Free text in electronic health records, with the help of natural language processing (NLP) technology, can be used to create accurate clinical decision support (CDS) tools, according to a study published this week in the Journal of the American Medical Informatics Association.
Researchers from the Mayo Clinic set out to develop a CDS system for cervical cancer screening geared specifically toward identifying patients with abnormal Papanicolaou (Pap) reports. According to the study's authors, providers often don't follow proper protocol for cervical cancer screenings; they believed that the use of CDS would help providers in their screening efforts.
Pap reports, however, are compiled in non-computer-friendly free text in patient records; free text, according to the authors, has been "largely underutilized" in CDS efforts, due to accuracy concerns. To that end, the authors opted to create various NLP tools for text extraction, with an ultimate goal in mind of creating a CDS system using a free-text rulebase--which would pull information from EHRs--and a guidelines rulebase--which would pull from national cervical cancer screening guidelines.
Overall, screening recommendations from the CDS system for 74 patients were examined, with appropriate recommendations made for all but one of the patients. That led the authors to conclude that NLP was a reliable method for retrieving EHR information for such uses
"In the near future," the authors added, "our approach would be replicated for other free text-based decision problems such as colon cancer screening, sleep disorder management and asthma management."
This is not the first study to tout the use of NLP on free text. A study published last summer in the Journal of the American Medical Association found that use of NLP on free text in EHRs identified post-operative complications with better accuracy than claims data.
Additionally, NLP is at the forefront of IBM's healthcare efforts to bring predictive analytics to providers.