An Active Database Approach to Computerised

Clinical Guideline Management

Kudakwashe Dube and Bing Wu

Department of Computer Science, School of Computing, Dublin Institute of Technology,

Kevin Street, Dublin 8, Ireland.

Abstract. This paper presents a generic approach, and a case study practising

the approach, based on a unified framework harnessing the event-condition-

action (ECA) rule paradigm and the active database for the management of

computer-based clinical practice guidelines and protocols (CPGs). The CPG

management is cast into three perspectives: specification, execution and

manipulation, making up three management planes of our framework. The ECA

rule paradigm is the core CPG representational formalism while the active

database serves as the kernel within the CPG management environment

facilitating integration with electronic healthcare records and clinical workflow.

The benefits of the approach are: flexibility of CPG management; integration of

CPGs with electronic patient records and clinical workflows; and incorporation

of CPG management system into computerised healthcare systems.

1 Introduction

Clinical practice guidelines (CPGs) encapsulate medical knowledge about when

clinicians and patient should carry out specific activities in disease management.

CPGs are effectively “knowledge models of preferred processes of care” [1], which

can guide patient care activities and clinical workflow or care flow [2]. In practice, a

CPG is “a method that identifies actions that are to be performed and that specifies

conditions that govern when it is appropriate to perform them” [3]. Thus, a CPG

embodies the core compositional primitives of the event-condition-action (ECA) rule

paradigm. ECA rules are specified by an event, a condition and an action with the

semantic that the event must occur for the action to be executed subject to the

condition being satisfied [4]. An active database management system (ADBMS) is a

DBMS that incorporates the ECA rule mechanism and provides functionalities as

stipulated in the Active Database System Manifesto [5]. The ECA rule paradigm

could be a strong basis for formalising CPGs while an active database system could

be a promising management environment for computerised CPG. The authors’

hypothesise that, in combination, these two could facilitate a solution to the problem

of the integration of computer-based CPGs with the electronic patient record and

computerised clinical workflow. This paper presents a computerised CPG

management approach and a case study which uses a generic, unified management

framework and harnesses the ECA rule paradigm and the active database in the CPG

management task. The rest of this paper is organised as follows: Section 2 presents a

Dube K. and Wu B. (2006).

An Active Database Approach to Computerised Clinical Guideline Management.

In Proceedings of the 3rd International Workshop on Computer Supported Activity Coordination, pages 107-115

DOI: 10.5220/0002474001070115

 SciTePress

review of the ECA rule paradigm in computer-based CPG management support;

Section 3 presents the generic framework for CPG management; Section 4 presents

the method developed in this work for managing the CPGs using the ECA rule

paradigm and active databases within the SpEM framework; Section 5 presents a case

study in which the proof-of-concepts system is developed and used to manage the

microalbuminuria CPG for diabetes patients; and Section 7 summarises and concludes

this paper.

2 Relevant Work in Supporting Clinical Guideline Management

This section reviews relevant work in supporting computerised CPG management.

Research in computer-based support for the management of CPGs has been on-going

for over a decade. A number of approaches and systems have emerged during this

period. For a more general review, readers are referred to de Clercq et al [6]. Most

approaches focus on the formal representation and execution of computerised CPGs.

Among them, the Arden Syntax [7], an established HL7 standard [8], and HyperCare

[9] are most relevant to this work. While most approaches use control structure

primitives [6] to represent guidelines, the Arden Syntax and HyperCare use the ECA

rule paradigm. However, the Arden Syntax focuses on CPG specification only and

HyperCare focuses on execution only, using an active database system in its

implementation. Both the Arden Syntax and HyperCare do not create patient-specific

CPG instances. Instead, all rules operate with a global scope covering all patients.

Furthermore, these two approaches offer no support for the management of

information on both CPG specifications and their instances. Hence, they do not fully

exploit all the advantages of the database approach to realise CPG information

management. Departing from others, authors of this paper have developed a unified

CPG management framework which places equal emphasis on the CPG management

perspectives of specification, execution and manipulation and harnesses the active

database for supporting all aspects of this framework. In this framework, generic CPG

specifications are created and held in the database together with execution state and

effects for easy management. The CPG execution aspect supports patient-specific

CPG instances and harnesses the active mechanism of the DBMS. Furthermore,

database manipulation and query features are exploited for the computerised CPG

information management.

