Automated Identification of Common Disease-Specific Outcomes for Comparative Effectiveness Research Using...

Area of Science:

• Clinical informatics
• Biomedical data science
• Translational medicine

Background:

• Establishing common disease-specific outcomes is crucial for clinical trial data comparability and meta-analyses.
• Traditional expert panel methods for identifying common outcomes are time-consuming and laborious.

Purpose of the Study:

• To develop and evaluate a generalized pipeline for automatically identifying disease-specific common outcomes.
• To streamline the process of finding, downloading, and analyzing clinical trial data for outcome identification.

Main Methods:

• Developed an automated pipeline interfacing with ClinicalTrials.gov API to download relevant trial data.
• Parsed and grouped primary/secondary outcomes by text similarity and ranked them by frequency.
• Assessed pipeline output for chronic obstructive pulmonary disease (COPD) against manually abstracted outcomes from literature reviews.

Main Results:

• The pipeline processed 3876 COPD-related studies, matching manual download counts.
• Achieved 92% recall and 79% precision in identifying common outcomes compared to manual abstraction.
• Identified relevant COPD outcomes not previously covered in literature reviews, suggesting potential for future research.

Conclusions:

• An automated, evidence-based pipeline can identify common clinical trial outcomes with comparable quality and breadth to literature reviews.
• This automated approach offers an efficient alternative to traditional methods and can uncover novel relevant outcomes.