Human language developed at least 100,000 years ago, and has evolved into an amazingly complex and subtle mechanism for communicating ideas.
Computers emerged on the scene a mere half-century ago. And yet, we often find ourselves trying to structure our clinical communications around “talking” to computers, rather than talking to other caregivers. The impact to quality of care is potentially large, and often ignored.
Imagine a physician in your organization wants to communicate something like this:
I considered prescribing W, but decided that, given the patient’s family history of condition X and test results Y, medication Z would be more appropriate.
Would your EHR allow this subtle line of reasoning to be captured in a template? If the only way your clinician can say this is in a free-text field, does the presence of the template discourage communicating this at all? What is lost when a future caregiver cannot see the reasoning behind earlier provider choices?
At its core, the purpose of clinical documentation is to allow human caregivers to communicate information to human caregivers. And yet we often forget this in our rush to gain the benefits of quantitative data capture.
Discussing realistic expectations for EHRs, Dr. David Trachtenbarg reports that “many physicians preferred to use a point-and-click method to document normal findings and to type or dictate abnormal findings.”
It is at the very point that unusual (and likely important!) information needs to be communicated that we naturally fall back to human language instead of templates and quantitative data fields.
One promise of Natural Language Processing is to mitigate the conflict implicit in today’s EHRs: using language for robust communication vs. using quantitative data for computer consumption. As NLP improves, we will gain the ability to run quality and other analytics directly on clinical text, reducing the pressure to “talk” directly to the computer instead of each other.
Richard Wolniewicz is a Division Scientist, Natural Language Processing at 3M Health Information Systems.