Clinical Scorecard: AI-Selected Trial Candidates are More Diverse

At a Glance

Key Highlights

• AI system processed 1,476 patient records in six days
• Achieved 96.2% accuracy in answering trial-relevant questions
• Identified 30 eligible patients, 37% of whom were Black
• 93% of AI-identified patients were not registered with a cardiologist
• 99% accuracy in excluding ineligible patients

Guideline-Based Recommendations

Diagnosis

• Utilize AI to analyze EMRs for identifying eligible patients

Management

• Implement AI systems to streamline clinical trial recruitment processes

Monitoring & Follow-up

• Regularly assess AI performance against traditional methods

Risks

• Ensure AI systems are validated for accuracy and reliability

Patient & Prescribing Data

Patients with transthyretin amyloid cardiomyopathy

AI may enhance recruitment from under-represented populations

Clinical Best Practices

• Integrate AI systems for efficient patient identification
• Provide traceable explanations for AI decisions
• Validate AI findings with clinician review

References

• Journal of Cardiac Failure