Natural Language Processing For Enhanced EMR Data [transcript] [audio]
Guest: David Fusari
Presenter: Neal Howard
Guest Bio: David Fusari has held multiple technology leadership positions in healthcare IT companies, including ABILITY Network, Microsoft and Sentillion. He is an inventor on 19 patents and a member of nationally recognized clinical informatics organizations. David earned his BSC in computer information systems from Bryant College.
Segment Overview: In this health supplier segment, David Fusari, CTO and Co-founder of TriNetX returns to talk about their new service that combines Electronic Medical Record (EMR) and Natural Language Processing (NLP) data to deliver the industry’s most comprehensive data set for analysis and making clinical trial design more efficient and thus, much more time and cost-effective.
Transcript
Neal Howard: Welcome to the program. I’m your host Neal Howard here on Health Professional Radio. Glad that you could join us. Our guest today is returning to speak with us Mr. David Fusari. He’s CTO and co-Founder of TriNetX and he’s returning to talk with us about their brand-new service that combines EMR or Electronic Medical Records with Natural Language Processing and they’re doing that in order to deliver some very efficient and cost-effective clinical trial design. Welcome back to the program David. How are you?
David Fusari: Great. Thanks for being here again, Neal.
N: For our listeners who may not be familiar with you from when you were here on other segments, give us just a brief background about yourself.
D: Sure. So I’ve been in the healthcare technology space for 20-plus years and being very technology centered and healthcare centered, I’ve always strived to make sure that we find ways to advance the delivery of healthcare within our ecosystem both here and now globally. With TriNetX, our plan has been to develop a global oriented research network to really help organizations optimize clinical trials so that we can bring new therapies to market more quickly, so that we can help those people who need those therapies especially around rare diseases and other kinds of opportunities where the science is advancing so quickly that we get those kinds of opportunities and those kinds of treatments in place for patients who need them.
N: So to that end, what is TriNetX doing? I mean you’ve been in a business for a while, what exactly are you involved in other than this EMR and NLP combination?
D: So let me give you the foundation of what TriNetX has been all about since it started. It started just about three and a half years ago and what we’ve been focused on is helping healthcare organizations and large pharmaceutical companies and sponsoring organizations to better optimize their clinical trial design so that they can ensure that when they’re looking for patients that fit their clinical profile that those patients quote actually exists. You’ll see that in the healthcare industry. Many, many clinical trials tend to fail or under recruit or take significantly longer to recruit a patient and thus elongating the life cycle and increasing the cost it takes to bring new therapies to market. The ability to better optimize your trial, given the fact the trials are more complex these days and that requires more clinical sites to try to organize to execute a clinical trial are just areas that we are helping to optimize and we’ve been doing that with the creation of this research network where we’ve recruited 67 healthcare organizations so far which represents hundreds and hundreds of hospitals and facilities to work with patient, de-identified patient data in a way that allows sponsoring organizations to better target patients for a clinical trial.
N: Now this network, this global network that you’re speaking of, it entails biopharmaceutical companies, all sorts not just hospitals, correct?
D: Correct. So the data is predominantly coming from hospitals but our users are global users representing biopharmaceutical companies that are international in nature along with our healthcare, our hospitals members which are really the data asset into our platform are international as well. So we have contracts in Israel, and Singapore, and Germany, and Malaysia, Hungary, UK, Italy, so throughout Europe and obviously the U.S.
N: When it comes to designing these trials, what is the problem with getting most or all of your information from the electronic medical records?
D: Great question Neal. What we find is that the electronic medical record is a great start to identifying patients but electronic medical records don’t always contain the depth of a clinical understanding in a clinical context of individual patients. There is still a huge swath of healthcare information about patients that is locked away in some kind of textual documents. And because it’s locked away in textual documents, it doesn’t make it easy to analyze or understand to identify patients for a clinical trial. So when we think about certain elements, think about a cardiology trial, one of the primary elements that people tend to look for is what is the level of ejection fractions usually the left ventricular ejection fraction or LDEF so that people understand what is the flow of blood going through that patients heart. They’re having a blockage or they’re having an issue. And when they’re trying to identify patients for a trial, for a cardiac trial that can be one of the key elements that people are looking for. That is not typically a data element structured in an EMR. It’s sitting in a cardiology report somewhere. In addition if we think about pathology reports for cancer patients, there’s a lot of content that’s sitting in a pathology report that’s not regularly structured content in an EMR, that is extremely beneficial for identifying the right patients for a trial. That can include any kind of genomic variance about the patient’s cancer. It can include staging and morphology along with treatment patterns that the patient having ongoing to see if they’re a right fit for a clinical trial. So that additional content really helps fine-tune and brings together this structured data that we call the EMR
data because it’s already coded and structured in a sort of a database friendly way along with this, what we call “unstructured data” and having a way to map those two worlds together gives us the opportunity to get a much deeper understanding of our patient population that we’re looking to study.
N: Now you have mapped those two together successfully, yes?
D: Correct. We started earlier this year with a data program working with three of our hospital organizations to integrate that data into our platform, mine through that content and make that data available through our TriNetX solution. We’ve now been rolling that out and there are organ healthcare we’re working with our healthcare organizations now get this broadly deployed.
N: Now Natural Language Processing, it sounds like you’re translating different languages. Medical terminology can’t be considered a different language especially if a lot of the information isn’t readily available. How exactly does natural language processing work to integrate itself with EMR?
D: We’ve been partnered with a company called the “AVerbis”. They’re based out of Germany founded by a physician who built some of the core technology, developed a core technology engine and how do you process documents and more specifically, how do you process clinical documents and try to understand so clinical ideas and notions within those documents. And it’s through that Averbis core technology that we integrate and build the pipeline into our platform that allows us to look at the text and the content of these documents to pull out the meaningful information that’s important. So for example how do you know something is a lab and how do you know something is a diagnosis based on the text of the document? And they have developed and fine-tuned that technology and being able to map that technology so that you look at clinical terminology within the document. In addition, it’s not just finding the clinical terminology you also have to put a context around it. Then the context around any clinical terminology you find in a document you want to know what was that related to the patient or were they documenting that relative to family history? What does it actually say the patient have this diagnosis or does the document actually say, “Oh there’s actually no presence of this cancer versus a presence of the cancer?”. So you have the what we call “negated terms”. So all of that intelligence has to be brought in to the processing of a document so that you can then make meaningful use out of it and that’s Averbis helps us do.
N: You mentioned meaningful use it also goes a long way toward physicians accomplishing meaningful use as a requirement in their daily operations once the information is there readily accurately and efficiently.
D: Absolutely you can. When I was mentioning meaningful use it was not necessarily in the Affordable Care Act context of meaningful use, but I meant making it usable, meaningful usability. But you are correct to that being able to better drive and understand a patient, their ailments, so when you’re dealing with care management settings where you have to provide specific care around a certain patient population because there are high risk. Now you have more data and information to better identify those patients that truly are at risk.
N: Where can we go online and learn more about TriNetX?
D: Sure. You can go to our website, www.trinetx.com and through our website you can get to our NLP pages or you can just hit /NLP and you’ll learn a little bit more about that or learn more about TriNetX in general.
N: Great. David, it’s always a pleasure.
D: Good talking with you again.
N: You as well. You’ve been listening to Health Professional Radio for this Health Supplier Segment. I’m your host Neal Howard. Transcripts and audio of the program are available at hpr.fm and also at healthprofessionalradio.com.au. Be sure and visit our affiliates page when you visit our platform at hpr.fm.
0 Comments