Measures of sleep and cardiac functioning during sleep using a multi-sensory commercially-available wristband in adolescents

Abstract

To validate measures of sleep and heart rate (HR) during sleep generated by a commercially-available activity tracker against those derived from polysomnography (PSG) in healthy adolescents. Sleep data were concurrently recorded using FitbitChargeHR™ and PSG, including electrocardiography (ECG), during an overnight laboratory sleep recording in 32 healthy adolescents (15 females; age, mean±SD: 17.3±2.5years). Sleep and HR measures were compared between FitbitChargeHR™ and PSG using paired t-tests and Bland-Altman plots. Epoch-by-epoch analysis showed that FitbitChargeHR™ had high overall accuracy (91%), high sensitivity (97%) in detecting sleep, and poor specificity (42%) in detecting wake on a min-to-min basis. On average, FitbitChargeHR™ significantly but negligibly overestimated total sleep time by 8min and sleep efficiency by 1.8%, and underestimated wake after sleep onset by 5.6min (p<0.05). Within FitbitChargeHR™ epochs of sleep, the average HR was 59.3±7.5bpm, which was significantly but negligibly lower than that calculated from ECG (60.2±7.6bpm, p<0.001), with no change in mean discrepancies throughout the night. FitbitChargeHR™ showed good agreement with PSG and ECG in measuring sleep and HR during sleep, supporting its use in assessing sleep and cardiac function in healthy adolescents. Further validation is needed to assess its reliability over prolonged periods of time in ecological settings and in clinical populations.

Keywords: Actigraphy; Adolescence; Fitbit; Heart rate; Polysomnography; Wearables.

Fig 1

Minute-by-minute heart rate (HR) across the night for a representative participant. Upper panel: HR (bpm) derived from the electrocardiogram (ECG, black) is superimposed on FitbitChargeHR™ HR (grey). Vertical bars show periods of wakefulness detected by FitbitChargeHR™ and are excluded from the analyses. Lower panel: Differences in HR (ECG minus FitbitChargeHR™) across the night.

Fig 2

Bland-Altman plots for polysomnographic (PSG) and FitbitChargeHR™ total sleep time (TST), sleep onset latency (SOL), wake after sleep onset (WASO) and sleep efficiency (SE) in a sample of 30 healthy adolescents. Bias (thick solid line), upper and lower agreement limits (mean difference ±1.96SD, thin dotted lines) are shown. The PSG measure is shown on the x-axis and the difference (PSG minus FitbitChargeHR™) is shown on the y-axis.

Fig 3

Bland-Altman plot for electrocardiographic (ECG) and FitbitChargeHR™ heart rate (HR) in 29 adolescents. Bias (ECG minus Fitbit HR, thick solid line) and upper and lower agreement limits (mean difference ±1.96SD, thin dotted lines) are shown.