“Don’t take your eye off the electronic health information ball.” That was the advice of Micky Tripathi, National Coordinator for Health IT under the Biden administration, when asked about the number one piece of advice[1] he would give to payers to prepare for the Office of the National Coordinator’s (ONC) Cures Act Final Rule,[2] which includes a provision requiring that patients are free to access all of their electronic health information, structured and/or unstructured, at no cost by October 2022.
Compliance with the Final Rule allows for the structured data requirement to be addressed first, but Tripathi’s statement sent a clear message that stakeholders would be wise to take a long view on electronic health information. Once they have a better grasp on standardized patient data definitions over the next 18 months, they will quickly be asked to move on to unstructured data.
“Although 18 months might seem like a long time, it’s vital for healthcare stakeholders to begin that complicated learning process right away,” Tripathi recommended.
“Once that 18 months is over, it is everything. It's text notes, transcriptions, and other kinds of documents. The only way that we'll be able to get our arms around that is using algorithms, machine learning, and other kinds of approaches, such as natural language processing, to be able to take advantage of on behalf of the patient, on behalf of better quality, to be able to take advantage of that broader, comprehensive information that's available.”
And just like that, Tripathi shifted natural language processing (NLP), from a nice-to-have solution to an essential asset in 2022 and beyond. If the urgency to adopt NLP feels overwhelming to you, don’t worry. Today, I’ll walk you through what NLP is and why the Final Rule necessitating NLP is actually a tremendous opportunity for payers. This is part one of a two-part series on NLP and payers; the second part of the series will guide payers in successfully embedding market validated NLP into their existing workflows to drive efficiency and rich insights.
The great opportunity of unexplored, unstructured data
Let’s start by exploring what is it about the unstructured data that has caused it to be specifically called out by Trepathi. Think about the massive volumes of member data you currently process from a variety of sources that are both structured (meaning it is clearly defined and searchable) and unstructured (meaning it is qualitative, in its native format). This includes everything from claims to admitting notes and more. While not all member data payers are processing is directly from health systems, a lot of it is, and health systems generate a huge amount of information. In fact, a single healthcare system or hospital generates an average of 50 petabytes of data per year, and 97 percent of it is never used after its initial creation.[3]. Per the Final Rule, a large portion of this previously ignored data must be accessible by patients. While that is great news for patients, it also offers significant opportunity for payers.
That’s because these unexplored, unstructured data often hold the richest member insights; they span a multitude of application or business areas, from gaps in care to risk stratification to quality measures, and unlocked, they could be used to reimagine and optimize many payers’ processes. We’ve long known that the information contained in this unstructured content is valuable, but in the past, it simply wasn’t feasible to attain. Imagine the manual labor required to sift and sort through petabytes upon petabytes of narratives and try to extract useful insights. It was unsustainable — until recent advances in NLP technology that have made it possible for computers to read, normalize, and understand human language with incredible speed and precision.
How NLP can crystallize your patient view
To understand the utility of NLP for payers, let’s take, as one example, an admitting note from a hospital. In the structured data, we’d find things like the diagnosis, medications, and potentially some lab values. That gives us a helpful, but flat, one-dimensional view of our patient with a few useful facts and little else. It’s like trying to build a jigsaw, but more than half the pieces are missing. When we apply NLP over that same admitting note, we can suddenly see our patient members with clarity.
Layering on insights from written notes and other text, we can now see symptoms, lifestyle factors, and medications that provide a real sense of who this patient is. As payers know, nuance is so important: one patient with diabetes that's well controlled and asymptomatic is very different from another patient with diabetes, poorly managed blood sugars, renal insufficiency and mobility difficulties. Using traditional structured search methods, these critical differences are often missed. With NLP, the subpopulations within different diseases are distinctly stratified.
I should also note that NLP does not just surface these details, it actually extracts and normalizes them to a terminology or an ontology of choice. For instance, in the context of risk adjustment, NLP can normalize the information it extracts to HCC codes, HEDIS measures, and more. The implications for payers here are clear. Applied at scale, payers can use NLP to unlock predictive models, identify high-risk members with tremendous precision, reduce manual chart review and streamline business audit processes that require extensive medical review.
Planning for the future
To really extract the most out of NLP, payers may want to consider making the investment in an enterprise-wide approach that can address the deluge of data that is not just coming now, but also in the future. Healthcare data is exploding and will only continue to grow exponentially. With a single patient generating nearly 80 megabytes of data each year in imaging and EMR data alone, RBC Capital Market projects[4] that by 2025, the compound annual growth rate of data for healthcare will reach 36 percent. This growth rate is notably faster than the rates projected for many other massive industries, including manufacturing, financial services and media and entertainment. While adopting an enterprise strategy for unstructured data can feel like a daunting task, with modern NLP solutions that are flexible in use case and deployment model, it is practically possible, and there is a clear and strong business case for doing so. From forecasting shifts in membership to understanding call center trends and prescription choices, enterprise NLP can be a meaningful differentiator.
For any organization that was taking a wait-and-see approach to adopting NLP, now is the time to make the shift. That’s why at Linguamatics, the customer ask has quite rapidly gone from, “Do I need NLP?” to “How do implement NLP?” With that in mind, the second blog in this series will serve as a guide to NLP for payers, giving you potential application areas as well as use cases of how organizations are using NLP to transform and optimize their decision-making. Be on the lookout for that in the coming weeks.
If you are interested in what NLP can do for your organization, reach out to Linguamatics today for a demo or watch our webinar to learn how payers are leveraging NLP for competitive advantage.
References:
[1] https://ehrintelligence.com/news/onc-leader-tripathi-offers-tips-for-interoperability..
[2] https://www.healthit.gov/curesrule/
[3] https://www.weforum.org/agenda/2019/12/four-ways-data-is-improving-healthcare
[4] https://www.rbccm.com/en/gib/healthcare/episode/the_healthcare_data_explosion
