Natural language processing (NLP) has become a buzzword in healthcare. Organizations across the industry have moved into the space, and, for the most part, we support these moves. As an organization, we believe in the value of using artificial intelligence to support human experts in boosting the administrative stability and quality of care. Healthcare is, after all, ultimately about helping people. That said, there are a lot of misconceptions and misunderstandings about what NLP actually is and how it works. Today, we are going to help clear up the confusion and offer a deep dive with part one of two in our latest ebook series.
How NLP Works
Natural language processing (NLP) is one segment of the broader field of artificial intelligence. Even in its most advanced state, artificial intelligence is a series of algorithms that are layered in a specific way to accomplish a given task, typically assimilating and organizing enormous data sets that have traditionally been difficult-to-impossible for a machine to interpret. Artificial intelligence, more than anything, makes enormous volumes of data accessible to human experts, so those experts can then draw conclusions. It’s a complex topic as an area of research with its greatest potential still ahead of it despite decades of brilliant people working to unlock what it takes to teach a machine to understand medical documentation.
How Natural Language Processing in Healthcare is Used
In simple terms, NLP is a branch of artificial intelligence capable of parsing and understanding unstructured data to enable an additional layer of logic that can extract the semantics of the extracted data. For example, an EHR contains form fields with patient demographic data broken out line by line, name, age, ICD-10 codes, etc. Traditional software integrated into that EHR can act upon those designated data points because a name is a name; it’s a 1:1 match. The specific nature of the data entry method determines the usability by other software. NLP can read, comprehend, and even recommend actions based on information from unstructured narratives in clinical documentation. Physicians’ notes, pharmacy orders, administrative data; all of it can be digested by the NLP without the necessity of 1:1 form fields. This is because NLP can, as the name implies, process natural language.
How Health Fidelity Supports Your Goals with NLP
Health Fidelity’s Lumanent Insights NLP engine is both typical and atypical. It is typical in the sense that most applied NLP engines in healthcare are attempting to parse unstructured data and pass their findings onto a second layer of functionality, post processing logic, also known as the inference engine. The inference engine makes relevant, prioritized, verifiable suggestions to provider or payer staff in a number of different ways, depending on the recipient. Coders, for example, in both payer settings and in post-encounter deployments for providers receive direct suggestion based on care already provided. At the pre-encounter and point-of-care levels, Lumanent Insights offers suspected conditions based on the longitudinal clinical data. This covers conditions currently not being addressed but possibly present and untreated or at least under-documented.
What sets Health Fidelity’s NLP apart is its maturity, clinical specificity, and its effectiveness. The NLP engine within Lumanent Insights has been evolving under the guidance and effort of bioinformatics professor Dr. Carol Friedman at Columbia University since the mid 1990s. It has exclusively developed as a clinical engine, having processed nearly 300 million clinical records, each contributing to its refinement. While technological obsolescence has taught us to be concerned about older applications continuing to function, in the field of AI, the maturity and volume represent an insurmountable lead in development.
Its entire purpose is to support our clients and partners in their ongoing efforts by helping do what was previously impossible due to the sheer volume of data. Whether by serving a dramatic productivity multiplier for coding teams, by delivering clinical support before the encounter to help providers address patient risk factors, or by providing audit-risk and second pass opportunity recognition at the payer level, Lumanent is there.
It’s working, too. For example, in the first six months of the partnership alone, Change Healthcare processed 1.8 million encounters via over 300,000 separate patient records through our NLP recognizing a significant ROI and productivity lift.
Diving Deeper on NLP and Healthcare
This is a major inflection point, a “golden moment” as our CCO Robin Lloyd would say, between data gathering technology and actionable tech. Without NLP, it’s impossible to effectively staff to comb through the total volume of medical data and accurately extract actionable insights on anything from quality measures to risk adjustment.
What we have covered here is only scratching the surface of what NLP really does. If you’d like to know more, we have a new ebook going into much greater detail into how it works, what constitutes quality in the NLP engine, and how providers and payers alike benefit from this technology. To view our ebook, Natural Language Processing in Healthcare: Characteristics, Strengths, and Misconceptions, please go here.
