AI Could Match Patients with Clinical Trials

A survey by UPMC's Center for Connected Medicine finds enthusiasm for the idea.

Most healthcare executives believe that artificial intelligence could be tapped to identify and recruit patients for clinical trials, according to a survey by UPMC's Center for Connected Medicine (CCM).

The survey of 58 payer and provider senior executives -- many of whom specialize in AI and analytics, and more than half of whom work at larger hospitals and academic medical centers -- was compiled in February and published this week in a CCM report.

"Clinical trials are an important part of providing life-changing care to our patients. However, it can be a significant challenge to match patients to studies," says Oscar Marroquin, MD, chief healthcare and data analytics officer at UPMC, and a founding partner of the CCM. "I'm excited by the potential for AI to help medical centers do a better job of finding and recruiting trial participants."

Patient recruitment is widely viewed as cumbersome and expensive. The CCM report cites a 2020 study in the Journal of Medical Internet Research which found that 80% of trials don't meet initial enrollment targets of timelines, with delays resulting in lost revenues of as much as $8 million a day for drug development companies.

Nearly two-thirds (64%) of respondents in the CCM survey say that the most time-consuming, labor-intensive, and frustrating hurdle for launching clinical trials is finding suitable patients, which often requires a manual review of medical records. Using AI to quickly scan medical records to find eligible patients was seen by 61% of respondents as playing a "critical" role.

"AI is seen as having the potential to allow organizations to complete matching, identification, and recruitment tasks at scale, enabling them to participate in more trials," the report says, although noting that "smaller organizations are more likely to be unsure what role AI technology will play."

Marroquin says CCM has "already seen the benefits of using natural language processing to harness and analyze the vast quantities of unstructured data in healthcare to better understand the conditions of our patients."

"Applying these techniques for clinical trial matching would be an advantage for health systems, industry and patients," he says.

The survey found that respondents who don't see AI as playing a critical role in the future "mention that existing analytics capabilities could potentially provide the same outcomes."

However, while some IT vendors offer AI for clinical trials, the survey found that the perception among stakeholders is that the readiness of these systems is "mixed."

One clinical researcher told CCM that "high-fidelity matching benefits greatly from structured data, and a significant amount of needed data is currently in the EHR as unstructured data. This is a significant gap."

The CCM report also identifies disease management and prediction as the top use for AI at health systems. Over the next two years, survey respondents say their AI investments will be aimed primarily around operational optimization.