=Paper= {{Paper |id=Vol-1320/paper_23 |storemode=property |title=A Work Domain Ontology for Modeling Emergency Department Workflow |pdfUrl=https://ceur-ws.org/Vol-1320/paper_23.pdf |volume=Vol-1320 |dblpUrl=https://dblp.org/rec/conf/swat4ls/TaoOMMRKZF14 }} ==A Work Domain Ontology for Modeling Emergency Department Workflow== https://ceur-ws.org/Vol-1320/paper_23.pdf

A Work Domain Ontology for Modeling Emergency De-
partment Workflow
Cui Tao PhD1, Nnaemeka Okafor MD, MS2, Amit Mehta MD2,
Charles Maddow MD2, David Robinson MD MS2, Brent King MD MMM3,
Jiajie Zhang PhD1 and Amy Franklin PhD1
1
School of Biomedical Informatics,
The University of Texas Health Science Center at Houston, Houston, TX,
2
Department of Emergency Medicine, School of Medicine,
The University of Texas Health Science Center at Houston, Houston, TX
3
Nemours Alfred I. DuPont Hospital for Children, Wilmington, DE

Abstract Emergency Department clinicians perform life-critical tasks that require acquisi-
tion, processing, transmission, distribution, integration, search, and archiving of significant
amount of data in a distributed team environment in a timely manner. In order to better re-
veal the complexity of emergency care and reflect such a complexity in information system
design, we need an abstract description of the clinical and cognitive work performed by cli-
nicians, independent of how the clinical setting is implemented with specific technology, ar-
tifacts, and environmental variables. For this purpose, we developed a work domain ontolo-
gy for the ED (ED-WDO). We evaluated the semantics of the ED-WDO with domain ex-
perts and its application and usage using an emergency nurse assessment use case. From the
evaluation results, we can conclude that the lexical and semantic definitions of the classes,
the hierarchical structure, as well as the semantic relation definitions in the ED-WDO are
well defined and can faithfully represent the ED work domain.
Introduction
Emergency Department (ED) clinicians perform life-critical tasks that require acquisition,
processing, transmission, distribution, integration, search, and archiving of significant
amount of data in a distributed team environment in a timely manner. ED clinicians
monitor their constantly changing information environment, respond to unpredictably
occurring issues, collaborate and communicate with other people in the system as issues
arise, and prioritize and solve multiple issues as they occur. Managing information needs
and supporting clinical decision making in ED is of great importance for patient safety
and healthcare quality [2]. Rather than focusing on a single task at a time, ED clinicians
are forced to switch between multiple tasks and usually multiple patients. Many of these
switching decisions are based on unplanned, unorganized, and unpredictable environmen-
tal factors. This high level of complexity in the ED is one major factor that contributes to
potentially preventable adverse events [3]. Recent studies show that the complexity of
critical care can be addressed in a systematical way from a cognitive perspective [2, 4].
One fundamental step towards reducing the complexity of the ED is to recognize what
information is needed and processed by clinicians, the activities they perform with these
information and decisions they make regarding these information items and activities.
In order to better reveal system complexity and reflect such a complexity in information
system design, we need an abstract description of the clinical and cognitive work per-
formed by clinicians, independent of how the clinical setting is implemented with specific
technology, artifacts, and environmental variables. The work domain ontology (WDO) is
a framework for this purpose [5]. In this paper, we introduce our effort on developing a
work domain ontology for the emergency department (ED-WDO).
The ED-WDO is represented in the Web Ontology Language (OWL) [6]. OWL is a
standard ontology language that allows data and knowledge to be represented in a ma-
chine-understandable way (an ontology), which enables automatic intelligent queries and
semantic reasoning for the data. Successfully representing the ED work domain in OWL
will provide a standard and explicit ontological model for (ii) ED clinical & information
management processes, hospital business rules & resources planning; and (ii) triage and
decision support in the ED.
ED Work Domain Ontology (ED-WDO)
A Work Domain Ontology (WDO) outlines the basic structure of the work that the system
together with its human users will perform [7-9]. It is an explicit, abstract, implementa-
tion-independent description of that work. It describes the essential requirements inde-
pendent of any technology systems, strategies, or work procedures. It tells us the inherent
complexity of work; it separates work context (physical, organizational, computational,