3 A Generic Framework for Clinical Guideline Management

This section presents a generic framework, SpEM [10] (Specification, Execution and

anipulation) developed by the authors for supporting the computerised CPG

management. As illustrated in Figure 1, the SpEM framework contains 3 planes:

specification plane; execution plane and; manipulation plane with the active database

as the integrating factor among the three planes.

108

Specification

Plane

Manipulation Plane

Customisation,

instantiation and

change propagation

Enforcement

Plane

Query, dynamic

interaction and

manipulation

Query, manipulation

and version maintenance

Application

Domain Database

System with

Active Mechanism

Specification

Plane

Manipulation PlaneManipulation Plane

Customisation,

instantiation and

change propagation

Enforcement

Plane

Query, dynamic

interaction and

manipulation

Query, manipulation

and version maintenance

Application

Domain Database

System with

Active Mechanism

Fig. 1. The SpEM framework.

The CPG management process is fitted into the three planes of the framework. In the

specification plane, CPGs are translated into formal specifications and held

permanently in the database in a manageable form. In the execution plane, the stored

CPG specifications are used to create patient-specific CPG instances that are

executable by an engine based on the active database. In the manipulation plane, the

specifications and the executing CPG instances are manipulated using supported

operations and queries based on the SQL features of the underlying database system.

The active database is a base for an execution engine for CPGs. It provides facilities

for querying and manipulating information. It can also permanently hold the

electronic patient record in addition to CPG information and running processes.

Furthermore, it can guarantee future sharing of information through the generic nature

of databases and the standard language, the SQL. Table 1. summarises the support for

the SpEM framework in major existing CPG approaches and systems. However, these

approaches only provide guideline support mainly in terms of the specification and

execution of CPGs. Not much attention seems to have been paid to the support for the

manipulation and querying of the CPG information and to harnessing of database

features for CPG management. Manipulation of the static and dynamic aspects of

CPGs is important to allow flexibility and ease-of-use [19], which are some of the

major determining factors in the acceptability of guideline systems by clinicians [20].

Table 1. Support for the SpEM framework in existing CPG management systems.

SpEM Framework Support

Key: √ - full support, * - partially supported,

X – no support,

Manipulation

Guideline System

Specification Execution

Operation Query

Use of

Database

Features

DILEMMA/PRESTIGE

[11]

√ √ X X X

EON/Dharmma [12] √ √ X X X

PROforma [13]

√ √ X X X

GLIF [14] √ * X X X

Asgaard/ Asbru [15] √ √ * * X

GUIDE [2] √ √ X X X

PROGIDY [16] √ √ X X X

Arden Syntax [7]

√ X X X X

HyperCare [9]

X √ X X *

GASTON [17] √ √ * X X

GLARE [18] √ √ X X X

109

4 A Method for Computerised CPG Management

This section presents the method for managing computerised CPGs through the SpEM

framework. Fig. 2 illustrates how the method permeates the three planes of the

framework and allows CPGs to be formally specified, stored, executed with respect to

a given patient, and manipulated through querying and operations. Several aspects

need to be incorporated as components of the management process for supporting

CPG management. Domain knowledge, i.e., expertise and literature on recent

advances in medical knowledge, is the main source of CPGs. Its translation into CPGs

is done by clinicians and is outside the scope of SpEM.. Formal representation and

creation of CPG specifications and their storage are important aspects of the

computerisation of CPGs. Customisation and execution of computerised CPGs with

respect to specific individual patient are vital aspects of the management of

computerised CPGs. The manipulation of both the formal specifications and the

running process consists of the two aspects: querying; and performing manipulation

operations on the CPGs. The process illustrated in Fig. 2 is to ensure that most aspects

of the computerised CPGs can be managed. Fig. 2 also illustrates the main

functionality for supporting CPG management. The main requirement of the

specification plane is a declarative language, the clinical P

rotocol LANguage, PLAN,

together with its model [21]. PLAN was designed to use the ECA rule paradigm as

the core representation construct for specifying CPGs. Fig. 3. illustrates the general

syntax of PLAN in the Backus-Nuar Form (BNF). A PLAN specification consists of a

descriptive header; a set of schedules; and the protocol rule set. Just like a protocol, a

schedule is named and consists of an entry criteria and a list of rules each of which is

either a dynamic or static rule. For a more detailed discussion of PLAN, readers are

referred to Wu and Dube [21]. In the execution plane, the relevant PLAN

specification is customised and installed for each patient as an instance that runs

within the ECA rule mechanism of a database system. The execution of the guideline

instance follows the ECA mechanism. Within the manipulation plane, provision is

made to perform operations and to issue queries using a query and manipulation

language that is based on the SQL.

