Predictive Performance of a Model of Anaesthetic Uptake with Desflurane

We have previously shown that a model of anaesthetic uptake and distribution, developed for use as a teaching tool, is able to predict end-tidal isoflurane and sevoflurane concentrations at least as well as commonly used propofol models predict blood levels of propofol. Models with good predictive performance may be useful as part of real-time prediction systems. The aim of this study was to assess the performance of this model with desflurane. Twenty adult patients undergoing routine anaesthesia were studied. The total fresh gas flow and vaporizor settings were collected at 10-second intervals from the anaesthetic machine. These data were used as inputs to the model, which had been initialized for patient weight and desflurane. Output of the model is a predicted end-tidal value at each point in time. These values were compared with measured end-tidal desflurane using a standard statistical technique of Varvel and colleagues. Data was analysed from 19 patients. Median performance error was 78% (95% CI 8-147), median absolute performance error 77% (6-149), divergence 10.6%/h (-80-101) and wobble 8.9% (-6-24). The predictive performance of this model with desflurane was poor, with considerable variability between patients. The reasons for the difference between desflurane and our previous results with isoflurane and sevoflurane are not obvious, but may provide important clues to the necessary components for such models. The data collected in this study may assist in the development and evaluation of improved models.