Figure 1: Overview of the Work Domain Ontology [1]

etc.) from the nature or functions of the work itself. A WDO is composed of goals, opera-
tions (or actions), objects, and the constraints that capture the functions of work. Figure 1
shows the four fundamental components (goal, object, operation, and constraint) of WDO
and their definition, scopes, attributes, and relations. The WDO is represented in OWL
for a standard and formal representation, where Goal, Object, and Operation are defined
as OWL classes, constraints among them are defined as object properties, and attributes
are defined as data properties.
On top of the WDO, detailed ontologies can be defined for specific work domains. Each
detailed WDO outlines the basic structure of the work that a system for that work domain
together with its human users required for the work. It provides an explicit, abstract, im-
plementation-independent description of the specific domain of work. In the next section,
we introduce our implementation of a work domain ontology for the ED work domain.
The ED work domain ontology (ED-WDO) includes classes that define operations and
objects, as well as the goal for each operation. The ED-WDO is built on top of the WDO
and it adopts all the concepts, properties, and constraints defined in the WDO. The ED-
WDO was built with the additions of the essential classes and their constraints specifically
for the ED work domain. Following the American College of Emergency Physician Defi-
nition, the practice of emergency medicine includes “the initial evaluation, diagnosis,
treatment, and disposition of any patient
requiring expeditious medical, surgical,
or psychiatric care.” [10]
Define the Emergency Department Staff
Objects: We first defined the ED staff
object classes in the ED-WDO. Figure 2
shows these classes and their hierarchical
information in the protégé ontology edi-
tor. As Figure 2 shows, ED_Staff is an
OWL class which is a subclass of the
Object class defined in WDO. We further
classified ED staff into four categories:
Administrative Staff, Clinicians, Emer-
gency Room Technicians, and Nurses,
each of which is defined as an OWL
class. Under each of these category clas- Figure 2: ED staff objects defined in ED-WDO
ses, further classes can be defined. For ex-
ample, a provider can be an attending physician, resident, or an Advance Practice Profes-
sional (nurse practitioner, physician assistant). For each class, a textual definition can be
defined. Acronyms and alternative labels of each object class can also be defined if appli-
cable. For example, “Advance Practice Professional” can also be called as “APP” or “lim-
ited license provider”. These can be defined as alternative labels of the class.
Define the Emergency Department Operations: We then defined the operation compo-
nents for a typical ED patient from arrival to departure. Column 1 in Error! Reference
source not found. shows the meta-level operations in the ED-WDO. Each operation is
defined as an OWL class (also a subclass of the general Operation class in WDO). Each
operation can be further classified into different subclasses. For example, we can further
define different disposition types such as Admit, AMA (Against Medical Advice), LWBS
(Leave Without Being Seen), Transfer, and Home/Self-care according to the American
College of Emergency Physicians Emergency - Department Medical Record Elements
[11].
Define Goals: We also defined major (intermediate) goals for ED visit. Column 3 in Er-
ror! Reference source not found. shows the details. Each goal is defined as an OWL
Table 1: High level ER-WDO components
class and a subclass of the Goal class de-
Operation Required Object Goal
Arrival Medical Receptionist Check in
Triage Triage Nurse Determine Emergency Severity Index Category
Nurse Assessment Care Area Nurse Collect initial encounter data
MSE Clinician Determine whether an emergency medical condi-
tion (EMC) exists
Administration Admin Staff Billing
Provider Assess- Clinician or Nurse Diagnosis for treatment
ment
Test Ordered by clinician or nurse Obtain Information for Assessment and diagnosis

Treatment Clinician or Nurse Provide initial treatment/stabilize the patient
Disposition Clinician Sign to discharge the patient
Departure Care area nurse Instruct the patient for departure
fined in the WDO.
Define the Required Objects and Goals: Error! Reference source not found. shows the
details. We have defined 10 meta-level operations, their required objects (medical profes-
sionals), and the major goal for these operations. For example, the hospital must provide
an appropriate medical screening examination to determine if an emergency medical con-
dition exists [10]. Therefore for the operation MSE (Medical Screening Exam), the re-
quired object is Clinician since an MSE can only be done by a health care provider or an
advanced practice professional. We use an OWL restriction to define this condition:
MSE
requireObject
some
Clinician

where requireObject is an OWL object property defined in the WDO for specifying any
required object for a given operation; some represents owl:someValuesFrom axiom which
formally defines that MSE requires at least one Clinician to be its object. Please note that
in some clinical settings, MSE can be performed by nurses. This constraint can be adjust-
ed to allow nurses to be associated with the requirements for this object.
We also defined relationships between operations and goals using the requiresOperation
property defined in WDO. For example, the goal “Determine whether an emergency med-
ical condition (EMC) exists” in ED requires operation MSE. We can use an OWL re-
striction to define this condition:
“Determine
whether
an
emergency
medical
condition
(EMC)
exists”

