Virtual Centre for Health Informatics, Aalborg University
Editor :
Andersen, Stig Kjær, Aalborg University, DK
Contributions from:
Andersen, Stig Kjær,Aalborg University, DK, Quaglini, Silvana,
University of Pavia, IT, Ying Lie O.,Buskens, Erik, University of Utrecht,,
NL, Guus Schreiber, University of Amsterdam NL,Motta, Enrico, The
Open University, UK, Gentner, Carsten PROMATIS AG, D
Date: March 14, 2000
Version number: 1.01

HC 4017 P AT M AN FINAL P AGE 2
PART I: EXECUTIVE SUMMARY....................................................................................................................... 3
PART II: SETTING THE SCENE.......................................................................................................................... 5
A PPROACH ............................................................................................................................................................... 7
Knowledge and technology used...................................................................................................................... 7
Validation and evaluation ................................................................................................................................ 9
RESULTS AND ACHIEVEMENTS..................................................................................................................... 14
D ESCRIPTION OF THE TOOLS ................................................................................................................................. 16
Tool no. 1 - The Guideline Editor.................................................................................................................. 17
Tool no. 2 - The WPDL Translator ............................................................................................................... 17
Tool no. 3 - The Organisation Ontology Building Tool .............................................................................. 17
Tool no. 4 - INCOME..................................................................................................................................... 18
Tool no. 5 - Oracle Workflow ........................................................................................................................ 18
Tool no. 6 - The Guideline Discussion Forum ............................................................................................. 18
D ESCRIPTION OF DEMONSTRATORS ...................................................................................................................... 19
The "guideline management tool" Demonstrator ....................................................................................... 19
The INCOME Tool Demonstrator................................................................................................................. 20
The CAREFLOW tool demonstrator............................................................................................................. 21
The Discussion Forum for knowledge sharing ............................................................................................ 22
The PatMan Project web page ....................................................................................................................... 24
CONCLUSION AND FUTURE PLANS .............................................................................................................. 25
CONTACT DETAILS............................................................................................................................................. 28
ABSTRACTS OF DELIVERABLES.................................................................................................................... 29
D1 - T HE G UIDELINE /P ROTOCOL E DITOR ............................................................................................................ 29
D2 - T HE G UIDELINE BASED P ATIENT W ORKFLOW M ANAGEMENT ................................................................... 29
D3 – T HE L OW LEVEL W ORKFLOW B UILDER D OCUMENT DOMAIN ANALYSIS AND REFERENCE MODELS ........ 30
D4 – T HE P ETRINET L IBRARY P ROTOTYPE .......................................................................................................... 30
D5 - WEB S ERVERS FOR D ATABASE I NTERFACING ............................................................................................ 31
D6 - O NTOLOGY -D RIVEN M ANAGEMENT OF M EDICAL G UIDELINES .................................................................. 31
D7 – O NGOING APPLICATIONS REPORT ................................................................................................................ 32
D8 – D EMONSTRATION PHASE RESULTS ............................................................................................................... 32
LIST OF PUBLICATIONS INCLUDING ABSTRACTS ................................................................................. 33
A RTIFICIAL I NTELLIGENCE FOR B UILDING L EARNING H EALTH C ARE O RGANIZATIONS .................................... 33
G UIDELINES BASED WORKFLOW SYSTEMS ........................................................................................................... 35
A C OMPUTERISED G UIDELINE FOR P RESSURE U LCER P REVENTION ................................................................... 36
D IFFERENT INTERACTION MODALITIES WITH COMPUTERISED GUIDELINES ......................................................... 37
P LANET -O NTO : F ROM N EWS P UBLISHING TO I NTEGRATED K NOWLEDGE M ANAGEMENT S UPPORT ................. 40
O NTOLOGY -D RIVEN D OCUMENT E NRICHMENT : P RINCIPLES AND C ASE S TUDIES ............................................. 41
D ISCUSSING , B ROWSING , AND E DITING O NTOLOGIES ON THE W EB ................................................................... 42
A K NOWLEDGE -B ASED N EWS S ERVER S UPPORTING O NTOLOGY -D RIVEN S TORY E NRICHMENT AND
K NOWLEDGE ......................................................................................................................................................... 43
R EPRESENTING S CHOLARLY C LAIMS IN I NTERNET D IGITAL L IBRARIES : A K NOWLEDGE M ODELLING
A PPROACH ............................................................................................................................................................. 44
A CTING AN U LCER P REVENTION P LAN THROUGH G UIDELINE BASED T OOLS .................................................... 45

HC 4017 P AT M AN FINAL P AGE 3
Short description and figure
PatMan proposes a methodology for building decision support systems that integrate the two key aspects of health care delivery
1.
clinical procedures and
2.
organisational functioning thereby integrating guideline-based care with patient workflow management. The figure shows the relation between the main steps and components followed by
PatMan for providing health care professionals with tools for building a decision support system.
Setting the Scene
The motivation for this project lies in the observation that hospital information systems often oversee the organisational knowledge that is essential to guarantee efficient delivery of quality care. Much effort has been put into developing clinical guidelines, which are intended to prevent dissimilarities, malpractice and waste of resources in patient treatment.
However, to be effectively implemented within an organisation such as a hospital or an outpatient department, a guideline (GL) requires a customisation process, suiting it to specific organisational and patient contexts. In sharp contrast to industrial and commercial environments, where the concept of Workflow Management
(WfM) is an established and crucial element of management information systems, existing health care information systems have often neglected to consider organisational issues and characteristics.
Approach
PatMan proposes a methodology for building decision support systems that integrate the two key aspects of health care delivery – clinical procedures and
organisational functioning.
The aims:
to provide methodological and software support for site customisation of generic healthcare GLs.
to produce tools supporting the acquisition, modelling and management of both medical and organisational knowledge, based on conceptual models (i.e.ontologies).
PatMan TOOLS
Tool #6 Guideline Discussion
Forum: Management of medical and organisational knowledge
Medical knowledge
Organisational knowledge
Tool #1
GUIDE
Tool #2
WPDL Translator
Guideline formalisation
WPDL model
Org Model
Patient
Database
Patient Workflow Model
WPDL – Petri Net
Tools # 4-
INCOME for simulations
Tools # 5 Oracle
Workflow: Web service for database integration and GL implementation
to assist health practitioners in the management of health care protocols by means of patient workflow management systems (WfMS).
The activities:
1) Development of a representation formalism for clinical guidelines and protocols.
2) Development of a general methodology for the translation of a
GL into a Petri net, in order to simulate GL implementation.
3) Development of a set of tools supporting the deployment of medical guidelines and protocols in the context of work practices in specific health care settings.
4) Empirical evaluation of these tools in specific health care domains, namely in relation to
management of patients with ischemic stroke
prevention of pressure ulcers
5) Theoretical evaluation of the same tools in other domains, namely a surgery guideline for the management
Database describing
Organisational knowledge
Resource allocation
Careflow Management System
Results and Achievements
The principle objective of this project was to demonstrate that health care organisations can benefit by the application of a careflow management
system, which is a WfMS applied to specific patient situations. The project results show the benefits of providing a dynamic model of the workflow process in question, having all the involved entities explicitly defined.
The dynamic model makes it possible: a) to simulate the process evolution to clarify where and when information and/or resources are produced and consumed, b) to implement GLs in the real clinical setting, by efficiently allocating the necessary resources.
The components of the guidelinebased careflow management system can be accessed through an internet browser.
Tool #3
Organisation
Ontology Builder
Applications :
Stroke
Pressure Ulcer
Diabetes
Aorta Aneurysm of abneurysm aorta abdominialis and a follow-up guideline for diabetes.

