Mapping Free Text into MedDRA by Natural Language Processing: A Modular Approach in Designing and Evaluating Software Extensions

MagiCoder is a Natural Language Processing application designed to extract MedDRA terms from narrative clinical text. MagiCoder has been developed to support the work of people responsible for pharmacovigilance. Given a narrative description, MagiCoder proposes an automatic encoding; the pharmacologist reviews, (possibly) corrects, and then validates the solution. This drastically reduces the time needed for the validation of reports with respect to a completely manual encoding. In this paper we extend in a modular way and analyse MagiCoder, comparing its different new extensions. We designed a benchmark consisting of a representative set of adverse drug reaction reports that also includes long and badly written descriptions. We measured an average precision and recall of 68.74% and 70.19%, respectively. On descriptions up to 100 characters, both precision and recall exceeded 75%, i.e., 77.97% and 75.78%, respectively.