Back in the days when patient charts were largely written observations created with pen and paper, it was pretty much impossible to get a comprehensive picture of a patient’s healthcare data and how it all connected. Even now. With the advent of EMRs and using computers to enter information into specific applications and databases, it has still been difficult to get a big picture perspective.
All that is about to change. Natural language processing (NLP) is providing powerful software and processes that enable mining and analyzing of charts, notes, lab results and more, transforming a wide range of unstructured data into information that can be quickly understood and acted upon. When applied to the deluge of healthcare data sources and processes, NLP has the potential to dramatically improve patient care outcomes as well as enable smarter risk management for doctors and payers.
Imagine being able to improve healthcare results by mapping social media posts against environmental data. That’s exactly what is happening in Texas. Researchers at Parkland Hospital in Dallas conducted a study that looked at tweets that included the word “asthma” and used natural language processing to map them to both information collected by air-quality sensors and patient electronic health records. By connecting these dots, doctors were able to predict with 75% accuracy the relative number of asthma-related visits that they could expect on any given day.
Another area of focus has been in heart failure patients. At Intermountain Healthcare in Salt Lake City, doctors have been using NLP to identify potential heart failure patients using data from 25 different types of free text documents stored in electronic health records. According to Scott Evans, Director of Medical Informatics at Intermountain, heart failure patients often come into the hospital or visit a healthcare provider for a non-heart related issue such as a hip or knee replacement, respiratory issues, diabetes or any of a number of other reasons. But NLP provides caregivers with much broader overview, allowing them to identify potential heart failure patients by looking across their overall healthcare data set.