HC 4017 P AT M AN FINAL P AGE 4
Users of the services are:
medical experts that are responsible for the guideline/ protocol formulation.
physicians and other health staff that are responsible for guideline implementation in daily routine.
resource managers, such as ward chiefs and administrative staff, who are responsible for the best allocation of resources.
The PatMan tools:
The PatMan tools comprise the following tools 1-3 that were completely developed under PatMan, tools 4-5 adapted within the PatMan project, and tool 6 that was adapted and refined from the previous HC-
REMA project.
Tool no. 1: The Guideline editor,
GUIDE, which computerises the representation of a guideline or protocol. The editor is written in Java.
A GL written with GUIDE may also embed decision analysis models, such as decision trees.
Tool no. 2: The WPDL translator is a program that produces a representation of the Guideline as a WPDL
(Workflow Process Definition
Language) file. The translator is written in C++.
Tool no. 3: The Organisation Ontology
Building tool is used for modelling the organisational aspects of the careflow.
A prototypical ontology for health care organisations has been embedded. The tool is written in Visual Basic.
Tool no. 4: INCOME is a commercial product from the partner PROMATIS.
This tool exploits a relational database.
INCOME is able to simulate a Petri net representing the guideline/protocols. It embeds facilities to represent the behaviour model, the organisation model, and the information model.
Tool no. 5: Oracle workflow is a commercial product from ORACLE, which permits describing and implementing a workflow model on a real world setting. Three main modules compose this tool.
3.
The Builder allows the user to build the model as a sequence of tasks.
4.
The Monitor allows for viewing and administering the status of a specific instance of a workflow process.
5.
The Notification Mailer lets people receive notifications of work-items awaiting their attention via E-mail, and acts based on their E-mail responses.
Tool no. 6: Tool no. 6: The Guideline
Discussion Forum (GDF) integrates several technologies supporting organizational management and customises them for supporting medical guidelines. The web-based support discussion and debate provides a web environment where both guideline developers and users can discuss and inspect guidelines. This facility is tightly integrated with knowledge modelling tools, so that semantic information retrieval is also supported. particular, users can use the
LOIS query tool, to find information about a guideline through the associated knowledge model.
Demonstrators
To enable demonstration of the different functionalities of the tools and their interaction with data and knowledge stores, three demonstrators were assembled.
These PatMan results have been collected into a set of self-explanatory movies, enriched with sound
(downloadable from the PatMan Web page). They concern
6.
the construction of a GL with the guideline editor (Guideline
Management Tool applied to prevention of pressure ulcers),
7.
the simulation of the same GL through INCOME (INCOME
Tool demonstrated on a specific ischemic stroke session),
8.
the implementation of the same
GL into a real medical setting, once the GL has been integrated into the careflow management system (Careflow Tool
Demonstrator developed for patients with acute ischemic stroke).
Particular attention has been paid to the management of possible exceptions that can arise in clinical routine.
Plans for the Future
The major current efforts of international co-operative research groups on GL development and diffusion are focussed on establishing standards for both representation and implementation. Future refinement of the PatMan tools will go in that direction, for example by exploiting the XML language in various phases of the modelling process.
Managing guidelines is a specific instance of the generic problem of managing organizational knowledge.
The tools developed in the PatMan project are now being used in several organizational knowledge management projects. Hence, future developments will extend also beyond the healthcare domain and we will look at ways to integrate the PatMan tools into an integrated support infrastructure to manage the whole life-cycle of organizational knowledge.
WfMS developers and vendors are addressing the problem of exception management, and not only in the health care sector. Within PatMan this problem has only been partially faced, simply because it was not one of the primary project aims. However in the future it will be seriously considered, because careflow systems will be accepted by users only if they are flexible enough to tackle tasks that may not belong to a predefined GL. In addition, actual WfMS mainly manage synchronous communications, while it should be very important to manage also asynchronous communications, still with the aim of improving flexibility.
Contact Details
Project Name: PatMan – Patient
Management Workflow System
Research Area: Health Care
Timescale:
01.03.98 - 29.02.00
Budget:
Overall cost: 650 KECU
European Commission contribution: 425 KECU
Keywords: workflow, careflow, guidelines, decision support, resource allocation, ontology, internet service
Key Project Participants:
Stig Kjær Andersen, DK
Aalborg University – V-CHI
Fredrik Bajers Vej 7D1, 9220 Aalborg
Tel: +45 96358751, ska@v-chi.dk
Silvana Quaglini, IT
University of Pavia-CBIM
Via Ferrata 1, 27100 Pavia
Tel: +39 382505679 sil@aim.unipv.it
Erik Buskens, NL
University of Utrecht
P.O. Box 80035, 3508 TA Utrecht
Tel: +31302538180, e.buskens@med.ruu.nl
Enrico Motta, UK
The Open University
Walton Hall, MK7 6 AA Milton
Keynes
Tel: +44 1908653506, e.motta@open.ac.uk
Carsten Gentner, D
PROMATIS AG
Descostrasse 10, 76307 Karlsbad
Tel: +49 724891450, gentner@sun03.promatis.de
Project Coordinator:
Stig Kjaer Andersen
Project Web Address: http:/aim.unipv.it/projects/patman

HC 4017 P AT M AN FINAL P AGE 5
A major challenge for medical organisations is to increase productivity and reduce costs without adversely affecting the quality of patient care. Health care professionals now generally agree that the quality of care depends not only on the providers' professional skills but also on the level of collaboration within the organisation providing the health services.
During the last years, supporters of evidence-based medicine have focussed primarily on improving clinical practice by means of Guidelines (GLs). GLs are usually developed by authoritative panels of experts who reach consensus after a complex process involving careful literature reviews, personal communication and eventually a
Consensus Conference. GLs should prevent dissimilarities, malpractice and waste of resources in patient treatment.
However, to be effectively implemented within an organisation such as a hospital or an outpatient department, a GL requires a customisation process. Organisational characteristics such as structures, commitments, roles, policies and preferences have to be taken into account. However, existing health care information systems have often not considered organisational aspects. This is in clear contrast to industrial and commercial environments, where the concept of Workflow Management (WfM) is an established and crucial element of management information systems.
A Workflow Management System (WfMS), as defined by the Workflow Management
Coalition, is "a system that completely defines, manages, and executes workflow through execution of software whose order of execution is driven by a computer representation of the workflow logic". A WfMS provides a dynamic model of the process in question with all the involved entities explicitly defined, so as to make it possible among others to simulate the process evolution and to clarify where and when information and/or resources are produced and consumed.
The PatMan project attempts to bridge this gap in visions and practice by proposing a methodology for building decision support systems that integrate the two key aspects of health care delivery:
•
clinical procedures and
•
organisational functioning.
This methodology is suitable when clinical procedures may be described through clinical practice GLs and protocols (which provide the workflow process logic) and when a model of the organisation is available which specifies roles, responsibilities, resources, etc..
The resulting system, a Careflow Management System, can be defined as a Patient
Workflow Management System (PWfMS) applied to a specific patient situation. Since patient care in general is a distributed activity, the system is designed to be implemented on the Internet/Intranet.

HC 4017 P AT M AN FINAL P AGE 6
Therefore, the PatMan project aims are
• to provide methodological and software support for site customisation of generic healthcare guidelines.
• to assist health practitioners in the management of health care protocols by means of patient workflow management systems.
• to produce tools supporting acquisition, modelling and management of organisational knowledge and to construct specific models (i.e.ontologies) and tool configurations to support the management of medical guidelines.
Users of the services can be
•
medical experts that are responsible for the guideline/protocol formulation,
•
physicians that are responsible for guideline implementation in daily routine, and
•
resource managers, such as ward chiefs and administrative staff, who are responsible for the best allocation of resources.
The tools developed within the PatMan project aim at facilitating developers and users during the entire “life-cycle” of a guideline for clinical practice.
The project team, under the coordination of the Virtual Centre of Health Informatics at Aalborg University (DK), comprised the Knowledge Media Institute at the Open
University (UK), the Consorzio di Bioingegneria e Informatica Medica including the
University of Pavia (IT), the Department of General Health Care and Epidemiology at the University of Utrecht (NL) and Promatis Informatik AG(D).
The KMI-Open University team focused on providing tools which allow users who are not experts in knowledge modelling to take part in the collaborative construction of a knowledge model and to interact with it at a level which abstracts from the underlying language. In addition, they integrated the knowledge modelling technology with in-house (but publicly available) web-based technology to support documentcentred discussion and debate and customised the resulting integrated set of tools for medical guidelines.
The University of Pavia team developed tools for the graphical representation of a guideline, for its integration into a careflow system, and for the analysis of the guideline impact on clinical and cost outcomes. The Promatis team provided expertise and tools for simulating the careflow process through a Petri net representation of the guideline and a formal representation of the health care environment organisation.
The Aalborg University team and the Utrecht team provided support for detailing health care organisation resources.
Medical groups provided expertise for building the exemplars for the system evaluation (for the site applications of the PatMan tools). In particular, the four hospitals IRCCS Instituto Neurologico C. Mondino in Pavia, Ospedale Civile di
Brescia, Ospedali Riuniti di Bergamo and Ospedale San Gerardo di Monza (MI) were involved in collecting data about patients with ischemic stroke.

HC 4017 P AT M AN FINAL P AGE 7
The objective of this project is to demonstrate that health care organisations can benefit from introducing the concept of workflow management in their policy. The rationale behind this proposal lies in the observation that hospital information systems often lack the exploitation of organisational knowledge that is essential to guarantee efficient delivery of quality care. PatMan has developed a methodology for building decision support systems that integrate the two different aspects of health care delivery, clinical procedures and organisational functioning.
The three objectives of the project were:
9.
To develop a representation formalism for clinical guidelines and protocols, capturing medical knowledge,
10.
To develop tools for embedding into previously represented guidelines/protocols the organisational knowledge necessary to carry out tasks in specific health care environments.
11.
To evaluate these tools in one or more health care centres.
Knowledge and technology used
Patient workflow management. A WfMS is a knowledge based system using knowledge about procedures and rules that apply in different locations within the organisation. The knowledge about Workflow activities, participants and resources is used together with knowledge about their relationships in order to make an appropriate description of the health care processes. This “knowledge distribution” is reflected by the multi-layer structure as shown in Figure 1 Each layer provides a different kind of knowledge which can be used either as an input for the WfMS or to structure the overlying knowledge layer.
Workflow layer
Information layer
BM Aspirate
BM_State=NormoCellular
AND Blast<10%
PLT and PMN count
PLT and PMN count
PLT<80000 OR PMN<1000
Wait 3 days
PLT>80000 AND PMN>1000
2nd cycle
(2+3+7)
BM_State=NormoCellular
AND Blast>10%
2nd cycle
(3+5+10)
Wait 7 days
BM Aspirate
HOSPITAL INFORMATION SYSTEM PETRI NETS LIBRARY
NURSE
RESOURCE
MANAGER
Request resource
Assigned resource
Professional-Nurse-request patient
Blood drawing
Release resource
Analyst-request
Request resource
(patient,PLT,PMN)
Results reporting
Chief-Analyst-request
Release resource
Assigned resource
NURSE
RESOURCE
MANAGER
PATIENTS
RECORDS
ORG. CHART
Organisation layer
RESOURCES ROLES
Figure 1: The multi-layer structure of the Patient Workdflow Management system
Figure 1 illustrates the different elements and layers of the patient workflow management system which is the foundation of the PatMan project.

