"The big change that's occurred – and the big positive boost in all this – is natural language processing," says van Terheyden, "or as we term it, clinical language understanding."

Technology such as Nuance's "analyzes the content real-time with the physician's documentation," he says. "One of the things that happens in most CDI processes is the feedback to the clinician occurs sometime – could be hours, could be days, even weeks – later."

But most clinicians, they're on this diminishing logarithmic curve of knowledge: 'If you ask me now, I'll know. If you ask me in a few hours or days, I'm decreasingly likely to know the answer to the question you're asking me. I've got to reactivate all that knowledge and remember.'

By offering "real-time feedback that educates," says van Terheyden, "physicians are getting education at point of documentation – the best form of education you could ask for." Moreover, "it frees up the resources we've got – the very limited CDI specialists and coding specialists – to focus on the more complex cases we have that our technology may not be able to detail," he says.

NLP technology has traditionally been less than ideal – requiring repeated emendation and revision to ensure the spoken word is translated correctly into text. Has the technology evolved in the past few years?

"Oh, it's almost indescribably so much better," says van Terheyden. "One of the great things about this technology is the massive learning curve." In Nuance's case, that's thanks to a cloud-based platform.

"Every time someone uses our technology, it gets uploaded securely in the cloud, we process it and return it," he says. If a clinician enters a note or code erroneously, "we know instantly because if it's wrong somebody doesn't use it, they don't take action on it."

As of now, Nuance is processing "more than a billion transactions per month in the cloud, and we're learning from that interaction," says van Terheyden. "If it's right and you use it, you know that it's right, and if it's not, you either re-dictate or discard it and we input that into the system.

"Our learning curve has really escalated: We're certainly north of about 93 percent accuracy on average, across a number of modalities," he says. That's "helping people on the back-end, that we focus all the high-quality, high-accuracy feedback to the clinician and the evidence suggests that the response rate and the agreement rate is all very high."

Ultimately, van Terheyden says he wants his session to emphasize that "ICD-10 is not a negative implementation. It's a positive. I think we should see it as an improvement in the whole coding and understanding of what we're doing. I think it's a benefit to everybody in the system – patients, coders, physicians alike.

"Importantly," he adds, "you have to work together as a team to deal with it."

Not unlike, perhaps, the brotherhood of Night's Watch, guarding the northern border from White Walkers?