Phenotype Instance Verification and Evaluation Tool (PIVET): A Scaled Phenotype Evidence Generation Framework Using...

Area of Science:

• Computational biology
• Medical informatics
• Health data science

Background:

• Automated extraction of clinically relevant patient characteristics, termed computational phenotypes, is crucial for healthcare decision-making.
• Validation of these data-driven phenotypes is essential before clinical implementation.
• Existing methods require robust validation to ensure clinical meaningfulness.

Purpose of the Study:

• Introduce the Phenotype Instance Verification and Evaluation Tool (PIVET), a novel framework for validating computational phenotypes.
• Enhance the speed and automation of phenotype validation using co-occurrence analysis on a large medical corpus.
• Improve upon existing tools like PheKnow-Cloud for phenotype validation.

Main Methods:

• Utilized co-occurrence analysis on an online corpus of medical journal articles to build clinical relevance evidence sets.
• Employed NoSQL database indexing and an optimized co-occurrence algorithm for efficient evidence generation.
• Developed a statistical model trained on expert-verified phenotypes for automated clinical relevance classification.

Main Results:

• PIVET demonstrated an order of magnitude improvement in speed compared to PheKnow-Cloud while maintaining comparable accuracy.
• The framework is scalable to larger corpora without sacrificing performance.
• Ridge regression achieved the best performance, with an average F1 score of 0.91 in predicting clinically relevant phenotypes.

Conclusions:

• PIVET significantly enhances the speed and automation of computational phenotype validation.
• The tool offers comparable accuracy to existing methods while being substantially faster and more scalable.
• PIVET represents a major advancement in validating computational phenotypes for clinical use.