HC 4017 P AT M AN FINAL P AGE 8
The Wf Layer stores High Level Petri Nets (PNs) used to represent health care processes.
The Information Layer is represented by two information repositories: the HIS and a
Petri Net (PN) Library. The latter contains a detailed description of medical activities by means of the PNs formalism.
The Organisational Layer stores an ontological description of the health care unit organisation, where roles, resources, preferences, policies are made explicit.
The WfMS uses the resulting PN as its input model.
The guideline representation which was developed in PatMan is a modular, top down structuring of the health care process: each procedure can either be atomic or it can be decomposed into a certain number of sequential or parallel sub-procedures. This representation captures the general prescriptive aspects of Guidelines, but a PWfMS needs to be specific enough to provide precise indications for each patient and to keep track of the patient management in terms of both efficacy and efficiency. Petri Nets have proven to be a formalism suitable for this purpose, since they provide a computational framework for managing parallel and sequential processes, considering triggers, constraints and timing which govern the process execution. PNs have an intuitive graphical representation, well-defined semantics and a formal background which allows system analysis and simulation.
The project exploited existing products dealing with Petri Net representation of the workflow and its simulation and database-centred management. (Existence of databases is a fundamental requirement for the project for data regarding both the patients and the hospital resources). The project required software development for what concerns guideline and protocol formalisation, translation of these formalised models into Petri Nets and, in general, for the user interface. Since we used HTML and Java languages widely, existing WEB servers were used for the interface with the database.
In order to embed the tools in specific clinical environments, the guidelines that were represented within the PatMan project concerned three clinical problems: (1) management of children with acute myeloid leukaemia, (2) management of patients with ischemic stroke, and (3) prevention of pressure ulcers. Therefore three major user groups representing these environments worked within the project: Problem (1) was only used in the development of a GL representation, whereas (2) and (3) were used to test the complete Careflow Management System. More detail about the site applications is given in the following section on Validation and Evaluation.
Problems encountered and changes in the project plan
In the original specification the project was to produce a set of web based software tools to support the generation and evaluation of optimal job schedules. However, early in the project it became apparent that a much better focus for this task would be to provide intelligent support for knowledge sharing and retrieval, tailored to the needs of guideline users and developers.

HC 4017 P AT M AN FINAL P AGE 9
There were several reasons for this change. First of all, extensive discussions with health care workers who develop and use guidelines highlighted the need to support dialogue, knowledge sharing and knowledge management processes centred on guidelines' documents. Secondly, the Knowledge Media Institute is a centre of excellence in the area of knowledge management and this was a valuable opportunity for matching clearly defined needs to the competences of the consortium members.
No particular problems were encountered from the methodological or technical point of view during the project. However, some problems related to Java applet performance were solved by developing a Java application.
Some organisational problems were encountered at the beginning of the implementation of the guideline for the pressure ulcer prevention. Difficulties concerned the training of the nurses for use of the system, because they were few and very overloaded. The problems were solved through some meetings with the nurses themselves together with the ward chief in charge. The ensuing discussions also produced useful feedback for making the system more user-friendly and quick-to-use.
Validation and evaluation
The types and aims of the validation activities were:
1.
Validation of the different tools as they were developed to check that the results met representation requirements before moving to the next level.
2.
Validation of the translation from the GL to the PetriNet to ensure consistency between the two representation forms.
3.
Validation of user-friendliness and system utility from the users’ point of view.
Ad.1.Validation of tools:
Since the project required the development of different tools, each of them was tested as soon as it was developed. For example, to test the completeness of the semantics chosen for the guidelines protocol editor, some real world guidelines and protocols were formalised by using the tool developed. Validating this tool meant to verify if it was able to represent and run the guidelines that we were interested in.
We used guidelines provided by the AHCPR, the American Agency for Health Care
Policy Research, because they are easily accessible through the Internet and are knowledge sources certified by the National Library of Medicine. In addition they represent a wide spectrum of possible guideline formats, ranging from those very descriptive and general to those very prescriptive and detailed.
Ad. 2. Validation of the guidelines translation:
The second validation type concerned the translation from the guideline to the Petri
Net, and this was verified by medical user groups. The validation concerned not only the translating tool, but also and more importantly, the description of the organisation.
The PetriNet must provide the input to the Workflow system for what concerns the use and release of the various resources.
The guidelines that were represented within the PatMan project concern three clinical problems: (1) management of children with acute myeloid leukaemia, (2)

HC 4017 P AT M AN FINAL P AGE 10 management of patients with ischemic stroke, and (3) prevention of pressure ulcers.
Accordingly, there were three major user representation groups working with the project, chosen because of their close working relationship with the Pavia consortium:
1.
the paediatric division of the IRCCS San Matteo University Hospital in Pavia
2.
the stroke unit staff of the IRCCS Mondino Neurological Hospital in Pavia, which co-ordinates the work about stroke in other three main hospitals in the Lombardia
Region, namely in Monza, Bergamo and Brescia
3.
the General Medicine ward of the IRCCS Fondazione S. Maugeri-Clinica del
Lavoro, Pavia.
Ad. 3. User Representation Groups:
Finally the applications were the validation test from the users’ point of view. In this last phase the user-friendliness of the interface and the system utility perception by the user were tested.
As soon as the prototypes were available, the approach consisted of (1) a training activity about the system use, and (2) use of the system for a pre-defined, short time, and (3) report of claims and suggestions by the users.
The motivations of each user group from the hospitals mentioned above for implementing a guideline were different. Pediatricians were interested in the protocol representation, to be used as an intelligent browser and an educational tool, while the stroke unit team was interested in the improvement of the therapeutic process management (workflow aspect) and in the validation of the guideline (correlation between compliance and health/economic outcome). Finally the general medicine team was interested in improving the nurses’ information exchange (another workflow aspect) and in assessing the nurse workload in an objective way.
User groups greatly contributed to the development of the project by stimulating discussions about both the guideline formalisation and the optimal guideline implementation in the clinical routine. The latter issue is very important, because without a user consensus on the user/guideline interaction it is impossible to produce a real-world application. In particular, the stroke unit staff also provided all the information about the unit organisation (resources, timing, personnel, etc), required to implement the workflow prototype.
Major efforts were focussed on improving user acceptance:
•
The possibility of gleaning textual information from the web was added to the guideline task (very useful for education).
•
The “silent” modality was implemented for the stroke management guideline, i.e.
the guideline does not suggest anything until a non-compliance is evident. This was done because the users of this guideline are expert neurologists, and it would be too annoying for them to continuously see the guideline suggestions.
•
Some rules of the ulcer prevention guideline that were judged too restrictive by the nurses were reformulated in such a way that the AHCPR guideline was made site-specific. This was necessary for organisational constraints.

HC 4017 P AT M AN FINAL P AGE 11
Specific user feedback from the implementation of guidelines in two clinical
situations:
Ischemic stroke:
The four hospitals IRCCS Istituto Neurologico “C. Mondino” in Pavia, Ospedale
Civile di Brescia, Ospedali Riuniti di Bergamo, Ospedale San Gerardo di Monza (MI) were involved in collecting data about patients with ischemic stroke. Each hospital has a stroke unit, and the same guideline was implemented at each unit. The guideline, originally proposed by the American Heart Association, was site-specified according to the differences between the American and the Italian systems and populations. As an example, some examinations, like VDRL, that are mandatory in the US, due to the incidence of sexually-related diseases, in Italy are not part of the routine, because they are not cost-effective. In addition, the four medical centres used the same schema for the patient database (MS-ACCESS 2 with a Visual Basic user interface), which is linked to the guideline inference engine.
Modifications with respect to the original guidelines were decided by professionals of the four centres during three ad-hoc meetings, two of them before the guideline implementation, and one after one month of experimentation.
The system was used on the average by three health care operators at each site. 485 patients were collected during the second year of the project. After a first phase (three months of preliminary experimentation) during which program bugs were corrected and some adjustments made according to the user suggestions, no particular problems were encountered. Medical users were particularly happy with the system, because it allowed them to perform statistics on a population of patients that is "homogeneous", because patients are treated with the same guideline, independently from the site. In this application this is very important, because pooling different sets of data is the only way to have enough data to perform significant analyses.
In the ongoing application report the statistical analysis was performed on the collected data. The analysis highlights the need for a workflow system, mainly for organising first aid to the stroke patients. The system is found to be useful as a signalling system, hence the field “do not forget”. The most important part in the treatment is the “need of speed” after onset, and proper treatment as soon as possible to enable good recovery. Evaluation of these goals was performed through statistical analysis and preliminary results are available, that demonstrate the cost-effectiveness of the guideline implementation. Statistics were performed on the parameters that were chosen at the beginning of the study as the indicators for the guideline impact.
They concern survival, quality of life and direct hospital costs.
The pressure ulcer prevention system:
The GL implementation took into special account that it might adversely affect or interfere with daily work. At the beginning of their shift, nurses ask for a print-out of the workplan for the day. The workplan is produced by the GL inference engine according to the GL rules and fired with data stored in the electronic patient record
(EPR). At the end of their shift, nurses store data about the executed workplan, using the same interface that they have been using for many years for the HIS. Of course nurses adopting the GL are not the only users accessing the EPR. Additional information is stored also by physicians and laboratorists. The difference lies in the

