=Paper=
{{Paper
|id=Vol-131/paper-5
|storemode=property
|title=Process Oriented Knowledge Management to Support Clinical Pathway Execution
|pdfUrl=https://ceur-ws.org/Vol-131/KMM05_No04_Jablonski.pdf
|volume=Vol-131
}}
==Process Oriented Knowledge Management to Support Clinical Pathway Execution==
https://ceur-ws.org/Vol-131/KMM05_No04_Jablonski.pdf
Process Oriented Knowledge Management to Support
Clinical Pathway Execution1
Stefan Jablonski, Christian Meiler, Sascha Müller, Rainer Lay
Institute for Computer Science – Department of Database Systems
University of Erlangen-Nuernberg
Martensstr. 3, D- 91058 Erlangen
{stefan.jablonski, christian.meiler, sascha.mueller, rainer.lay}
@informatik.uni-erlangen.de
Abstract. Processes play an important role in modern hospital settings. Both
cost pressure and increasing quality requirements force hospitals to redesign
their processes and to document their execution and results. Clinical Pathways
are evidence based and interdisciplinary treatment processes for specific
diagnosis. Along such treatment processes a lot of documents – both patient
specific and unspecific – are produced. All these documents contain knowledge
which is interesting for further reuse. This paper presents an approach for
process oriented knowledge management along clinical pathways by using
semantic document tagging generated by rules in the pathway models.
1 Situation
The public health system in Germany is currently undergoing major changes caused
by several factors [3]. The goal of this effort is to reduce cost and at the same time
improve efficiency and quality of patient treatment. Both goals can be achieved by
using information technology in medical environments [2].
In a typical hospital setting, the medical staff has to deal with several different
information systems such as hospital information systems (HIS) [7], tool specific
software or laboratory systems. The ideal conception - to collect, store and analyze all
information in a single big HIS - is often not put into practice and the medical staff
has to spend time and effort to collect the required information from several IT
systems [6].
The requested information can be divided into various categories by certain criteria
such as type of the disease, date of the patient treatment, clinical ward or responsible
physician. In order to improve clinical processes, this type of information has to be
easily and quickly accessible. In practice, this is problematic because relating data
across system boundaries requires either a lot of time or detailed IT knowledge. In
consequence, the medical staff has to be supported by a knowledge management
system (KMS) that allows performing similarity searches in the existing clinical data.
1 Acknowledgement: Part of this work is funded by the Deutsche Forschungsgemeinschaft
(DFG)
Clinical data is often contained in documents2 like for example findings, patient
records, measurements, etc. Most of these documents are created, maintained and
analyzed by the medical staff using existing clinical IT systems. Thus, the medical
staff knows this kind of documents very well and can effectively extract the needed
information. The remaining documents are from other sources and often used for
reference, like medical guidelines [1], lookup tables, medical terminologies or
checklists. This kind of documents must be easily accessible for the medical staff, as
well.
A clinical pathway [4] describes the patient’s treatment process within a hospital
and therefore is a very characteristic and well known means of orientation for the
medical staff. Thus, a clinical pathway can be regarded as one common criterion to
structure the medical knowledge. Clinical pathways are built on evidence based
medicine [4] and clinical practice knowledge adapted to a specific organizational
environment. Documents play an important role in both stages of a clinical pathway:
Design and execution. For example, medical guidelines are needed to set up a valid
clinical path in the design phase. During the execution of a clinical pathway
documents are stored and requested as well, e.g. a physician needs to take a look at a
radiograph during the examination or a nurse needs a checklist to prepare the
operation room for the next surgery. The documents provide both background
information and context data.
2 Semantic tagging along the clinical pathway
The situation described in the first section implies that a knowledge management
solution for a hospital should be based upon the existing medical documents that are
used along the patient treatment process. In fact, using existing documents has several
advantages: In contrary to a newly generated representation of extracted knowledge, a
more efficient comprehension of the presented knowledge is guaranteed. Accordingly,
the acceptance is raised as there is no need for the medical staff to understand the
structure of new documents. Furthermore, existing documents can simply be reused
and the expensive creation of new documents can be omitted.
The medical documents originate from very different sources and need to be
organized and related to each other. A basic principle to structure the knowledge is to
tag the documents with different characteristic attributes (e.g. key words) [5]. By
adding predefined attributes, the documents gain semantic information that can be
automatically processed by a KMS. Typical examples for such attributes are the type
of the disease or the responsible physician. What criteria are to be used throughout the
KMS depends on the individual needs of the organization. It is very important to
define a common understanding of the classification throughout the organization.
The tags cannot be assigned to every document manually. Tagging rules have to be
defined that describe what type of attribute has to be assigned in which way to what
kind of document. This can be done during the design of the clinical pathway, i.e.
knowledge management has to be considered already at the conceptual level (cf. Fig.
2 In this context, the term “documents“ stands for all kinds of knowledge carriers, like for
example forms, diagrams, images, documents, multimedia content, etc.
