activity monitor , medicine , sleep stages , wrist , sleep quality , physical medicine and rehabilitation , analysis of variance , physical therapy , physical activity , psychology , polysomnography , surgery , insomnia , computer science , psychiatry , operating system
In addition to measuring steps and other physical activity metrics, body‐worn activity monitors are designed to measure sleep quantity and quality. However, little research has assessed activity monitors' sleep tracking accuracy or compared accuracy among different brands and models of monitors. PURPOSE To validate and compare several activity monitors for assessment of total sleep time and stages of sleep throughout a night. METHODS Eleven participants, aged 18–31 years old, wore four activity monitors on their wrists. The Misfit Flash (MF) and Fitbit Charge (FC) were worn on the left wrist, while the Fitbit Flex (FF) and Jawbone UP 4 (JU) were worn on the right wrist. Three electroencephalogram (EEG) electrodes were attached to the head (one behind left ear, one on each side of the upper parietal portion of the skull), and the electrodes were connected via wires to a BIOPAC system and computer. The study took place overnight in a sleep laboratory, with participants going to bed and waking up as they would on a typical day. Total sleep, light sleep, deep sleep, REM sleep, and number of awakenings were recorded from the four activity monitors and compared to criterion‐measured values from the EEG. Repeated measures ANOVA tests were used to determine differences between predicted and measured values for each variable. RESULTS For total sleep time, participants averaged 369.3±46.8 minutes/night. Compared to the EEG, the FC (p=0.005) and MF (p=0.002) significantly overestimated total sleep by 32.4–38.1 minutes, while the FF and JU were not significantly different from the EEG (p>0.05). For minutes of light sleep, the FC (p<0.001) and FF (p<0.001) underestimated by 176.5–191.2 minutes when compared to the EEG's time (206.0±58.0 minutes), while the MF and JU were not significantly different from the EEG (p>0.05). For deep sleep, all the monitors except the JU significantly overestimated time when compared to the EEG (163.3±39.0 minutes), by 72.8–214.4 minutes (p<0.01 for all). For REM sleep, the JU was not significantly different from the EEG (88.5±32.2 minutes), while the other monitors did not assess REM sleep. The number of times the participants woke up throughout the night averaged 3.1±2.2 times. Compared to the EEG, all monitors except the FF significantly underestimated by 1.3–2.0 times (p=0.001–0.036). CONCLUSION Overall, the JU was the most accurate for all stages of sleep and total sleep time, but significantly underestimated the number of times awake. All other monitors had large differences from the EEG in at least one sleep variable. While reliability of these monitors still needs to be established, large inaccuracies seen by some of these activity monitors suggest that sleep data from the wearable activity monitors tested in this study should be interpreted cautiously. Support or Funding Information Alma College CORE Research Program This abstract is from the Experimental Biology 2018 Meeting. There is no full text article associated with this abstract published in The FASEB Journal .