HC 4017 P AT M AN FINAL P AGE 12 fact that nurses must fill the EPR, because the GL will not produce the next day workplan if the previous day’s workplan has not been discharged. Also storing the non-compliance is mandatory in order to continue with next workplans: if some of the suggested actions have not been performed, nurses must give a justification, and the same actions will be suggested again the day after.
The interaction with the system is then at specific time intervals at the beginning and at the end of the daily work. All day long, nurses annotate information on the paper workplan. This was considered the optimal system/user interaction in our environment, because a real-time storing of data about actions performed on the patient is impossible for logistical reasons. Nurses may browse the guideline represented through the graphical editor, and by clicking on the action boxes, they may obtain explanations about the suggestions. Texts are derived by the AHCPR prose guideline and by the AISLEC text (the text modified by the Italian nurse association).
The computerised GL were routinely used for some months. Feedback from nurses can be divided into pro’s and con’s:
Positive comments were:
• improved planning - the printed workplan is useful for the nurses, particularly in the moment of the change of the shift from one nurse to another.
• workplan documentation - at the moment of patient discharge, all the preventive actions are documented and printable.
• workload assessment - the system facilitates estimating in an objective way the workload that the ulcer prevention activity implies for the nurses.
Further requirements for a more effective use of the guideline:
• more flexibility in the risk assessment – nurses would like to be able, even if the case is rare, to put a patient at risk, even if the rules are not met.
• more flexibility in the time of actions – repositioning every patient at a fixed frequency is not considered effective and efficient; nurses want to be able to tailor the repositioning frequency to the single patient.
• more facilities for quick data input – the number of nurses in the ward is not very high, so time is very important. They ask facilities such as default answers that can be confirmed by a simple mouse click.
In conclusion, we designed and implemented the decision support system for the pressure ulcer prevention, integrating it into the existing electronic patient record of a general medicine ward.
Evaluation of general project experience
Not surprisingly, a major lesson learned in relation to general project experience was the importance of a balanced profile of the project partners. Moreover each member of the project team has to have well defined roles. In PatMan the partner composition included small industry, medical informatics researchers with close connections to hospital networks, knowledge management expertise, and medical experts. This composition profile is well suited for pursuing most of the project aims. The PatMan experience is that the collaboration works well when there is clarity about the division

HC 4017 P AT M AN FINAL P AGE 13 of expertise among the partners and clarity about the interfaces between the competency areas of the partners. In addition there must be a common platform for communication and terminology, mutual respect and interest. When this situation was not present, the project experienced firstly, that the workload of the partners become unbalanced and secondly that the given task was not satisfactory completed the first time around. The success of the project depends on the ability to create the first situation – and on the support the tools developed in the project can provide.

HC 4017 P AT M AN FINAL P AGE 14
The PatMan project aims at promoting an effective and efficient health care delivery by providing integrated support for tackling medical and organisational issues during patient management. A major objective of this project is to demonstrate that health care organisations can benefit by the application of a careflow management system, which is a workflow management system (WfMS) applied to specific patient situations.
Summarising, the PatMan project aims have been:
• to provide methodological and software support for site customisation of generic healthcare guidelines.
• to assist health practitioners in the management of health care protocols by means of patient workflow management systems.
• to produce tools supporting acquisition, modelling and management of
organizational knowledge and to construct specific models (i.e.ontologies) and tool configurations to support the management of medical guidelines.
These aims have been met by the following activities:
•
Development of a representation formalism for clinical guidelines and protocols.
•
Development of a set of tools supporting the deployment of medical guidelines and protocols in the context of work practices in specific health care settings.
•
Empirical evaluation of these tools in health care centres.
The PatMan tools comprise tools that were completely developed under PatMan, tools
adapted within the PatMan project, and tools that were adapted and refined from the previous HC-REMA project.
The tools developed within the PatMan project comprise:
•
Tool no. 1: The Guideline Editor, GUIDE, which computerises the representation of a guideline or protocol. The editor is written in Java and made available as a web-service.
•
Tool no. 2: The WPDL Translator is a program that produces a representation of the Guideline as a WPDL (Workflow Process Definition Language) file. The translator is written in C++.
•
Tool no. 3: The Organisation Ontology Building Tool is used for modelling the organisational aspects of the careflow. A prototypical ontology for health care organisations has been embedded. The tool is written in Visual Basic. Oracle Web service is used for storing and assessing data.
The commercial tools adapted within the PatMan project comprise:
•
Tool no. 4: INCOME is a commercial product from the partner PROMATIS. This tool exploits a relational database. INCOME is able to simulate a Petri Net representing the guideline/protocols. It embeds facilities to represent the behaviour model, the organisation model and the information model.
•
Tool no. 5: Oracle Workflow is a commercial product from ORACLE which permits describing and implementing a workflow model on a real world setting.

HC 4017 P AT M AN FINAL P AGE 15
Three main modules compose this tool which is accesseble through ORACLE web server.
•
The Builder allows the user to build the model as a sequence of tasks. Each task can be provided with the opportune functionality through PL/SQL code.
•
The Monitor allows for viewing and administering the status of a specific instance of a workflow process.
•
The Notification Mailer lets people receive notifications of work items awaiting their attention via E-mail, and acts based on their E-mail responses .
1.
The final tool is actually an integrated set of technologies supporting knowledge management processes, which have been customised for the management of medical guidelines. Some of these technologies have been developed in PatMan, some in the related project HC-REMA (by the same consortium) and same constitute baseline technology developed by OU-KMI outside the PatMan project.:
Tool no. 6: The Guideline Discussion Forum (GDF) integrates several technologies supporting organizational management and customises them for supporting medical guidelines. The web-based support for discussion and debate provides an environment where both guideline developers and users can discuss and inspect guidelines. This facility is tightly integrated with knowledge modelling tools, so that semantic information retrieval is also supported. In particular, users can use the LOIS query tool, to find information about a guideline through the associated
knowledge model, constructed by instantiating for a particular guideline the generic
PatMan ontology for medical guidelines. GDF is also integrated with the WebOnto, a web-based editor/browser for knowledge models.
.
A patient workflow management system, supported by the above mentioned modular tools, produces a guideline-based careflow management
system, as illustrated in
Figure 2.
PatMan TOOLS
Tool #6 Guideline Discussion
Forum: Management of medical and organisational knowledge
Medical knowledge
Tool #1
GUIDE
Tool #2
WPDL Translator
Guideline formalisation
WPDL model
Organisational knowledge
Org Model
Tool #3
Organisation
Ontology Builder
The tools and the demonstrator applications can be used in the analysis, design and implementation steps in the development life cycle of a Careflow
Patient Workflow Model
WPDL – Petri Net
Tools # 4-
INCOME for simulations
Database describing
Organisational knowledge
Resource allocation
Management System. The tools and applications that have been developed are presented in the table that follows.
Tools # 5 Oracle
Workflow: Web service for database integration and GL implementation
Patient
Database Careflow Management System
Applications :
Stroke
Pressure Ulcer
Diabetes
Aorta Aneurysm
Figure 2 – Tools and tool relationships for the development of a careflow system
Empirical evaluation of the tools and applications took place in the following application domains:

HC 4017 P AT M AN FINAL P AGE 16
4.
Stroke: emergency guideline for ischemic stroke management
5.
Ulcer: guideline for pressure ulcer prevention
6.
AAA: surgery guideline for aneurysm aorta abdominalis (acute abdominal aneurysm)
7.
Diabetes: follow-up guideline for diabetes
The table indicates how the development of these guidelines has been performed differently in each case, depending on the context and application. The modularity of the tools makes it possible to make appropriate choices during the entire development cycle. The development process entails three steps: (1) a domain analysis leading to a business process model, (2) workflow design with the help of the PatMan guideline editor, resulting in a patient-workflow model, and (3) a workflow simulation with the help of Petri Net tools. The domain analysis gives a first semi-formal diagrammatic representation of the target process. This is subsequently refined with the help of the other PatMan tools. Typically, the scope of domain analysis is broad, and subsequent design steps focus on a selected subpart of this process.
Domain analysis
Worksheets
Tools
Preliminary analysis
Outline design Business Process Model tools
Guideline analysis
Guideline Editor
Discussion Forum
Domains
AAA, diabetes,
Stroke
AAA, diabetes
Stroke, ulcer
Workflow design
Guideline Editor
Translator
Stroke, ulcer
Design
Implementation
Workflow simulation
INCOME Petri Net
Patient Workflow Model
Stroke
Stroke, ulcer
Workflow management system
Careflow System
WebOnto
Database Integration
Stroke, ulcer
Ulcer
Stroke, ulcer
In the following a brief description of each tool is given, including a summary description, the necessary software platform, and a description of the tools’ functionality, input and output. Since the tools are internet based, no hardware requirement is given.

HC 4017 P AT M AN FINAL P AGE 17
Tool no. 1 - The Guideline Editor
Description: The Guideline Editor is a graphical editor, GUIDE, which has been developed within the PatMan project. It is used for the computerised representation of a guideline or protocol.
Functionality: The Guideline Editor provides the semantics for the graphical guideline/protocol representation. It allows representing and browsing the guideline at different detail levels, and it embeds an inference engine for producing advice in connection with the patient record. In addition it is linked to a decision model builder and analyser, developed within the previous project HC-REMA. It can be used as an educational tool. The system functionality is then illustrated through a fully developed medical exemplar, concerning an official guideline for managing patients in a stroke unit. A demo version is available on the PatMan homepage.
Software platform: The Guideline Editor is written in Java, and it is available as an application and as a Web service. It runs on Unix and Windows platforms.
Input: a formal description of a textual guideline in terms of tasks and subtasks, and rules for the task execution. The SNOMED server, another input to GUIDE, is useful for maintaining a homogeneous terminology.
Resulting output: an internal representation of the guideline into a relational database.
Tool no. 2 - The WPDL Translator
Description: The guideline contains only medical knowledge, but to achieve an efficient and effective care delivery, both medical and organisational knowledge must be represented through computational formalisms. The WPDL Translator is a program that produces a representation of the guideline as a WPDL (Workflow Process
Definition Language) file.
Functionality: The WPDL Translator is used as an interface from GUIDE to
INCOME (the tool for the Careflow simulation). INCOME is the system used for the workflow modelling and simulation. To be imported into INCOME, the guideline is firstly translated into a high-level Petri Net and then into the WPDL language.
Resources, both human and technological, that are necessary for performing the activities are represented within an organisational model in INCOME.
Software platform: It is written in C++.
Input: the relational database describing the guideline.
Resulting output: WPDL description of the guideline in terms of activities and transitions.
Tool no. 3 - The Organisation Ontology Building Tool
Description: The Organisation Ontology Building Tool is a program used for modelling the organisational aspects of the careflow. A prototypical ontology for health care organisations has been embedded.
Functionality: The Organisation Ontology Building tool provides a module for the graphical editing of the ontology, a module for the creation of normalised tables according to the previously specified ontology, and a module for the ontology class instantiation.
Software platform: It is written in Visual Basic. Oracle Web service is used for storing and assessing data.
Input: organisational knowledge in terms of organisational units, roles, functions, etc..
Resulting output: a relational database describing the specific organisation; this database is an input for the careflow manager.

