Allowing natural language processing to pore over disparate data stored in electronic health records, researchers in Canada have shown the AI-based technology can reveal real-world experiences and outcomes of patients with stage III breast cancer.
Mark Levine, MD, of McMaster University and colleagues had their work published online Aug. 1 in JCO Clinical Cancer Informatics.
After drawing and anonymizing data from 50 relevant patients in their health system’s EHR, the team developed specialized NLP annotators to mine unstructured clinical text and render it in a structured format.
To validate their approach, they applied the annotators to 19 more patients.
Their tool successfully extracted information on tumor stage, patient age, initial treatment and other factors that made it possible to build timelines showing key steps along patients’ care journeys.
Further, comparing the tool’s results with the gold standard—clinicians’ notes in medical charts—they found that, for 171 data elements, NLP and the chart agreed 76% of the time.
Moreover, with “additional manipulation using simple logic,” the disagreement was reduced to only six elements, the authors report.
“It is possible to extract, read and combine data from the EHR to view the patient journey,” Levine et al. concluded.