Study objectives: Actigraphy is generally compared to polysomnography (PSG), which has been considered the gold standard for sleep studies. The objective of the present study was to evaluate the concordance between PSG and two previously proposed algorithms (Cole et al, 1992; Sadeh et al, 1994) to analyze actigraphic recordings. The minute-by-minute agreement rate was evaluated through calculation of sensitivity, specificity, and accuracy. Regarding the sleep parameters, the concordance was performed through the Bland and Altman technique.
Design: A night of adaptation to the sleep laboratory followed by simultaneous polysomnographic and actigraphic recordings throughout the night.
Participants: 21 healthy volunteers.
Setting: A sleep laboratory
Interventions: None.
Results: Ninety-one percent of all PSG epochs were correctly identified by both algorithms, and this accuracy is reasonably satisfactory. The actigraphy was a sensitive method, with values of 99% and 97% for Cole's and Sadeh's algorithms, respectively. However, actigraphy had a low specificity: 34% and 44% for Cole's and Sadeh's algorithms, respectively. The Bland and Altman technique showed that actigraphy systematically overestimated Sleep Latency, Total Sleep Time and Sleep Efficiency while it underestimated Intermittent Awakenings.
Conclusions: The results of this study show the utility of actigraphy as a useful method for assessment of sleep, despite its limitations regarding identification of waking epochs during sleep. The Bland and Altman concordance technique was revealed to be a powerful tool to evaluate how well actigraphy agreed with polysomnography. This technique, combined with calculations of sensitivity and specificity, appears to be the most adequate procedure for the assessment of concordance.