Sramana Mitra: How is your system architected to be able to take unstructured data to structured data? Can you talk a little bit about how you do that from a technical point of view?
John Harrison: We have a platform that handles the exchange of a lot of documents today. We handle well over a billion documents flowing through our platform every year. That’s a core fax infrastructure. In the interpretation of the data, we leveraged a variety of AI technologies to do a few things.
The first is document classification. It finds a structure for medical documentation classification. We will automatically classify documents against that. The next thing that we do is take a document and run it through OCR.
Then, we run two separate sets of analysis on the document; one is imaged-based and one text-based. The image-based analysis is looking for visual clues of the document – distance-based correlation of particular data elements, overall design of the documents, the concentration of text on the document, and the number of pages.
The second set of analysis is looking for textual clues about the data and its role in the information. We’re able to look for tags and names, but also for much more clever things like making interpretations from a text even if a piece of data isn’t directly referenced in the text.
In its very basic form, what we’re doing is running all the data through a variety of different machine learning models that analyze that document and its content from different perspectives and then disambiguate the data that’s returned from the different models to provide us the platform’s best guess to the question.
Sramana Mitra: Very interesting. What white spaces do you see out there that you would steer new entrepreneurs to start companies in?
John Harrison: If you think about our perspective of the market, we’re looking at the overall healthcare market from a vision of the challenges of interoperability of data and through the lens of the value that could be delivered to clinical analytics by being able to surface so much data that is locked up.
As I think about the white space through that biased lens, I see two things. At a macro level, there’s no doubt that healthcare is moving more and more towards patient-specific treatment programs where we’re tailoring the treatment delivered to patients much more to the patients themselves.
Immunotherapy is a great example of that where we’re customizing a treatment based on the unique sample. One of the white spaces that exists is the logistics and workflow for managing that environment. When you’re in a situation where you need to deliver a million different therapies from a million different suppliers, it creates a very complex logistics structure that needs to be tightly controlled and managed.
From what we’re seeing in the market today, we’re not seeing great progress. That’s one area where I see an interesting white space. That’s not a place where we’re going but something we, as humans, could benefit from enormously.
The second thing that I came up with and that I wanted to share was while a lot of work has been made on the development of decision support systems for physicians, I believe that there is an enormous amount of opportunity there that’s untapped. Today, we expect a physician to provide holistic, informed, and effective treatment for a patient in the 8 to 10 minutes that they have while a patient is sitting in the room with them.
There’s value that can be derived by finding better ways to deliver patient summaries that highlight the relevant data about the patient across massive amounts of data sources, and be able to surface the data that matters, and match the various values about that patient to all of the most current thinking about various treatment regimens that are available.
Sramana Mitra: I can’t wait for an AI doctor.
John Harrson: You and me both.
Sramana Mitra: Thank you for your time.
This segment is part 4 in the series : Thought Leaders in Healthcare IT: John Harrison, Chief Commercial Officer of Concord Technologies
1 2 3 4