The use of epochs to stage sleep results in incorrect computer-generated AHI values

The objective of this study was to determine what impact the omission of respiratory events scored during wake epochs has on the generation of the apnea/hypopnea index (AHI) using computerised polysomnography (PSG).

Sixty diagnostic PSG recordings were scored using the recommended method of sleep staging with epochs. In addition, absolute sleep scoring criteria was used to calculate the total sleep time (TST). The TST, respiratory events and AHI generated from both scoring methods were compared.

The AHI from epoch scoring (AHI(EP)) was significantly less than that derived from absolute sleep staging (AHI(ABS)) [AHI(EP) = 15.8 events/hour (7.3, 38.2), AHI(ABS) = 18.8 events/hour (8.7, 40.2); p < 0.0001]. From a total of 8,820 legitimately scored respiratory events ≥10 s duration, 723 (8%) were excluded from the AHI(EP) calculation as they were identified as having occurred during wake, but 400 of these (57%) started in an epoch of wake and continued into epoch(s) of sleep. These omitted events were not trivial [median 16.6 s (13.0, 24.0)] and were frequently associated with oxygen desaturation events [median 4% (3.0, 7.0)] with a range of 3% to 39%.

The use of arbitrarily defined epochs for sleep staging has a direct impact on legitimately scored respiratory events resulting in the systematically incorrect reporting of the AHI. This artefact is due to computerised data reduction excluding all respiratory events that are contained within an epoch(s) of wake as well as those that start in an epoch scored as wake.