Background: Actigraphy has become an important tool in sleep research, however, most actigraphic models have had little evaluation of their sleep scoring software. Some reports have described poor agreement of actigraph and polysomnogram (PSG) results. In this study, we examined the accuracy of the Actiwatch-L instrument with Actiware((R)) software version 5.0 (Minimitter-Respironics, Bend, Oregon).
Methods: We analyzed the sleep latencies and total sleep times in 33 actigraph recordings and compared performance to simultaneous polysomnography. For scoring sleep latency, the default criterion for scoring sleep onset (10min of immobility) was compared with criteria of 4, 5, 6, and 15min of immobility, and low, middle, and high activity thresholds were compared.
Results: Of 4, 5, 6, 10 and 15min of actigraphic immobility, the 5min criterion yielded the most accurate sleep latencies. The 5min criterion also yielded near-optimal estimates for total sleep time and wake after sleep onset.
Conclusions: We found that a default 10-min immobility parameter for sleep onset was not as accurate as 5min for scoring sleep latency and total sleep time in this clinical sample. There is a need for expanded samples to further evaluate algorithm scoring parameters and the search for superior alternatives.