pyDeid

pyDeid deidentifies free-text medical records using a rule-based approach with configurable tradeoffs between performance and speed, as well as precision and recall.

Key Features:

• Rule-based deidentification: pyDeid uses regular expression-based rules combined with fixed exclusion lists and inclusion lists to deidentify free-text data.
• Optional named entity recognition: Additional configurations include optional named entity recognition for enhanced deidentification.
• Custom name lists: Users can configure custom name lists to tailor deidentification to specific contexts or workflows.
• Performance-speed flexibility: pyDeid allows prioritization between deidentification performance and processing speed.
• Precision-recall flexibility: pyDeid enables configuration to prioritize either precision or recall based on use case requirements.
• Cross-domain generalizability: pyDeid is generalizable to domains outside of clinical data and customizable for specific contexts.

Scientific Applications:

• Clinical data deidentification: pyDeid removes personally identifiable information from free-text clinical data to enable broader research availability.
• Hospital admission notes processing: pyDeid deidentifies admission notes from hospital datasets for secondary research use.
• Discharge note deidentification: pyDeid processes discharge notes for deidentification benchmarking and research applications.
• Synthetic clinical data evaluation: pyDeid can be applied to artificial intelligence generated admission notes to evaluate deidentification performance on synthetic data.

Methodology:

pyDeid uses a combination of regular expression-based rules, fixed exclusion lists and inclusion lists to deidentify free-text data, with optional named entity recognition and custom name lists as additional configurations.