Fig. 2 The management support process in SpEM.

110

<static_rule>::= ECA rule with time events

<protocol>::=PROTOCOL<protocol_body>END PROTOCOL.

<protocol_body>::=<protocol_header>SCHEDULE_SET<schedule_ist>END

SCHEDULE_SET<protocol_rule_set>

<protocol_header>::=<name>;<description>,<creator_detail>;<category>;

<schedule_ist>::=<schedule> | <schedule>;<schedule_list>

<schedule>::=SCHEDULE<schedule-header>;<entry_criteria>;<rule_list>END SCHEDULE

<protocol_rule_set>::=RULE_SET<rule_list>END RULE_SET;

<rule_list>:=<rule> | <rule>;<rule_list>

<dynamic_rule>::= ECA rule

Fig. 3. The high-level BNF syntax of PLAN.

5 Case Study: TOPS and the Management of the

Microalbuminuria CPG for Diabetes Patients

This section outlines the design of the proof-of-concepts system, TOPS, and a case

study in which TOPS is used to manage the CPG for microalbuminuria in diabetes

patients. Fig. 4 illustrates the generic CPG management method that is used in TOPS.

Specification

Phase

Customisation

Phase

Installation

Phase

Execution

Phase

Management

(querying &

modification)

atient Clinical

ata

Test

Results

Test orders, alerts,

alarms, etc

Category

rotocol specification

atient test

ordering plan

nstantiated patient

test plan

Fig. 4. The CPG Management Process for clinical protocols.

External

Communications

Module

Manipulation

Module

Specification

Module

Protocol

Management

Layer

Query

Module

Re-Play

Module

Instantiation

& Execution

Module

Management Components

External Systems Users

Database Access Manager

Rule Activity

Listener

Time Event

Generator

Dynamic SQL Generator

ECA Rule

Extension

Module

Execution Mechanism

and Database Access

Extension Layer

Rule Activity Notifier

Local

Patient

Record

Protocol

and Plan

Specifications

Plan

Execution

Logs

ECA

Mechanism

Protocol Execution

Engine (

DBMS)

DB Connectivity

External

Communications

Module

Manipulation

Module

Specification

Module

Protocol

Management

Layer

Query

Module

Re-Play

Module

Instantiation

& Execution

Module

Management Components

External Systems Users

Database Access Manager

Rule Activity

Listener

Time Event

Generator

Dynamic SQL Generator

ECA Rule

Extension

Module

Execution Mechanism

and Database Access

Extension Layer

Rule Activity Notifier

Local

Patient

Record

Protocol

and Plan

Specifications

Plan

Execution

Logs

ECA

Mechanism

Protocol Execution

Engine (

DBMS)

DB Connectivity

Fig. 5. Architecture of the proof-of-concepts system, TOPS.

111

In the specification phase, CPGs are formally represented and specified. Resulting

specification are stored in a database. The domain expert needs to be involved in this

process. In the customisation phase, the computerised CPG is customised to suit the

specific clinical problem suffered by the patient. This phase produces patient-specific

CPG specification instances as illustrated in Fig. 4. In the installation phase, all the

ECA rules in the CPG are added to the active database resulting in an active CPG

instance. In the execution phase, the execution process proceeds in accordance with

the ECA mechanism. The manipulation phase includes querying and manipulation

operations and permeates all the above phases. The architecture of TOPS has three

layers as illustrated in

Fig.

5. External to TOPS, are users and external systems. The top layer is the clinical

protocol management functionality for specifying, storing, executing, manipulating

and querying CPGs. The middle layer provides services that extend the ECA rule

execution mechanism of the underlying DBMS and handles database connectivity.

The bottom layer is the CPG execution engine as well as the system database, which

is currently based on the Oracle9i DBMS. The architecture provides support for the

three planes SpEM. Issues of concurrency and efficiency in rule execution are

handled by the Oracle9i DBMS with the exception of ECA rule extensions that are

implemented externally.

