The Cleveland Clinic has partnered with newcomer Dyania Health to identify a broader array of patients for clinical trials using artificial intelligence.
The Cleveland Clinic runs a massive research operation, and clinical trials for pharmaceutical products comprise a major slice of it. Dyania Health will now be the clinic’s core tech partner to support recruitment for clinical trials and potentially more applications in the future.
Before partnering with Dyania, the Cleveland Clinic relied on manual chart review to identify patients for clinical trials. Not only is the process time-consuming, it is limited to specific points in time when the review occurred. The clinic tried out some assistive technology, but it was ultimately too inaccurate and expensive to continue to justify, executives said.
Dyania’s propriety LLM-based platform Synapsis AI combs the medical records of the clinic’s millions of patients to identify eligibility for its spate of clinical trials. It can also run queries daily to recheck whether patients can qualify for trials based on changes to their health.
“How we were recruiting for clinical trials, it was a very painful, archaic, inefficient exercise that was frustrating to anybody who is involved in that cycle of clinical trials, not just in Cleveland Clinic, but nationally,” said Lara Jehi, M.D., chief research information officer at The Cleveland Clinic.
Jehi described that research coordinators manually review the medical charts of the patients in the office of the physician they’re serving and identify patients that match the inclusion and exclusion criteria. Reviewing one chart typically takes a human between 30 minutes and two hours, Eirini Schlosser, founder and CEO of Dyania Health, said.
“The problem with that is that besides the time that is being consumed, this is a very limited, narrow view of the whole pool of patients that could potentially benefit from the intervention,” Jehi said.
With Synapsis AI, customers can get new reports every day on eligible patients for their trials.
Dyania launched out of stealth in October 2024 and raised a $10 million series A led by HealthX Ventures. Tech Square Ventures and Cleveland Clinic Ventures also backed the round. The company has been working on using AI for automated chart review since 2020, annotating a very large data set of medical records to train a large language model. Schlosser said the company is flexible on use cases for the chart review technology.
Identifying eligible participants for clinical trials is one of the most difficult use cases because patients’ characteristics are changing over time across multiple data sources. Schlosser said Dyania has eight AI models working simultaneously on the issue.
Before the enterprisewide rollout of Synapsis AI, announced Wednesday, Dyania and The Cleveland Clinic ran a series of pilots to test the technology. Dyania used its technology to identify candidates for real clinical trials in oncology, cardiology and neurology. For a melanoma trial, Synapsis AI identified all eligible patients in 2.5 minutes with 96% accuracy, whereas nurses spent 400 minutes on the task.
In one week of searching for candidates for a cardiology trial, Synapsis AI reviewed 1.2 million patient records and identified twice as many eligible patients as in three months of traditional recruitment.
While Synapsis AI will be available to all researchers that want to use it, Schlosser explained that the complex generative AI is best used for clinical trials where the inclusion and exclusion criteria are complex. She explained that finding patients with Type 2 diabetes in a certain age range, for example, isn’t difficult and thus is not a good application of the technology.
“Where the technology can add the most value is where you have a decent-sized patient population, but within those medical records, the characteristics of the patients are changing quite quickly,” Schlosser explained. “So they've finished one therapy, not yet started another. And so what makes them difficult to find is actually that window in time where the patient exactly has their stars aligning to meet the inclusion-exclusion criteria.”
“One patient today might not meet criteria, but tomorrow they do, and so it's automatically finding those differences over time,” Schlosser said.
The AI system has to look across free text in notes, lab results, structured and discrete data to deduce conclusions about the patients and compare them against the trial criteria. It can’t rely on ICD-10 billing codes, which are not confirmed diagnoses, but go into the note to see if later the diagnosis was confirmed.
The Cleveland Clinic Research employs close to 2,000 individuals, has over 200 labs and spent more than $500 million last year. The research arm of the clinic runs about 1,000 clinical trials across disciplines, and the heaviest disease areas researched are in cancer, cardiovascular and neurology.
One major benefit of the technology is its ability to identify eligible patients across the health system. The Cleveland Clinic has more than 220 locations in Ohio alone and more around the world. This means that the technology can find individuals at community clinics who are often left out of trials because they can’t be easily identified by human reviewers.
In one of the cancer studies for which the clinic partnered with Dyania, 80% of the participants came from community clinics. The increased capacity of AI to query huge volumes of records across a large health system can increase the diversity of trial participants as well.
“[Research coordinators are] often missing patients who are not going in to see a specialist, who are maybe being seen in a community site that's not at the main campus,” Schlosser said.
In a clinical trial related to cardiovascular disease, the Cleveland Clinic became one of the highest enrolling sites in the U.S. and had 37% higher diversity in the patient population because it used Dyania to find eligible participants. As a result, the study sponsors reopened enrollment and gave the clinic additional slots.
Synapsis AI will be rolled out in a phased approach, starting with certain disciplines and then extending to others, executives said.