Clinical data miner: an electronic case report form system with integrated data preprocessing and machine-learning...

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Clinical Research Software

Background:

• Clinical diagnostic model research traditionally faces challenges due to the lack of integration between electronic Case Report Form (eCRF) systems and mathematical software.
• This disconnect results in complex, error-prone processes for extracting diagnostic models and limits model performance insights during data collection.

Purpose of the Study:

• To introduce the Clinical Data Miner (CDM) software framework, an integrated eCRF system.
• To enhance the efficiency of clinical diagnostic model research workflows by incorporating data preprocessing and machine-learning libraries.
• To enable optimization of patient inclusion numbers through real-time study performance monitoring.

Main Methods:

• The CDM software framework was developed using a test-driven development (TDD) approach to ensure high software quality.
• CDM features a modular architecture designed for future extensibility and integration.

Main Results:

• The TDD approach successfully delivered high-quality software, with the CDM eCRF Web interface currently used in studies with over 4000 patients.
• CDM's integrated libraries simplify previously manual and error-prone steps in diagnostic model research.
• Study coordinators can monitor predictive performance and optimize patient inclusion using CDM's monitoring capabilities.

Conclusions:

• CDM is uniquely positioned as the only eCRF system with integrated data preprocessing and machine-learning libraries.
• This integration significantly improves the efficiency of clinical diagnostic model research.
• CDM facilitates accurate assessment of data collection termination points, leading to improved models and reduced patient recruitment costs.