Holly Fletcher | The Tennessean
A Franklin-based tech firm is looking to unlock troves of health care data — an effort that’s won a key relationship with industry heavyweight HCA.
Digital Reasoning specializes in software that helps its clients to more easily parse troves of emails or other unwieldy receptacles of information — and without relying on the use of key words. Computers can use context to yield results similar to a person reading text could, only much faster.
The company, which counts among its clients the U.S. government and investment banks, is expanding its presence in the health care space, said Hal Andrews, Digital Reasoning’s president of the health care division.
Simply put, Digital Reasoning is teaching machines to look for the subtlety and familiarity that exists in conversation, shorthand and subtext, which humans can pick out.
For example, the software can understand the shorthand used to communicate between people familiar with a subject — perhaps, a nurse and doctor exchanging comments over email about a patient — or detect if a code is being used.
The software, which can be installed for each client and does not require sending the information back to Digital Reasoning, is an effort to declutter the amount of information health care workers are inundated with each day.
“To the extent we eliminate the noise around you and say, ‘Here are the things we know are interesting to you,’” then workers, from clinicians to administrators, can be freed up from repetitive, yet necessary, tasks, said Andrews.
In health care, the uses for technology are nearly unending.
Patients, for instance, could benefit from secondary diagnoses from a test such as a CT scan in an emergency room, perhaps performed because of an injury from a car accident. That test could result in a finding that wasn’t the visit’s impetus but requires follow-up. Yet, that result, even if, may get lost in the file.
Cognitive computing can flag the finding and highlight it for the best-suited person or team for follow-up using natural language processing that looks for context and information using smarter search parameters that reflect the nuance a person uses when reading, said Andrews.
Scouring roughly 33.000 CT scans over 13 months from emergency room visits in Nashville, HCA came up with more than 1,000 incidental findings. Natural language search was responsible for 55, said Dr. Jonathan Perlin, HCA’s president of clinical services and chief medical officer.
“That’s really exciting because what kind of disease do we want to see?” said Perlin. “Curable.”
Data is an inevitable byproduct of care, said Perlin. A single visit results in both diagnostic codes and clinician notes, which could prove useful to the patient’s care or to understanding the results of care.
Yet most of it is hidden in patient files, whether electronic or paper. An estimated 80 percent of health care data is unstructured, meaning an untold amount of information is unanalyzed and largely unknown.
Even the structured data — entered in codes or trackable formats — is a challenge to harness, said Perlin.
But the dormant information is tantalizing and important to all corners of the industry, ranging from diagnostic teams and operations managers to researchers.
“It opens up the opportunity for us to move from creating data to using data,” said Perlin.
The partnership with Digital Reasoning — HCA is both an investor and a customer — opens the possibility of being able to tap into reserves of data that exist, and will continue to multiply. According to Perlin, the relationship is “really a match made in heaven — or Middle Tennessee.”
“It’s our history, it’s our memory. If we could reliably reach into our memory and know what works best, then imagine what we could do,” said Perlin. “That’s fundamentally why we’re so excited about the era of cognitive computing.”
For HCA, which operates a vast network of hospitals, and others, assimilating data from multiple platforms is essential. Most of HCA’s facilities use Meditech software, but there are facilities that use Epic and Cerner, said Perlin.
The company’s data science efforts have resulted in a clinical data warehouse with information from facilities providing care in-patient or outpatient. The next generation of computing will not only help decode the information sitting in files but also blend multiple types of data for a more detailed picture of care both for the individual and in the system as a whole, said Perlin.
“Today it’s a brute force exercise,” said Perlin.