Unattended sleep study in screening for sleep apnea in the acute phase of ischemic stroke

Objectives: We aimed to compare the feasibility of automatically and manually scored cardiorespiratory polygraphy results for sleep apnea screening in the acute phase of ischemic stroke.

Methods: Prospective study included 204 ischemic stroke patients, who underwent nocturnal unattended portable monitoring with three-channel device during 48 h after onset of stroke symptoms. Sleep apnea diagnosis was determined as a respiratory event index (REI) ≥5/hour.

Results: The mean age of the patients was 67.7 years and 62.7% were males. Automatic scoring identified sleep apnea in 148/72.5% and manual scoring in 186/91.2% of patients, showing moderate agreement (Kappa value 0.407). Of the cohort, 39 (19.1%) patients fulfilled the criteria for diagnosis of central sleep apnea (CSA) (>50% of events central) and 165 (80.9%) had obstructive sleep apnea (OSA). Automatic scoring identified poorly the type of events, and missed 18.6% of recordings. Oxygen desaturation 4% index (ODI4), interclass correlation coefficient (ICC) value 0.993, (95% CI 0.990-0.994), lowest saturation (ICC value 0.989, 95% CI 0.985-0.991) and percent of time spent <90% (ICC value 0.987, 95% CI 0.982-0.990) showed excellent agreement. Interclass correlation value for REI was 0.869 (95% CI 0.828 to 0.901) and ICC value 0.848 (95% CI 0.800 to 0.885) for obstructive apnea index, both indicating good agreement. Automatic scoring underestimated the severity of sleep apnea.

Conclusions: Both automatically and manually scored results recognized sleep apnea, albeit manual scoring identified more sleep apnea. Sleep apnea diagnosis agreement was moderate. Underestimation of the severity of sleep apnea by automatic scoring may result in undertreatment of sleep apnea.

Clinical trial registration: URL:http://www.clinicaltrials.cov. Unique identifier: NCT01861275.