Looking broadly at the industry, how can NLP dramatically reshape outcomes for patients, providers and payers? What are some key uses in healthcare?
SyTrue’s NLP OS ™ is already providing value to stakeholders across the care continuum through a variety of critical real-life situations, including providing data for predictive analytics. Xerox, one of our clients, takes data created at the point of care and predicts whether a patient admitted at the Emergency Department will go home or will end up in intensive care. They’re creating real-time risk models to assess concurrent risk. Clients like vRad, the largest radiology company, create radiology indices to analyze and benchmark daily imaging outcomes. With the benchmarks, radiologists can see whether their reports are in line with similar-type benchmarked reports, or whether they’re under-diagnosing. It’s pretty clear, I think, that NLP improves patient safety and outcomes and helps practices become data driven.
NLP, as it’s evolved, now offers a unique level of accuracy in data capture. This means better diagnoses, better care decisions and sharper insights for any provider group.
Can you address popular misconceptions about NLP — what should be understood better so that NLP can have its proper niche?
I see this often: Many misunderstand what it takes to create a scalable, extensible, and accurate NLP platform that drives value for clients and stakeholders. They think it’s an academic black box (like long ago). But it demands a whole enterprise level solution with multiple user roles for it to succeed. Our competitors take 4 to 6 months to implement one solution per facility. SyTrue can implement thousands in that time. SyTrue has deep understanding of the pitfalls around free text, giving us a great advantage. We fully grasp the challenges our partners face when trying to roll out NLP solutions and we’ve built an NLP OS ™ to address these same challenges. When you’re looking for an NLP solution to unlock the context inside an unstructured text, you can’t use technology that remains a black box. You need an enterprise NLP OS ™ developed for users who need to create and manage information extraction accurately.
NLP brings both benefits and issues to the table. Data refinement and workflow come to mind. How do you tackle these and help clients tackle these issues?
We’re experienced in tackling tough data refining and workflow issues. SyTrue’s Smart Data Platform boosts your HIT initiatives in several ways. Say you want a solution to convert ICD-9 to ICD-10. That means you must be able to extract data accurately, easily and efficiently in real time. And you need a workflow solution that’s easily managed. With our NLP operating system, you gain semantic interoperability and a workflow solution all at once. You use your applications with our APIs, through SyTrue’s Smart Data Platform. So you easily power up the content and data layer of your application, and customize it to your needs.
Are there serious criticisms of NLP, or would you say it has shortcomings?
One key problem we’re tackling is language. There is no “Easy” button when you deal with language usage, especially variations in usage among the 700,000 US physicians practicing today. But if you’ve architected a system correctly, NLP will account for all shortcomings out of the box and include built in tools to solve other challenges in processing unstructured data. An NLP system should already know how to solve most or all challenges you’ll face when implementing the NLP solution. Most vendors have not addressed this key challenge. As a result, they face shortcomings, delays, cost overruns and unsuccessful outcomes. But if you’ve been in the business as long as I have, there’s rarely a question I’ve not faced earlier or that we’ve not accounted for earlier. So there’s a difference between having an NLP engine, which many have, and an NLP OS ™ operating system, which SyTrue has.