Pauline Anderson

December 07, 2016

HOUSTON — Bioinformatics, the use of computers to help drive medical decisions, is changing the face of epilepsy care, an expert says.

Tracy A. Glauser, MD, director, Comprehensive Epilepsy Center, co-director Genetic Pharmacology Service, Cincinnati Children's Hospital, and professor, University of Cincinnati Department of Pediatrics, Ohio, explored this idea during the Presidential Symposium at the American Epilepsy Society (AES) 2016 Annual Meeting.

"We embed students with lots of knowledge in medical school, turn them into information wizards, and then we want them to gain knowledge and wisdom over the decades," Dr Glauser told his audience. "We are trying to do that with computers."

While professional judgement will always be vital to shaping patient care, medicine has reached the point where the amount of information needed to make decisions "is moving beyond unassisted human capacity," he said. "So we need to accept that we have to partner with the bioinformatics experts to drive better care."

Bioinformatics spans a very large spectrum of data — for example, imaging, clinical informatics, technology and public health informatics — and uses algorithms and data structures to turn information into knowledge.

Institutionalized Silos

Electronic medical records make up one part of the bioinformatics framework. Most of the US population now has an electronic medical record, but one of the problems has been that each healthcare institution has its own personalized system.

"This has created institutional silos of data," making it "tough" to do clinical research, said Dr Glauser.

To meet that challenge, experts have developed computer tools to gather information from health records, integrate these in a simple format, and identify cohorts for clinical research.

Dr Glauser talked about the National Patient-Centered Clinical Research Network (PCORnet), the mission of which is to provide better data and create an infrastructure across institutions. There are 20 patient-powered research networks, including the rare epilepsy network.

"We will see more of these patient-focused innovations that gather the biomedical information directly from the patients and link that all together," said Dr Glauser.

He also talked about "wearable tech" products that predict and detect seizures, games that educate while maintaining cognition, and virtual reality tools.

Dr Glauser described the bioinformatics steps he and his colleagues have taken at Cincinnati Children's Hospital.

"The first step on this path was to create an infrastructure where we had our computers and our machines learn about epilepsy, just like we teach our fellows and our residents."

The next step was to optimize this process. Researchers trained the computers to "grab" structured data, such as blood pressure, weight, and medical history, and put them with the unstructured data, such as doctors' notes.

Computers learned to use physician notes to identify candidates for epilepsy surgery or for clinical trials in a timely manner, said Dr Glauser.

Dr Glauser and his colleague are now focusing on epilepsy surgery. "We all know how hard it is to get our colleagues to send us the patients who are good candidates for epilepsy surgery."

An internal study at Cincinnati found that it took 6 years to get a referral.

Researchers taught the computer to evaluate words, tone, sentence construction, and other aspects of physician notes to identify patients who could be referred for surgery.

Using computers, they determined that the doctor's language started to change for patients who were going to be intractable vs those who were not going to be intractable.

"We found that it was at about the 4-year mark of therapy that the language started to diverge, but it took another 2 years before physicians sent their patients for epilepsy surgery evaluation."

Now, the system sends a message to physicians alerting them that a particular patient is a potential epilepsy surgery candidate.

Comorbidities

One of the most important roles of bioinformatics is to pick up comorbidities among patients with epilepsy. Dr Glauser used as an example the case of a 9-year-old female patient enrolled in a double-blind medication trial who became seizure free. However, she experienced sadness and depression and ended up taking her own life.

"We realized that what we thought was a benign epilepsy was not so benign," he said. So he and his colleagues taught the computer to identify increased suicidality.

What patients say, how they say it, and the body language and facial expressions they use provide important clues about suicidality. For example, people who are suicidal have longer pauses in their speech compared with other people and different vowel spacing.

The computer is now able to spot these differences to identify suicidal patients.

In the future, Dr Glauser envisions that clinics will have cameras and listening devices, screens, and monitors that issue information and flags about incoming patients.

"Over the last 20 or 30 years, we have done a very nice job of taking data and putting it into the form of what's called 'information,'" he said. "Our goal over the next 10 to 15 years should be to take that information and drive it into knowledge…and eventually drive it into wisdom."

Asked his views on the bioinformatics presentation, Daniel Goldenholz, MD, PhD, advanced clinical fellow, National Institutes of Health, Bethesda, Maryland, said it was "wonderful."

"In my personal view, and this is still kind of fringe, we are going to realize that doctor decision-making is very flawed, and needs to take a back seat to computational algorithms that look not at one patient but at populations."

When looking at the entire population, experts can begin to discover patterns, said Dr Goldenholz. "You can then use the knowledge you gain from that and feed that into an individual patient."

Doctors, he said, are "excellent collectors of information," are "excellent explainers" of what needs to be done and why, and are good at providing support. "But we are going to be a whole lot less involved in deciding that this drug or that drug is right for you because we're not that good at that."

Michael R. Sperling, MD, Baldwin Keyes Professor of Neurology and director, Comprehensive Epilepsy Center, Thomas Jefferson University, Philadelphia, Pennsylvania, was somewhat more skeptical about the role of bioinformatics in epilepsy.

An experienced physician "knows what to ignore and what not to ignore from the data," he told Medscape Medical News. "There's the nonverbal communication that a computer is never going to capture."

Doctors have to consider a large amount of information and determine what to use and what to "throw out," added Dr Sperling. "I don't see how computers are going to figure out so easily what to throw out."

And when doctors make a mistake, they can go back and ask themselves where they went wrong. "How do you build that self-questioning, that self-doubt, into computers?"

As for alerting physicians that a patient may be eligible for surgery, that's not telling them anything they don't already know, said Dr Sperling.

"They just don't want to give up" trying new therapies and generally don't want to "acknowledge failure," he said.

American Epilepsy Society (AES) 2016 Annual Meeting. Presidential Symposium: "Harnessing the Power of Bioinformatics in Epilepsy." Presented December 3, 2016.

For more Medscape Neurology news, join us on Facebook and Twitter

Comments

3090D553-9492-4563-8681-AD288FA52ACE
Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as:

processing....