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 conditions in a cost-efficient way. Objective: We conducted a systematic review of publications reporting on the 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 the Cumulative Index to Nursing and Allied Health Literature (CINAHL), Cochrane, Embase, MEDLINE, PubMed, PsycINFO, and Web of Science databases using the keyword Fitbit to identify relevant publications meeting predefined inclusion and exclusion criteria. Results: The search yielded 3085 candidate articles. After eliminating duplicates and in compliance with inclusion and exclusion criteria, 22 articles qualified for systematic review, with 8 providing quantitative data for meta-analysis. In reference to polysomnography (PSG), nonsleep-staging Fitbit models tended to overestimate total sleep time (TST; range from approximately 7 to 67 mins; effect size=-0.51, P<.001; heterogenicity: I2=8.8%, P=.36) and sleep efficiency (SE; range from approximately 2% to 15%; effect size=-0.74, P<.001; heterogenicity: I2=24.0%, P=.25), and underestimate wake after sleep onset (WASO; range from approximately 6 to 44 mins; effect size=0.60, P<.001; heterogenicity: I2=0%, P=.92) and there was no significant difference in sleep onset latency (SOL; P=.37; heterogenicity: I2=0%, P=.92). In reference to PSG, nonsleep-staging Fitbit models correctly identified sleep epochs with accuracy values between 0.81 and 0.91, sensitivity values between 0.87 and 0.99, and specificity values 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 nonsleep-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=.25; heterogenicity: I2=0%, P=.92), TST (P=.29; heterogenicity: I2=0%, P=.98), and SE (P=.19) but they underestimated SOL (P=.03; heterogenicity: I2=0%, P=.66). Sleep-staging Fitbit models showed higher sensitivity (0.95-0.96) and specificity (0.58-0.69) values in detecting sleep epochs than nonsleep-staging models and those reported in the literature for regular wrist actigraphy. Conclusions: Sleep-staging Fitbit models showed promising performance, especially in differentiating wake from sleep. However, although these models are 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 are not a substitute for PSG.

The rapid emergence of new measurement instruments and methods requires personnel and researchers of different disciplines to know the correct statistical methods to utilize to compare their performance with reference ones and properly interpret findings. We discuss the often-made mistake of applying the inappropriate correlation and regression statistical approaches to compare methods and then explain the concepts of agreement and reliability. Then, we introduce the intraclass correlation as a measure of inter-rater reliability, and the Bland–Altman plot as a measure of agreement, and we provide formulae to calculate them along with illustrative examples for different types of study designs, specifically single measurement per subject, repeated measurement while the true value is constant, and repeated measurement when the true value is not constant. We emphasize the requirement to validate the assumptions of these statistical approaches, and also how to deal with violations and provide formulae on how to calculate the confidence interval for estimated values of agreement and intraclass correlation. Finally, we explain how to interpret and report the findings of these statistical analyses.

We compared performance in deriving sleep variables by both Fitbit Charge 2™, which couples body movement (accelerometry) and heart rate variability (HRV) in combination with its proprietary interpretative algorithm (IA), and standard actigraphy (Motionlogger® Micro Watch Actigraph: MMWA), which relies solely on accelerometry in combination with its best performing ‘Sadeh’ IA, to electroencephalography (EEG: Zmachine® Insight+ and its proprietary IA) used as reference. We conducted home sleep studies on 35 healthy adults, 33 of whom provided complete datasets of the three simultaneously assessed technologies. Relative to the Zmachine EEG method, Fitbit showed an overall Kappa agreement of 54% in distinguishing wake/sleep epochs and sensitivity of 95% and specificity of 57% in detecting sleep epochs. Fitbit, relative to EEG, underestimated sleep onset latency (SOL) by ~11 min and overestimated sleep efficiency (SE) by ~4%. There was no statistically significant difference between Fitbit and EEG methods in measuring wake after sleep onset (WASO) and total sleep time (TST). Fitbit showed substantial agreement with EEG in detecting rapid eye movement and deep sleep, but only moderate agreement in detecting light sleep. The MMWA method showed 51% overall Kappa agreement with the EEG one in detecting wake/sleep epochs, with sensitivity of 94% and specificity of 53% in detecting sleep epochs. MMWA, relative to EEG, underestimated SOL by ~10 min. There was no significant difference between Fitbit and MMWA methods in amount of bias in estimating SOL, WASO, TST, and SE; however, the minimum detectable change (MDC) per sleep variable with Fitbit was better (smaller) than with MMWA, respectively, by ~10 min, ~16 min, ~22 min, and ~8%. Overall, performance of Fitbit accelerometry and HRV technology in conjunction with its proprietary IA to detect sleep vs. wake episodes is slightly better than wrist actigraphy that relies solely on accelerometry and best performing Sadeh IA. Moreover, the smaller MDC of Fitbit technology in deriving sleep parameters in comparison to wrist actigraphy makes it a suitable option for assessing changes in sleep quality over time, longitudinally, and/or in response to interventions.

