An Electronic Health Record Based on Structured Narrative

From Clinfowiki
Jump to: navigation, search

An Electronic Health Record Based on Structured Narrative

S. Johnson, S. Bakken, D. Dine, S. Hyun, E. Mendonca, F. Morrison, R. Bright, T. Van Vleck, J. Wrenn, P. Stetson

Journal of the American Medical Informatics Association 15:1 January/February 2008


Motivated by the lack of penetration of electronic health records into national health care practice, a group of researchers from Columbia University set out to explore possible solutions to the conflict between the clinician’s preference for data entry as free text and the computer’s need for data to be coded for reuse and analysis. The authors propose a modification of the current models of the electronic health record with the intention of facilitating both data entry and retrieval. Their model, which they describe as “structured narrative,” allows the clinician to enter information into the record in free text but in a highly organized format, to which standardized coding is then applied by a natural language processing component, making the data reusable.

The authors developed their model after reasoning that widespread use of electronic health records—and the benefits that would accrue thereof—will not occur unless entering information into the record is both quick and easy for the clinician but also structured enough to allow automated analysis. A literature review convinced them that allowing clinicians to describe their assessment of the patient’s health in natural language resulted in much more robust information than did wholly coded reports. In order for the full potential of the electronic health record to be enjoyed, however, enough structure must be imposed on the data within it to allow the computer to process it.

With these conclusions, the authors developed their model of structured narrative with the text divided into named and coded sections, within which the clinician enters free text. This text is then analyzed by the natural language processing module and represented as documents in extensible markup language; the documents, in turn, are represented by the HL7 clinical document architecture. The document ontology is one part in a larger controlled terminology system, which allows linkage to any other coding system. While they acknowledge that their model requires more evaluation, the authors do seem to have succeeded in creating a versatile and viable method of integrating structured and free text data.