Thursday, August 25, 2011

Natural Language Processing: Implications for Population Health and Disease Management

While the topic of "natural language processing" may seem esoteric and far afield from the science of  population health management (PHM), the Disease Management Care Blog thinks this JAMA article on the topic may be important to the industry.

Briefly, the authors tested an automated "Multi-threaded Clinical Vocabulary Server natural language processor system" to scour the free text (such as progress notes and discharge summaries) of the electronic health records of  a sample of Veterans Affairs patients for evidence of post-operative complications.  Recall that up until now, the only way to measure complications is to pay someone to go through every page of the record or rely on largely self-reported and voluntary "coding" billing systems.  Compared to expert nurse reviewers, the automated natural language processing system in this study correctly identified - depending on the type of post-operative complication - about 60 to 90% of the cases.

This is important to the PHM industry because its care and case managers typically enter their interactions with patients into their own electronic documentation systems using a combination of 1) close-ended and goal-oriented check boxes ("importance of daily self-weighing discussed with client") in combination with free text that documents more open-ended conversations.  While buyers of PHM services pay to have every patient get every intervention all the time and every time, the industry has been criticized for imposing one-size-fits-all care management protocols on its patients.  That's why they force their nurses to use those check-boxes.

While detecting the occurrence of complications following surgery is not the same as measuring the content of patient coaching, the DMCB predicts the technology can be readily adapted to PHM.  When that happens, coaching nurses won't have to put up with as many of those check boxes and be able to focus on having "real" open ended conversations with their clients. Free-text systems will be able to count whether "weighing" was included in the documentation and nurses will be able to focus more on the patient and less on data entry.

That's a good thing.

And by the way, the DMCB thinks it's just a matter of time until the next step: analysis of voice recordings between nurse-coaches and their patients.

No comments: