JMIR Publications

ABSTRACT

Background:

Wearable sleep monitors are of high interest to consumers and researchers because of their ability to provide estimation of sleep patterns in free-living condition in a cost-efficient way.

Objective:

We conducted a systematic review of publications reporting performance of wristband Fitbit models in assessing sleep parameters and stages.

Methods:

In adherence with the Preferred Reporting Items for Systematic Reviews and Meta Analyses (PRISMA) statement, we comprehensively searched CINAHL, Cochran, Embase, Medline, PubMed, PsycINFO, and Web of Science databases using keyword ‘Fitbit’ to identify relevant publications meeting predefined inclusion/exclusion criteria.

Results:

The search yielded 3085 candidate articles. After eliminating duplicates and in compliance with inclusion/exclusion criteria, 22 qualified for systematic review, with 8 providing quantitative data for meta-analysis. In reference to polysomnography (PSG), non-sleep-staging Fitbit models tended to overestimate Total Sleep Time (TST: range ~7 to 67 min; Effect Size (ES)=-0.51, P<0.001; Heterogenicity: I2=8.8%, P=0.361) and Sleep Efficiency (SE: range ~2 to 15%; ES=-0.74, P<0.001; Heterogenicity: I2=24.0%, P=0.254) and tended to underestimate Wake After Sleep Onset (WASO: range ~6 to 44 min; ES=0.60, P<0.001; Heterogenicity: I2=0%, P=0.920), but without significant difference in Sleep Onset Latency (SOL: P=0.368; Heterogenicity: I2=0%, P=0.918). In reference to polysomnography (PSG), non-sleep-staging Fitbit models correctly identified sleep epochs with accuracy between 0.81 and 0.91, sensitivity between 0.87 and 0.99, and specificity between 0.10 and 0.52. Recent-generation Fitbit models that collectively utilize heart rate variability and body movement to assess sleep-stages performed better than early-generation non-sleep-staging ones that utilize only body movement. Sleep-staging Fitbit models in comparison to PSG showed no significant difference in measured values of WASO (P=0.251; Heterogenicity: I2=0%, P=0.920), TST (P=0.289; Heterogenicity: I2=0%, P=0.980), and SE (P=0.189), but they underestimated SOL (P=0.026; Heterogenicity: I2=0%, P=0.663). Sleep-staging Fitbit models show higher sensitivity (0.95-0.96) and specificity (0.58-0.69) in detecting sleep epochs in comparison to non-sleep-staging ones and also that reported for regular wrist actigraphy.

Conclusions:

Sleep-staging Fitbit models show promising performance, especially in differentiating wake from sleep. However, although a convenient and economical means for consumers to obtain gross estimates of sleep parameters and time spent in sleep stages, they are of limited specificity and not a substitute for PSG.