0313 Validation of a Consumer Activity Tracker Against Polysomnography and Actigraphy in Insomnia.

Abstract

Introduction: Consumer activity trackers claiming to measure sleep/wake patterns are ubiquitous within clinical and consumer settings. However, validation of these devices in sleep disorder populations are lacking. Methods: We examined one night of sleep in 28 individuals with insomnia (19-82 years-old; mean=45.2±17.3) using polysomnography, a standard wrist actigraph (Actiwatch Spectrum Pro: ACT) and a consumer activity tracker (Fitbit Alta HR: FB). Epoch-byepoch analysis was used to determine agreement between each device and polysomnography for overall sleep/wake detection and sleep variables, including total sleep time (TST), sleep efficiency (SE), wake after sleep onset (WASO) and sleep latency (SL). Clinically meaningful limits of agreement were set a priori at ±30 minutes for TST, SL, and WASO, and ±5% for SE. Analyses compared detection of light sleep (N1+N2), deep sleep (N3), rapid eye movement (REM) and wake by FB, relative to polysomnography. Results: Compared to polysomnography, both activity trackers displayed high sensitivity (96.79% and 95.73%, respectively) but low specificity (38.21% and 43.86%, respectively). Both devices overestimated TST and SE and underestimated SL and WASO, and all variables except SL exceeded clinical cut-offs. FB demonstrated sensitivity and specificity rates of 78% and 59%, respectively, in light sleep, 49% and 95% in deep sleep, 65% and 90% in REM, and 41% and 96% in stage wake. Conclusion: Consistent with prior models of research-grade and consumer activity trackers, both devices in this study were more accurate in detecting sleep than wake, and sensitivity/specificity values were statistically equivalent. Thus, this model of FB could serve as a low-cost substitute for actigraphy in insomnia. Moreover, data would be as useful clinically as research-grade actigraphs (which is, admittedly, debatable). This FB model missed several occurrences of specific sleep stages, though when it did identify N3, REM or wake, it was generally accurate. Thus, it underestimates these sleep stages and overestimates light sleep, providing a picture of less quality sleep than actually obtained. Future research should examine night-to-night variability in binary sleep/wake outcomes as well as sleep variables over several days within this population.