The SpEM framework in TOPS has been demonstrated through a case study

involving the CPG for the diagnosis and management of microalbuminuria (MA) in

diabetes mellitus as interpreted by a practising clinician at the local Diabetes Day

Clinic in Dublin. Fig. 6. illustrates the state chart for the microalbuminuria protocol

(MAP). For each state and its associated transitions, rules are designed to perform

what must be done when the patient enters, stays and exits from the state. Other rules

serve to trigger patient state transitions. Thus, the state chart guides creation of MAP

specification. Fig. 7. illustrates the outline structure and content of the resulting MAP

specification. The MAP specification used to create a patient plan when a user

requests the creation of a MAP instance for a specified patient. Patient plan execution

proceeds in an event-driven manner according to the set of ECA rules making up the

MAP instance. In the query illustrated in Fig. 8., a patient-specific MAP instance’s

snapshot at a given time or interval is retrieved. The MAP instance snapshot refers to

@PROTOCOL@ MAP;

DESCRIPTION: This is a protocol for the diagnosis and management of microalbuminuria in

diabetes patients;

CREATOR: DR JOHN NOLAN;

CATEGORY: DIABETIC_NEPHROPATHY;

#SCHEDULE_SET#

^SCHEDULE^ AUS, <AUS_rules> ^END SCHEDULE^

^SCHEDULE^ OIS, <OIS_rules> ^END SCHEDULE^

^SCHEDULE^ MAS, <MAS_rules> ^END SCHEDULE^

^SCHEDULE^ CMA, <CMA_rules> ^END SCHEDULE^

^SCHEDULE^ NPH, <NPH_rules> ^END SCHEDULE^

#END SCHEDULE_SET#

~RULE_SET~

STATIC_RULE AUS1,<body of rule AUS1>;

RULE OIS1,<body of rule OIS1>;

RULE MAS1,<body of rule MAS1>;

RULE CMA1,<body of rule CMA1>;

RULE CMA2,<body of rule CMA2>;

RULE CMA3,<body of rule CMA3>;

RULE CMA4,<body of rule CMA4>;

RULE NPH1, <body of rule NPH1>;

~END RULE_SET~

@END PROTOCOL@

Abbreviations:

AUS – Annual Urine Screening, OIS – Other-Infections Screening, MAS – MicroAlbuminuria

Screening, CMA – Confirmed MicroAlbuminuria, NPH – Nephrology referral

Fig. 6. State chart for the microalbuminuria

protocol.

Fig. 7. Structure of the Microalbuminuria

Protocol in PLAN.

112

the instance’s current state in terms of rule composition and the status of its rules at

that time or interval. The query illustrated in Fig. 9. provides information on what

tests were ordered with respect to the specified patient during the given time interval.

The term order in the query can be generalised to rule-action so that one can obtain

information on rule actions that have been performed during the specified time

interval. In this case study, the use of the highly intuitive UML state chart brings easy

communication with domain experts during CPG knowledge elicitation, capture and

specification. Subsequent extraction of the relevant ECA rules is also made easier

since the state chart naturally supports the ECA rule paradigm [22] and is also easily

understood by domain experts. The database offered a uniform and flexible way to

access, manipulate and query all information from specification, to executing process

state, to data in the patient record. The generation of SQL trigger code implementing

ECA rules of the MAP was automatically supported by TOPS and required minimal

intervention. This makes it easy for application domain experts to use TOPS with no

knowledge of the SQL trigger specification language. However, domain experts still

needed to be familiar with the specification language, PLAN, which is closer to their

domain language than the SQL. The execution of the rule actions was subject to the

availability of the appropriate software module that implements the action. Thus, rule

actions in the microalbuminuria CPG needed to be predefined and any new action

requires that the module to implement such an action be developed. However, rule

actions were designed to be generic and re-usable by other rules in other CPGs. Using

the database permits operations and queries on various aspects of the MAP through an

SQL-based manipulation language. It was shown that the MAP can be modeled and

specified by using the ECA rule paradigm guided by the state chart. This case study

demonstrated the applicability of the SpEM framework and the active database in

enabling the support for the management of the MAP for diabetes patients.

