Open Source Tool for Deidentifying Pathology Reports
Author Information
Author(s): Beckwith Bruce A, Mahaadevan Rajeshwarri, Balis Ulysses J, Kuo Frank
Primary Institution: Beth Israel Deaconess Medical Center
Hypothesis
Can an open source software tool effectively deidentify pathology reports while maintaining the integrity of the text?
Conclusion
The study successfully demonstrated the creation of an open-source deidentification program that performs well on free-text pathology reports.
Supporting Evidence
- 1254 out of 1800 pathology reports contained identifiers.
- 3439 out of 3499 unique identifiers were successfully removed.
- Performance varied by institution, with removal rates between 94.7% and 99%.
Takeaway
The researchers made a computer program that can hide personal information in medical reports so that people can use the reports for research without knowing who the patients are.
Methodology
The software was developed using Java, MySQL, and JDOM, and tested on 1800 pathology reports from three institutions.
Potential Biases
The scrubber may over-scrub non-identifying phrases, leading to potential loss of important information.
Limitations
The scrubber's performance varied by institution, and some identifiers, especially misspelled names and certain numeric identifiers, were not removed.
Participant Demographics
Pathology reports from three institutions were used, including in-house and external consultation cases.
Statistical Information
P-Value
p<0.001
Statistical Significance
p<0.001
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website