JMIR Publications

ABSTRACT

Background:

Pharmacovigilance and safety reporting, which involves processes for monitoring the use of medicines in clinical trials, plays a critical role in the identification of previously unrecognized adverse events or changes in the patterns of adverse events.

Objective:

This study aimed to demonstrate feasibility of automating the coding of adverse events described in the narrative section of the serious adverse event report forms to enable a statistical analysis of the aforementioned patterns.

Methods:

We used the Unified Medical Language System (UMLS) as the coding scheme, which integrates 217 source vocabularies, thus enabling coding against other relevant terminologies such as ICD-10, MedDRA and SNOMED. We used MetaMap, highly configurable dictionary lookup software, to identify mentions of the UMLS concepts. We trained a binary classifier using Bidirectional Encoder Representations from Transformer (BERT), a transformer-based language model that captures contextual relationships, to differentiate between mentions of the UMLS concepts that represent adverse events and those that do not.

Results:

The model achieved a high F1 score of 0.8741 despite the class imbalance.

Conclusions:

These results confirmed that automated coding of adverse events described in the narrative section of the serious adverse event reports is feasible.