Routine Use of Integrated Knowledge Based Functions - Some Prerequisites and Results

Tafazzoli AG, Altmann U, Katz FR, Wächter W, Hölzer S, Dudeck J

Institut für Medizinische Informatik, Justus-Liebig-Universität Gießen,
ali.g.tafazzoli@informatik.med.uni-giessen.de

Introduction

In the context of clinical cancer registries, there are a lot of useful ways to help staff with relatively unsophisticated knowledge based functions.
The objectives have been the developement of an Arden-Syntax based shell for integrated knowledge based functions, and its clinical use in the setting of the "Gießener Tumordokumentationssystem" (GTDS). First domain of application has been helping the documentation personell in recognizing data inconsistencies and incompleteness by confronting them with appropriate messages timely to the documentation process. This is intended to reduce the time needed for data correction, because the documentation personell is still involved in the potentially incorrect case.

Methods and Results

The ARDEN-Syntax permits the definition of Medical Logic Modules (MLM) via a framelike mechanism. Due to the content types of an MLM, it can be used for integrated knowledge based functions (IKBF). Specifics of these IKBSs are the automated triggering, the reusing of existing data for reasoning and the automated presentation of messages. A typical phenomenon of IKBF is the emergence of "unjustified messages" caused by the automated triggering in conjunction with an asynchronous presentation of eventually generated messages. An analogous phenomenon has been observed in the Brigham Integrated Computing System [Kuperman 1997]. This phenomenon has to be addressed by retract methods.
Since mid 1998 the shell is in routine at now three settings. In a 6 months monitoring period in one setting almost 250 messages were presented to the documentating staff. Due to the implemented retract method half of the generated messages were recognized as "unjustified messages". About 70% of the judged messages were considered as correct by them. Other settings using the GTDS have requested for the IKBFs.

Conclusion

IKBF combined with a retract method promises user-accepted knowledge based functions. Additionnally, restricting to relatively unsophisticated IKBFs, knowledge-engineer acceptance can be obtained by structuring the knowledge base in strictly independent modules. Further potential domains are study management, and if documentation moves closer to the process of patient treatment the circulation of clinical guidelines. Finally we hope that - via useful IKBF - the clinical acceptance for structured documentation can be improved.

References

Kuperman GJ, Hiltz FL, Teich JM: Advanced Alerting Features : Displaying New Relevant Data and Retracting Alerts. Proc AMIA 1997, S247-249