Decision support of inspired oxygen selection based on Bayesian learning of pulmonary gas exchange parameters

To investigate if the real-time Bayesian learning of physiological model parameters can be used to support and improve the selection of inspired oxygen fraction. Supporting the selection of inspired oxygen fraction relies on predictions of arterial oxygen saturation. The efficacy of using these predictions to select inspired oxygen was tested retrospectively in a system for estimating gas exchange parameters of the lung (Automatic Lung Parameter Estimator, ALPE). For the predictions to offer effective decision support they need to be accurate and above all safe. These qualities were tested with data from 16 post-operative cardiac patients, using two different tests. The aim of the first test was to assess retrospectively if the predictions could have supported clinical decisions. The second test sought to establish if the predictions could support improving the efficiency of inspired oxygen selection during an ALPE oxygen titration. The predictions were found to be reasonably accurate, and most importantly safe in both of the tests. The method described can be used to support the selection of inspired oxygen fraction, and it has the potential to improve the efficiency of inspired oxygen selection during an oxygen titration.