The AI-assisted screening tool identified eligible candidates faster and with a higher eligibility rate (20.4% vs 12.7%) than manual screening.

Compared with manual screening, an artificial intelligence (AI)-assisted patient screening tool can improve the rate of eligibility determination and enrolment for randomized clinical trials, according to a research letter published online Feb. 17 in the Journal of the American Medical Association.

Ozan Unlu, M.D., from Mass General Brigham in Somerville, Massachusetts, and colleagues developed the Retrieval Augmented Generation Enabled Clinical Trial Infrastructure for Inclusion Exclusion Review (RECTIFIER) large language model tool and tested it in a prospective, blind, randomized clinical trial within the Mass General Brigham Health System. The trial was completed within an ongoing heart failure implementation trial and included 4,476 patients who were randomly assigned. The primary outcome was time to trial eligibility determination, which was assessed using a survival analysis framework.

The researchers found that compared with patients screened manually, patients screened with the AI-assisted method achieved the primary end point of eligibility determination significantly more quickly (subdistribution hazard ratio [HR], 1.78). The eligibility rate was 20.4 and 12.7 percent for the AI-assisted screening and manual screening methods, respectively. At the end of the trial, there were 35 and 19 enrollments (1.6 and 0.9 percent, respectively) using the AI-assisted screening and manual screening methods (subdistribution HR, 1.79).

“Implementing AI-assisted screening tools like RECTIFIER enhances clinical trial efficiency and expedites the screening and recruitment process, which may lead to faster trial completion and earlier access to novel therapies for patients,” the authors write.