Consumer wrist sleep trackers explained only 2.5-16.2% of the variance in how well adults said they slept, agreed less in people with insomnia (39.4% vs 82.4% concordance in good sleepers), and against polysomnography over-read sleep efficiency by 1.75-7.9% and under-read night waking by 7-30 min: a systematic review of 5 observational studies (2006 adults, athletes not reported; GRADE low to moderate) with no pooled estimate

Srivali and Cheungpasitporn asked how well consumer wrist trackers agree with what adults report about their own sleep, measured with validated questionnaires and diaries. They searched three databases to October 2025 and found five observational studies with 2006 participants. Device metrics explained only 2.5-16.2% of the variance in subjective sleep quality.

Total sleep time correlated moderately with same-day diaries (r = 0.367), but devices did not capture Pittsburgh Sleep Quality Index scores. Agreement depended on who wore the device: 82.4% concordance in good sleepers against 39.4% in people with insomnia (p = 0.006), and poor in clinical and older groups.

Against polysomnography, sleep efficiency agreed only fairly (ICC 0.478-0.570) and was over-read by 1.75-7.9%, sleep-onset latency barely correlated (r = 0.033) and wake after sleep onset was under-read by 7-30 min. GRADE certainty ranged from low to moderate. The authors conclude that device data should complement, not replace, validated subjective sleep assessment.