Natural language processing is a sub-discipline of artificial intelligence – and one that can be of great use in healthcare, extracting clinical nuggets from all the free text in electronic health records and data warehouses.
Marty Elisco, CEO of Augintel, a healthcare NLP company, believes that NLP will be mainstream in 2023 for three reasons: the problems have been solved, the value has been proven, and the time is right.
Health Informatics News sat down with Elisco to get him to expand on these reasons and help healthcare CIOs and other healthcare IT leaders understand why 2023 might just be the year of NLP.
Q. One of the reasons you suggest more healthcare provider organizations will adopt natural language processing technology in 2023 is that tthere creases have been ironed out. Please talk about the defects that you say have been taken care of and how this will encourage adoption.
A. First, let’s level the definition of NLP. NLP refers to the branch of computer science that aims to give computers the ability to understand text and spoken words in the same way as human beings.
NLP can be applied in many contexts. It may refer to voice-to-text recognition. It can also be used for handwriting recognition. But in our segment, and in the context of this discussion, we use NLP for content intelligence – or information extraction – from the written word.
About five years ago, machine learning technology took a giant leap forward. It has become possible to cost-effectively train algorithms with massive amounts of data. This innovation enabled NLP for content intelligence – machine learning was beginning to be applied to massive amounts of narrative data to create NLP models capable of identifying key concepts described in text.
Over the past two years, as the cost of developing a model has come down, it has become economically feasible to develop industry-specific models.
For example, in the legal industry, NLP has been used for electronic discovery. Lawyers use NLP to extract documentation provided during the discovery phase to facilitate the consumption of relevant content. And there have been advancements more recently in the use of NLP in healthcare – behavioral health and health and social services, more specifically.
Initial content intelligence efforts in health and social services were typically custom projects intended to analyze point-in-time data rather than providing a tool that could be accessed daily. The expertise and effort required to “teach” the context of deep healthcare was too much for many and resulted in the project failing – or not getting started at all.
Over the past year or so, industry-specific solutions have become commercially available because the pilots to prove them have been completed. These pilots have benefited from the collaboration between data scientists and customers/users who have refined the language model for the needs of this industry.
Thus, the creases have been ironed out. The technology is mature and stable, innovative tech companies have built easily accessible mission-specific SaaS solutions with deep context, and customers are now reaping the rewards.
Q. You also say that the value of NLP has been proven. Please give some examples of NLP proving its value.
A. The return on investment made by organizations leveraging NLP has been achieved.
As an example, social workers in Allegheny County continued to find that so much rich information was buried in case notes and unstructured data. With information overload, it took so long for social workers to find relevant data.
They wanted to solve this challenge – the challenge of quickly accessing important data at the right time with the ultimate goal of helping improve services for the families and children they serve. They knew that being able to quickly and easily access better information would help paint the big picture of a case, without having to spend hours flipping through notes.
One social worker in particular claimed that the NLP platform alone saved her five hours a week in administrative tasks.
An NLP platform has also helped Allegheny County better understand the social determinants of health. Typically, it would take a careful review of the entire background to understand things like history of drug use or housing insecurity – two SDOH factors that have a significant impact on overall well-being. But with all the colors, details, and deeper descriptions living in unstructured data, an NLP tool allows case workers to see the warning signs in real time.
Needless to say, it’s extremely helpful for families when caseworkers can extract information like this from unstructured data earlier in the process.
Q. And finally, you say that with the year 2023, the time has come for NLP in healthcare. Please elaborate.
A. It’s no secret that staffing shortages and burnouts have proven to be a real challenge for healthcare organizations at all levels in recent years. According to a study published in Mayo Clinic Proceedings, the rate of clinician burnout among U.S. physicians increased significantly during the first two years of the COVID-19 pandemic after six years of decline.
Additionally, the study found that clinician burnout was 62.8% in 2021, compared to 38.2% in 2020. The trend is clear.
Additional research has shown that 64% of burnout is attributed to administrative burden, which certainly contributes to social worker pain points. With social workers in such high demand, attrition remains high.
Some organizations report 30% attrition per quarter. There is a loss of case knowledge that occurs with attrition and this loss has a direct impact on the bottom line. When new caregivers are assigned, they simply don’t have time to read entire charts, which can lead to disruptions in the continuum of care, especially in complex cases.
So you have overworked social workers and clinicians, spending too much time away from the people they care for, and they’ve had enough. Coupled with the impact on outcomes of the loss of case knowledge, it is clear that the status quo simply cannot continue if we are to maintain a reliable and functional health system.
At the same time, significant progress has been made in cost-effective machine learning tools, particularly NLP, which can alleviate some of this stress. Now is the time for healthcare providers to build on the tools available. Therefore, I think 2023 will be the year NLP takes off.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email the author: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.
#reasons #NLP #mainstream #healthcare