Overview of the 2022 n2c2 shared task on contextualized medication event extraction in clinical notes

Area of Science:

• Natural Language Processing (NLP)
• Clinical Informatics
• Biomedical Data Science

Background:

• Accurate medication histories are crucial for quality medical care.
• Understanding medication change events in clinical notes is essential but challenging.
• Extracting medication changes without clinical context is insufficient for practical use.

Purpose of the Study:

• To evaluate NLP systems for extracting medication change events and their context from clinical notes.
• To advance the understanding of contextual information surrounding medication events.
• To support the development of real-world clinical applications.

Main Methods:

• The 2022 National NLP Clinical Challenges Track 1 focused on contextualized medication events.
• Three subtasks were defined: Named Entity Recognition (NER), Event detection, and Context extraction (action, negation, temporality, certainty, actor).
• Transformer-based large language models were predominantly used by 32 participating teams.

Main Results:

• High performance was observed for NER systems.
• Performance for Event and Context extraction was significantly lower.
• Challenges included indirectly stated events, reliance on distant textual clues, and multiple events per mention.

Conclusions:

• NLP for medication extraction is mature, but contextual understanding is an open research problem.
• Further research is needed to bridge the gap between NLP capabilities and clinical application requirements.
• Improving context extraction is key to supporting accurate medication history taking in clinical practice.