Text-mining the data in electronic health records can help identify patients with multiple sclerosis (MS) and track the key clinical traits in the course of their disease, according to a new study in the Journal of the American Medical Informatics Association (JAMIA).
The researchers, from Vanderbilt University Medical Center, noted that MS is a "highly variable" and "poorly understood" disease; most studies, they said, have focused on its origins, as opposed to the disease's course, such as type and frequency of symptoms, which measures the progression of the disability. The authors evaluated natural language processing techniques applied to EHRs, and used four algorithms based on ICD-9 codes, text keywords, and medications.
The algorithms identified 5,789 patients with MS, and were highly precise for clinical trait extraction, with precision rates ranging from 87-99 percent.
"This dataset provides a rich resource for better understanding MS and also shows that extraction of detailed disease states and markers of prognosis in patients with chronic disease is possible and may yield a powerful tool in chronic disease research," the researchers said. "We have shown that detailed clinical information valuable to research studies is recorded in medical records of individuals with MS, and that this information can be extracted in a highly reliable manner."
EHRs continue to prove their potential worth in research and predictive modeling to identify at-risk patients or those with undiagnosed medical conditions, such as growth disorders. Increased refinement of the data stored in EHRs will improve this research and the resulting patient care even further.
To learn more:
- read the study