Is natural language processing the key to the future of MTSOs?

By Lynn Jusinski

The first thing Brenda Hurley, CMT, AHDI-F, noticed was "language" in the company's title. As director of industry relations and compliance for Medware Inc., Maitland, FL, she knows all about language, working for a medical transcription service organization (MTSO). So she headed up to the booth at the Toward an Electronic Patient Record (TEPR) conference and asked Language and Computing's Vice President Kyle Silvestro just what it is the company does.

The answer got her wheels spinning. She immediately phoned Kim Buchanan, CMT, AHDI-F, director of credentialing and education with the Association for Healthcare Documentation Integrity (AHDI) to see if the company's vice president could be added to the roster for AHDI's annual conference. The technology in question is natural language processing (NLP), and Language and Computing isn't the only company to offer it. No stranger to HIM, it's used in computer-assisted coding applications, clinical information extraction and semantic search.ÊThe software analyzes text documents, processes the results and provides the discrete data for population of EHR/HIM or billing systems. Hurley immediately saw a connection between NLP and medical transcription, a field with plenty of free text.

For years, transcribed documents have been decried as somewhat useless in an electronic world. The document appears as a blob of text that discrete data can't be extracted from. Using NLP, those documents may just garner more value in a computerized HIM world, according to Hurley.

A Good Match?

NLP allows for a computer to go through a blob of free text, and the engine then identifies all the clinically relevant data. Information can be pulled out of the free text and then presented to the clients of MTSOs.

For example, if a facility wanted to know which patients had open heart surgery and a history of smoking, the NLP system would be able to go through the trans-

cribed documents and pull out the number of patients. That data can then be used as a value add for the MTSO, and the provider will have the information handy to do what it wishes with it.

Hurley pointed out that a craving for this data extraction is what drives many providers to seek out EHRs. Transcribed documents are left out in the cold when it comes to data extraction, and with NLP, transcription services may be able to add value to what was just a blob of text. "We could use it for better patient care. And that's just amazing," Hurley said. "This is why the EHRs are out there being touted, because, really, what they're trying to do is improve patient care. If people aren't focused on that, they're focused on the wrong thing."

Along the lines of quality of care, Silvestro noted that an NLP system can also provide data back to MTSOs' clients about present on admission conditions that can lead to identification of misdiagnoses, such as congestive heart failure. If something isn't diagnosed properly, the provider could lose not only money, but could also end up mistreating a patient. In addition, NLP can be used by

MTSOs for automating the coding process. Also, with more discrete data about treatments, physicians could have a better idea of what treatments are working and what mistakes may have been made in the past.

Many Unknowns

The idea of using NLP at MTSOs is still a rather new one. Hurley noted that even though NLP is used in other areas of health care, there's no real established model for using the technology in transcription, so the cost factors are unknown. Also up in the air is how much a client would be willing to pay for an NLP service such as data extraction. MTSOs might not have a lot of up-front money to use to purchase an NLP system, which could be another uncertainty for the future of the technology in the industry. "There has to be some costs related to this," Hurley said. "Now what's it worth to a client? That's the unknown here."

Silvestro noted that after his presentation at the AHDI convention last month in Orlando, he has heard a lot of buzz in the industry about NLP. People asked what areas the technology could be utilized in, what skill set would be needed on the MTSO's part and how it would fit in to the MTSOs. That all depends on how NLP would be used, according to Silvestro. If an MTSO were to utilize NLP for auditing transcription, for example, where the NLP system checked for common mistakes made by MTs and then reported those to quality assurance, an expert in medical transcription, which an MTSO would already have on board, would be needed to populate a set of rules, and then an XML programmer would need to help fit the rules to the NLP system, meaning no large team of people would be needed to implement the technology.

Another question exists about what kinds of MTSOs will use NLP. At the moment,

Silvestro admitted that, because of lack of current infrastructure to deposit data into EHR systems within the industry, an NLP system might be out of reach for smaller MTSOs or independent contrac tors. Silvestro said that his company is trying to develop more NLP uses that are specific to medical transcription, and some of those uses might put NLP more in reach for smaller services.

What the Future Holds

Silvestro sees medical transcription as a perfect time in the documentation process to use NLP, and he also believes that it empowers MTs and MTSOs, because it can increase accuracy and add value to MTSOs. Silvestro also noted that no matter how good the NLP engine might be, if the documentation doesn't make sense or is inaccurate, it could affect what the NLP system outputs. "That's really the biggest challenge of NLP, is the documentation," Silvestro said. "If you have garbage in, we might be able to read it, but if a human can't read it, you're going to get garbage out, too."

Hurley noted that as someone who's been in the medical transcription industry for over three decades, she has always thought that the information put out by MTSOs is rich and important to patient care. MTSOs also often keep documents for quite some time, and she said that an MTSO keeping client documentation for the length of the contract wasn't uncommon, meaning that there would be a wealth of text documents to extract data from. She's still not sure of how NLP will play out in the industry, however. "I'd like to see it, but will I see it? I think it will demonstrate a new value to transcribed documents, and this is exciting for a person like me who's been in this field for, well, a long, long time," Hurley said.

Silvestro views NLP as an opportunity for MTSOs that shouldn't be passed up. He added that NLP will allow MTSOs to add to the number of services they offer to clients, and the more services offered, the more likely clients are to stay with them. If they don't offer this type of service, someone else downstream will, he said. "You're the ones who are creating the document," Silvestro said. "So if you're the creator, you might as well segment that data and actually provide more useful information back to the physician—not that that piece of paper isn't useful, just that it gets filed away, or it gets scanned [MTSOs] can add value along with coding. There's so much more that they can add."

Lynn Jusinski is an associate editor with ADVANCE.