|"Crawling" an EHR for text?|
For example, if the DMCB types that the patient "has a 10 year history of diabetes" in a clinical encounter note without otherwise officially documenting that condition in a "diagnosis field" or "billing code," the presence of diabetes won't be electronically recognizable. The good news is that NLP is getting to the point where, much like a classic "web crawler," it can be used to "scan" free text and look for phrasing that identifies otherwise silent and undocumented diagnoses or treatments.
The good news is that there's this article courtesy of Fierce Healthcare that, in addition to saying nice things about the DMCB, gives a lot more background on the art and science of NLP or "computer assisted coding." It turns out that its been under development for more than ten years, can generate physician reminders and by inferring a diagnosis, automate the tedious exercise of coding a patient visit.
The DMCB likes the notion and would make two other points:
1. It doubts current EHR vendors will develop an NLP capability because they can't commercialize it and it's not in the meaningful use criteria. As a result, this functionality will need to be "bolted on" by a third party. The companies that can develop a single solution that not only fits multiple EHRs but can adapt to a health information exchange will win.
2. Speaking of meaningful use criteria, it's time to include NLM. Its ability to find patients at risk, facilitate population-based program planning, assist in predictive modeling, infer and automate diagnosis coding, generate reminders and act as an early warning system for outbreaks of disease make its potential too great to ignore.
Image from Wikipedia