CAMBRIDGE, MD — Like many people I know, work friends can’t be Facebook friends. I save my Facebook account for personal screeds on politics and religion and for sharing family pictures with interested friends and relatives separated by geography.
Occasionally, however, the discussion turns to health IT matters. A couple weeks ago, I posted about the new Massachusetts HIE going live for all my pals in the state. Demand your data! I said, which begat a comment string that led to our friend Abigail, a nurse and instructor from Rhode Island, bemoaning the fact that “HIE,” for many patients, adds up to unintelligible page upon unintelligible page of stat-laden PDFs because of the lack of interoperability between EHR systems.
At that point, it was game on. Inspired by the recent town-hall-format presidential debate, I asked her if she had anything in particular she’d like to ask or tell Doug Fridsma, M.D., director of the Office of Standards and Interoperability and the acting chief scientist in the Office of the National Coordinator for Health Information Technology — our country’s health data interoperability czar – knowing I’d be running into him at the HealthTech Council meeting.
“Well I think you might want to point out that the charting from a single day on the ICU can now run to 30 pages of printouts that obfuscate more than they inform – we need creative new data presentations that could be developed by a team of MDs, informatics specialists, librarians and graphic artists.” As opposed to programmers or other vendor groupthinkers.
What would Dr. Fridsma have to say to that? Metadata and annotations hold the key, he said, although initiatives to standardize and make these usable are still in early development. However, he sees a time where, for instance, the metadata to a record would indicate “ICU stay” and annotations would help aggregate key points of data (such as high blood pressure and low blood pressure during the stay) for later use. In effect, although all of the patient’s data – all those continuous blood-pressure readings, etc. – would remain in the complete record, filters could sort the key points in a patient’s report so those 30-page PDFs could turn into more usable snapshots of an episode of care.
The trick is to understand information decay, which may only be pulled off with the help of human interaction, Fridsma said. As in, data from one ICU stay might have intensely important data, such as a hypertension event that has long-lasting health consequences. A different stay might yield unremarkable test readings that turn out to be insignificant in the big picture of a patient’s health status. It’s hard for a machine to differentiate them.
“The aggregation issue is probably, at this point, still in its infancy,” Fridsma said. “Particularly, time-series data aggregation is tough. [It] requires medical knowledge to do it right, and I’m not sure that we are always going to be able to do that automatically, but if you annotate the data with how it was collected, who it was collected by, and at what point in their hospital stay it was…smart people should be able to do the things that need to be done, like ‘should we aggregate this with mean, mode, and all these things we care about?’ There are some things that we’re always going to require clinicians to interact with. Our job, then, is to figure out what information we can provide to them so they can summarize that ICU stay in a way that makes sense to [patient and next care provider down the line].”
Good question, nurse Abigail. Looks like no matter how well EHR design is sharpened in the future, people will still be whipping a patient’s data into shape to make it comprehensible – and usable.