Development and Evaluation of Perl-Based Algorithms to Classify Neoplasms From Pathology Records in Synoptic Report...

Area of Science:

• Computational Pathology
• Bioinformatics
• Medical Informatics

Background:

• Synoptic reporting standardizes pathology diagnostics.
• Electronic pathology reports present classification challenges due to linguistic variation.

Purpose of the Study:

• To demonstrate Perl-based algorithms for classifying electronic pathology reports.
• To validate algorithms for specific neoplasms including melanoma, lymphomas, leukemia, and breast, ovarian, testicular, and thyroid cancers.

Main Methods:

• Developed case-finding cohorts using diagnostic codes and unique identifiers.
• Applied Perl algorithms for record classification (malignant, in situ, suspect, nonapplicable).
• Validated algorithm accuracy through hand-review, calculating interrater reliability, sensitivity, specificity, PPV, and NPV.

Main Results:

• Perl algorithms achieved high specificity (>98%).
• Most benign results were correctly classified; only leukemia had a positive predictive value below 95% (91.9%).
• Sensitivity varied, with ovarian cancer (33.3%), leukemia (52.8%), and non-Hodgkin lymphoma (62.9%) below 80% due to linguistic variation.

Conclusions:

• Algorithm logic built around synoptic reporting is valuable for identifying neoplasms in electronic pathology results.
• Perl-based coding in SAS offers a strength for developing highly specific algorithms amidst institutional documentation variations.