Content uploaded by Danny Ammon
Author content
All content in this area was uploaded by Danny Ammon on Jun 29, 2015
Content may be subject to copyright.
Management of Knowledge-intensive
Healthcare Processes on the Example of
General Medical Documentation
Danny Ammon1, Dirk Hoffmann2, Tobias Jakob2, Ekkehard Finkeissen2,
Vesko Detschew1, Thomas Wetter3
1 Technische Universität Ilmenau, Gustav-Kirchhoff-Str. 2, 98693 Ilmenau, Germany
2 medrapid GmbH, Schweizertalstr. 35, 69118 Heidelberg, Germany
3 Universitätsklinkum Heidelberg, Im Neuenheimer Feld 672, 69120 Heidelberg
{danny.ammon, vesko.detschew}@tu-ilmenau.de
{dirk.hoffmann, tobias.jakob, ekkehard.finkeissen}@medrapid.info
thomas.wetter@med.uni-heidelberg.de
Abstract. Healthcare Processes are characterized by knowledge-intensive tasks.
In contrast to this, most of the efforts for business process management in
healthcare do not refer to this quality, and software engineering in healthcare
relies on an unspecific process-oriented approach. In this contribution, we
present a method of capturing and analysis of a knowledge-intensive process,
from which we derive requirements to a knowledge- as well as process-oriented
information system for the example of general medical documentation. We
present the resulting implementation of a knowledge-based electronic patient
record and discuss the potentials and open issues for our proposal.
Topics: process modeling in healthcare, process-oriented system architectures
in healthcare, integrating healthcare processes with electronic medical records
1 Introduction
1.1 Motivation
Successful approaches of process management in the branches of healthcare as well
as in any other sector will have to reflect the special characteristics of the domain they
deal with. For healthcare this is the medical domain, applied in everyday patient
treatment. The domain of medicine is very complex, subdivided into many specialized
fields, and is not only changing rapidly, but also growing exponentially. Furthermore,
the demand of up-to-date medical knowledge in patient supply is highly critical in
terms of time. We can therefore classify healthcare as a knowledge-intensive sector.
Knowledge utilized in patient treatment can at least be divided into three types:
1. General medical knowledge is a result of research and independent of individuals.
2. Institutional knowledge refers to the properties and functions of the special
healthcare institution where a clinician is acting.
26/103
A-PDF Split DEMO : Purchase from www.A-PDF.com to remove the watermark
2 Danny Ammon et al.
3. Patient-specific knowledge is generated during the consultation and the diagnostic
measures of the clinician.
It is the task of every professional working in the sector of healthcare, to map
general medical knowledge onto the information gathered about a specific patient,
and to combine this with the institutional knowledge to determine the optimal
diagnostic or therapeutic procedures for the ongoing process of treatment.
The leading questions for this contribution are now: How can the knowledge
intensity of healthcare be taken into account for a suitable business process
management in healthcare? How is it possible to add process knowledge to the
institutional knowledge of the clinician, where this important role and use of
knowledge in the processes themselves is incorporated? What are the advantages and
potentials of a knowledge-oriented business process management in healthcare? And,
last but not least: Can these potentials be utilized for deriving requirements and
specifications of an adequate process- and knowledge-oriented IT support? How
could such an information system look like? We will consider these questions here by
the exemplary process of general medical documentation and the associated
information system, the electronic patient record.
1.2 Overview of the Contents
In this contribution, we define a knowledge-oriented healthcare process management
approach using the example of general medical documentation. We begin by
providing a background and reviewing related work, including the suggestion of an
appropriate language for the modeling of knowledge-intensive business processes, the
Knowledge Modeling and Description Language (KDML). This language is then used
for establishing exemplary and generic models of patient treatment including
documentation (process model), and of clinical documentation itself (activity model).
In the next step, out of the models we present, requirements for a knowledge-
oriented information system for the process of clinical documentation are derived. As
a possible result of our approach, we present a knowledge-based electronic patient
record and its characteristics in the subsequent paragraph.
Finally, we discuss the potentials as well as open questions for the here-proposed
type of process management and information system.
