Context Aware Workflow as Autopilot

The effective use of adaptive workflow engines in the health care enterprise can be thought of as an autopilot. A simple autopilot does not fly the plane, and it does not perform the tasks of aircraft subsystems. It monitors and observes that a process is veering off an established course. It alerts the pilot to unusual events and makes minor adjustments in trajectory by gently tweaking subsystems to push the aircraft toward level flight on a predetermined heading. In this way, the autopilot prevents major errors by handling multiple small errors and correcting them. More advanced autopilots use terrain following radar and other subsystems to automatically fly the plane. However, even then, they rely on a hierarchy of subsystems to actually execute flight.

For example, in designing a robotic system like the DARPA Trauma Pod, which is planned to autonomously perform surgery on the battlefield, a workflow engine will not do the surgery. It will check that the

Trauma Pod is ready for surgery, handle the logistic aspects of assuring supplies and instruments are available, cleanup after robotic surgery is complete, restock, sanitize, and prepare for the next surgery. In addition, it will assure no sponges are left inside the patient, using Radio Frequency IDentification (RFID) technology. When it sees exceptions, it will alert hierarchical subsystems or external systems to take appropriate action. This approach, which MIT Prof. Rodney Brooks calls a "subsumption" architecture, is a way to take a large collection of dumb subsystems and orchestrate them to exhibit intelligent behavior in robotic design [8]. The same approach can be taken to monitor heterogeneous distributed systems in a health care enterprise, which are "dumb" in the sense they cannot communicate well with one another or adjust well to one another. A higher-level workflow engine operating like an autopilot can alert subsystems or clinicians to perform adjustments to patient processes before major problems occur. Hence, in this manner, a large collection of dumb subsystems can be made to appear "smart."

An adaptive workflow engine can be used to orchestrate the behavior of the many disparate health care systems in a surgery center through direct integration via Web services, HL7 standards-based messaging, a hybrid solution, or proprietary adaptors [32]. The ability to "capture" the function of legacy components in an enterprise is a standard complex adaptive systems strategy and the basis for introduction of intelligent agents into advanced software systems [4].

Another important aspect of an autopilot is total situational awareness of what is happening with subsystems and using that awareness to unobtrusively alter aircraft behavior. The pilot of an aircraft wants an autopilot to do its task so well that its operations are transparent. In order to promote user adoption of new clinical processes in a health care enterprise, the workflow engine must be transparent to routine operations and only become visible when a critical event occurs. The introduction of RFID technology for capture of critical data on patients, staff, instruments, and supplies helps to make this possible.

0 0

Post a comment