EHR data-mining using NLP can improve cancer pain management

Using natural-language processing tools to mine electronic medical records for information on patient pain can lead to better pain-management strategies for terminal cancer patients, concludes a study published online today in the current issue of the Journal of the American Medical Informatics Association.

The study was intended to see how well NLP could depict the experience of patients with metastatic prostate cancer, to identify novel pain phenotypes, and develop ways to visualize pain status. The researchers used text from 4,409 clinical encounters for 33 men enrolled in a 15-year study.

The plan was to develop a four-tiered pain scale and identifying factors correlating with severe pain month-by-month, including the impact of drugs and palliative radiation, according to the project abstract.

The researchers said they identified pain patterns that were "undetectable without the use of NLP" to comb through the records, and were consistent with what would be expected in those patients. They also said the findings suggested opportunities to study the molecular basis of cancer pain.

The research was conducted using proprietary NLP software, but the researchers said the lessons are applicable to other NLP software.

"Electronic health records have greatly facilitated detection and understanding of disease phenotypes and their relationship with genetic and non-genetic factors," the authors concluded, recommending that future studies focus on comparing NLP with other types of pain survey tools "and on practical integration of the two methods in settings where electronic health records are in routine use."

Natural language processing also is being used to develop a clinical decision-support tool for diagnosing and treating cancer. The project combines IBM's Watson natural-language processing capabilities with clinical knowledge and data provided by Memorial-Sloan Kettering Cancer Center In New York City.

To learn more:
- here's the study