2 Background and Related Work
2.1 Knowledge in Healthcare
The analysis of medical knowledge is a research topic both of theoretical medicine
and of healthcare knowledge management. Additional and more profound
differentiations of types of medical knowledge, have been accomplished there
[4][11][1]. The resultant knowledge dilemmas in medicine have been expressed
similarly: At first the impossibility for a modern clinician even in medical subareas to
access all available general scientific cognitions right at the moment of a clinical
decision in a special case [1]. Moreover, the difficulties which arise from the
27/103
Management of Knowledge-intensive Healthcare Processes 3
appliance of abstract knowledge to individuals [19]. An associated characteristic is
the retroactive effect of the invention of new therapeutic options on the differentiation
of disease patterns which adds to medicine’s exponential growth [6].
An additional classification of knowledge in general has been established by
Polanyi, who distinguished between formal, verbalized, communicatable, explicit
knowledge and human-bound, experiential, believed, sensed, tacit knowledge [15].
Both types of knowledge play an important role in healthcare, explicit knowledge as
general textbook knowledge and tacit knowledge as clinical experience both affecting
every clinical decision.
Management of knowledge in companies, as a different subject, will have to rely
on the mentioned findings on general and medical knowledge. An accordant
approach, the well-known model of the dynamics of knowledge creation has been
developed by Nonaka and Takeuchi [13], which in turn influenced the modeling
language we suggest for healthcare process management later in this and in the
following section. They divided the conversion of knowledge in business processes in
four types: internalization, which creates tacit knowledge, externalization, which
creates explicit information objects, socialization which refers to the communication
of tacit knowledge, and combination which means deriving new information from the
use of two or more information objects.
2.2 Formalization of Medical Knowledge
Explicit medical knowledge as a part of medicine has been subject to various kinds of
formalization, with or without IT support. Medical terminologies, such as the ICD
[22] or SNOMED [17] have been developed for coding, statistical analysis,
reimbursement or decision support. A representation of medical semantics and
clinical information [3] has been achieved through the creation of medical ontologies,
e.g. UMLS [21] or OpenGALEN [14]. Finally, the application of medical knowledge
bases for encyclopediae, didactics, as terminology server or as decision support
system, has been established as a result of medical knowledge engineering. An
example for a modern knowledge-based system is medrapid, a professional healthcare
internet portal for the distribution and exchange of general medical information in
health care [5]. The medrapid knowledge base serves as fundament for the exemplary
application of the findings in this paper to a process-oriented, knowledge-based
information system for general medical documentation, in the fifth section.
2.3 Models of Clinical Practice
Not only medical knowledge, but clinical practice as well has been subject to analyses
and modeling. The generic model of medical treatment which is shown in the next
section, relies on the proposal made in [12]. The author distinguishes two loops of
acting of a clinician: The diagnostic loop which is repeated until the findings for a
certain patient are sufficient to deduce a diagnosis, and the therapeutic loop which is
repeated until a special therapy for the given diagnosis has proven effective. This
model of the clinical action is very abstract and idealized (what happens if a given
28/103
4 Danny Ammon et al.
diagnosis has to be revised because of therapeutic findings?) but shall serve as a basis
for the simplified example proposed in this contribution. An overview and assessment
of generic models of clinical practice and their and direct utilization for the design of
electronic patient record systems can be found in [18].
2.4 Modeling Languages for Knowledge-intensive Business Processes
A first step in the management of business processes is their identification and
documentation. For this task, various process modeling languages have been
developed. In the last years, several approaches of extensions or additions to these
languages have been introduced, which allow a knowledge-oriented analysis of
processes [9]. Furthermore, there are several types of healthcare modeling languages
used for documentation of clinical practice guidelines or clinical pathways [1].
Out of these options, for this contribution we chose the general knowledge-process
modeling language KMDL (Knowledge Modeling and Description Language) for the
following reasons: Traditional business process modeling languages do not or only in
a static view regard knowledge objects in business activities [8]. Healthcare process
modeling languages, on the other hand, directly represent medical knowledge but do
not link it to objects or subjects and, in case of clinical practice guidelines, do not
incorporate institutional knowledge [1].
