The effect of measurement error on the hierarchical identification of a gas exchange model

Abstract

Using mathematical models allows simulation of a patient’s individual physiology and can thus be employed for predicting reactions to changes in therapy settings. For a model to be utilized at the bedside it must be identifiable from the available data within a reasonable timeframe. A previously presented approach using a hierarchy of models that is independent of initial parameter estimates showed promising results. This work evaluates how the hierarchical method behaves under noisy data. The approach was evaluated using data from twelve synthetic patients with added noise of different amplitude. The results were compared to the standard method of identifying the model directly with arbitrary initial estimates. Results show that while the model identified with the direct approach leads to a lower prediction error than the hierarchical approach when the initial estimates are close to the parameter values used to create the data, they become higher than the prediction error produces by the model identified with the hierarchical approach when the initial estimates are drawn from a wider range around the true model parameters.