A New Way of Data Analysis and Rating of the Blood Oxygen Saturation Variability

Abstract

This report continues the series of researches on the data set for peripheral blood oxygenation. Here we confirm the separation of the healthy adults into three subgroups by variabilities of oxygenation. Two new measures (descriptors) of variability are suggested, which correlate with the statistical analogs and Recurrence Ratios. These measures are handier in clinic praxis because of simpler computing comparing with Recurrence Ratios. We found the lower height of modes (i.e., of most probable oxygenation levels) and their higher variabilities for the riskiest subgroup concerning tor COVID-19..

Development of a method for differential analysis of data on the arterial blood oxygenation in healthy adults

“…Practice requires that new ways of describing variability, alternative to traditional statistical methods [15], should be reliable enough and easy for clinicians to understand and apply. In addition, such methods, of course, can be purely empirical, as in [14,16]. However, it is better if they have certain statistical justification, as in [17], although this research outlines only the contours of such a description.…”

“…Well-known statistical characteristics of variability are standard deviations (dispersions), interquartile ranges, and total data spreads [15]. However, direct application of these methods, as well as variability descriptors from Poincar graphs [12], faces certain difficulties that are analyzed in [16]. The difference in distributions of probability of real data from the normal Gauss distribution (multimodality, large share of emissions, asymmetry, etc.…”

“…The difference in distributions of probability of real data from the normal Gauss distribution (multimodality, large share of emissions, asymmetry, etc. ), makes us look for new, unconventional ways to assess the variability of blood oxygenation data for both SpO 2 [14,16], and for SaO 2 [17].…”

“…Specifically, work [16] offers new variability indicators, such as the number of observed levels of blood saturation, or the probability of top (most likely) levels. The authors of paper [17] applied the technique of differential analysis of primary blood oxygenation records and proposed the mixed Poisson distribution to describe the number of switches between observed levels of saturation.…”

“…They rely on the statistical distribution of data on the quartile, showing the interquartile range in the form of the boxes' height, and the full data spread in the form of vertical lines with dashes («whiskers»). Emissions can be shown at separate points [15,16]. Fig.…”