Measurement error in visually scored electrophysiological data: respiration during sleep

Area of Science:

• Sleep Medicine
• Neuroscience
• Biomedical Engineering

Background:

• Electrophysiological sleep studies involve simultaneous recording of multiple physiological parameters.
• These sleep data are typically scored and analyzed visually by trained human scorers.
• Assessing the accuracy of visual scoring is crucial for reliable sleep disorder diagnosis.

Purpose of the Study:

• To investigate the measurement error associated with the visual evaluation of respiratory disturbances during human sleep.
• To determine the reliability and agreement among trained scorers when analyzing sleep recordings.
• To identify factors influencing discrepancies in visual sleep data scoring.

Main Methods:

• Three experienced scorers independently evaluated 28 all-night polysomnography recordings.
• Scorers assessed various parameters, focusing on respiratory disturbances.
• Inter-rater reliability and agreement were statistically analyzed.

Main Results:

• Scorers demonstrated high agreement on only a subset of the evaluated variables.
• The experience level of the scorers did not significantly impact the reliability of their judgments.
• Independent assessments of recording quality did not explain the observed discrepancies in scoring.

Conclusions:

• Visual scoring of certain sleep data, particularly respiratory disturbances, exhibits inherent unreliability.
• These findings underscore the potential for measurement error in current sleep analysis paradigms.
• The study suggests a need for improved standardization or objective methods in sleep data interpretation.