AI software accelerates identification of patients for clinical trials

A Los Angeles-area start-up called Deep 6 AI is shaking up the clinical trials marketplace by applying a new brand of AI to the mix.

Company founder and CEO Wout Brusselaers explained in a meeting at HIMSS, in Orlando, that traditionally, drug companies have found it very hard to recruit patients for the clinical trials needed to test their new medications.

They often work with hospitals and contract research organizations, which comb through records to find the right patients for the right trial.

Problem is, it’s a laborious process and it can take a long time to find even just one patient who meets the criteria of the trial.

Overall, drug industry stats show that pharma manufacturers were hoping to recruit 6.7 million patients in the United States alone for clinical trials in 2016. However, they were only able to find 2.2 million, a shortfall of 4.5 million.

“About 9 out of 10 trials are slowed down by this, as a result,” said Brusselaers.

Enter Deep 6 AI, which has automated most of the process of qualifying patients.

Its first partner was Cedars-Sinai Medical Center, in Los Angeles. “In one instance, using traditional methods, research staff for a cardio study at Cedars-Sinai found and recruited just two patients in six months,” said Brusselaers. “Using Deep AI 6, they identified 16 patients in about half an hour. Three weeks later, eight of them had been recruited onto the study.”

These productivity gains allow small departments to drastically increase the number of studies they can successfully perform. Again at Cedars-Sinai, a small team of three recruiters is now finding patients for 30 clinical trials in a year. “Before Deep 6, they had only one research coordinator, who did only one study a year.”

The software is making such a difference, that the hospital is hiring more recruiters and launching more studies.

Not only does this benefit patients, but it’s bringing new revenues into the hospital.

“Most hospitals do clinical research as a loss-leader,” said Brusselaers. “It brings them prestige and helps attract talent and patients, but they often lose money doing it.” He said that using the Deep 6 AI software, they can find more patients for trial, “exponentially faster”, and earn money from the big pharma companies by partnering with them.

The software is proving to be a hit with large, research-oriented hospitals.

Deep 6 AI is now rolled out at 15 different hospital centres, and more are in the works, said Brusselaers.

This year, the company expanded to Canada, and hired Raj Sharma as Director, Clinical & Academic Partnerships. Sharma is a veteran of the health technology sector both as a clinician and a business leader, and has an understanding of the clinical and technological capacity of hospitals, universities and their research infrastructure.

His mission is to bring the Deep 6 AI software into the Canadian healthcare sector, to the benefit of patients and hospitals themselves, which stand to earn money from the trials. Raj Sharma can be reached at: raj@deep6.ai.

Overall, Deep 6 AI now employs 30 people. Brusselaers says the company is doubling in size each year, and will hit 60 employees later in 2019.

He explained that it was formed by some like-minded people who worked at different companies, while socializing over “Monty Python, Key & Peele and beers.” Brusselaers himself is originally from Belgium, and speaks English, French and Flemish.

For its part, Deep 6 AI has gained the support of Cedars-Sinai, where it was launched in the hospital tech accelerator, called the Cedars-Sinai Accelerator Powered by Techstars. The company has also participated in Stanford University’s StartX accelerator and TMCx in Houston.

The Deep 6 AI software is quite different than other solutions on the market. Brusselaers explained that the problem in recruiting patients is that the data needed from their health records is usually in an unstructured format, such as physician notes or pathology reports.

Normally, people are needed to manually read through the records and extract the needed data, which can take weeks, if not months.

“In EMRs, usually only 20 percent of structured data is machine-searchable,” said Brusselaers. “We make the entire record rapidly searchable.”

Using AI and natural language understanding (NLU), Deep 6 AI devised a way to convert the unstructured information into data points on a chart. “We take the structured and unstructured data and turn it into a graph. Each patient becomes a multi-dimensional vector, with many thousands of data points.”

These points include standard items like blood pressure, temperature, weight, but also more complicated factors, such as tumour histories and Gleason scores, genetic information and mutations, and even lifestyle information, such as smoking history.

According to the company, Deep 6 AI’s software’s use of graph analyses can also help identify patients with conditions not explicitly mentioned in medical records. As a result, the software finds more patients who better match trial criteria in a fraction of the time.

“This is very difficult for people to do, but it’s perfect for machine analysis,” said Brusselaers.

Deploying the Deep 6 software to a health system typically takes about 60 days or less. “Most of that time is spent understanding the system’s data structure, setting up the historical ingestion and QC’ing the data, with very limited hands-on time involvement from their staff,” said Brusselaers.