6 Summary and Conclusion

This paper has presented a unified CPG management framework, SpEM, for

computerized CPG management. The paper also presented a generic method with a

case study to harness the ECA rule paradigm and active databases to provide

computerized CPG support, by following the SpEM. Active databases combine the

ECA rule paradigm with data management to present a promising environment for

TOPS:\> query

QUERY:\> --->SELECT PLAN FROM SNAPSHOT WHERE TARGET:2005-7-19

01:55:02,2005-7-19 01:55:58; SOURCE:PATIENT_ID=25

processing query ...

…

PLAN [ PL$25$1$ ] SNAPSHOT @[2005-07-19 22:30:26.91]

[rule 1]--->[ 72, PL$25$1$AUS2, DYNAMIC, READY ]

[rule 2]--->[ 73, PL$25$1$AUS3, DYNAMIC, READY ]

[rule 3]--->[ 74, PL$25$1$OIS2, DYNAMIC, READY ]

[rule 4]--->[ 75, PL$25$1$OIS3, DYNAMIC, READY ]

[rule 5]--->[ 76, PL$25$1$OIS4, DYNAMIC, READY ]

…

[rule 19]--->[ 90, PL$25$1$CMA4a, DYNAMIC, READY ]

[rule 20]--->[ 91, PL$25$1$CMA4b, DYNAMIC, READY ]

[rule 21]--->[ 92, PL$25$1$NPH1, DYNAMIC, READY ]

[rule 22]--->[ 93, PL$25$1$main$AUS1, STATIC, EXECUTING ]

END SNAPSHOT FOR PLAN PL$25$1$.

QUERY:\> --->exit

TOPS:\>

TOPS:\> query

QUERY:\> --->SELECT ORDER FOR PATIENT WHERE TARGET:2005-7-16 17:48:30,2005-

7-16 17:51:25; SOURCE:PATIENT_ID=61

processing query ...

launching specialised query handler ...

processing ORDER query [ 2005-7-16 17:48:30,2005-7-16 17:51:25 ] for [ PATIENT ]

...

Tests ordered for [PATIENT_ID=61] during time interval [2005-7-16 17:48:30,2005-7-16

17:51:25]

Dip_stick_urine Profile, DSU, 2005-07-16 17:49:28.0

Urinary_Tract_Infection Profile, UTI, 2005-07-16 17:50:06.0

-------------------

End test listing.

QUERY:\> --->EXIT

TOPS:\>

Fig. 8. A query for a snapshot of the com-

position of a CPG instance in TOPS.

Fig. 9. A query to find out what clinical tes

orders where made during the specified time

interval.

113

supporting CGPs and their integration with the electronic medical record and clinical

workflow. This work contributes a generic approach with a framework to unify the

three CPG management dimensions and an active database method for computational

support. The benefits of the approach are: flexibility based on CPG information

management; ease of integration of CPGs with electronic patient records and clinical

workflows due to the active database approach; and ease of incorporation of CPG

management system into the healthcare systems due to ubiquity of database systems

within most institutions. Our future work will focus on: 1) finding ways to deploy and

evaluate the SpEM and TOPS in a real patient care setting without infringing on

confidentiality and proprietary license restrictions on APIs and schemas in existing

hospital systems; 2) improving our framework though enriched specification model

and language; 3) more efficient methods of exploiting and enhancing the active

mechanism by hybridising it with other paradigms; and 4) useful concepts and

methods for information manipulation, query and replay.

Acknowledgements

The Arnold F Graves Trustees and the Dublin Institute of Technology are gratefully

acknowledged for sponsoring this research work.

References

1. OpenClinical, The Medical Knowledge Crisis and its solution Through Knowledge

Management. 2001, OpenClinical.org.

2. Quaglini, S., Stefanelli, M., Cavallini, A., Micieli, G., Fassino, C., and Mossa, C.,

Guideline-based careflow systems. Artif Intell Med, 2000. 20(1): p. 5-22.

3. Pattison-Gordon, E., Cimino, J.J., Hripcsak, G., Tu, S.W., Gennari, J.H., Jain, N.L., and

Greenes, R.A., Requirements of a sharable guideline representation for computer

applications. 1996, Stanford University: Stanford, California, USA.

4. Widom, J. and Ceri, S., eds. Active Database Systems: Triggers and Rules for Advanced

Database Processing. 1996, Morgan Kaufmann: San Francisco, USA.