Knowledge process modeling languages like KMDL focus instead on knowledge
types, on conversion of knowledge in the context of business processes and therefore
provide a link between the utilization of domain-specific knowledge as it is
represented in medical guidelines, and the illustration of general business processes as
well as healthcare-specific activities. KMDL is directly based on the knowledge
management approach by Nonaka and Takeuchi [7][8]. The current KMDL version
2.1 provides two perspectives onto the sequence of actions, where the process view
shows the business process as an execution of single tasks, whereas the activity view
is used to consider the knowledge conversions during the fulfilling of a special tasks.
The defined entities for modeling of a process view as well as an activity view can be
found in fig. 1.
3 Modeling Knowledge-Intensive Healthcare Processes
In this section, we use the KMDL approach of capturing knowledge-intensive
business processes for the design of a generic model which focuses on clinical
treatment as a whole in the process view, whereas in the activity view we focus on the
general medical documentation as a type of knowledge conversion during the
treatment process.
We concentrate on clinical documentation as special knowledge-intensive process
because of the fact that every clinical action is the result of an interpretation of data
or information by the clinician. This generation of knowledge by interpretation itself
is one of the main reasons of the legal obligation for documentation in healthcare in
many countries. Therefore, the fundament of the activity of documentation is clinical
29/103
Management of Knowledge-intensive Healthcare Processes 5
control flow
IS
task
role
XOR OR AND
PI
affiliation
process interface
informationsystem
interconnection operators
process
socialization
(green
)
externalization
(blue
)
internalization
(red
)
combination
(gray)
person
Person
team
information
object
undefinded conversion
(black
)
listener
unde-
fined
Funktion
Ffunction
conversion
singular n times
singular n times
singular n times
-
singular n times
singular n times conversion
affiliation
knowledge object
task
requirement
Funktion
Mconversion
-
method
KMDLEntities: Process View
KMDL Entities: Activity View
Optional Objects
knowledge and its characteristics itself, distinguishing at as one of the basic activities
directly and completely referring to knowledge in healthcare as we defined it.
3.1 Process View of Clinical Treatment
The first approach of characterizing clinical documentation is a generic model of
clinical treatment itself, which provides information about the specific time when
documentation is necessary as well as the concrete circumstances which are to be
documented. A KMDL process view of clinical treatment, based on the statements in
[12] is shown in fig. 2.
Roles have been omitted in the process view since for this idealized model there is
only one actor—the clinician. The relevant information system for the further
consideration is the electronic patient record, which has been assigned to every task of
documentation.
Fig. 1. KMDL Entities for Process View (above) and Activity View (below).
30/103
6 Danny Ammon et al.
clinical treatment
determine patient-
specific data
document data
and interpretation
develop diagnostic
hypotheses
take diagnostic
measures
take therapeutic
measures
monitor
progression
interpret
progression
XOR XOR
treatment
complete
start of
treatment
interpret
determined data
document
measures
document
measures
document
progression and
interpretation
patient recordpatient record
patient record
patient record
diagnostic loop
therapeutic loop
Fig. 2. KMDL Model of Generic Medical Treatment (Process View)
We can now derive the following assumptions from this abstract model:
general medical documentation is related to diagnostic and therapeutic actions
two types of documentation can be found, affecting on one hand concrete data or
executed measures, on the other hand interpretations of the clinician (to which the
diagnosis itself belongs as well)
The process model has thus already been helpful in analyzing the documentation
process; in the following we will deduce further assumptions from the activity view.
3.2 Activity View of Clinical Documentation
In fig. 3 the model of clinical documentation is presented as KMDL activity view. We
now introduce the role or person clinician, which utilizes as least the three different
knowledge objects as we mentioned them, for the externalization of his tacit
knowledge about medicine in general, the institution he or she is working in, and
about the special patient he is treating at the moment. At the same time, this
knowledge as a whole is a requirement for the correct completion of the
documentation task: One who lacks medical knowledge, is not aware of the
31/103
Management of Knowledge-intensive Healthcare Processes 7
characteristics of his or her institution, or does not know the patient, cannot correctly
document any finding, action or interpretation.
