Health Experts Say AI Is a Time Saver for Clinical Trial Recruitment

Health Experts Say AI Is a Time Saver for Clinical Trial Recruitment

Clinical researchers often face strong headwinds when it comes to clinical trial recruitment due to overburdened staff, complex protocols with stringent eligibility criteria, high rates of protocol deviations, and high dropout rates. Using artificial intelligence (AI) to precision match patients to clinical trials can help de-risk and accelerate recruitment. Today, health systems across the country are investing in AI to mine their structured and unstructured electronic medical record (EMR) data to better assess trial feasibility, precisely identify patients for trials, and reduce the time their staff spend on chart reviews. 

In this blog, we share perspectives from two leaders in clinical research, Robert (Robb) Stillman, MA, RN, CPHIMS, FHIMSS, at The James Comprehensive Cancer Center at The Ohio State University and Suneet Mittal, MD, at Valley Health System. We discuss their deployment approaches, the impact AI has had on their clinical trial recruitment so far, and lessons learned along the way.

AI for Clinical Trials at The James Comprehensive Cancer Center

Increasing Clinical Trial Enrollment with AI

Robb Stillman is the Director of Clinical Research Informatics at The James Comprehensive Cancer Center at The Ohio State University (OSU), the third largest cancer program in the country in terms of space. He is an oncology nurse by background and in his role at the Clinical Trials Office (CTO) he bridges the gap between IT research professionals and clinicians by providing strategic leadership for clinical research informaticists and advising senior leaders on information technology initiatives.

On any given day, OSU is running hundreds of cancer clinical trials, many which are initiated by the providers themselves. As an NCI-Designated Cancer Center, Robb said they do a good job of meeting clinical trials enrollment targets—about 30% of patients that come through the cancer program are on trials. The challenge, however, is that matching patients to trials is labor intensive. According to Robb, the most immediate reason for launching an AI program was to decrease the time their staff spend screening patients.

“We didn’t want to hire people just for the purpose of screening. We wanted to reduce the time of the people that we already had and we wanted to minimize the need for additional resources.” 
– Robb Stillman, MA, RN, CPHIMS, FHIMSS

Additionally, Robb wants to continue to increase enrollment, particularly among women minority and underserved populations. He also wants AI to help them improve the precision of their feasibility assessments, enhance matching, and identify patients with specific genetic markers.

A Centralized Deployment Approach

Currently, OSU’s CTO centrally manages the AI program, helping researchers at The James Comprehensive Cancer Center with pre-screening, screening validation, and recruitment of patients. Eventually, they will roll the AI program out to their investigators. 

OSU has a data feed from their EMR and four different OMICS vendors that goes directly into the Deep 6 AI platform, allowing them to precisely find patients for trials in minutes. Today, 15 disease group teams use Deep 6 AI to match patients based on upcoming appointments. Most of their clinical research coordinators (CRCs) are embedded in the clinic. Instead of having to read through charts, CRCs upload a clinic schedule into the AI software and get back a list of all the eligible trials available for each patient with an upcoming appointment. Then, they can quickly validate patients with evidence in the chart. 

“That’s where the real magic happens—in matching patients to trials that are sitting right in front of them. And that is actually the component that we are hoping to be able to put in front of our physicians very shortly as well.”
– Robb Stillman, MA, RN, CPHIMS, FHIMSS

Robb said he is also in the midst of a project to embed the Deep 6 AI functionality in the EMR system. A ‘Deep 6 AI’ tab would live on the patient’s chart and list all trials for which that patient is eligible. Robb is optimistic about Deep 6 AI being incorporated into clinical workflows so that all clinicians, even those not on the research team, can help find trials for patients.

The Advantages of an AI Solution Over Traditional EMR Searches 

Using tools that come with the EMR system are sufficient for searching structured data, however, a lot of information is an unstructured format (e.g., clinician notes or pathology reports). Additionally, performing keyword searches to find patients can be tricky when physician within the same network have different ways of documenting their care.

“It takes a lot of time and effort to read through every patient record, especially for patients that have received care for a considerable amount of time or have a complex disease. So, it’s really like hitting the lottery if you find someone who actually meets the trial criteria.” 
– Robb Stillman, MA, RN, CPHIMS, FHIMSS

The advantage of using Deep 6 AI is that it allows for considerably faster screening. The AI mines structured and unstructured data to match patients more precisely, and it does so in a matter of minutes. The AI can also identify patients with specific genetic markers, even when the genomics data are buried in a clinician note or hidden in a disparate lab report. 

Robb said that at OSU they’re also working with sponsors and investigators to leverage the AI earlier in the process—for protocol design and feasibility—to hopefully limit downstream issues during recruitment.   

