Automated Analysis of Data Is Inferior to Visual Analysis of Ambulatory Sleep Apnea Monitoring

Abstract

BACKGROUND: Many ambulatory sleep apnea monitoring devices are equipped with software which allows an automated analysis of data as well as a visual analysis. OBJECTIVE: The Merlin system which records heart rate, snoring sound, efforts, oronasal flow, body position and oxygen saturation was investigated to identify proper parameter settings for the automated analysis and to compare the automated with the visual analysis in patients with mild obstructive sleep apnea syndrome (OSAS). Sensitivity and specificity of the visual and automated analysis of ambulatory monitoring in comparison with visual polysomnographic (PSG) analysis were determined. METHODS AND RESULTS: First, we tried to find the optimal parameters for the automated analysis, using 7 different settings in 17 OSAS patients. Furthermore, we applied the optimized setting to 66 OSAS patients who were admitted (age 50.9 +/- 9.9 years, BMI 32.9 +/- 5 kg/m(2)), and compared the results with the visual analysis of raw data. The patients slept for one night in the sleep laboratory with Merlin and PSG simultaneously to compare the visual and automated analysis of Merlin data with results from the visual analysis of PSG. Automated analysis leads to an underestimation of the respiratory disturbance index (RDI; p < 0.001) compared with both the visual analysis and results of PSG. Using a cutoff level of 5 apneas and hypopneas/h for the diagnosis of OSAS, the sensitivity of Merlin with the automated analysis is 40.6% and the specificity is 100%. With a cutoff level of 15/h, sensitivity and specificity rose to 91.3 and 100%, respectively, which is comparable to the visual analysis. CONCLUSION: Merlin is a reliable device for detection of sleep-related breathing disorders, but recordings should be analyzed visually, especially in patients with a low RDI.