Using Latent Mixed Markov Models for the choice of the best pharmacological treatment.

The choice of the best pharmacological treatment for an individual patient is crucial to optimize convalescence. Due to their effects on pharmacokinetics variables like gender and age are important factors when the pharmacological regimen is planned. By means of an example from anaesthesiology the usefulness of Latent Mixed Markov Models for choosing the optimal anaesthetic considering patient characteristics is demonstrated. Latent Mixed Markov models allow to predict and compare the quality of recovery from anaesthesia for different patient groups (defined by age and gender and treated with different anaesthetic regimens) in a multivariate non-parametric approach. On the basis of observed symptoms immediately after surgery and a few days later the probabilities for the respective dynamic latent status (like health or illness) and the probabilities for transition from one status to another are estimated depending on latent class membership (patient group).