Driving Engagement and Adoption Internally

Robb has spent a lot of time setting realistic expectations with internal teams. He explained that it takes time for the AI to be trained. Patient-matching AI is not the same sort of AI that we hear about in the media. It’s not Siri or ChatGPT. Building queries requires clinical knowledge and the queries need to be iterated on. What they’ve learned at OSU along the way is that there’s a lot of art and science, clinical knowledge, and deep protocol understanding required to be able to convert a protocol’s inclusion and exclusion criteria into something that AI can easily mine the EMR data for.

Robb said the key to their success has been relying on the clinical team at Deep 6 AI to build queries and then involving the OSU study team during validation. His experience has shown that the implementation of AI won’t work unless experts on the study team are actively engaged and helping iterate the queries by validating the patients to make sure that the queries are working. 

“I’ve been a nurse for a long time, and I’ve been involved in research for a long time. Initially, I thought that my small team of about five people could build the queries. But one of our directors in our clinical trials office was like, ‘do you realize we have about 400 active trials and this is nearly an impossible task.’ So, we have partnered with Deep 6 AI to build them. To date, they’ve helped us build almost 300 trials. It’s not something that they do in a vacuum. They have clinical folks that work with me or they work with our disease teams to iterate on those queries.”
– Robb Stillman, MA, RN, CPHIMS, FHIMSS

AI for Clinical Trials at Valley Health System  

The Valley Hospital is a fully accredited acute care non-for-profit hospital that’s located in Bergen County, New Jersey. It serves nearly half a million people in 32 adjacent cities in Bergen County. Dr. Suneet Mittal is the Chair of the Cardiovascular Service Line for Valley Health System, and Director of Cardiac Research for Valley Health System.

Since deploying the Deep 6 AI platform in 2021, Valley Health System has built 56 unique trials in the software to recruit patients faster and improve diversity. Dr. Mittal believes that AI helps you identify patients very quickly, especially for trials that have complex inclusion and exclusion criteria that require looking at the unstructured data to find the right patients. He also has seen that it saves his staff time because the team is screening from a narrower, more precise patient list. 

“It’s a lot easier to find a patient out of 10 charts that have been AI-matched as opposed to going through a hundred or a thousand charts to find those 10 patients that may be a potential fit for your study.” 
– Suneet Mittal, MD

According to Dr. Mittal, the team is already seeing the impact AI can have. For example, Deep 6 AI was used to recruit chronic heart failure patients with reduced ejection fraction for a trial for a drug called Vericiguat. The study team at Valley Health System had been recruiting patients for a year prior to using AI without success. They had hoped to be able to recruit two to three patients per month but ended up finding only six patients that first year (and five of them failed screening because of the trial’s very complex inclusion and exclusion criteria). The trial sponsor ran a social media campaign, but it didn’t help. Within two months of turning the trial over to Deep 6 AI, they found 126 patients that were potentially eligible for the study. Additionally, they found 12 African American patients that were eligible, of which two-thirds were validated by clinicians after evidence review. The research team estimated that the AI software saved them up to an hour of screening time per patient. 

At Valley Health System, the team is also using AI to identify candidates for devices and flag patients for follow up and screening appointments. For example, they have used AI to find patients who are eligible for MitraClip™ so they could have conversations with the patient and the referring providers regarding their eligibility. Valley Health System is also using an algorithm developed by Graticule in partnership with Deep 6 AI to identify and prioritize candidates for a life-saving heart device trial. Additionally, it has been helping them ensure patients with pulmonary nodules found on their CT scans of the chest and heart don’t fall through the cracks.

“AI allows us to identify those patients who may have these incidental findings that require follow-up and create an infrastructure that allows us to better manage these patients going forward.”
– Suneet Mittal, MD

Selecting the Right Trials for AI-Powered Recruitment 

Dr. Mittal commented on his approach to selecting the right trials for AI-powered recruitment: “We’re early into our experience…and right now we base it on the priority of the study to our institution and how well-suited we think a trial is to be able to sort through inclusion and exclusion criteria based on the information that’s available readily within the EMR system. And over time we would love to get to the point where it’s synonymous with every trial, but we have a little bit more learning to do before we get there.”

The Future of AI at Valley Health System 

In the future, Dr. Mittal believes we’ll see more AI-driven recruitment success stories. He compared today’s AI revolution to the move from local IRBs to central IRBs. 

“…there was a day that everyone had a local IRB and then they went to centralized IRBs and they saw the value in doing that. And there’s going to be no doubt in my mind that partners will see this as the next evolution of how to get the right patient enrolled in clinical trials.”
– Suneet Mittal, MD

To hear more from Robb Stillman and Dr. Suneet Mittal, check out this webinar on ‘Accelerating Patient Recruitment with Artificial Intelligence,’ hosted by the Greater Houston Area Association of Clinical Research Professionals. Learn more here about how healthcare organizations are using AI to recruit patients faster.

About The Author

Nour Malki