Granularity of medical software agents in ICU -trade-off performance versus flexibility

Abstract

Intelligent computing opens new opportunities to assist physicians with automatic medical decision support. Today, physicians still experience manual time-consuming data analysis of medical records while evaluating the patient's outcome. Especially in the Intensive Care Unit (ICU), large amounts of data per patient per day are generated, making manual data analysis a very complex task. Early detection of changes in patients' conditions is a major challenge in health care. To achieve this, agent technology and a service-oriented architecture have been deployed for medical decision support. In this paper, the choice of the level of service granularity is studied in detail. Fine-grained granularity of services seems an attractive idea to establish extreme reuse and flexibility but we need to be aware of possible performance implications. The modular design of an antibiotic switch therapy agent is presented and trade-offs between performance and flexibility are thoroughly evaluated. © 2008 Springer-Verlag Berlin Heidelberg.