October 17, 2018 | Gordon Moore
Okay, so the title of this blog reads like an obscure academic paper, but here’s the point: For computers to aid us in finding needles in haystacks, we have to teach computers to make the appropriate meaning of the words we use, and computers are blunt and literal.
For example, what does “cold” mean? A word search finds “cold” in the medical record. Is this person chilled? Do they have an upper respiratory infection? Are they emotionally detached? We must teach the computer how to recognize the context so that we can be sure the computer output has the meaning we expect if, for instance, we are looking to tally up how many people had colds this past January.
We live in an era of information technology that holds great promise, but that promise comes with risks we must understand and mitigate. I came across the research of Andrea Gilmore-Bykovskyi, PhD, RN, an Assistant Professor and nurse researcher at the University of Wisconsin-Madison School of Nursing and Alzheimer’s Disease Research Center. Dr. Gilmore-Bykovskyi has been studying the impact of communication and the meaning of words that clinicians use in the medical record. As a geriatrician, her work has focused on clinical care delivery and health disparities among people living with and at risk for Alzheimer’s disease and related dementias.
People with dementia are much more likely to experience complications and have worse outcomes that those without dementia. Knowing that a person has cognitive impairment should lead to specific interventions to mitigate risk. For example, a person going home from the hospital after a procedure that requires blood thinning with Warfarin must follow a complex set of instructions. Too much, too little or bad timing of the medication can lead to devastating and even life-threatening complications. Failure to recognize cognitive impairment puts that person at much greater risk of post-hospital complications.
Dr. Gilmore-Bykovskyi’s research explores the risks of this failure to recognize cognitive impairment, describes communication failures and how they related to hospital readmission, and explores how clinicians use words and phrases in the medical record that can expose cognitive impairment even if it is not formally diagnosed.
This work is part of the foundation of how we make meaning from words so that we can be confident in the results of computer augmentation of analytics. Listen to our conversation on the Inside Angle podcast.
L. Gordon Moore, MD, is Senior Medical Director, Clinical Strategy and Value-based Care for 3M Health Information Systems.
Get blog and podcast insights from our Solventum experts at the intersection of data science and healthcare.