A knowledge-extraction experience in anticoagulation for early postoperative cardiac valvular surgery

Cardiac valvular prosthesis surgery requires an anticoagulant treatment whose management is particularly difficult in early postoperative surgery. The capacity to avoid both thromboembolic and haemorrhagic risks relies on the integration of numerous parameters. In this paper, we present an experience of Knowledge Extraction from clinical and biological data from a cardiac surgery department of the University Hospital of Grenoble. The purpose of our study is to analyse the different types of decision, to detect possible anomalies in decisions, and to try to provide some computer-assisted help for these decisions. Using the Decision Tree approach provided final decisions that are consistent with clinical results and made explicit some implicit parameters used in the prescription task. This study also provided a better understanding of the behaviour of the prescribers, with some unexpected results.