japolia - Fotolia
Many health IT experts believe that natural language processing (NLP) in healthcare can be used effectively for clinical decision support in healthcare. Healthcare is rife with unstructured data and the goal of NLP in healthcare is to extract value or meaning from this unstructured data using algorithms, artificial intelligence and other machine learning technologies.
Population health is one area where NLP in healthcare is promising. Rasu Shrestha, chief innovation officer, University of Pittsburgh Medical Center, and Scott Evans, medical Informatics director at Intermountain Healthcare, Salt Lake City, Utah, discuss the relationship between NLP and population health management.
What role does NLP play in population health?
Rasu Shrestha: [Population health] for us, you know, it really is about better managing risk across populations of patients that we're dealing with. Both on the clinical side as well as on the payer or the health plan side.
It's leveraging data to get at intelligence that would allow for us to risk-stratify our patient populations and really determine the best path forward around mitigating risk and managing chronic conditions -- and ensuring that we're able to get ahead of some of these challenging conditions before they become challenging or possible. And a lot of that is driven by NLP.
So when we talk about, you know, risk management … it really is about better … being able to decipher what's oftentimes locked away, these unstructured documents that I was referencing earlier.
Population health is absolutely critical. It's the biggest driver of moving from volume-based practices of health care, the value-based practice of health care. But let's move away from the buzzwords and let's leverage the right sets of technologies, NLP included. To make sure that we're actually doing what needs to be done. And in this case, it's really about better managing risk.
Scott Evans: Population health is … looking at individuals [as] a population … I think anything that has dictated data, whether it's population health, [or] individual [health] -- I see that as exactly the same thing -- there's data in there that's found only in dictated reports, or dictated documents. If you want to find that data, you're going to have to use NLP. So whether it's data mining using dictated report, which can be a part of population health, you need to use NLP. But population health [depends] on how you want to define it. If you're looking for outcomes, what is best? You know medicine changes almost every day. We hear all these things that, "This is good for you. This is bad for you." And that one year you're supposed to be taking this vitamin. The next time they say, "That vitamin's bad, don't be taking this medication, etc."
So anything that you can read, you can pull the data out of using NLP. You can use individual data to generate populations of interest.
How natural language processing, big data will affect patients
Interoperability to be driven by next-gen EHRs
The role of medical transcriptionist is changed by use of EHRs