requiresOperation
some
MSE
where requiresOperation is an OWL object property in the WDO for specifying any re-
quired operation for a given goal; some represents the owl:someValuesFrom axiom which
defines that the goal requires at least one MSE to be its operation in the ED workflow.
ED-WDO Expert Evaluation
We followed the ontology evaluation criteria introduced by Brank et al [12] to evaluate
the WDO-ED meta ontology. The ontology evaluation criteria cover several levels: syn-
tactic, lexical, hierarchical, semantic relations, and context and application. The OWL
ontology has been validated using HermiT reasoner v1.3.8 embedded in Protégé 4.3
(http://protege.stanford.edu/) for syntactic and consistency checking. For lexical, hierar-
chical, and semantic relations defined in the ontology, we interviewed four ED clinicians
from two different hospital systems for manual evaluations of the ontology. The members
of the review panel are not involved in the development of the ontology. We refined the
ED-WDO according to the review panel’s feedback until the experts agreed that the on-
tology reasonably represents the ED work domain. For context and application evalua-
tion, we evaluated the ontology on an emergency nurse assessment use case, which we
will discuss in the next section.
ED-WDO Use Case Evaluation
The ED-WDO models the basic backbone structure of the ED work domain which is in-
dependent of any artifacts, healthcare settings, or implementations. It also provides the
flexibility to be extended for any specific settings, requirements, or focus. The Emergency
Nurses Association (ENA), for example, provides a guideline for the workflow of nurse
assessment and documentation for different patients [13]. Here we use it as a use case to
evaluate and illustrate how the ED-WDO can be applied and extended to model the spe-
cific work domain for nurse assessment.
Initial assessment:
For the initial assessment, patients with different emergency severity levels need to follow
different workflows. We first need to model patients with different levels of emergency
severity as different object classes (subclasses of the Object class in WDO) in the extend-
ed ED-WDO for nurse assessment. Figure 4 shows the patient classification using the
Emergency Severity Index (ESI) per ENA. As specified in the ED-WDO, the Emergency
Severity category for each patient is decided on the triage stage. Only patients classified
as levels 3-5 needs to complete a full nurse assessment. Patients classified as levels 1-2,
on the other hand, require immediate medical interventions and will not be delayed in
order to complete a full nurse assessment. In this case, we can add a new constraint to the
“Nurse Assessment” Operation,
“Nurse
Assessment”
requireObject
only
(Patient_Level3
or
Patient_Level4
or
Patient_Level5)
which indicates that this operation only requires patients classified as level 3, 4, or 5.
Figure 4 Patient Category using the Emergency Severity Index (ESI)

• Level 1 = Critical: Every 5-15 minutes as needed and no less frequently than every hour for
the first four hours, then every 2 hours if clinically stable.
• Level II = Emergent - vital signs no less frequently than every hour for the first four hours, then
every 2 hours if clinically stable.

• Level III = Acute – vital signs no less frequently than every two hours for the first four hours, then
every four hours if clinically stable.

• Level IV = Urgent – vital signs per acuity and clinical assessment, but no less than every four hours.
• Level V = Minor - vital signs per acuity and clinical assessment, but no less than every four hours.

Figure 3: Reassessment Guidelines for Patients with Different Levels of Emergency Severity

Reassessment:
The ENA also specified guidelines for reassess patients according to their Emer-
gency Severity category.
In order to represent different temporal factors for reassessment for patients in different
categories, we adopted the time representation from the clinical narrative temporal rela-
tion ontology (CNTRO) [14]. CNTRO specifies how to represent different kinds of tem-
poral relations and expressions including repeated events with frequencies. Figure 5
shows an example of how to represent the reassessment operation for level 2 patients.
There are two stages involved in this operation: Reassessment Critical Stage 1 and Reas-
sessment Critical Stage 2. These stages can also be represented in OWL. For example, we
represented “Reassessment Critical Stage 1” as follow:
“Reassessment
Critical
Stage
1”
frequency
some

(Frequency
and
(hasUnitOfMeasure
value
hour)

and
(hasValue
some
int[<
1]))

“Reassessment
Critical
Stage
1”
duration
some

(Duration
and
(hasUnitOfMeasure
value
hour)

and
(hasValue
value
4))
Figure 5: Example of representing reassessment temporal pattern

Assessment for patients with different ages:
Additional assessments may be required for patients with different ages. For example,
patients under 18 months of age will have a head circumference measured. We also need
to be able to represent this kind of constraints for operations. For this example, we first
need to represent the objects that satisfy the constraint “under 18 months of age”. We use
an object property hasAge to represent a patent’s age. We then define the constraint has-
Value some int[<18] with unit of measures as month to represent “under 18 months”. The
relation between the operation and the object can be specified as:
“Head
Circumference
Measure”
requireObject
(Patient
and
hasAge
some
(Age
and
(hasUnitOfMeasure
value
month)