HC 4017 P AT M AN FINAL P AGE 18
Tool no. 4 - INCOME
Description: INCOME is a commercial product from PROMATIS, one of the partners of the PatMan consortium. This tool exploits a relational database.
Functionality: INCOME is able to simulate a Petri Net representing the guideline/protocols. It embeds facilities to represent the behaviour model, the organisation model, and the information model.
Software platform: Windows and Unix.
Input: the WPDL file obtained through the tool no. 2, organisational knowledge in terms of organisational units, roles, functions, etc., and simulated patient data.
Resulting output: the Petri Net version of the guideline, and reports from simulation concerning resource consumption, in terms of both costs and physical resources.
Tool no. 5 - Oracle Workflow
Description: Oracle Workflow is a commercial product from ORACLE, which permits describing and implementing a workflow model in a real world setting.
Functionality: Three main modules compose this tool. The Builder allows the user to build the model as a sequence of tasks. Each task can be provided with the opportune functionality through PL/SQL code. The Monitor allows for viewing and administering the status of a specific instance of a workflow process. The Notification
Mailer lets people receive notifications of work-items awaiting their attention via Email, and acts based on their E-mail responses.
Software platform: Windows and Unix
Input: the WPDL file obtained through the tool nr. 2, organisational knowledge in terms of organisational units, roles, functions, etc., provided by Tool nr. 3, and the patient record.
Resulting output: the careflow management system.
Tool no. 6 - The Guideline Discussion Forum
Description: The Guideline Discussion Forum is a tool primarily developed within other projects, and refined within PatMan. It exploits WebOnto, a tool developed within the previous project HC-REMA by the same consortium. It provides a web environment where experts can discuss guideline specification, and users can ask for particular parts of the guideline prose. The tool comprises of (1) a web-based browser/editor allowing collaborative model development over the web (WebOnto),
(2) a user-friendly front-end interface, which allows users to query a knowledge base at a level which abstracts from the underlying modelling language (Lois), (3) a knowledge acquisition interface which takes advantage of an underlying ontology to support a model-based approach to knowledge acquisition (Knote), and (4) an ontology for characterizing medical guidelines and (5) a web-based environment, customised for medical guidelines, which support guideline-centred discussion and debate.
Functionality: Both guideline developers and users can discuss and inspect guidelines, using both lexical and semantic retrieval mechanisms The discussion is organized around the guideline specification document and is structured in terms of a simple set of argument-forming relations, such as agree, disagree, comment about and
question about. Semantic retrieval is supported by guideline-specific instantiations of the generic PatMan ontology for medical guidelines. This ontology defines formally the notion of medical guideline and identifies the various types of knowledge associated with a guideline (e.g., associated-condition, population-user-type,

HC 4017 P AT M AN FINAL P AGE 19
location-constraints, etc.). The PatMan ontology builds on a generic medical ontology developed by the same consortium in the earlier HC-ReMa project and on other ontologies available at the OU-KMI ontology server.
Software platform: The tool is web-based and accessible through any web browser.
The server technology itself has been developed by the University of Pavia and OU-
KMI and can be installed on Unix, Windows and Linux platforms.
Input: the guideline ontology, ontology and instantiations with specific guidelines.
Resulting output: the discussion forum about guidelines.
To enable demonstration of the different functionalities of the developed tools and their interaction with data and knowledge stores, self consistent demonstrators for the
Windows platform were assembled. These are available on CD-ROM and as downloads from the project website. http://aim.unipv.it/projects/patman
The "guideline management tool" Demonstrator
The guideline management tool offers health care professionals the possibility of editing and browsing models stored in different servers, and to use these models as a decision support in connection with the electronic patient record. The tool is designed to work in a web environment.
The fundamental components of the demonstrator are:
Tool #1 GUIDE ( Gui deline Userfr iendl y I nt egr ated Developm ent
Environm ent ), which is a program to edit, save and browse guidelines
Connections to knowledge bases (see fig. 3)
GL archive
Tool #1
GUIDE
Medical knowledge
Guideline editor
Simulator
Decision models
Patient
Database
Interpreter for the GL representation in the GL archive for application in a real situation
(Inference Engine).
Applications
Pressure Ulcer GL
Figur 3 GUIDE in context
Condition Interpreter, which chooses among different paths of the guideline on the basis of actual patient conditions.
Decision-theoretic model to be used when the physician must choose among different alternatives and the guideline does not indicate a preferred path.
The GUIDE-related software is written in Java, using Jdk1.1.7, Swing 1.0 for graphical environments, JDCB for connections to databases and RMI for connections to Engine and Condition Interpreter. The tool has been tested on Solaris 2.5 and 2.7,
Win9x and Windows NT. The GUIDE has to connect to an Oracle database for loading and saving guidelines. The JDCB connection has been successfully tested in the presence of a firewall. Unfortunately this is not true for connections to the Engine

HC 4017 P AT M AN FINAL P AGE 20 and Condition Interpreter. For this reason, the tool functionalities are maintained separately, in such a way that a user, according to his own configuration, can use one or more of them.
In the PatMan context “Medical knowledge” is mainly the set of clinical practice suggestions, contained in the GLs text. We provide a meta-knowledge, represented by an ontological description of the possible tasks that can be included into a GL. This guarantees a common terminology and task management among different GLs that possibly run in the same environment. The same conceptual model should be shared by the database storing the electronic patient record, to allow an easy integration with the GLs. The task ontology has been drawn from UMLS (Unified Medical Language
System), in particular taking into consideration the SNOMED and Mesh thesaurus.
In GUIDE, the GL is hierarchically structured in “pages” representing different abstraction levels: there are non-atomic tasks, whose expansion will be in an inner page. Every task is labelled wit h a bull et indicati ng the strength of the scient ifi c evidence suppor ting the indi cat ion cont ained in the task i tself . A deci sion task m ay al so call a model based on deci si on theor y, in or der to handle the deci si on maker /patient pr ef erences. A GL is internally represented by a relational database.
The GUIDE demonstrator session contains a re-writing of the Pressure Ulcer Guideline including a customisation and simulation with a set of real patient data. A demo version of the editor is available as a link at the project website. (GLDem
A LOTUS SCREENCAM recording of the complete session is available on the
PatMan CD-ROM.
The INCOME Tool Demonstrator
The INCOME Tool demonstrator demonstrates how the INCOME tool takes the
WPDL description of the guideline as input and is able to build the corresponding
Petri Net. The Petri Net is the “behavioural model” of the workflow, because it represents the sequence of activities to be performed. Furthermore, INCOME requires the representation of resources necessary to perform these activities. Resources have been represented in the organisational model.
We can run simulations by using
INCOME. Tokens passing through activities may be seen during simulation running, by means of the so-called “animation” opportunity provided by INCOME.
Fig 4. shows the related databases and representations
Patient Workflow Model
WPDL – Petri Net
Tools # 4 INCOME
Simulation
Database describing
Organisation knowledge
As a demonstration of the animation functionality, the network shown in Fig. 5 becomes animated, with tokens and colours indicating the activity flow.
Patient
Database
Applications
Ischemic Stroke
Figure 4. INCOME in context
Particular pictures may be assigned to passing tokens, in such a way as to give an immediate understanding of what is happening. The left side of Fig. 5 shows the higher level of the hierarchic Petri Net as

HC 4017 P AT M AN FINAL P AGE 21 generated by INCOME by reading the WPDL file. The right side of Fig. 5 shows the expansion of the sub-acute phase.
Once the simulation run is finished,
INCOME maintains the results in its tables and views. Some reports are given by default, such as resources allocated at any particular time. Other reports, which are particularly interesting for our applications, have been designed through SQL queries on the INCOME
Figure 5. Petri Net generated by INCOME database. On the basis of the retrieved variables, different graphical outputs have been produced using
MATLAB.
The demonstrator consists of a set-up allowing for traversing with the tool a hypothetical session. The demonstrator which may be used through the web at a remote site.
A specific Ischemic Stroke session is recorded with LOTUS SCREENCAM and available as a download and on the PatMan CD-ROM.
The CAREFLOW tool demonstrator
Through Oracle Workflow, we have developed a prototype of the careflow system for patients with acute ischemic stroke. The main modules of the system interacting with the Oracle Workflow Engine are:
12.
The Monitor, that allows viewing and administering the status of a specific instance of a workflow process (work-item). The status of any set of events that have been traversed appears differently from an untraversed set of events.
By this tool the administrator is able to control whether the workflow runs correctly.
Patient Workflow Model
WPDL – Petri Net
Database describing
Organisation knowledge
Tools # 5 Oracle
Workflow
Web service for database integration
Patient
Database
Monitor
Notifier
Activities
Applications
Ischemic Stroke
Careflow Management System
Patient Workflow
Management System
Figure 6. Careflow Tool

