Polysomnography scoring for sleep apnea. Use of a sampling method

The all-night polysomnogram is used to assess the severity of sleep apnea by the quantitative analysis of a number of respiratory and sleep parameters. Because precise scoring of these parameters is a time-consuming process, we studied the effect of scoring varying percentages of polysomnograms on the accuracy of estimating sleep apnea severity. Twenty adult all-night polysomnograms with apnea-hypopnea (AH) indexes ranging from 1.0 to 67.8 were scored in the following manner: every 30-s epoch (100% scoring), every other epoch (50% scoring), every fourth epoch (25% scoring), every sixth epoch (16.7% scoring), every eight epoch (12.5% scoring), and every tenth epoch (10% scoring). Each method was then compared with 100% scoring. For each parameter we established criteria for classifying studies as having a mild, moderate, or severe degree of abnormality. The number of studies correctly categorized for the following parameters was: (1) AH index, 20/20 for 50 and 25% scoring, 19/20 for 16.7 and 10% scoring, and 18/20 for 12.5% scoring; (2) AH time, 20/20 for 50, 25, and 16.7% scoring, 17/20 for 12.5% scoring, and 18/20 for 10% scoring; (3) mean AH duration, 20/20 for 50, 25, and 12.5% scoring, 18/20 for 16.7% scoring, and 15/20 for 10% scoring; (4) mean event percent O₂ saturation, 20/20 for 50 and 25% scoring, 18/20 for 16.7 and 12.5% scoring, and 17/20 for 10% scoring; (5) longest AH, 19/20 for 50% scoring, 18/20 for 25% scoring, and 15/20 or less for 16.7, 12.5, and 10% scoring; (6) lowest percent O₂ saturation, 19/20 for 50% scoring, 17/20 for 25% scoring, and 15/20 or less for 16.7, 12.5, and 10% scoring. The correlation coefficients (r) of the mean sleep stage percentages and mean sleep efficiency indexes of the 20 polysomnograms for all 5 levels of scoring compared with 100% scoring were greater than 0.94. We conclude that scoring by sampling is a highly accurate method for estimating AH index, AH time, mean AH duration, mean event percent O₂ saturation, sleep efficiency index, and for sleep stage quantification. It is less accurate for identifying isolated events such as the longest respiratory event duration and the lowest percent O₂ saturation.