Parsing Clinical Trial Eligibility Criteria for Cohort Query by a Multi-Input Multi-Output Sequence Labeling Model

Area of Science:

• Biomedical Informatics
• Computational Linguistics
• Clinical Trial Management

Background:

• Clinical trial recruitment relies on complex eligibility criteria often written in unstructured free text.
• Automating the interpretation of these criteria is essential for efficient patient screening and trial enrollment.

Purpose of the Study:

• To investigate the use of a supervised multi-input multi-output (MIMO) sequence labeling model for parsing clinical trial eligibility criteria.
• To assess the model's ability to translate free-text criteria into a computable format.

Main Methods:

• A supervised multi-input multi-output (MIMO) sequence labeling model was employed.
• A BERT-based encoder was utilized within the MIMO framework.
• Experiments were conducted on a small, manually annotated training dataset.

Main Results:

• The MIMO framework achieved an overall lenient-level AUROC of 0.61.
• Performance was suboptimal, indicating room for improvement in parsing accuracy.
• The model demonstrated the potential to represent eligibility criteria as logical and semantically clear tuples.

Conclusions:

• Representing clinical trial eligibility criteria as structured tuples is a promising approach for database query translation.
• Further development of NLP models is needed to improve the accuracy and reliability of automated criterion parsing.
• This method could enhance the efficiency of electronic patient screening for clinical trials.