Optimizing clinical trials recruitment via deep learning

Area of Science:

• Biomedical research
• Clinical trial management
• Artificial intelligence in healthcare

Background:

• Clinical trials are crucial for new treatment development but are complex and expensive.
• Selecting high-enrolling investigators is vital for efficient trial execution and cost control.
• Current methods for investigator selection may not be optimal.

Purpose of the Study:

• To introduce DeepMatch (DM), a novel deep learning approach for optimizing clinical trial investigator selection.
• To rank investigators based on their predicted enrollment performance for new clinical trials.
• To improve the efficiency and reduce the cost of clinical trial execution.

Main Methods:

• DeepMatch (DM) utilizes deep learning to analyze heterogeneous data from investigators and trials.
• The approach learns from existing data to predict investigator enrollment performance.
• Investigators are ranked based on their expected performance in upcoming clinical trials.

Main Results:

• A large-scale evaluation on 2618 studies demonstrated DM's effectiveness.
• DM improved investigator ranking by up to 19% compared to state-of-the-art methods.
• DM enhanced the detection of top/bottom performing investigators by up to 10%.

Conclusions:

• DeepMatch (DM) offers substantial improvements over current industry standards for investigator selection.
• DM enhances the assessment of investigator enrollment potential and speeds up list generation.
• The approach facilitates data-informed decisions for selecting investigators, optimizing trials, and reducing therapy costs.