“…The main advantages of the proposed method compared to the ODD and ODI are: (1) the identification of oxygen desaturations is dependent on the definition of a baseline, which in most practical situations may be particularly difficult to determine, even more so for patients with comorbidities that can affect their Oxygen Saturation (as is the case of dialysis patients) 11 , 19 ; (2) the ODI definition varies between authors, and they are usually tailored for the specific application of identifying sleep apnea events, by looking at a very specific shape in the oxygen saturation time series 37 ; (3) since the definition of desaturations is so specific, deviations from most common scenarios can jeopardize its predictive ability, situations like dialysis patients or patients in high altitude already present reduced average of oxygen saturation, which can impact the ODI through the baseline calculation 19 , 49 ; (4) some of the ODI definitions specify the desaturation minimal duration (usually 10 s), making them more difficult to identify in time series with low sampling frequency (e.g., 0.1 Hz) 20 , 40 , 46 . The proposed method is more robust against these shortcomings for the ODI: it does not depend on any baseline definition; it is also insensitive to the averages of the Oxygen Saturation; its definition is more precise, and it is robust against the sampling frequency of the input time series, and the only free parameter ( ) is directly interpretable in terms of the timescale of the problem.…”