1). At the modeling stage all necessary data is available: Document source and target
systems, data conversion rules, data types, organizational assignments, etc. Using this
information, it is possible to define what data affects the attributes of a document. In
general this relation is not straightforward, i.e. a simple mapping between the
attending physician and the attribute “responsible physician”, is possible but it is not a
typical example. Often more complex combinations have to be reflected in these
rules, e.g. weight, size and a formula describe the body mass index and are thereby
part of the risk classification of a person. The rules need not only to define how the
data is transformed, but also how the data is extracted from the document and
attached to it afterwards. In order to foster reuse of these rules, they can be stored in a
repository and applied to other compatible document types as well. The tagging rules
are applied during the execution of the clinical pathway (cf. Fig. 1).
Automatic generation of classification data is, especially in the medical domain,
Tagging rule
Rule: Document Rule:
Set interviewer Set physician
Rule: Rule:
Set tool Set tool
Computer Tonometer
Patient record Measurement Data
MTA Physician Application
Type
Role
Clinical Pathway Put patient
on pathway
Instance Dataflow
PC am03 NC Tonometer
Patient: Smith IOP: 5 mm Hg
Mr. Green Dr. Myers
Process step Attributes: Attributes:
Set interviewer Set physician
Interviewer: Physician:
Set tool Mr. Green Set tool Dr. Myers
Applied Tool: Tool:
tagging rule PC am03 NC Tonometer
Attributes/Tags
Fig. 1. Tagging rules are defined at the conceptual level of a clinical pathway
restricted to certain areas and must not replace the decisions of qualified medical
personnel. Accordingly, it has to be possible for the medical staff to change the
attributes of documents if necessary, i.e. the attribute assignment of the KMS must
not be static. In order to maximize acceptance and minimize the necessary human
interference, the attribute to value assignment of the documents has to consider more
than just the last few steps of the clinical pathway. For example, the attribute
diagnosis, which might be relevant for most of the collected documents, is normally
only known at the end of the clinical pathway. The context of the documents, i.e. the
clinical pathway, has to be considered when assigning the attribute values of the
documents. This can be achieved by different strategies, for example by assigning the
attributes of all documents not before the end of the execution of the clinical pathway
or by using hierarchical attribute sets.
The attribute types define dimensions of a multidimensional classification space.
The allowed values of the attributes are the marks within a dimension. The
dimensions should be orthogonal to each other in order to avoid misclassification of
documents. The structure of the clinical pathways is a well known and very
significant means of orientation for the medical staff and should be emphasized.
Further eligible candidates for dimensions are for example diagnosis, responsible
physician, ward or type of examination. According to our experience, it turns out that
four to seven categories (i.e. attribute types) are sufficient to reach a reasonable
tradeoff between usability and expressiveness of the KMS.
3 Benefits and Outlook
The basic idea of our approach is to collect and tag documents used during the
execution of a clinical pathway. Reusing documents increases the acceptance by
presenting the knowledge in a familiar style to the medical staff. The documents are
semantically tagged by the system according to rules defined once during the design
of the clinical pathway. These semantically enriched documents can be directly linked
to the process based KMS described in [5] or used in the context of semantic web
applications to support clinical work.
References
[1] AWMF, ÄZQ: Leitlinien Manual. In: Zeitschrift für ärztliche Fortbildung und
Qualitätssicherung. Urban & Fischer Verlag, Jena, Januar 2001.
[2] Beyer M., Kuhn K., Meiler C., Jablonski S., Lenz R.: Towards a flexible, process-oriented
IT architecture for an integrated healthcare network. SAC 2004, 2004.
[3] Bundesministerium der Justiz: Gesetz zur Reform der gesetzlichen Krankenversicherung ab
dem Jahr 2000. Stand: 17.12.1999.
[4] Hellmann, W. (Hrsg.): Klinische Pfade. Konzepte, Umsetzung, Erfahrungen. Ecomed
Verlagsgesellschaft, Landsberg/Lech, 2002.
[5] Jablonski, S., Horn S., Schlundt M.: Prozessorientiertes Wissensmanagement mit der
i>WorkBench. In: Abecker, A.; Hinkelmann, K.; Maus, H.; Müller, H.J. (Hrsg.):
Geschäftsprozessorientiertes Wissensmanagement, 2002.
[6] Jablonski, S., Lay, R., Meiler, C., Müller, S.: Process Based Data Logistics: A solution for
Clinical Integration Problems. In: Rahm, E. (Hrsg.): First International Workshop on Data
Integration in the Life Sciences (DILS 2004), LNBI 2994: Springer Verlag, 2004.
[7] Prokosch, H.U.: Hospital Information Systems: A Pragmatic Definition. In: Prokosch, H.U.,
Dudeck, J. (Hrsg.): Hospital Information Systems: Design and Development Characteristics;
Impact and Future Architecture. Elsevier, Amsterdam, XIII-XVI, 1995.
[8] Sacket, D.L.; Richardson, W.L.; Rosenberg W.: Evidenzbasierte Medizin - EbM –
Umsetzung und Vermittlung. Deutsche Ausgaben, München, 1999.