So which findings can we derive from the activity model?
three types of knowledge are required for clinical documentation
both data resp. measures and interpretations are externalized with this knowledge
in background
the knowledge included in a patient record is individual-bound to the treating
clinician
The KMDL model of clinical documentation has therefore provided additional
insights to the actual process, based here only on a generic approach which contains
no institutional details or deviations.
4 Deriving Requirements and Potentials for Supporting IT
In the previous section, we have documented and analyzed a generic healthcare
process by means of knowledge-oriented business process modeling. In this chapter,
we will continue the short example by transforming the resulting findings into
requirements to an information system designed to support the process. Since we
chose clinical documentation, the information system needed is an electronic patient
record where the knowledge the clinician externalizes to is case-specifically
structured.
Another necessary structure in the resulting system has to be the special date of
treatment (called session), based on the sequential or time-specific tasks the process
model is built of. Furthermore, when we look at the process model as a whole, we
clinical documentation
patient-related
knowledge
general medical
knowledge
clinician documentation patient record
Mdocumentation
system
Institutional
knowledge
patient-related
knowledge
general medical
knowledge
institutional
knowledge
Fig. 3. KDML Model of General Medical Documentation (Activity View)
32/103
8 Danny Ammon et al.
find this long-term process describes what clinicians call an episode of care—―a
series of temporally contiguous health care services related to treatment of a given
spell of illness or provided in response to a specific request by the patient or other
relevant entity‖ [10]. An electronic patient record designed on the basis of the
knowledge-oriented process models we presented, must therefore structure the
inserted data patient-, episode- and session-based.
Directly adopting from the assumptions based on the model, the supporting system
has to allow the classification of certain diagnostic as well as therapeutic measures
and findings. Furthermore, personal interpretations of a clinician will have to be
marked as such, including a diagnosis.
Specifically, the potential of an electronic patient record lies in the support of
general, patient-independent knowledge that is relevant to clinical documentation.
Such general knowledge need not be individual-bound and can therefore be stored as
explicit knowledge in the information system itself. This change in knowledge-
utilizing during the process of documentation can avoid errors prevailing whenever a
single person has to rely on his or her own mind for the externalization of actually
already explicit, general knowledge.
In the following section, we propose an information system for general medical
documentation which directly derives from the here-mentioned knowledge-oriented
requirements of the documentation process.
5 Example: Knowledge-based Clinical Documentation
5.1 Specifications of a Knowledge-oriented Documentation System
In this section, we present the specifications of an electronic patient record which
convert the above-developed assumptions and requirements concerning the process of
clinical documentation in general.
Starting with the proposed potential of a system which provides knowledge
support, the electronic patient record we suggest is knowledge-based. There are two
potential types of knowledge which can be stored in the system, the more complex
one being general medical knowledge, and the one referring to this knowledge being
institution-specific knowledge. Supporting functions result in certain input guidelines
which, for example, suggest diagnostic measures for a certain kind of suspected
diagnosis and, out of these, highlight the ones which can be executed in the institution
the clinician works in.
Furthermore, we distinguish between four structuring concepts as components of
the software:
1. The patient, which forms the upper classification of data since all documentation
data is patient-specific.
2. The episode which is described by the process of treatment shown in the process
model; when this process starts again for a certain patient, a new episode begins.
3. The session as time-specific structuring criterion which keeps the option of a
chronological view at the record.
33/103
Management of Knowledge-intensive Healthcare Processes 9
4. The clinical condition resp. the executed measures. While the clinical condition
represents the clinician’s interpretation on the information provided by the patient,
the executed measures represent explicit knowledge about the actions implemented
during treatment, including and differentiating diagnostic as well as therapeutic
measures.
In the following, we present details of a reference implementation of these
specifications.
