How accurately does wrist actigraphy identify the states of sleep and wakefulness?

Because sleep and wakefulness differ from each other by the amount of body movement, it has been claimed that the two states can be accurately distinguished by wrist actigraphy. Our objective was to test this claim in lengthy polysomnographic (psg) and actigraphic (acf) samples that included night and day components. Fourteen healthy young (21-35 years) and old (70-72 years) men and women lived in a laboratory without temporal cues for 7 days. Each subject continuously wore sleep-recording electrodes as well as 2 wrist-movement recorders. Act measurements were converted to predictions of sleep and wakefulness by simple-threshold and multiple-regression methods. Psg served as the gold standard for calculation of predictive values (PV, the probability that an act prediction is correct by psg criteria). N/A. N/A. N/A. The 7-day act recordings showed clear circadian cycles of high and low activity that respectively corresponded to subjective days, when subjects were wakeful, and subjective nights when they slept. Lower act levels corresponded to deeper states of psg sleep. Logistic regression on a 20-minute moving average of act gave the highest overall PV's. Nevertheless, the mean PV for sleep (PVS) was only 62.2% in complete, day + night samples. PVS was 86.6% in night samples. Act successfully predicted wakefulness during subjective nights (PVW = 89.6) and accurately measured circadian period length and the extent of sleep-wake consolidation, but it overestimated sleep rate and sleep efficiency. Act systematically decreased before sleep onset and increased before awakening, but reliable transitions among joint psg/act states (the Markov-1 property) were not demonstrated. Low PV's and overestimation of sleep currently disqualify actigraphy as an accurate sleep-wake indicator. Actigraphy may, however, by useful for measuring circadian period and sleep-wake consolidation and has face validity as a measure of rest/activity.