Zero-Shot Clinical Trial Patient Matching with LLMs
Problem
Manually matching patients to clinical trial eligibility criteria takes up to 1 hour per patient, slowing drug development.
Why it matters
Faster, cheaper trial matching accelerates drug development and connects patients to potentially life-saving treatments.
What the AI did
Built a zero-shot LLM-based system that reads unstructured clinical text to evaluate trial inclusion criteria, plus a cost-efficient prompting strategy and a two-stage retrieval pipeline.
Results
State-of-the-art scores on the n2c2 2018 cohort selection benchmark; matching an order of magnitude faster and cheaper than status quo
Impact
Benchmark-leading automated trial matching demonstrated on a public benchmark; broad real-world adoption not stated in source.
Record
| Venue | NEJM AI |
|---|---|
| Published | 2024 |
| Work type | article |
| Authors | Michael Wornow, Alejandro Lozano, Dev Dash, Jenelle Jindal, Kenneth W. Mahaffey, Nigam Haresh Shah |
| Citations | 57 (OpenAlex, at import time) |
| References | 16 |
| Model or method | LLM-based system (specific model not stated in source) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1056/aics2400360 |
| Open access | green |
| Keywords | clinical trial matching, LLM, zero-shot, n2c2 benchmark, patient screening, retrieval pipeline |
Importance rating
major — millions of lives, an industry or a field
Abstract frames trial matching as a key unsolved challenge in bringing new drugs to market, with manual screening taking up to 1 hour per patient; affects the entire drug development pipeline.