Elevated asleep heart rate (HR) is a risk factor for cardiovascular disease and other-cause morbidity and mortality. We assessed the accuracy of Fitbit Inc. PurePulse® photoplethysmography with reference to three-lead electrocardiography (ECG) in determining HR during sleep. HR of 35 (17 female) healthy adults 25.1 ± 10.6 years of age (mean ± SD) was continuously recorded throughout a single night of sleep. There was no significant difference in asleep HR mean (0.09 beats per minute [bpm], P = 0.426) between Fitbit photoplethysmography and ECG; plus, there was excellent intraclass correlation (0.998) and narrow Bland–Altman agreement range (2.67 bpm). The regression analysis of Bland–Altman plot of mean asleep HR indicates Fitbit tends to slightly overestimate reference values in the lower range of HR (HR < 50 bpm) by 0.51 bpm and slightly underestimate reference values in the higher range of HR (HR > 80 bpm) by 0.63 bpm. Mixed model analysis of epoch-by-epoch (5-min epochs) asleep HR showed significant “U” shape trend (P < 0.001) in amount of Fitbit error (absolute amount of difference between ECG and Fitbit values regardless of overestimation or underestimation) in regard to HR, i.e. smaller error in the medium range of HR (60–80 bpm) and slightly larger error for lower (80 bpm) ranges of HR. However, effect of age, body mass index, gender, and subjective sleep quality measured by Pittsburgh sleep quality index (good/poor sleepers) on error in estimating HR by the Fitbit method was not significant. It is concluded that Fitbit photoplethysmography suitably tracks HR during sleep in healthy young adults.

Background: The relationship between 24-hour rest-activity rhythms (RARs) and risk for dementia or mild cognitive impairment (MCI) remains an area of growing interest. Previous studies were often limited by small sample sizes, short follow-ups, and older participants. More studies are required to fully explore the link between disrupted RARs and dementia or MCI in middle-aged and older adults. Objective: We leveraged the UK Biobank data to examine how RAR disturbances correlate with the risk of developing dementia and MCI in middle-aged and older adults. Methods: We analyzed the data of 91,517 UK Biobank participants aged between 43 and 79 years. Wrist actigraphy recordings were used to derive nonparametric RAR metrics, including the activity level of the most active 10-hour period (M10) and its midpoint, the activity level of the least active 5-hour period (L5) and its midpoint, relative amplitude (RA) of the 24-hour cycle [RA=(M10-L5)/(M10+L5)], interdaily stability, and intradaily variability, as well as the amplitude and acrophase of 24-hour rhythms (cosinor analysis). We used Cox proportional hazards models to examine the associations between baseline RAR and subsequent incidence of dementia or MCI, adjusting for demographic characteristics, comorbidities, lifestyle factors, shiftwork status, and genetic risk for Alzheimer's disease. Results: During the follow-up of up to 7.5 years, 555 participants developed MCI or dementia. The dementia or MCI risk increased for those with lower M10 activity (hazard ratio [HR] 1.28, 95% CI 1.14-1.44, per 1-SD decrease), higher L5 activity (HR 1.15, 95% CI 1.10-1.21, per 1-SD increase), lower RA (HR 1.23, 95% CI 1.16-1.29, per 1-SD decrease), lower amplitude (HR 1.32, 95% CI 1.17-1.49, per 1-SD decrease), and higher intradaily variability (HR 1.14, 95% CI 1.05-1.24, per 1-SD increase) as well as advanced L5 midpoint (HR 0.92, 95% CI 0.85-0.99, per 1-SD advance). These associations were similar in people aged 70 years, and in non–shift workers, and they were independent of genetic and cardiovascular risk factors. No significant associations were observed for M10 midpoint, interdaily stability, or acrophase. Conclusions: Based on findings from a large sample of middle-to-older adults with objective RAR assessment and almost 8-years of follow-up, we suggest that suppressed and fragmented daily activity rhythms precede the onset of dementia or MCI and may serve as risk biomarkers for preclinical dementia in middle-aged and older adults.