Systèmes informatiques d'aide à la décision en médecine : panorama des approches utilisant les données et les connaissances

Clinical decision support systems (CDSSs) are computerized tools capable of handling all the characteristics of a given patient to provide the most probable diagnoses (diagnostic decision-making) or best treatments (therapeutic decision-making). Today, CDSSs cover a wide range of tools developed for different purposes and of variable complexity. This article provides an overview of CDSSs and of the different approaches implemented. Numerical approaches are mainly based on the analysis of data sets. They were initially applied to diagnosis decision support. The deployment of computerized patient records makes large amounts of data available for predictive models to support therapeutic decision-making. Symbolic approaches rely on the modeling of decision-related knowledge. These approaches were originally implemented for expert systems expected to reproduce the reasoning of expert clinicians engaged in a decision-making process. Recently, following the principle of evidence-based medicine, the subjective expertise of expert systems has been replaced by the objective knowledge of clinical practice guidelines in modern CDSSs. Different CDSSs, whether historical or current, are described, various types of knowledge exploitation, either in automatic mode or documentary mode, are presented. Formerly confined to research laboratories and pilot experiments, it is likely that, given the demands of modern societies for quality care, CDSSs will enter routine use in the near future to improve clinical practices and enable performance evaluation.