The InterMed Collaboratory

InterMed began (on May 1, 1994), as a collaboration among the Stanford Medical Informatics (SMI) at Stanford University, the Decision Systems Group (DSG) at Brigham and Women's Hospital (BWH), and the Department of Medical Informatics at Columbia University. Researchers in the Laboratory of Computer Science at Massachusetts General Hospital (MGH), and the Center for Medical Education at McGill University became closely involved at later dates. One of the central objectives of InterMed has been to develop sets of tools and resources for disseminating clinical guidelines across medical disciplines and settings. The development of the GuideLine Interchange Language (GLIF 2.0), a computer-based format that can be used to distribute guidelines across different institutions and systems has been the primary product of this undertaking. More about InterMed, 1994-1998 can be found here.

Since June 1999, the InterMed collaboratory, now including the Stanford Medical Informatics (SMI) at Stanford University, the Decision Systems Group at Brigham and Women's Hospital (BWH), and the Department of Medical Informatics at Columbia University, has been working on the development of the new version of GLIF, called GLIF 3.0

The funding of the InterMed project has ended on December 2003, but the research on GLIF and its implementation will continue. One route by which research on GLIF will continue is through the HL7 Clinical Guidelines Special Interest Group.

The principal investigators for the InterMed Collaboratory are:
Edward H. Shortliffe, MD, PhD, Professor and Chair, Department of Medical Informatics, Columbia University, New York, NY
Robert A. Greenes, MD, PhD, Brigham and Women's Hospital, Boston, Massachusetts
Vimla L. Patel, PhD, DSc, Professor, Department of Medical Informatics and Department of Psychiatry, Columbia University, New York, NY
Samson Tu, MS, Section on Medical Informatics, Stanford University, Stanford, California

The entire InterMed group includes:

Stanford Medical Informatics, Stanford University:
bbbbbSamson Tu, MS
bbbbbMor Peleg, PhD [now with Haifa University]
bbbbbElmer Bernstam, MD, MS [now with University of Texas at Houston]

Department of Medical Informatics,Columbia University:
bbbbbEdward H. Shortliffe, MD, PhD
          Vimla L. Patel, PhD, DSc
bbbbbDongwen Wang, BE

Decision Systems Group, Harvard University:
bbbbbRobert Greenes , MD, PhD
bbbbbAziz Boxwala, MBBS, PhD [now with Eclipsys Corporation]
bbbbbLucila Ohno-Machado, MD, PhD
bbbbbOmolola Ogunyemi, PhD
bbbbbQing Zeng, PhD

 

Background

Recent tends in health care delivery have led to an increase emphasis on the development of guidelines for prevention, diagnosis work-up, treatment, and patient-management processes (clinical pathways). Such guidelines are motivated by concerns about marked variation in clinical practice and are designed to help to provide a common standard care both within a health care organization and among different organizations. Because the guidelines can be complex and often are designed for use where several levels of providers are working together to take optimal care of patients, researchers have sought methods of computer-aided support of the administration of such guidelines. Significant work is required to put the guidelines, typically created as text documents, into a useful format to support this computer assistance. Much of the work is related to resolving ambiguities in the paper versions of the documents. Another approach,m providing computer-based authoring environments, solves some of the problems, but highlights the need for an accepted standard for the computer-based representation of these guidelines and the way in which they are to be used. Thus a guideline converted to a computer-based representation at one institution is typically unusable by the clinical information systems at other organizations. A significant goal of our proposed research is to develop a common standard which will allow for the computer-based representation of a guideline at one institution, and its storage on a central server on the Internet. The purpose is to provide the ability to have the content of the guideline made available either to provide direct execution of guidelines in response to network-based queries or to serve as a repository of standardized, encoded guidelines that may be downloaded, translated and used as part of computer-based patient record system at institutions beyond the site at which they were originally encoded.

The primary product of the InterMed Collaboratory in the years 1994-1998 was the GLIF language for sharing computer representations of guidelines. We have found that the process of encoding guidelines in the GLIF language discovers ambiguities in the documents which can in turn be fed back to the guideline's creators. This research proposal defines a set if steps for extending the expressiveness of the GLIF language and for developing methods to translate GLIF into local execution languages, such as the Arden Syntax.

Research Tasks

In response to the motivations described in the previous section, we seek to address the following specific aims:

1) Extending the GLIF language for representing the description of guideline knowledge. GLIF 2.0 supports the description of the relationships among different steps in the guideline. The primary extension are to create standard descriptions of the steps themselves and to have a well-defined meaning associated with each of the recommended actions in order to aid in the copmuter-based execution of the guideline.

2) Examining, characterizing, and encoding representative guidelines. We will use a subset of the ACP-developed guidelines along with federally developed guidelines for our testbed. The encoded guidelines will become available as part of the guideline server, described below.

3) Creating a guideline server (accessible over the Internet) that can provide support for distributing and sharing encoded versions of guidelines. The server will allow: (a) browsing the encoded guidelines using viewing tools (adapted from PROTEGE and GEODE).; (b) accessing text versions of the guidelines; (c) computing eligibility criteria for guidelines and execution of some categories of protocols; (d) downloading guidelines in GLIF format; and (e) request information about other individuals or organizations which are utilizing the same guidelines. We will also produce a set of Application Program Interfaces (APIs) to allow network-based applications to have remote access to capabilities of the server. Program access to the server's execution model requires that a communication protocol be worked out for getting data elements from the remote program to the server and for transferring advice back from the server to the remote site. In some cases, it will not be possible to obtain sufficient data to make definite decision at the server, and thus the server must be able to specify conditional information back to the remote site for evaluation by the user.

4) Extending guideline-encoding and viewing tools: providing syntax checkers to ensure guideline integrity and semantic checkers (writing programs that will validate a GLIF3 specification of a guideline: for example, (a) answer reachability problems (is node B reachable from node A); (b) trying to find deadlock situations, or missing "elses" (a branch with only one branch destination, a decision with only one next step); and (c) if the guideline contains a branch criterion that utilizes certain data elements, the checker can make sure that branch criteria are not redundant, or result in situations where the logic of the criteria prohibits traversal to certain parts of the guideline. Ted also mentions an integrity checker, but does not explain what he means.

5) Developing methods for turning the GLIF representation into an executable guideline representation for application by the central guideline server in response to remote-procedure calls over the Internet.

6) Developing tools for generating explanations of recommendations

7) Studying cognitive aspects of creating guidelines, of the encoding tasks described above, and of the clinical use of guidelines in computer-assisted environments. The results of these studies will provide feedback to the ACP and other organizations about ways to improve guideline design and provide better understanding about problems that can occur as the textual format of the guidelines are turned into the structured representations needed for online storage and execution.

Last update on December 18, 2002 by Peleg@stanford.EDU