Patient Recruitment (Clinical Trials)
Patient recruitment is expensive (15-20% of cost for typical oncology trials)
Synth Approach:
A combination of edge analytics / federated learning with large language models to analyze patient medical records at the trial sites (i.e., no PHI data leaves trial site)
The Benefits:
- Up to 50% faster and cheaper patient recruitment
- Improved diversity and trial quality
- Reduced risk of data breaches and HIPAA violations
Diving deeper into the problem
80% of trials fail to meet enrollment timelines; 50% of clinical research sites enroll 1 or no patients
Recruiting 1 patient can cost $6,500–$15,000, depending on therapeutic area
Trials often rely on manual EMR screening, missing eligible candidates. This delays going to market, making the trial more costly due to more days on site and potentially losing out on being first to market
Inadequate diversity: less than 5% of cancer trial participants are Black, despite higher incidence in some groups, and FDA guidance requiring biodiversity
Recruitment Leakage
How do we reach the right patients to achieve a higher enrollment rate?
Disconnected systems, manual processes, and privacy barriers make recruitment slow, expensive and biased
Solution Overview:
- Federated learning and edge analytics with large language models to analyze patient medical records at the trial sites
- No PHI data leaves the trial site, ensuring privacy and compliance
- AI models trained on-site to identify eligible patients based on trial criteria
- Automated patient matching and recruitment process
Technical Details
- Fine-tuned Large Language Models (LLM) that analyze Inclusion/Exclusion criteria and assess patient match by reading patient EMR
- Uses homomorphic encryption, differential privacy to ensure compliance
- Can integrate across FIS, Cerner, Epic, and other EMR systems with adapters
- Compatible with HL7/FHIR for secure semantic interoperability
ROI — Faster, Cheaper, Smarter Recruitment
Federated Learning can save $20M+ per trial, reduce recruitment time by 50%, and drive better trial diversity and quality through smarter matching
Other Benefits:
- Improves sponsor-site relationships with automated matching tools
- Minimizes site burden by automating EMR screening workflows