5.2 Electronic Patient Record Based on medrapid
The above-gathered specifications of a knowledge-based electronic patient record
have been implemented as a documentation system based on the medrapid knowledge
base [2]. Since medrapid is an institution-independent web portal, there is no
institutional knowledge stored up to now. medrapid allows searching, navigating and
communicating of around 9,000 disease patterns, 47,000 diagnostic measures and
16,000 therapeutic alternatives. An example of a clinical condition referring to the
hepatitis disease is shown in the screenshot in fig. 4.
The electronic patient record relating to this knowledge base has been designed and
implemented utilizing the CommonKADS methodology for knowledge-based
applications [16], which has also shown to be suitable for capturing knowledge-
intensive business processes [20]. The resulting web-based application consists of
three components: a patient table, a session and episode diagram, and the file view
showing clinical conditions and executed measures for the chosen patient at the
chosen time. Further details on the architecture of the software are provided in [2],
here the emphasis lies on the implementation of the model-based specifications: As a
visual example, the knowledge-based documentation of a clinical condition is shown
in the tree-type view in fig. 5. The general medical knowledge of hepatitis is adopted
Fig. 4. Example of General Medical Knowledge Stored in medrapid
34/103
10 Danny Ammon et al.
from the knowledge base, suggesting a superset of possible clinical conditions, from
which the clinician has to specify the patient’s condition (patient-related knowledge)
by activating the suggested items (red / green square before the item). Furthe rmore,
additions to the active items are possible in free text fields, building a set of patient-
specific knowledge independent from but linked to the general medical knowledge. In
the exemplary screenshot, the clinician can specify the localization of hepatitis for the
patient, e.g. to describe the current status of the liver as primarily affected organ, or to
insert special secondary affections
As a third way of documentation, additional files or documents belonging to a
patient’s clinical condition resp. diagnostic or therapeutic measures, can also be
uploaded (document icon behind the text field).
The resulting application in all is thus designed as a proposal for an episode-
oriented, knowledge-based electronic patient record implemented according to the
specifications derived above.
In the following last section, we summarize the results of our research, raise open
questions and develop possible approaches for future tasks in this field.
7 Conclusions and Future Work
In this contribution, we have used a generic medical process as an simple example to
show the possibility of a knowledge-oriented healthcare process management
approach. We have proposed a KMDL process and activity view of the clinical
documentation and concluded several statements from the analysis of these models.
We were able to transform these statements into requirements to a knowledge-
oriented optimization of clinical documentation, and have taken these requirements as
a basis for the specification of documentation-supporting information system. Finally,
Fig. 5. Possible User Interface of a Knowledge-based Electronic Patient Record
35/103
Management of Knowledge-intensive Healthcare Processes 11
we have presented the medrapid knowledge-based electronic patient record as an
exemplary implementation of the established specifications.
The process example is of course a short and abstract one, which does not refer to a
concrete, institution-specific business process referring to specific personnel,
activities, or surroundings. As a result, institutional knowledge has been omitted from
the specifications and the implementation, keeping the approach a generic one.
Furthermore, general medical documentation, too, is only a part of real documentation
processes, which also must refer to or initiate additional administrative processes like
patient transport or drug order, and in some cases are supplemented by very specific,
technically-detailed documentation types like an operation summary, which bases on
a very detailed, new quality of medical knowledge.
A good start for future work is therefore the application of the proposed method to
more practical process examples, where two approaches alone—business process
modeling and process-oriented software-engineering—have already proven effective.
Besides that, the spectrum of process analysis must be broadened from clinical
treatment resp. general medical documentation alone to a variety of other clinical
processes, including administrative and care activities.
Additionally, the knowledge-based electronic patient record as it was implemented,
has yet to be practically used to examine in detail the effect of a knowledge support to
the documenting clinician. This could be possible by an evaluation period where
―traditional‖ electronic methods of general medical documentation are compared to
the knowledge-based method in terms of effectiveness and efficiency.
The completion of the suggested future tasks would provide a good support for the
preliminary unified approach presented here, which leads from a capture of
knowledge-intensive processes in healthcare over the generation of requirements for
an optimized variant of these processes to the transformation into details of process-
and knowledge-oriented information systems. The resulting software has to be best-
suited for the support of clinicians executing time-critical, knowledge-intensive duties
in their everyday clinical work.