and
(hasValue
some
int[<
18]))
Conclusion and Future Direction
In this paper, we introduce our preliminary effort to create an ontology for the Emergency
Department work domain (ED-WDO). The ED-WDO includes necessary ED operations
and objects, as well as the goal for each operation. It outlines the basic structure of the ED
work that the system together with its human users will perform. It is an explicit, abstract,
implementation-independent description of the ED work.
The primary purpose of ED-WDO is to serve as an abstract ED model for understanding,
measuring, and designing cognitive work to increase care quality and patient safety. By
identifying the ED-WDO, we will know the work that has to be done. All other factors,
including how the work is implemented, how it is performed procedurally by users and
machines, and how different designs affect user performance, can then be examined. In
other words, with ED-WDO, we can explore how decisions are made, care given, and
information sought in EDs that vary in the degree to which they have adopted electronic
health records, use health information technology, or vary according to implementation
specific idiosyncrasies. In our project, we are currently using the WDO-ED to design am
information visualization system with multiple levels of details to support opportunistic
decision making by clinicians.
We evaluated the semantics of the ED-WDO with domain experts. From the evaluation
results, we can conclude that the lexical and semantic definitions of the classes, the hierar-
chical structure, as well as the semantic relation definitions in the ED-WDO are well de-
fined and can faithfully represent the ED work domain. For the context and application
criterion, we evaluated the usage of the classes and properties, on an emergency nurse
assessment use case. The results also indicated that the ED-WDO can be used to represent
the use case.
Several future directions we would like to pursue to extend and improve the ED-WDO.
First, the WDO focuses on operations and goals. It declares constraints such as required
objects for an operation, or required operations for a goal. We would like to provide more
flexibility on defining the constraints in ED-WDO, e.g., to be able to model required op-
erations or required goals (of operations) for different categories of patients. We then need
to propose new properties to define this kind of relations. Second, an ontology develop-
ment process is usually iterative. We plan to evaluate the ED-WDO using more use cases
and data. Based on the results, further improvement can be done to the ontology itself.
Acknowledgement:
This research is supported by AHRQ under grant 5RO1HS021236-02.
References
[1] A. Franklin, D. J. Robinson, and J. Zhang, "Characterizing the Nature of Work and Forces
for Decision Making in Emergency Care," in Cognitive Informatics in Health and
Biomedicine
Case Studies on Critical Care, Complexity and Errors, V. L. Patel, D. R. Kaufman, and T. Cohen,
Eds., ed, 2014.
[2] M. Smith and C. Feied. The emergency deaprtment as a complex system. Available:
http://nesciorg/projects/yaneer/emergencydeptexpdf.
[3] P. Croskerry, K. Cosby, S. Schenkel, and R. Wears, Patient Safety in Emergency Medicine
2008.
[4] P. Nugus, K. Carroll, D. G. Hewett, A. Short, R. Forero, and J. Braithwaite, "Integrated
care in the emergency department: a complex adaptive systems perspective," Soc Sci Med,
vol. 71, pp. 1997-2004, Dec 2010.
[5] E. Markowitz, E. V. Bernstam, J. Herskovic, J. Zhang, B. Shneiderman, C. Plaisant, et al.,
"Medication Reconciliation: Work Domain Ontology, prototype development, and a
predictive model," AMIA Annu Symp Proc, vol. 2011, pp. 878-87, 2011.
[6] . OWL Web Ontology Language. Available: http://www.w3.org/TR/owl-features/
[7] K. Butler, J. Zhang, C. Esposito, A. Bahrami, R. Hebron, and D. Kieras, "Work-centered
design: A case study of a mixed initiative scheduler " in CHI 2007 Proceedings, ed, 2007,
pp. 747-756.
[8] J. Zhang and K. Butler, "UFuRT: A work-centered framework for the design and
evaluation of information systems," in Proceedings of HCI International, 2007.
[9] J. Zhang and M. Walji, "TURF: toward a unified framework of EHR usability," Journal of
Biomedical Informatics, vol. 44, pp. 1056-1067, 2011.
[10] H. M. McBreen and M. A. Jack, "Evaluating humanoid synthetic agents in e-retail
applications," Systems, Man and Cybernetics, Part A: Systems and Humans, IEEE
Transactions on, vol. 31, pp. 394-405, 2001.
[11] American College of Emergency Physicians. Emergency Department Medical Record
Elements. Available: http://www.acep.org/WorkArea/DownloadAsset.aspx?id=29880
[12] J. Brank, M. Grobelnik, and D. Mladenic, "A Survey of Ontology Evaluation
Techniques," in Proc. of 8th Int. multi-conf. Information Society, 2005.
[13] . Emergency Nurses Association Emergency Department Assessment and Documentation
Standards. Available:
http://www.ena.org/membership/document_share/policies/Documents/EDAssessmentDoc
Standards.pdf
[14] C. Tao, W. Q. Wei, H. R. Solbrig, G. Savova, and C. G. Chute, "CNTRO: A Semantic
Web Ontology for Temporal Relation Inferencing in Clinical Narratives," AMIA Annu
Symp Proc, vol. 2010, pp. 787-91, 2010.