Clinical Semantics - Computational Intelligence in Healthcare: Advanced Methodologies - page 14

(semi)automatic feedback from these observations to agent goals (i.e., functionality)
supported by high-abstraction level information fusion. The physical sensor interfacer
component handles syntactic language constraints in natural language user interactions
through the generation of all possible structural phenomena from the interactions on
hand. The physical goal prioritizer and explanation organizer components provide
high-abstraction level prioritization, ordering and information fusion functionality. The
physical action interfacer component serves as a semantic interface to various physical
information processing actions, including the physical user interface for an interacting
user. The physical, scheduled heartbeat component initiates actions based on the state
of the explanations. As a starting point, work towards a prototype of an open clinical
agent for emergency medicine has been presented in [8-11] as discussed next.
3.2 The Chief Complaint Agent
Emergency medicine represents a particularly challenging care setting for clinical
information management. Rapid and effortless information submission and retrieval
are essential to match the pace of a busy emergency department (ED). It is the point-
of-entry for a diverse population of patients into the health care system. ED's see
patients from newborns through old age, and treat the full range of illnesses and con-
cerns. Patients with acute trauma, chronic disease, potential exposures and social
problems can all be encountered on any given night. This wide range of patients and
ailments makes it more challenging to develop an efficient mechanism for the entry of
discrete medical data for ED encounters. [9]
The chief complaint (CC) is one of the most important components of ED triage
decision making. It is a key determinant of the direction and history taking, physical
examination, and diagnostic testing in the ED [17]. Still, the chief complaint has a
number of flaws. For example, the patient may not always be able to accurately de-
scribe their concern (e.g., “I don't feel well”), the same chief complaint can map to a
wide range of severity (e.g., 'chest pain' can be either indigestion or a heart attack).
Despite its flaws, the chief complaint is the first available description of what's wrong
with the patient, and is often electronically accessible in many hospitals. [9]
A number of studies (e.g., [18] and [19]) report activities towards a standard, lim-
ited set of encodings for the structured entry of chief complaint information in clinical
ED information systems. However, no consensus exists on this matter and it remains a
challenging and elusive goal. The purpose of the CC is to record the patient's primary
problem, and it is often recorded in his or her own words. Obviously, the implied
diversity of expression is difficult to compress into a set of strictly problem-oriented,
structured encodings without losing the nuances of original message. This is espe-
cially true if one wants to accommodate the rich and informative use of almost an
unbounded set of anatomic (e.g., problem related to which body part?) and functional
(e.g., problem related to which bodily function?) descriptions and positional and other
qualifiers (e.g., “left-sided”, “possible”, “sports-related”, etc.) all of which can be
expressed in a number of ways and on varying levels of abstraction with sometimes
subtly different semantics. [9]
There are many reasons for gaining understanding of the ED chief complaints in an
automated, computable context. Recently, the use of chief complaint information in
syndromic surveillance has gathered attention in medical informatics research. This