Electronic health records (EHR) can be a valuable tool for clinical and genome research studies, allowing the identification of large groups of patients with a given disease. However, their utility hinges on the ability to identify patients with the same disease across different institutions using different record systems.
Robert Carroll, Joshua Denny, M.D., assistant professor of Biomedical Informatics and Medicine, and colleagues evaluated the ability of an algorithm developed at one institution to identify patients with rheumatoid arthritis (RA) from electronic health records at two other institutions with different EHR systems.
They report, in the June 1 issue of the Journal of the American Medical Informatics Association, that the algorithm reliably identified RA cases at the two other institutions, and that minor modifications (“retraining”) further improved the sensitivity of the algorithm.
The results show that a method developed at one institution, while not specifically designed to be portable, can be implemented at other institutions with different EHR systems. This could significantly speed identification of patients for research studies and help link different EHR systems to facilitate large-scale research collaborations.
The project was supported by grants from the National Institute of General Medical Sciences (GM092691) and the National Library of Medicine (LM008748) of the National Institutes of Health.
