A transformation of oxygen saturation (the saturation virtual shunt) to improve clinical prediction model calibration and interpretation

ObjectivePeripheral oxygen saturation (SpO2) obtained from pulse oximetry is a widely used physiological measurement. Clinical interpretation is limited by the nonlinear relationship between SpO2, degree of impairment in gas exchange, and effect of altitude. SpO2 is frequently dichotomized to overcome these limitations during prediction modelling. Using the known physiological relationship between virtual shunt and SpO2, we propose the saturation gap as a transformation of SpO2.\n\nApproachWe computed the theoretical virtual shunt corresponding to various SpO2 values and derived an accurate approximation formula between virtual shunt and SpO2. The approximation was based on previously described empiric observations and known physiological relationships. We evaluated the utility of the saturation gap in a clinical study predicting the need for facility admission in children in a rural health-care setting.\n\nMain ResultsThe transformation was saturation gap = 49.314*log10(103.711 - SpO2) -37.315. The ability to predict hospital admission based on a continuous variable SpO2 or saturation gap produced an identical area under the curve of 0.71 (95% CI: 0.69-0.73), compared to only 0.57 (CI: 0.56-0.58) based on diagnosis of hypoxemia (defined as SpO2<90%). However, SpO2 demonstrated a lack of fit compared to saturation gap (goodness-of-fit test p-value <0.0001 versus 0.098). The observed admission rates varied linearly with saturation gap but varied nonlinearly with SpO2.\n\nSignificanceThe saturation gap estimates a continuous linearly interpretable disease severity from SpO2 and improves clinical prediction models. The saturation gap will also allow for straightforward incorporation of altitude in interpretation of measurements of SpO2.