Software thresholds alter the bias of actigraphy for monitoring sleep in team-sport athletes.

OBJECTIVES

Actical actigraphy is commonly used to monitor athlete sleep. The proprietary software, called Actiware, processes data with three different sleep-wake thresholds (Low, Medium or High), but there is no standardisation regarding their use. The purpose of this study was to examine validity and bias of the sleep-wake thresholds for processing Actical sleep data in team sport athletes.

DESIGN

Validation study comparing actigraph against accepted gold standard polysomnography (PSG).

METHODS

Sixty seven nights of sleep were recorded simultaneously with polysomnography and Actical devices. Individual night data was compared across five sleep measures for each sleep-wake threshold using Actiware software. Accuracy of each sleep-wake threshold compared with PSG was evaluated from mean bias with 95% confidence limits, Pearson moment-product correlation and associated standard error of estimate.

RESULTS

The Medium threshold generated the smallest mean bias compared with polysomnography for total sleep time (8.5min), sleep efficiency (1.8%) and wake after sleep onset (-4.1min); whereas the Low threshold had the smallest bias (7.5min) for wake bouts. Bias in sleep onset latency was the same across thresholds (-9.5min). The standard error of the estimate was similar across all thresholds; total sleep time ∼25min, sleep efficiency ∼4.5%, wake after sleep onset ∼21min, and wake bouts ∼8 counts.

CONCLUSIONS

Sleep parameters measured by the Actical device are greatly influenced by the sleep-wake threshold applied. In the present study the Medium threshold produced the smallest bias for most parameters compared with PSG. Given the magnitude of measurement variability, confidence limits should be employed when interpreting changes in sleep parameters.