Today, there are more than one billion images online for 7.5 million registered patients at the 1,447 VA sites of care across the United States. These images are stored and retrieved electronically, and can be accessed across the VA. To meet safety and regulatory requirements, all paper documents are scanned so the VA has a single source of health record information.

Information is only useful if it can be retrieved in a reasonable time period. As more and more data is available in the electronic health record, old paradigms for storing and retrieving information become less suitable. Record retention times for electronic patient records are increasing, as many “wired” medical institutions decide to keep electronic records for time periods approaching the life of the patient.

This paper discusses an approach that uses technology to reduce the manual burden of indexing records, while making use of human decision making when efficient. This implementation provides index terms along multiple axes, such as specialty, procedure, event, image source, kind of user, and body part.

The clinical display workstation allows users to view the multimedia electronic health record; a record which goes far beyond the paper chart in functionality and in readily available information. After selecting the patient, the user is presented with a list of image studies for that patient. With increasing numbers of studies to choose from, more automated mechanisms were needed to help users find the images of interest.

There are a wide variety of medical specialties that use images, and the needs of clinicians vary. Some clinicians routinely view or interpret images related to their specialty or procedures. Others view specific images created by other clinicians, often comparing them with other image studies and report findings. The ontology project described here handles a wide variety of images and documents, and provides tools for viewing them.

The VistA Imaging System allows users to view image studies captured and stored at different VA medical facilities. Therefore, the terms used for retrieval and sorting were standardized across the VA. Users can display images from any VA Medical Center in the United States on the same computer screen in an integrated manner.

Because the VistA Imaging System handles a wide variety of multimedia objects, the retrieval mechanisms are complex, and the requirements for retrieval keys varied. A multi-key approach to indexing and retrieval was determined to be most appropriate, allowing any multimedia data object to be classified along multiple different axes and, therefore, be located in different ways, depending on the user’s needs.

For example, instead of retrieving by a single title such as “CARDIAC CATH OUTSIDE REPORT,” a user can retrieve by any combination of “CARDIOLOGY,” “MEDICINE,” “CATHETERIZATION,” and “OUTSIDE MEDICAL REPORT.” Various combinations of terms will yield a more or less specific set of candidate records.

Originally, five new index axes were selected to implement:

The INSTITUTION index term indicates the medical facility where the record originated, for example a VA facility, a Dept. of Defense Medical Center, an Indian Health Service Center, or a private hospital.

The CLASS index term is used to identify components of the clinical or administrative patient chart. This index is used to restrict viewing of documents and images to the appropriate users.

The TYPE index is used to indicate a particular kind of document or image. These may be exact document titles, such as ADVANCE DIRECTIVE or CONSENT, or general document types such as PROCEDURE REPORT or OUTSIDE MEDICAL RECORD.

Because the VistA Imaging System deals with many images that are captured during medical procedures, an index for PROCEDURES is very important. There are also a number of medical events that occur at a particular point in time but are not classified as procedures, for example “HOME VISIT.” These event terms behave in the same manner as procedure terms, and are linked to specialties and subspecialties.

The SPECIALTY or SUBSPECIALTY index terms are typically linked in a hierarchical fashion, making searching and sorting easier.

A body part index is currently under development. This index will build on existing term lists, and will map to LOINC.

The user can search or sort the list of studies using any combination of index terms.

During the document scanning or image capture process, the user selects the index terms to be assigned. A graphical user interface was developed to streamline this process. Index term choices are limited by the terms previously selected for other axes. For example, if a user selects a specialty of CARDIOLOGY, the terms in the PROCEDURE/EVENT selection boxes will be limited to Cardiology procedure/events. Index terms are automatically assigned during DICOM capture.

To further streamline the assignment of index terms, capture workstation configuration buttons can be predefined. Clicking one of these buttons instantly allows the user to apply all workstation settings including capture device parameters and index terms associated with the document or image.

Configuration tools are provided for users viewing images to configure the study lists. They can limit the study lists by specifying the specialties, events, procedures, types, dates, etc., to be displayed. Tabs allow users to change configurations immediately.

The above table shows the growth of terms over time. A national index term board was established to govern additions and changes to the standard index terms.

This effort allows users to retrieve information based on descriptors along the axes of document type, procedure/event, specialty/subspecialty, body type, institution, and user class. Users have found this to be a good solution, and have contributed to the effort by proposing new or missing terms.