Call us toll-free: 800-878-7828 — Monday - Friday — 8AM - 5PM EST
By Kyle Murphy, PhD for EHR Intelligence
Few dispute the value of the knowledge comprising physician notes. The problem for healthcare organizations and providers is being able to extract meaningful and useful data from that unstructured free text for the purposes of care coordination, coding, and reporting — to name a few. And it’s about more than transforming the unstructured free-text format of physician notes into machine-consumable state, which natural language processing (NLP) is proving capable of, but having an endgame in mind.
“It’s not hard to get to the stuff. What’s hard is to know what to do with it,” says Dan Riskin, MD, CEO of Health Fidelity, which was recently named the recipient of a National Science Foundation grant for research in big data analytics and data-driven healthcare.
“In our mind, the natural language processing is a critical base because you need to make the information machine consumable and you can’t do that much off of the 20 percent of the discrete data. You need to incorporate the 80 percent of unstructured data,” explains Riskin. “With that said, it’s only a base. We leverage this for the things we build, but we don’t ever pretend that the base itself is sufficient.”
According to Riskin, the past few years have witnessed a move from technological to business challenges. “A few years ago, I thought it was a hard technical problem but overcomeable,” he explains. “But now what I’m seeing is the needs are so broad between revenue, analytics, quality measures, and predictive modeling that being able to infer or understand what the words mean and what the doctor meant is very, very hard. It requires very smart computing and that intelligence is what’s hard to get at.”
In terms of preparedness for making use of this brand of analytics, Riskin sees federal mandates as a significant driver, particularly the conversion to the International Classification of Diseases, Tenth Revision (ICD-10), whose deadline is set for Oct. 1, 2014:
The tip of the spear of what the market is ready for is revenue cycle stuff, sometimes called computer-assisted coding. There may be revenue analytics associated. The market is ready for that. They were ready for that anyway because they were getting frustrated with scaling the manual processes, but they’re even more ready for that with ICD-10 conversion. People are nervous. The market is ready at the tip of the spear. Now is the market ready in quality measures and predictive modeling and the things that will actually improve the quality of care? Probably not yet.
For those responsible for overseeing coding and clinical documentation improvement efforts, the value of physician notes is obvious. Pertinent information that cannot be coded easily in structured fields is generally there in free text if you know what to look for. And what NLP tools highlight is the potential for converting this unstructured data into appropriate ICD-9 or eventually ICD-10 codes.
Despite the emergence of these and other tools for computer-assisted coding, none is valuable without the willingness of healthcare organizations and providers to look beyond simply meeting mandates and down the road. “The entire country is in the weeds right now, no one more than the health systems. MU2 and ICD-10: It forces people to be in the weeds and work their hardest to address mandates,” argues Riskin.
“ICD-10 is not about going from a small number of codes to a moderate number of codes,” claims Riskin. “The patient doesn’t actually care about that. And Meaningful Use 2 is not about reporting smokers. The country doesn’t actually care about that. These things are about getting sufficiently detailed information and sufficiently broad information that actually gets used in analytics to improve care, and yet we’re not really thinking about that all that much.”
Within the next year, Stage 2 Meaningful Use and ICD-10 are going to raise the level of expectations for healthcare organizations and providers to document and report their care. Minus the wherewithal to get the most out of clinical information to meet the demands of the changing healthcare landscape, valuable information will remain untapped and diminish in its ability to improve a variety of outcomes.