HC 4017 P AT M AN FINAL P AGE 22
13.
The Notification Mailer, which lets people receive notifications of work-items awaiting their attention via E-mail, and acts based on their E-mail responses. Each involved operator can inspect his own to-do list, including necessary support information in order to perform the task(s).
14.
The Activities, which are real-world tasks. Activities may be either manual, semiautomatic or automatic. For manual activities the workflow engine only requires notifications for operator availability and for activity completion. For semiautomatic activities, the engine offers support such as html forms and allows executing the assigned tasks through a web browser. The automatic activities are transparent to the users and are completely executed by the computer.
With the use of these modules it is possible to check the workflow status and resource availability, to assign tasks to people according to their roles as defined in the organisation model, and to monitor task execution. Information about the performed tasks are stored in the workflow database in such a way that statistics can be performed concerning times, costs and workloads.
The session follows the executing of the Workflow Engine managing a set of tasks for emergency, anamnese and instrumental examination. The demonstrator shows how the system recruits an agent for a task execution, especially how to differentiate between a compliant and a non-compliant task.
The Careflow tool demonstrator specific for Ischemic Stroke is recorded with LOTUS
SCREENCAM and available as a download and on the PatMan CD-ROM.
The Discussion Forum for knowledge sharing
The PatMan knowledge management tools are generic: they can be applied to medical decision making, as well as to other domains. Over the past 12 months this generic suite of tools has been applied to several other knowledge management projects, with industrial, academic governmental organisations. To name just a few, we should mention our ongoing collaboration with British Aerospace in the domain of managing
‘best practices’ and the work with UK’s University of Industry, which aims to provide intelligent support for users looking for courseware addressing specific organisational and student-centred needs.
In the context of the PatMan project we focused on a specific guideline, “Pressure
Ulcer in Adults: Prediction and Prevention” guideline (delivered by the AHCPR
Association). The resulting application model is part of the demonstrator and is placed on a customised server - see http://enrich.open.ac.uk/patman/docs/pressureulcer/pressure-ucer.html (a link on the project web page).
The Discussion Forum demonstrator session is recorded with LOTUS SCREENCAM and available as a download and on the PatMan CD-ROM

HC 4017 P AT M AN FINAL P AGE 23
Figure 8. Snapshot of the PatMan Discussion Forum applied to Pressure Ulcer guideline
The pressure ulcer page shows the guideline specification document, automatically structured by means of OU-KMi’s D3E toolkit. This associates discussion spaces to document sections, according to criteria set by the user. For instance, in figure 8 we show the part of the guideline document related to the Risk Assessment activity. On the right hand side of the figure it is possible to see one particular thread of discussion, associated with this activity. In general, several threads can be associated with a particular topic: our demonstration in the domain of pressure ulcer contains both threads of discussion between physicians and nurses trying to apply the guideline in a particular healthcare setting, as well as discussion between modelling expert trying to formalise the guideline in terms of the
PatMan ontology.
The two buttons on the bottom left corner of the PatMan Discussion Forum allow users to view the pressure ulcer ontology in WebOnto (“View Ontology”) and to pose queries using the Lois interface (“Query”). Lois makes it possible to ask ‘semantic queries’, such as “what activities in the pressure ulcer guideline require the intervention of a physician?”. An important aspect is that it is possible to home in from the answers, say “pressure ulcer risk assessment is an activity which has to be carried out by a physician”, to the relevant part of the guideline document. Thus, it is possible to use Lois to access the guideline document semantically, rather than through lexical search.
In addition to the customisation of the Discussion Forum in the domain of pressure ulcer, we should also mention other results which are publicly available. The
WebOnto tool enables knowledge engineers to collaboratively browse and edit

HC 4017 P AT M AN FINAL P AGE 24 knowledge models over the web, using a standard web browser. Since WebOnto was made publicly available on the web in October 1999, we have had several hundred visitors to the site and the tool has attracted significant interest from both academic and industrial organisations. The University of Amsterdam carried out a comparative evaluation of several knowledge modelling tools, including WebOnto. Web Onto was evaluated very favourably, in particular being judged as the most user-friendly and as the knowledge modelling tool requiring the shortest learning curve. We have been informally approached by the chair of the W3C's RDF Interest Group enquiring about the possible use of WebOnto to support RDF specifications. Our library of OCML models can be accessed through the WebOnto browser at URL http://webonto.open.ac.uk
. (a link on the project web)
The PatMan Project web page
Figure 9 shows the web page with the information related to the PatMan Project. In particular, from this page it is possible
to identify the project partners and to visit their web sites
to download the Java applet for managing a graphical representation of guidelines
to download some movies illustrating the main PatMan tools, namely
a session where a guideline is graphically represented and then connected to a patient record.
a simulation of the workflow using a Petri Net within the INCOME environment.
a session where the workflow model is being designed within the ORACLE
WORKFLOW environment.
a session where users interact with the workflow system previously developed.
Movies have been produced by using LOTUS SCREENCAM, and the executable files have been put on the server for downloading.
Fig 9. The web page related to the demo

HC 4017 P AT M AN FINAL P AGE 25
The principle objective of this project was to demonstrate that health care organisations can benefit by the application of a careflow management system, which is a WfMS applied to specific patient situations.
The project results, both the prototype tools and the results of implementing the tools in real clinical situations for several health domains, show the benefits of providing a dynamic model of the workflow process in question where all the involved entities are explicitly defined. Experience has shown that the dynamic model makes it possible: a) to simulate the process evolution to clarify where and when information and/or resources are produced and consumed, b) to implement GLs in the real clinical setting, by efficiently allocating the necessary resources.
Further development
Despite the great emphasis on medical guidelines, it is widely recognised that their implementation in clinical routines is very difficult and impaired both by the way in which guidelines are represented and by organisational constraints. The guideline editor and the guideline inference engine, developed within the project to overcome some of the GL representation problems, are able to manage a library of guidelines, each represented unambiguously and enriched with explanatory text. This can be thought of as a service provided by a “centre of excellence”, available on the web.
International co-operative research groups on GL development and diffusion are currently focussing major efforts on establishing standards for both representation and implementation. Future refinement of the PatMan tools will go in that direction, for example by exploiting the XML language in various phases of the modelling process.
WfMS developers and vendors are addressing the problem of exception management, and not only in the health care sector. Within PatMan this problem has only been partially faced, simply because it was not one of the primary project aims. However in the future it will be seriouosly considered, because careflow systems will be accepted by users only if they will be flexible enough to tackle tasks that may not belong to a predefined GL. In addition, actual WfMS mainly manage synchronous communications, while it should be very important to manage also asynchronous communications, still with the aim of improving flexibility.
•
•
•
•
Real world implementations already exist for some PatMan tools which have explored various facets and identifies bottlenecks and some scale-up problems: the use of the guideline graphical representation as an educational tool, for health care professionals .
the link of the guideline with the patient record, in order to give suggestions to professional health care operators and in order to store non-compliances.
the production of a written work-plan for nurses in the field of pressure ulcer prevension in order to enhance communication among health care operators.

HC 4017 P AT M AN FINAL P AGE 26
However the workflow-engine, at present only available in the laboratory at a prototypical level, needs to be applied to a real-world implementation. Especially the presence in the project of a company that produces workflow management systems is a good premise for moving the workflow engine to an implementation phase, which we hope to reach in a future EU action.
In order to extend the range of applications , the GUIDE tool will be used for representing a guideline to manage rehabilitation of stroke patients. In this case the end-users of the result will be the citizens themselves, that will be able to continue the rehabilitation program at their home. Commercial exploitation requires a tool refinement for what concerns ODBC connection to different database environments.
For example making available a connection to MS-ACCESS will make the tool a stand-alone package usable on a single PC. Actually, the stand-alone part of GUIDE concerns the editing and browsing possibilities, while accessing the patient record and the whole GL library require an ODBC connection with the ORACLE database.
From a technical point of view, the translation of the guideline into WPDL is a guarantee of its portability to different workflow systems, because WPDL is the language recommended by the Workflow Management Coalition. PatMan translates this code, starting from a representation of the guideline in ORACLE tables.
For the Discussion Forum we have now available both a powerful suite of integrated tools supporting various knowledge management activities and a customization of this technology for the area of medical guidelines, applied to the domain of prevention and prediction of pressure ulcer in adults. The quality of our Discussion Forum technology can be ascertained from the fact that it is now used in over half a dozen projects, in domains ranging from managing organisational best practices, to delivering multi-media courseware to students. Partners include blue chip companies such as British Aerospace, British Petroleum and Siemens, as well as governmental organizations such as the UK’s University for Industry. The total value to OU-KMI of our ongoing projects in knowledge management, which are exploiting modelling technology developed in the course of the HC-ReMa and PatMan projects now exceeds Euros 1.8M. In addition, the OU-KMI has just been awarded a highly prestigious grant by the UK’s Engineering and Physical Sciences Research Council, in the context of the programme on Interdisciplinary Research Collaborations.
Although the final financial details have not been finalised yet, the award is expected to be worth Euros 2.12M to OU-KMi, over a six years span. The aim of this project, whose overall value will be in the region of Euros 11.4M is to produce the next generation of knowledge management technology, which will integrate results from various disciplines, including organizational science, knowledge acquisition and management, communication technology and language engineering, to name just a few. The technology developed in PatMan to support modelling and management of organizational knowledge will provide one of the main baseline technologies for the project. This highly prestigious award testifies not only to the importance of these issues for the present day knowledge economy, but also to the value and the maturity of the OU-Kmi technology, much of which has been developed in CEC-funded healthcare projects.

