As healthcare organizations look to leverage artificial intelligence (AI) in everything from billing, to clinical documentation, to patient care, one AI technology — natural language processing (NLP) — is already an important tool for clinicians and administrators alike. Here’s a primer on NLP and the value it brings to healthcare and health IT.
Table of Contents
What is natural language processing?
Natural language processing is a rapidly evolving branch of artificial intelligence that involves giving computers the capacity to understand spoken and written language. It’s “natural” because it doesn’t require humans to change how they communicate.
What are some everyday examples of NLP?
Creating and sending a text via voice, interacting with customer service through an online chatbot, or telling your smart speaker to play a favorite song are some common examples of NLP in action.
How is NLP used in healthcare?
When it comes to patient care, one of the most important applications for NLP may be in clinical documentation. Using voice recognition software allows a clinician to use voice transcription to record clinical details and notes in an electronic health record (EHR) and then immediately review the updated patient chart in written form on the screen.
Other uses of NLP in healthcare include digital assistants on provider websites that streamline communications with online visitors, along with programs that sift through reams of patient data to rapidly identify key information for use in clinical decision making at the point of care.
Why is NLP valuable to providers?
Since speaking is often faster than typing, using NLP technology speeds up the EHR workflow – allowing providers to spend less time interacting with the EHR interface and more time focused on direct patient care. With physician burnout driving many to leave medicine for other fields, NLP can be a useful tool to help reduce the clinician HIT burden.
How does NLP work?
All NLP applications work in roughly the same way: first, spoken or written language is recorded with a microphone or word-processing program, then that information is transformed into code a computer can understand.
It’s this processing of real-world input from a human being – with all its linguistical mistakes and variations – where artificial intelligence comes into play. Once the text data is prepared for analysis, many NLP systems use machine learning – a specific type of AI – to improve their algorithms, which rely on pre-programmed linguistic rules to break text down, determine what it means, and present it as actionable data, or words.
How does NLP relate to machine learning?
Many NLP systems use machine learning to improve their algorithms over time. Simply put, the more data these systems are fed, the better they “learn” to process future input the way the speaker or writer intended.
Why is it hard to implement NLP in healthcare?
The biggest challenge impeding adoption of NLP in the clinical setting has to do with the varied vocabulary of healthcare. Every specialty has its own lengthy list of disorders, diagnoses, treatments, and medications. On top of that, the extensive use of acronyms and abbreviations to document these clinical findings adds yet another level of complexity.
In many cases, despite these differences, NLP algorithms can determine the correct context and meaning of what was said. Sometimes, however — just like human interpreters — these tools can be prone to making mistakes.
This leads to a secondary hurdle preventing widespread NLP adoption in healthcare: provider mistrust of the technology. Some physicians worry, for example, that their dictated notes might be garbled by the software, which in turn could create other problems down the road when errors are not corrected. Most clinicians would gladly use NLP to reduce time spent on documentation, but only if they feel confident that those records will be accurate — and not potentially clinically misleading.
Looking ahead, one thing that’s certain is NLP is improving by the day. The technology may be a work in progress, but in healthcare, it’s here to stay.