References
1. Abidi, S.S.R.: Healthcare Knowledge Management: The Art of the Possible. In: Riaño, D.
(Ed.): K4CARE 2007. LNAI 4924, pp. 1–20. Springer, Berlin, Heidelberg (2008)
2. Ammon, D. et al.: Developing an Architecture of a Knowledge-Based Electronic Patient
Record. In: Proceedings of the Thirtieth International Conference on Software Engineering,
pp. 653–660. ACM, New York (2008)
3. Beale, T., Heard, S.: An Ontology-based Model of Clinical Information. In: Stud. Health
Technol. Inform. 129, 785–90 (2007)
4. Doroszewski, J.: Unity and Diversity of the Medical Action: A Review of its Components
and Their Interconnections. Metamed. 2, 155–168 (1981)
5. Finkeissen E., et al.: medrapid: Medical Community & Business Intelligence System. In:
Health Data in the Information Society: Proceedings of MIE2002. IOS Press, Amsterdam
(2002)
6. Gode, A.: Just Words. J. Amer. Med. Ass. 205, 352 (1968)
7. Gronau, N., Müller, C., Korf, R.: KMDL: Capturing, Analysing and Improving
Knowledge-Intensive Business Processes. J. Univ. Comp. Sci. 4, 452–472 (2005)
36/103
12 Danny Ammon et al.
8. Gronau, N., Weber, E.: Management of Knowledge Intensive Business Processes. In:
Desel, J., Pernici, B., Weske, M. (Eds.): BPM 2004. LNCS 3080, pp. 163–178. Springer,
Berlin, Heidelberg, 2004
9. Hädrich, T.: Contrary Positions About Modeling Knowledge Work. In: Althoff, K.-D. et al.
(Eds.): WM 2005. LNAI 3782, pp. 248–258. Springer, Berlin, Heidelberg (2005)
10. Hornbrook M., Hurtado A., Johnson R.: Health Care Episodes: Definition, Measurement
and Use. Med. Care Rev. 42, 163–217 (1985)
11. Hucklenbroich, P.: Klinisch-methodologische Aspekte medizinischer Expertensysteme. In:
Hucklenbroich, P., Toellner, R.: Künstliche Intelligenz in der Medizin, pp. 9 –36. Gustav
Fischer, Stuttgart, Jena, New York (1993)
12. Hucklenbroich, P.: Wissenschaftstheorie als Theorie der Medizin: Themen und Probleme.
In: Deppert, W. et al. (eds.): Wissenschaftstheorien in der Medizin: Ein Symposium, pp.
65–96. de Gruyter, Berlin (1992)
13. Nonaka, I., Takeuchi, H.: The Knowledge-Creating Company: How Japanese Companies
Create the Dynamics of Innovation. Oxford University Press, New York (1995)
14. OpenGALEN, http://www.opengalen.org/
15. Polanyi, M.: Personal Knowledge: Towards a Post-Critical Philosophy. The University of
Chicago Press, Chicago (1958)
16. Schreiber, G., et al.: Knowledge Engineering and Management: The CommonKADS
Methodology. MIT Press, Cambridge, MA (2000)
17. Systematized Nomenclature of Medicine, http://www.snomed.org/
18. Tange, H.J., Dietz, J.L.G., Hasman, A., de Vries Robbé, P.F.: A Generic Model of Clinical
Practice: A Common View of Individual and Collaborative Care. Methods Inf. Med. 42,
203–211 (2003)
19. Thomasma, D.C.: Applying General Medical Knowledge to Individuals: A Philosophical
Analysis. Theor. Med. 9, 187–200 (1988)
20. Trier, M., Müller, M.: Towards a Systematic Approach for Capturing Knowledge-Intensive
Business Processes. In: Karagiannis, D., Reimer, U. (eds.): PAKM 2004. LNAI 3336, pp.
239–250. Springer, Berlin, Heidelberg (2004)
21. Unified Medical Language System, http://www.nlm.nih.gov/research/umls/
22. WHO International Classification of Diseases, http://www.who.int/classifications/icd/en/
37/103