HC 4017 P AT M AN FINAL P AGE 27
Since the key users of the PatMan services are: a) medical experts that are responsible for the guideline/protocol formulation, b) physicians and other health staff who are responsible for guideline implementation in daily routine, and c) resource managers, such as ward chiefs and administrative staff, who are responsible for the best allocation of resources, then the potential markets for the PatMan tools are
• medical associations delivering clinical practice guidelines, which are interested in delivering a high-quality product and for whom the electronic version represents an added-value;
• hospital administrators interested in improving their organisations mainly for efficiency and economic reasons;
• software companies selling hospital information systems, interested in providing innovative products linked to the patient record.
Reaching these groups could be the focus of future activities.
In light of the results obtained with the guideline implementation in real world settings during the project, we can add to this list the main national and regional health care policy makers. The results of the stroke management system have already been sent to the Italian Ministry of Health. The role of these political entities should be to motivate local hospital administrators to use such systems, thereby helping to eliminate some of the organisational constraints of GL implementation.
The industrial partner PROMATIS AG is a company that produces and sells
Workflow Management Systems, while the Consorzio di Bioingegneria e Informatica
Medica with years’ experience in the health care domain, co-operates with the administrative staff of several hospitals. The synergy between the two partners guarantees the possibility to operate in the market area involved. KMI has known expertise in developing, promoting and commercializing web-based knowledgeintensive communication platforms.

HC 4017 P AT M AN FINAL P AGE 28
Project Name: PatMan – Patient Management Workflow System
Research Area: Health Care
Timescale: 01.03.98 - 29.02.00
Budget: Overall cost: 650 KECU
European Commission contribution: 425 KECU
Keywords::workflow, careflow, guidelines, decision support, resource allocation, ontology, internet service
Contractors:
Aalborg University – V-CHI, Stig Kjær Andersen, DK
Fredrik Bajers Vej 7D1, 9220 Aalborg
Tel: +45 96358751, ska@v-chi.dk
University of Pavia-CBIM, Silvana Quaglini, IT
Via Ferrata 1, 27100 Pavia
Tel: +39 382505679 sil@aim.unipv.it
University of Utrecht, Dep of General Health Care and Epidemiology, Erik Buskens, NL
P.O. Box 80035, 3508 TA Utrecht
Tel: +31302538180, e.buskens@med.ruu.nl
The Open University, KMI, Enrico Motta, UK
Walton Hall, MK7 6 AA Milton Keynes
Tel: +44 1908653506, e.motta@open.ac.uk
PROMATIS AG, Carsten Gentner, D
Descostrasse 10, 76307 Karlsbad
Tel: +49 724891450, gentner@sun03.promatis.de
Associate Contractors
University of Amsterdam, SWI, A.Th. Schreiber
Roetersstraat 154, 1018 WB Amsterdam
Tel: +31205256792, schreiber@swi.psy.uva.nl
Project Coordinator: Stig Kjaer Andersen
Project Web Address: http:/aim.unipv.it/projects/patman

HC 4017 P AT M AN FINAL P AGE 29
Authors: Silvana Quaglini, Clara Fassino, Mario Stefanelli - University of Pavia
Abstract In this deliverable we describe the work carried out within WP2 about the editor for representing guidelines and protocols in a formal way. We designed a system running on the internet, that offers to health care professionals the possibility of editing and browsing models stored in different servers. Both functional and computational architectures of the tool are described, as well as the underlying database. The system functionality is then illustrated through a fully developed medical exemplar, concerning an official guideline for managing patients in a stroke unit. A demo version of the editor is available at the following internet address: http://aim.unipv.it/~clara/java/Guide/GL.html
Date: January , 1999
Authors: Silvana Quaglini, Carlo Mossa, Clara Fassino, Mario Stefanelli -
University of Pavia
Abstract: The previous deliverable D1 described the graphical editor we developed to produce a computer-based representation of clinical practice guidelines and protocols.
The representation is internally based on relational tables. The present deliverable describes the next step to achieve a ìguideline-based patient workflow management systemî. The guideline contains only the medical knowledge, but to achieve an efficient and effective care delivery both medical and organisational knowledge must be represented through computational formalisms. INCOME is the system used for the workflow modelling and simulation. To be imported into INCOME, the guideline is firstly translated into a high-level Petri net and then into the WPDL language.
Resources, both human and technological, necessary for performing the activities, are represented within an organisational model in INCOME. This allows running the Petri net for simulating the implementation of the guideline in the clinical setting. The purpose of the simulation is to validate the workflow model and to find the optimal resource allocation, before the workflow system is installed. The guideline for the stroke management will be used for illustrating the methodology.
Date: January 1999

HC 4017 P AT M AN FINAL P AGE 30
Authors: Ying-Lie O, Erik Buskens -University of Utrecht - Guus Schreiber,
University of Amsterdam - Stig Kjær Andersen, Aalborg University. - Silvana
Quaglini, University of Pavia
Abstract: In this document we describe part of the methodological approach of
PatMan. We describe the process from written guideline or protocol to a software realisation as a support tool for medical staff. This process consists of a number of analysis and design steps. Attention for this process is important, as it is the only way to ensure that the automised guideline actually adds value to the health care process.
We describe three examples and generalise from these to get a number of reference models that can be used as templates for future users.
Date: July 5, 1999
Authors: Silvana Quaglini, Carlo Mossa, Clara Fassino, Mario Stefanelli- University of Pavia
Abstract As described in the previous deliverable D2, the internal representation of the guideline is used to produce the WPDL (Workflow Process Definition
Language) code. This code is the input to the Income package, that builds the corresponding Petri Net for simulation purposes. Simulations aim at validating the workflow model and finding the optimal resource allocation, before the workflow system is really implemented. The same WPDL code is the input of the Oracle workflow package. Deliverable D2 described the methodology developed to perform these steps and Deliverable 3 described the necessary steps for the organisation modelling. In the present deliverable, the work done on the Petri net refinement and
Petri net outcome analysis is presented. As a matter of fact, once the process has been described through the net structure, the specific tasks and resources needed to perform each activity must be described in the system, as well as the temporal and organisational constraints. In addition an effort has been done for the optimal exploitation of the Income results: general queries have been designed, in order to present the results of any simulation in a user-tailored fashion. The guideline for the stroke management will be used for illustrating the methodology.
Date: January 7, 2000

HC 4017 P AT M AN FINAL P AGE 31
Authors: Silvana Quaglini, Carlo Mossa, Clara Fassino, Mario Stefanelli -
University of Pavia
Abstract: This deliverable describes common data models for knowledge and information exchange and re-use. In this project we basically rely on the relational model for data storage and retrieval. This means that relational tables are used for guideline representation, for health care organisation description, and for the patient record. Firstly, the deliverable illustrates the web server for browsing the guideline internal structure. Secondly, it focuses on the organisation representation explaining how this is used on the web for the workflow management system. Finally, the patient record is shown. Oracle WEB Server is the commercial product used for storing and accessing data. The training exemplar concerns the management of patients with acute ischemic stroke.
Date: June 30, 1999
Authors: Enrico Motta, John Domingue, Marek Hatala and Simon Buckingham
Shum, Knowledge Media Institute, The Open University, UK
Silvana Quaglini - Medical Informatics Laboratory, University of Pavia, Italy
Abstract. Two main goals were set for this workpackage: i) to produce a web-based set of tools supporting the acquisition, modelling and management of organizational knowledge, and ii) to construct specific models (i.e., ontologies) and tool configurations to support the management of medical guidelines. Both goals have been achieved and in this document we describe the various models and technologies which we have developed, including: i) a web-based browser/editor allowing collaborative model development over the web (WebOnto); ii) a user-friendly frontend interface, which allows users to query a knowledge base at a level which abstracts from the underlying modelling language (Lois); iii) a knowledge acquisition interface which takes advantage of an underlying ontology to support a model-based approach to knowledge acquisition (Knote); and iv) an ontology for characterizing medical guidelines. These tools and models have been tried out for a specific guideline,
“Pressure Ulcer in Adults: Prediction and Prevention”, and the resulting application model is also described in this report. Finally, we also comment on the results of this workpackage from a broader perspective, in particular emphasizing the extent to which our technology for knowledge management and modelling has been taken up by industry and the excellent results in terms of publications, spin-off projects and invited appearances directly related to the work carried out in the PatMan project.
Date: December 3, 1999

HC 4017 P AT M AN FINAL P AGE 32
Authors: Silvana Quaglini, Carlo Mossa, Clara Fassino, Mario Stefanelli- University of Pavia
Abstract: The clinical applications actually under implementation concern the stroke management and the pressure ulcer prevention. In this deliverable the preliminary results of these implementations will be described. The stroke study is at a more advanced status, and we are able to provide statistics about 386 patients managed with the guideline help, with data collected from the four Italian centres where the guideline has been implemented. The pressure ulcer prevention guideline is implemented in the clinical routine since one month, and data about 40 patients have been collected. In addition to statistics, we will discuss about the impact on the system users
Date: December 1, 1999
Authors : Silvana Quaglini, Clara Fassino - University of Pavia, Enrico Motta, John
Domingue - University of Milton Keynes, Ying-Lie O - University of Utrecht, Stig
Kjær Andersen - Aalborg University
Abstract
The tools developed within the PatMan project aim at facilitating developers and users during the entire “life-cycle” of a guideline for clinical practice. In particular, the KMI-Open University team focused on providing tools which allow users who are not experts in knowledge modelling to take part in the collaborative construction of a knowledge model and to interact with it at a level which abstracts from the underlying language. The University of Pavia team developed tools for the graphical representation of a guideline, for its integration into a careflow system, and for the analysis of the impact of the guideline on clinical and cost outcomes. The Promatis team provided expertise and tools for simulating the careflow process through a Petri net representation of the guideline and a formal representation of the health care environment organisation. The Aalborg University team and the Utrecht team provided support for detailing health care organisation resources and for project evaluation methodology. All the tools are available for demonstration..
Date: February 28, 2000