5. Dittrich, K.R., Gatziu, S., and Geppert, A. The Active Database Management System

Manifesto: A Rulebase of ADBMS Features. in 2

Workshop on Rules in Databases

(RIDS). 1995. Athens, Greece: Springer.

6. de Clercq, P.A., Blom, J.A., Korsten, H.H., and Hasman, A., Approaches for creating

computer-interpretable guidelines that facilitate decision support. May. Artif Intell Med.,

2004. 31(1): p. 1-27.

7. Clayton, P.D., Pryor, T.A., Wgertz, O.B., and Hripcsak, G. Issues and Structures for

Sharing Medical Knowledge Among Decision-making Systems: The 1989 Arden

Homestead Retreat. in Annu Symp Comput Appl Med Care. 1989.

8. HL7, Arden Syntax for Medical Logic Modules., in Standards of the Health Level 7. 1999:

USA.

9. Caironi, P.V.C., Portoni, L., Combi, C., Pinciroli, F., and Ceri, S. HyperCare: a Prototype

of an Active Database for Compliance with Essential Hypertension Therapy Guidelines. in

AMIA Ann Fall Symposium. 1997. Philadelphia, PA: Hanley and Belfus.

114

10. Dube, K., A Generic Approach to Supporting the Management of Computerised Clinical

Guidelines and Protocols, PhD Thesis, in Computer Science. 2004, Dublin Institute of

Technology: Dublin. p. 324.

11. Gordon, C., Herbert, I., and Johnson, P. Knowledge Representation and Clinical Practice

Guidelines: the DILEMMA and PRESTIGE projects. in Medical Informatics Europe. 1996.

Copenhagen: IOS Press.

12. Musen, M.A., Tu, S.W., Das, A.K., and Shahar, Y., EON: A component-based approach to

automation of protocol-directed therapy. JAMIA, 1996. 3(6): p. 367-88.

13. Fox, J., Johns, N., Rahmanzadeh, A., and Thomson, R. PROforma: a method and language

for specifying clinical guidelines and protocols. in Medical Informatics Europe. 1996.

Amsterdam.

14. Ohno-Machado, L., Gennari, J.H., Murphy, S., Jain, N.H., Tu, S.W., Oliver, D.E., Pattison-

Gordon, E., Greenes, R.A., Shortliffe, E.H., and Barnett, G.O., The GuideLine Interchange

Format: A Model for Representing Guidelines. JAMIA, 1998. 5(4): p. 357-372.

15. Shahar, Y., Miksch, S., and Johnson, P., The Asgaard Project: A task-specific framework

for the application and critiquing of time-oriented clinical guidelines. Artificial Intelligence

in Medicine, 1998. 14: p. 29-51.

16. Johnson, P.D., Tu, S.W., Booth, N., Sugden, B., and Purves, I.N., A guideline model for

chronic disease management. 1999, Stanford Medical Informatics: Stanford.

17. de Clercq, P.A., Hasman, A., Blom, J.A., and Korsten, H.H.M., Design and implementation

of a framework to support the development of clinical guidelines. Int J Med Inf., 2001.

64(2-3): p. 285-318.

18. Terenziani, P., Molino, G., and Torchio, M., A modular approach for representing and

executing clinical guidelines. Artif Intell Med., 2001. 23(3): p. 249-76.

19. Quaglini, S., Stefanelli, M., Lanzola, G., Caporusso, V., and Panzarasa, S., Flexible

guideline-based patient careflow systems. Artif Intell Med, 2001. 22(1): p. 65-80.

20. Musen, M.A. Two decades of Models and Components: Why has Automation of Guideline-

Directed Care been so Difficult? in 1st European Workshop on Clinical Practice Guidelines

and Protocols (EWGLP 2000). 2000. Leipzig, Germany: IOS Press.

21. Wu, B. and Dube, K. PLAN: a Framework and Specification Language with an Event-

Condition-Action (ECA) Mechanism for Clinical Test Request Protocols. in 34th Hawaii

International Conference on System Sciences (HICSS-34): the Mini-Track in Information

Technology in Healthcare. 2001. Maui, Hawaii: IEEE Computer Society, Los Alamitos,

California.

22. Berndtsson, M. and Calestam, B., A Uniform Approach for Supporting Active Database

Features in UML and OMT. 2001, Computer Science Department, University of Skovde:

Skovde.

115