Towards a statistically oriented decision support system for the management of septicaemia

The first decision-support system designed for the management of septicaemia was MYCIN. Although MYCIN played a vital role in the conception of knowledge-based systems, it never became an established clinical system. This paper describes an alternative decision-support system for septicaemia management currently under development at St. Thomas' Hospital (London) where a large database of septicaemia episodes has been compiled. The three statistical approaches that have been considered are described. These are (i) relative frequencies, (ii) the naive Bayes method and (iii) logistic regression. We also discuss how the concept of probabilistic influence diagrams could be of benefit to the development and implementation of the decision-support system.