Natural language processing (NLP) shows “promise” in improving electronic health record documentation, recent research shows.
EHRs need clinical notes and reports to be in both human readable and machine readable formats, which requires more work and new computer skills for physicians. To that end, researchers from Arizona State University and elsewhere wanted to determine if NLP would be a viable solution for clinical data capture by using dictation rather than keyboard and mouse data entry.
In a study published in JMIR Medical Informatics, they evaluated an NLP-enabled solution, focusing on three data capture problem areas: efficiency, effectiveness/documentation quality and physician satisfaction/ease of use. The researchers compared four different protocols: the standard keyboard and mouse method; a fully NLP method; and two hybrids that used both standard and NLP methods.
Participating physicians, at Columbia University Medical Center were asked to document four notes using the four methods. A total of 118 notes were documented in the subject areas of cardiology, nephrology and neurology.
In each subject area, the protocols that used NLP required less documentation time than the standard keyboard and mouse protocols. There was no statistically significant difference in documentation quality among the protocols, although the standard protocol was more succinct and organized. The usability ratings for the NLP protocols were “significantly” higher than for the standard one.
The researchers opined that different parts of the clinical note should be documented in different ways; the optimal method for each part, they said, will require further study.
“Documentation methods using dictation and NLP have the potential to reduce some of the most egregious 'pain points' for EHR data entry," the researchers said. "These methods can facilitate capture and insertion of both structured data and transcribed text into the appropriate EHR sections, affording the user of the note the option of using one or both types of information. The structured data are ideal for interoperability and coding and may prove to be useful for analytics."