Is technologist review of raw data necessary after home studies for sleep apnea?

As the importance of portable monitors for detection of sleep apnea increases, efficient and cost-minimizing methods for data interpretation are needed. We sought to compare in stroke patients, for whom portable studies often have particular advantages, results from a cardiopulmonary monitoring device with and without manual edits by a polysomnographic technologist.

Participants in an ongoing stroke surveillance study in Corpus Christi, Texas, underwent sleep apnea assessments with the ApneaLink Plus device within 45 days of stroke onset. Recordings were analyzed by the device's software unedited, and again after edits were made to the raw data by a registered polysomnographic technologist. Sensitivity and specificity were calculated, with the edited data as the reference standard. Sleep apnea was defined by 3 different apnea-hypopnea index (AHI) thresholds: ≥ 5, ≥ 10, and ≥ 15.

Among 327 subjects, 54% were male, 59% were Hispanic, and the median age was 65 years (interquartile range: 57, 77). The median AHI for the unedited data was 9 (4, 22), and for the edited data was 13 (6, 27) (p < 0.01). Specificity was above 98% for each AHI cutoff, while sensitivity was 81% to 82%. For each cutoff threshold, the edited data yielded a higher proportion of positive sleep apnea screens (p < 0.01) by approximately 10% in each group.

For stroke patients assessed with a cardiopulmonary monitoring device, manual editing by a technologist appears likely to improve sensitivity, whereas specificity of unedited data is already excellent.