Clinical Semantics - Computational Intelligence in Healthcare: Advanced Methodologies - page 15

research is carried out in the context of ultimately developing practical applications
for monitoring the incidence of certain syndromes associated with bioterrorism and
other emerging infections with the goal of timely outbreak detection. In addition, we
want the computer to understand the chief complaint (CC) so that we can [9]:
- Quickly group similar cases for research purposes
- Expedite care by suggesting predefined care protocols to be initiated at triage
- Improve patient safety by alerting providers to conditions they might not have con-
sidered
- Improve efficiency by pre-selecting documentation templates for physicians
However, if we choose not to abandon collection of free-text chief complaint in-
formation, how do we optimize the capture and understanding process? We will next
briefly discuss and illustrate the ways to accomplish this with open information man-
agement. This is done with the specific goal of distilling categorical information from
chief complaints, for example, for automated syndromic surveillance, with a chief
complaint agent.
Figure 6 describe the architecture of a system for this purpose.
Fig. 6. Overview of the chief complaint agent architecture
The system provides for the capture and storage of chief complaints as natural lan-
guage compositions maintaining their full expressivity while a computable under-
standing of them is collaboratively built enabling distillation of categorical informa-
tion from the CC's. The core of the system is an unrestricted, associative linkage of
synonym and category definitions mapping features in free-text chief complaints into
higher abstraction level categories 1 as constrained by the operational domain. [9]
In practice, the associative linkage defines the production rules of a rewriting sys-
tem for a chief complaint language. Figure 7 illustrates the normalization of the
1 P: Problem
A: Anatomic location/function/property
Q: Qualifier
D: Diagnosis
I: Ignorable
U: Unknown