P048 The impact of variability in sleep scoring between manual and autoscoring on the measurement of OSA endotypes

Abstract Background Obstructive Sleep Apnoea (OSA) endotypic traits can be quantified using signals from routine polysomnography (PSG). However, it is unclear whether interrater variability in manual and auto sleep scoring influences the measurement of OSA endotypes and we aim to assess this. Methods Standard clinical variables and OSA endotype data were extracted from 133 PSGs scored first by two manually-scored methods (“A” and “B”) and a Compumedics autoscoring method (“C”). These scoring sets will be fully analysed with Phenotyping using Polysomnography (PUP) to deliver physiological endotype values and to assess native PUP scoring analyses. Differences were assessed between the scoring methods. Non-parametric tests were used due to non-normality of data with p<.05 taken as significant. Data are reported as median [IQR]. Progress to Date To date we have only analysed the impact the scoring methods have on the clinical PSG metrics. Lateral AHI was scored less by “A” than “C” (13.2 [31.4] vs. 20.5 [28.6], p=.040). Hypopnoeas/hour, total hypopnoeas, and NREM Sleep % were scored more by “C” but REM Sleep % and REM sleep mins were scored less by “C” while remaining similar between “A” and “B” (p<.001). Traditional vs PUP native scoring for standard PSG parameters differed in average SpO2 desaturation (“C”: 5% [2%] vs “C PUP”: 2.7% [1.4%] p<.001), ODI (3%) (“A”: 18.7 [31.7] vs “A PUP”: 26.2 [28.8], p<.024) and ODI (4%) (“B”: 12.6 [24.3] vs “B PUP”: 16.0 [25.9], p<.043). Intended outcome and impact Ongoing analysis will determine if autoscoring significantly impacts measurement of OSA endotypes.