Apixio's CTO on why it's taking so long to capitalize on EHR data

Not many industries can nimbly handle the kind of major technological change that healthcare did when, in less than a decade, EHR adoption shot up by 80 percent. Even adding federal oversight (not to mention some financial motivation through EHR Incentive Programs) to the mix seems unlikely to drive that degree of change. With expectations and demands for meaningful use of EHR data as high as they are, it's easy to forget that, by definition, healthcare is an early-stage industry when it comes to IT, one that's being asked to run a marathon when it's still wobbly on its feet.

"In that short time, doctor's offices and hospitals went from manila folders with notes handwritten on paper to electronic systems, and the industry is struggling with how to make this changeover more humane, because it's truly not humane," says John Schneider, chief technology officer of San Mateo-based Apixio, a cognitive computing firm that works to tease out meaningful data from EHRs and make it accessible. "You've got physicians up in arms because they don't have the time to do their jobs and they need to become data entry experts."

High expectations for EHR performance are no surprise, but when an industry struggles so mightily with what should be a basic changeover — paper to digital — growing pains and delays are no surprise, Mr. Schneider says. Due in part to the need to implement electronic records quickly at a time when providers were operating in a fee-for-service mindset, EHRs are still essentially point-of-sale systems. It's unlikely it could have happened any other way, as hospitals needed to transition to digital without disrupting their entire operations.

The problem with opening these digital records now and trying to pull actionable data and insights from them is that the information they contain is literally and figuratively unstructured. As much as 80 percent of the data found in medical records is classified as unstructured, meaning that essentially all that someone can do with the data is read it like a text document. That maybe isn't so bad except that medical records can be hundreds of pages long, and there can be tens of thousands of them to read. It's a tedious and expensive process at best.

"You've got a human being, a qualified medical professional, being paid to literally read a 200-page chart to understand what's going on with a patient," Mr. Schneider says. "A general practitioner might see 2,000 patients a year. Can you imagine having to go through this coding process for each of them? You'd go insane. And humans make mistakes."

This is just one of the many challenges presented by the sprint to adopt EHRs. It's also one that Apixio's Iris platform solution aims to tackle by combing through unstructured data to compile complete patient profiles that serve as a better foundation for analytics, according to Mr. Schneider. 

"There are these hidden gems sitting in doctors' notes that don't surface to the level you would need to do any sort of big data analysis," Mr. Schneider says.

The HCC Profiler uses a combination of machine learning and natural language processing to produce a set of predictions about a patient's status that are important to hospitals. These predictions are made up of details that humans might miss when they're processing unstructured information from a clinician's narrative and trying to properly code it into the patient's record.

A physician might know and note that a patient has a specific condition, but due to human or process oversight, that data is overlooked and isn't put into a coded form. The Profiler's job is to pull all those bits to the surface of a record and present them to a provider so they can integrate them into a record in a coded form if they wish.

"That gives them a significantly larger amount of transparency into what's going on in the health of their population," Mr. Schneider says. "For example, it could give an insurance company a much better understanding of the real risk that they're bearing because their models all depend on them accurately coding the specific conditions that their patients have."

All this is happening at a time when healthcare is still hesitant to invest significantly in IT, Mr. Schneider says..While the industry is acclimating bit by bit and year by year to allocating more funds to bigger IT budgets, achieving the kind of data analytics capability that other industries have is going to be incremental.

"Still, what's happened in healthcare over the past decade is truly remarkable," he says. "Ten years ago, EHRs had less than 10 percent penetration in healthcare, and even though systems to back up sexy terms like 'big data' aren't totally there yet, we are trying to challenge that orthodoxy of how you can use data in healthcare."

More articles on health IT:

Competing with the world: University of Utah Health Care's CIO on how hospital IT has changed
What the US can learn from Estonia's electronic health records
Putting Epic's EHR to work: Hawaii Pacific executives on the intersection of health IT & quality

© Copyright ASC COMMUNICATIONS 2020. Interested in LINKING to or REPRINTING this content? View our policies by clicking here.


Featured Webinars

Featured Whitepapers