THE USE OF A CASE-BASED REASONER IN THE MANAGEMENT OF RHEUMATOID ARTHRITIC PATIENTS

Nicole F. Gaskin, M. Frize
Department of Electrical Engineering, University of New Brunswick, P.O. Box 4400, Fredericton, NB, E3B 5A3

B. Nickerson
Faculty of Computer Science, University of New Brunswick, PO Box 4400, Fredericton, NB, E3B 5A3

F. Solven
Dr. Everett Chalmers Hospital, P.O. Box 9000, Fredericton, NB, E3B 5N5

ABSTRACT

A physician's decision on prescribing a treatment for a particular condition is partially based on knowledge of the treatment response observed in other patients. Important factors that must also be considered when selecting a treatment are patient similarity and symptom similarity. Remembering specific cases and then trying to find them amidst the patient files can be time consuming. The use of a case-based reasoner can facilitate this task.

INTRODUCTION

With the availability of computers, many physicians are creating databases with information about their patients. Their willingness to share this information provides the first step in creating a knowledge-base as a decision-aid tool for diagnosis and medical intervention. A rheumatologist has been keeping computerized records (dBase IV) for several years and has shared the data with our research group. The initial purpose of this particular research project was to develop a case-based system where a new patient could be quickly compared to the numerous cases in the database. The object is to find several closest matching cases that can be reviewed by the physician to develop a treatment plan based on past results.

Rheumatoid Arthritis (RA) is a temporal disease. RA is a chronic syndrome characterized by symmetric inflammation of the peripheral joints, potentially resulting in progressive destruction of articular and periarticular structures [1]. This cyclical disease needs to be monitored and treated over time. Thus, for more accurate patient matching, it is also necessary to match on the changes in the symptoms over that period of time.

(a) Methodology

Case-based reasoning consists in finding a solution to a problem based on prior examples [2]. In the most general terms, a case-base is the collection of all the information necessary for case storage and retrieval via case matching [3]. The knowledge-based reasoner used in this research was ART-IM (Automated Reasoning Tool for Information Management) developed by Inference Corporation [3][4]. This reasoner was used by K.Taylor in the development of a tool to assist in the management of Intensive Care Unit (ICU) patients [5]. It was his work that provided the inspiration for this project. A patient case is defined in the form of a schema. The schema contains features which describe a case. The database of information on RA patients contains many fields, some of which were included as schema features. The features chosen for this case-matching were based on discussions with the rheumatologist.

ART-IM has four types of feature matching: string, word, character and numeric matching. When a feature is defined, it is required that the matching method for that particular feature be specified along with the match-weight and mismatch-weight [4]. The weight assigned to a feature is dependent on the importance of the feature to a case. A feature considered very important to a case would be given a high match-weight and a feature considered not so important would be given a lower one. Default weights for each feature were set following a discussion with the rheumatologist. It was felt that the user should be able to change the weights at will, to allow a search tailored to a specific patient type. Each feature of the database case contributes a percentage of the assigned match-weight or mismatch-weight, depending on the degree of match or mismatch between the feature and the presented case's feature, to it's overall case score. The ten cases with the highest case score would be selected by the system and displayed for the user, along with the case score.

(b) User Interface

Acceptance of any medical knowledge-based system is as much related to the graphical user interface (GUI), as it is to the knowledge-base and expert system design themselves. Physicians have specific requirements when it comes to interfacing with computers, one of the main features being limited typing [5][6]. The GUI of IDEAS for ICU [5][7] served as an example of an effective GUI. The GUI developed for Rheumatoid Arthritic Patient Management was accomplished using Microsoft's Visual Basic Programming System. Several windows were created with various features. From the MENU window, a new patient could be added, an existing patient updated, a match performed between the new patient and the other patients in the database, and a patient could be deleted. Selecting new patient or update patient will open a window where the patient's personal information (time invariant data) may be entered. Upon completion of this, the next window opened allows the user to enter or change information about the patient's specific joints; the third window allows the user to enter clinical and lab information; this data can then be saved to the database. When the match patient option is selected, the ten cases most similar to the presented case are displayed. The physician can then go to the appropriate files to review the medications used in the treatment and study the outcomes for each patient.

DISCUSSION

Further discussion with the rheumatologist has resulted in a reduction in the number of features needed for matching. Different methods of representing the change in a feature over time, as a match feature, are currently being investigated. At present this application is tailored to the needs and existing database (which generates some unique data), of a particular rheumatologist. For acceptance on a larger scale, the opinions and suggestions of more rheumatologists are vital. The GUI as it is currently designed, will be modified to present the data in a more useful manner in the next phase of the project. The recent availability of a new inference engine, to replace the much slower ART-IM, has been suggested for use in this application and will also be incorporated into the system.

CONCLUSION

A case-based reasoner developed for the ICU was modified to fit the needs of a slower paced disease such as rheumatoid arthritis. The research to date is encouraging and already provides some benefits for the physician: a quick search of the database for similar patients, and easy retrieval and updating of patient files. The work done thus far, coupled with valuable suggestions and ideas, provides a good base for our future research.

REFERENCES

[1] Berkow, R., Talbott, J.H., The Merck Manual of Diagnosis and Therapy Thirteenth Edition, Merck Sharpe and Dohme Research Laboratories, Rahway, NJ, 1977.

[2] Riesbeck, Christopher K., Schank, Roger C., Inside Case-Based Reasoning, Lawrence Erlbaum Associates Inc., New Jersey, 1989.

[3] Inference Corporation, Case-Based Reasoning in ART-IM, Version 2.5, Inference Corporation, 550 N Continental Blvd., El Segundo, California, 1991.

[4] Inference Corporation, ART-IM Programming Language Reference, Version 2.5, Inference Corporation, 550 N Continental Blvd., El Segundo, California, 1991.

[5] Taylor, Kevin B., et al, Use of Case-Based Reasoning to Assist Patient Management in an Intensive Care Unit, Proc. of the Joint Conference of COMP and the CMBES, May 1993, Ottawa, pp.248-249.

[6] MacKinnon, Susan J., Gaskin, Nicole F., The Use of a Knowledge-Based System in the Management of Rheumatoid Arthritic Patients, Undergraduate Thesis, U.N.B., April 1995.

[7] Frize, M., et al, Computer-Assisted Decision Support System for Patient Management in an Intensive Care Unit, Proc. of the Eighth World Congress on Medical Informatics, July 1995, Vancouver, pp.1009-1012.