Here at pMD, our passion is applying technology to improve health care delivery and billing. Natural language processing, often abbreviated as NLP, is a class of technology with enormous potential for integrating into all aspects of health care. We’ll discuss the current state of NLP, its potential use in health care, and the challenges that lie ahead.
To understand Natural Language Processing (NLP), we must first understand the definition of natural language. Put simply, natural language is anything that people use to communicate with each other. This includes spoken language as well as written forms of language such as letters, emails, and text messages.
Natural language processing is then exactly what it sounds like: processing natural language for useful insights. In other words, NLP can be thought of as a pipeline. The beginning of the pipeline starts with the unprocessed natural language, and the product of the pipeline is the output of useful information for interpretation or analysis. The difficulty of NLP lies in the middle stage of this pipeline: engineering algorithms that are capable of processing the nuance of natural language into useful insights. As this is such a broad definition, everything from the humble autocorrect to sophisticated products such as Siri and Google Translate can be categorized as NLP.
Current applications of NLP to health care remain limited but promising. One type of NLP technology is speech-to-text. Providers speak into a dedicated device or mobile app, which records their voice. Then, either a human scribe or an automated voice recognition algorithm will transcribe their spoken words into text. This allows them to “write” documentation or messages much faster than typing. A natural evolution of this technology is the digital scribe, a program that not only records what the provider is saying but also analyzes entire provider-patient encounters to generate a condensed report.
On the documentation and billing side, NLP has found applications as well. One major area of application is EHR verification. Algorithms can scan through clinical notes and attempt to determine whether the provider’s documentation is sufficiently detailed. On the billing side, computer-assisted coding can perform a similar function by scanning documentation for billing-related information. These programs can help billers faster parse through long notes and suggest potentially billable codes.
How do these applications benefit health care? Firstly, they have the potential to provide a better patient experience by freeing the provider from staring at a screen. A fully operational digital scribe would handle documentation, allowing the provider to focus on the patient. Additionally, NLP products can streamline practice workflows, whether on the clinical or billing side. EHR verification can automatically flag missing information for correction, while computer-assisted coding can accelerate the billing process.
With much of NLP, the challenge lies in finding ways to analyze the nuances of natural language with algorithms. Although great strides have been made in improving machine understanding of natural language, there are still numerous problems with reliability as well as concerns about cost and interoperability. Ironically, many of the tasks described in this post would be relatively simple, albeit tedious, for somebody with medical knowledge. With NLP, the hope is that technology can offload the burden of these tedious tasks and allow clinicians, billers, staff, and patients to focus on health care.