Automatic retrieval of current evidence to support update of bibliography in clinical guidelines

摘要

This paper reports on a system developed to support medical experts in the process of updating clinical guidelines by automatically suggesting new articles suitable to the domain under consideration. It follows a comprehensive process based on several consecutive steps in order to (ⅰ) identify which articles from the current guideline are eligible to be updated; (ⅱ) retrieve and filter new related articles from medune; and (ⅲ) select the most relevant resulting articles by applying a scoring algorithm. Extensive validation is based on a set of experiments on 40 guidelines from multiple medical domains. The analysis of results shows a promising prospect as indicated by recall values greater than 90%.