Background. Recent efforts in the field of medical decision support systems focus on the creation of clinical care guideline representation formalisms and on the development of software compatible with those formalisms. The patient record is identified as the main repository, from which the data required by guidelines are retrieved. In order to ensure compatibility between guidelines and the record, several formalisms pose constraints on the contents and format of the latter. Other formalisms identify the problem of detecting in the patient record the data of interest, but do not provide solutions. In both cases the issue of coping with non-structured patient data representations is not systematically tackled. Natural Language Processing applications offer a solution to processing free text; additional efforts will, however, be required before they can achieve widespread use by healthcare practitioners.

We have developed an ActiveX software component, called the Guideline Step Control, which allows the creation of Clinical Care Guidelines capable of processing patient data, including free text, stored in electronic healthcare record software compatible with the Good European Health Record (GEHR) architecture1.

System. Guidelines are structured according to the Guideline Interchange Format (GLIF)2. Guideline Steps (actions, conditions, etc. in GLIF terminology) are represented by Guideline Step ActiveX Controls inserted in documents produced by ActiveX container applications. The controls are linked together with the aid of a specialized tool, Link Manager, which has been developed for this purpose.

Each conditional guideline step is assigned one or more Expert Systems, which are responsible for seeking in patient records those data that are necessary to determine the truth-value of the condition related to the particular step. Patient data of interest may include coded terms drawn from existing vocabularies, free text, numbers, date and time stamps, etc. Expert Systems are designed to recognise alternative ways in which users can represent a medical concept, thus offering an alternative to Natural Language Processing.

Reuse of expert systems is achieved with the aid of the Protégé II/Win Knowledge Based Systems Development Environment 3. An ontology, a knowledge acquisition tool and a library of problem solving methods have been developed to facilitate the creation of expert systems which operate on patient data as described above.

Compatibility with other existing guideline representation formalisms is achieved by means of exporting and importing modules, which tranform guidelines based on the Guideline Step Control to other formalisms and vice versa.

Evaluation. The initial implementation of the Guideline Step Control has been tested with HEALTH.oneTM 4, a widely used GEHR- and CEN-compatible Electronic Healthcare Record software. Two guidelines have been created and have been applied to a database of 150 patient records. The Guideline Step Control was found to identify and process alternative representations of medical concepts successfully. In cases that the required data were not contained in the record, the component interacted with the user as expected. The recommendations produced and the actions triggered during guideline processing were always the expected ones.

Conclusions. ActiveX technology and expert systems can be combined in a simple, yet architecturally open and powerful means of producing guidelines for processing real-life patient records and providing appropriate advice. Knowledge based systems development environments can be used to make the development and use of such guidelines feasible for the end users. Although further evaluation work is necessary, the initial results are encouraging.

References

1. CEN TC251/WG1. Medical Informatics: Electronic Healthcare Record Architecture. ENV 12265, 1995. http://www.centc251.org
2. Ohno-Machado L., PhD, Gennari J. H., Murphy S. N., et al. The GuideLine Interchange Format: a model for representing guidelines. JAMIA, Jul-Aug, Vol 5, No. 4, pp. 357-372
3. Musen M. A. Domain ontologies in software engineering: Use of Protege with the EON architecture. Methods of Information in Medicine 37:540-550, 1998
4. Maskens A, Geboers J. Electronic healthcare records: from European standards to products, the “pop” ehcr concept. In: Towards an Electronic Patient Record 97 - Proceedings Manual. Wagemann P (Ed.), Broadvue (Ill, USA): Kelvin Press, 1997

S. Deftereos, MD *, A. Maskens, MD, Ph.D. +, D. Sotiriou, Assoc. Prof. *
* Medical Physics Laboratory, School of Medicine, University of Athens, Greece
+ Health Data Management Partners s.a., Brussels, Belgium
Τελευταία αναθεώρηση : 1/1/2006