Tech & Health

September 23, 2015

Data diving for health

To most effectively use electronic health records for research, investigators should query multiple components of the record to identify patients with specific diseases.

by Henry H. Ong


Electronic health record (EHR) systems have not only improved patient care, but have also generated enormous amounts of data that are a rich resource for research into health care processes and basic biology.

EHR phenotyping, which is the process of identifying patients with a specific trait, is one of the crucial challenges for effective use of EHRs for research.

Reporting this month in Journal of the American Medical Informatics Association, Wei-Qi Wei, M.D., Ph.D., and colleagues studied de-identified Vanderbilt EHR data comprising over 2.3 million unique patients.

The investigators systematically evaluated the performance of three major EHR components often used in phenotyping – billing codes, medication, and clinical notes – to identify patients with 10 common diseases including Alzheimer’s disease and breast cancer.

The results showed that combining two or more EHR components provided a more consistent and higher performance than a single one for the selected phenotypes. Using multiple EHR components should be considered in future phenotyping design for the best performance, the researchers concluded.

The study was supported in part by National Institutes of Health grants LM010685, GM115305, GM103859 and HG006378.

Send suggestions for articles to highlight in Aliquots and any other feedback about the column to