Web-Based Health Care Guideline Agents

This paper describes efforts to develop and field an adaptive health planner and parallel agents that securely connect distributed users and data sets in a collaborative approach able to anticipate health items and to remind and alert about these items over the web. The planner and agents, reasoning temporally and non-monotonically, are able to handle changing health conditions, unknown health facts, user wellness and lifestyle preferences, and updates about each user's situations. A set of robust, open standards-based client server approaches (e.g., CORBA, ODBC, Java, and ANSI HL7 Common Object Model) are used to encapsulate and plug all these components into the web, and to handle the distributed agent and object communications. Extensions to these emerging standards are at times made due to engineering and domain considerations. From this perspective, the approach proposed in this paper is an experiment in an open standards framework for middleware in the healthcare field. This research also tries to reflect "lessons learned" about the extensions needed in these standards if healthcare middleware frameworks are to transparently support users over the web.