Finding appropriate clinical trials: evaluating encoded eligibility criteria with incomplete data

Area of Science:

• Biomedical Informatics
• Clinical Trial Management
• Artificial Intelligence in Medicine

Background:

• Automating the evaluation of clinical trial eligibility criteria is crucial for efficient patient recruitment.
• Breast cancer clinical trials present complex eligibility requirements that necessitate systematic evaluation.

Purpose of the Study:

• To develop and evaluate a system for automating the assessment of patient eligibility for breast cancer clinical trials.
• To enhance the process of matching patients with suitable clinical research protocols.

Main Methods:

• Developed a data model for breast cancer trial eligibility criteria, encoding them using standard vocabularies.
• Utilized Bayesian networks to manage missing patient data and calculate eligibility probabilities.
• Ranked clinical trial protocols based on the system's calculated probability of patient eligibility.

Main Results:

• The system demonstrated good agreement (kappa 0.86) with an independent physician for protocol selection.
• However, agreement was poor (kappa 0.24) for protocol ranking, indicating areas for improvement.
• The approach proved feasible in assisting with the initial selection of appropriate trials.

Conclusions:

• The automated system for evaluating clinical trial eligibility criteria is a feasible approach.
• This technology has the potential to significantly assist both physicians and patients in identifying suitable clinical trials.
• Further refinement is needed, particularly in the protocol ranking component, to maximize utility.