The work proposes a decision tree, a well-known supervised machine learning algorithm, aiming at formalizing clinical processes and figure out the implicit knowledge used by clinicians in order to support their decision making process. The experimental framework employs a dataset of 290 clinical records (with 39 attributes) of patients assisted in the Orthodontic School at the University of Naples "Federico II". The whole procedure has the dual purpose of improving orthodontists’ within and between agreeement so as to validate, from a clinical point of view, the heuristics generated by the decision tree. On the whole, the proposed approach, besides meeting the criteria proposed by Bodemer et al. for accuracy, ease of use and speed, produces an improvement of orthodontists agreement levels and a reduction of the attributes needed to take the clinical decision. Overall, from a managerial perspective, it turns to a reduction of costs in terms of both diagnostic and clinical examinations.
Keywords: Heuristics, Health care processes, Decision trees, Odontology, Efficiency