A successful technique for removing names in pathology reports using an augmented search and replace method.

Area of Science:

• Medical Informatics
• Clinical Data Management
• Epidemiologic Research

Background:

• Access to de-identified clinical data is crucial for epidemiologic and retrospective research.
• Existing de-identification methods often require natural language processing expertise or are not publicly available.
• Pathology reports contain a high frequency of proper names, often appearing in pairs.

Purpose of the Study:

• To develop an accessible and effective de-identification tool for pathology reports.
• To leverage the common occurrence of paired proper names for data anonymization.
• To improve the availability of de-identified clinical data for research.

Main Methods:

• Developed a substitution-based de-identification tool utilizing observations of proper name patterns.
• Created a Clinical and Common Usage Word (CCUW) list and a comprehensive proper name list.
• Employed publicly available data sources for tool development and list compilation.

Main Results:

• The method achieved high accuracy in de-identifying pathology reports, finding 98.7% of proper names.
• Only 0.3% of single proper names were missed in 1001 reports, with no identifiable name pairs.
• The tool demonstrated accuracy comparable to existing, more complex de-identification methods.

Conclusions:

• The developed method offers a practical and effective approach to de-identifying clinical data from pathology reports.
• This technique enhances the accessibility of de-identified data, supporting broader epidemiologic and retrospective research.
• Ongoing refinement of word lists aims to further improve de-identification accuracy and robustness.