HC 4017 P AT M AN FINAL P AGE 33
M. Stefanelli. Artificial Intelligence for Building Learning Health Care Organizations, in Lecture Notes in Artificial Intelligence, W. Horn, Y. Shahar, G. Lindberg, S.
Andreassen, J. Wyatt (eds), Proceedings of the Joint European Conference on
Artificial Intelligence in Medicine and Medical Decision Making, AIMDM’99,
Aalborg, Denmark, 1999, Springer, pp 13-29
(Invited lecture at the Artificial Intelligence in Medicine European Society
Conference, Aalborg (DK), June 1999)
Introduction
Information and Communication Technologies (ICT) offer society the opportunity to reengineering Health Care Organizations (HCO) into more value-driven, knowledgebased, and cost-effective enterprises. We are benefiting from astounding advances in medical knowledge and technology. At the same time many problems have to be faced by HCO even in developed countries: quality of care is very uneven and costs are rising. Citizens are unhappy with their care; health professionals are unhappy with the HCO they are working in; payers are unhappy with costs; and governments react by enacting regulations that will fail to introduce any substantial change. Financial resources will never be enough to do all we would like to do. Biomedical research will produce better knowledge and treatments, but these results will be gradual and likely offset by increased demand by an aging society. ICT revolution and better management of HCO offer promise of dramatic help.
To attain this goal HCO must be able to exploit ICT to become learning organizations.
Current theories of learning reveal that the process of acquiring knowledge cannot be separated from the process of applying it. Integrating working and learning is a fundamental requirement for HCO to increase the efficiency and effectiveness of their activities.
The integration of working and learning requires the development of tools to support these socially-based, process-oriented views of representations of work. In medicine representation of work may take many forms: patient record, instrumental and material resources, medical knowledge and expertise available within the organization, as well as the distribution of tasks among people involved in the management of different types of patients. Thus, patient management protocols provide models of medical processes which can be enriched by models of the organization to build models of patient workflows. These can be used as the core of
Workflow Management Systems (WfMS) which provide a model-based communication infrastructure for health care professionals collaborating in the patient management.
Medical work activities coordination involves a multidisciplinary research and goes beyond the current thinking in workflow management and business process

HC 4017 P AT M AN FINAL P AGE 34 engineering. In engineering, work processes can be pre-specified, organizational structure are stable, and exceptions are all modeled as adding work volume to activities. In HCO models, however, work processes are unpredictable and can change during execution.
Multidisciplinary research projects, based on mutual respect and willingness to learn from another discipline (such as organizational science, sociology, and computer science), can help to create a thriving research community that builds upon the strengths of different disciplines.
The great challenge for researchers in Artificial Intelligence in Medicine (AIM) is to exploit the astonishing capabilities of ICT to disseminate their tools to benefit HCO by assuring the conditions of organizational learning at the fullest extent possible. To do that the AIM community should activate new multidisciplinary research projects, based on mutual respect and willingness to integrate into its culture other disciplines, such as organizational science, sociology, and epistemology in order to design and develop decision support system smoothly integrated into a computer-supported collaborative work framework.
This paper seeks to stir debate, discussion, and action towards new research issues within AIM research community.

HC 4017 P AT M AN FINAL P AGE 35
S. Quaglini, C. Mossa, C. Fassino, M. Stefanelli, A. Cavallini, G. Micieli. Guideline-
Based Workflow Systems, Lecture Notes in Artificial Intelligence, W. Horn, Y.
Shahar, G. Lindberg, S. Andreassen, J. Wyatt (eds.), Proc. of the Joint European
Conference on Artificial Intelligence in Medicine and Medical Decision Making,
AIMDM 99, Aalborg, Denmark, June 1999, Springer, pp 65-75
S. Quaglini, C. Mossa*, C. Fassino, M. Stefanelli, A. Cavallini**, G. Micieli**
Dipartimento di Informatica e Sistemistica, Universit di Pavia
Via Ferrata 1, I-27100 Pavia, Italy sil@ipvaimed2.unipv.it
*Consorzio di Bioingegneria e Informatica Medica, Pavia, Italy
Via Ferrata 1, I-27100 Pavia, Italy mossa@aim.unipv.it
**Stroke Unit, IRCCS Istituto Neurologico "C. Mondino" Pavia, Italy
Via Palestro, 11 I-27100 Pavia, Italy micieli@cpbim1.unipv.it
Abstract.
This paper describes a methodology for achieving an efficient allocation of resources while using clinical practice guidelines. The resulting system can be classified as a guideline-based patient workflow management system . Both medical and organisational knowledge are represented through computational formalisms, from relational tables to
Petri net. Human and technological resources, necessary to guidelinebased activities, are represented within an organisational model. This allows running the Petri net for simulating the implementation of the guideline in the clinical setting, in such a way to validate the model and to suggest an optimal resource allocation, before the workflow system is installed. Finally, we are experimenting the real setting implementation.
For illustrating the methodology, an application concerning the management of acute ischemic stroke is presented.

HC 4017 P AT M AN FINAL P AGE 36
S. Quaglini, M. Grandi, P. Baiardi, M.C. Mazzoleni, C. Fassino, G. Franchi, S.
Melino. A Computerised Guideline for Pressure Ulcer Prevention, Proc. Medical
Informatics Europe 99 (MIE 99), P. Kokol, B. Zupan, J. Stare, M. Premik, R.
Engelbrecht Eds, IOS Press, 1999.
Silvana QUAGLINI*, Manuela GRANDI**, Paola BAIARDI**, M.Cristina
MAZZOLENI**, Clara FASSINO*, Giorgio FRANCHI**, Stefania MELINO**
*Dipartimento di Informatica e Sistemistica, Universit di Pavia
**IRCCS Fondazione S. Maugeri , Clin. del Lavoro e della Riabilitazione, Pavia,
Italy computerised guideline for the pressure ulcers prevention. In particular, we describe the site-specification of a guideline delivered by the Agency for Health Care Policy Research, its integration with the electronic patient record, and its introduction within the clinical routine. The system facilitates trained nurses in the patient management by producing daily work-plans, and novice nurses by running as an educational tool.

HC 4017 P AT M AN FINAL P AGE 37
Quaglini S, Grandi M, Fassino C, Stefanelli M.
Different interaction modalities with computerised guidelines, Proceed. Hospital without walls conference, London, UK,
March 1999.
There is no short abstract, because the work had to be presented as an extended abstract
Clinical practice guidelines (GLs) are more and more common in health care organisations.
They provide suggestions that are evidence-based and/or have received the consensus of leading medical experts. GLs have been introduced to decrease differences among institutions in treating similar patients and, at the same time, to improve the quality of the delivered care.
However, several problems are associated with GL diffusion: 1) the original GL is often modified to fit the organisational constraints of the clinical setting; in such a way, some GL portions are fully implemented, while other ones are implemented in less detail, or not implemented at all; 2) apart from organisational constraints, health care operators often do not fully comply with the GL, particularly when the GL introduction implies important behavioural changes; 3) outcomes to be used for evaluating the GL benefit are difficult to be defined, measured and compared; to this aim, the intention of the GL should be clearly defined, and statistical comparisons with control cases should be designed.
Information technology attempts to overcome these difficulties by offering
• formalisms for GL representation, allowing to avoid ambiguities, to verify the GL logical correctness, to see the GL at different levels of details, to evaluate non-compliances;
• the connection with the electronic patient record, that has been shown to enhance compliance, and that is mandatory for performing subsequent sound statistics;
• several possible modalities, for the user, of interacting with the same GL.
These facilities should help the correct diffusion of a GL, because they allow to minimize GL changes by adapting the GL to the actual clinical setting or, at least, to document the GL version implemented. The focus of this work is on the interaction modalities. First of all, we developed a graphical editor for the GL formal representation. The editor should be used by medical experts and/or knowledge engineers and the formalized GLs, graphically shown as an interactive flow-chart, should be stored into a GL server. In such a way, GLs are maintained and updated in centres of excellence, while different health care organisations can access them through the internet. Then, we propose different inference engines for running GLs: as a matter of fact, according to the application domain and to the organisation constraints, the same GL may assist operators differently. In other words, using this architecture, different organisations may use GLs in different ways. We distinguish three points of view for classifying the GL/user interaction.
From the advice production point of view:
1.
as soon as the GL detects conditions that can trigger suggestions an advice is produced.
We call this the explicit modality , and it can be suitable for novices or, in general, for non-skilled users;
2.
the user is advised only when a non-compliance is detected by the GL. We call this the silent modality and it can be worth for expert users.
From the non-compliance management point of view:
1.
the GL proceeds to the next task(s) if and only if the user fully complied with the GL for what concerns the previous tasks. This could be the case for mandatory procedures;
2 .
the GL proceeds also in case of non-compliance, but only if the user provides a justification for that;
3.
the GL proceeds in any case, storing the non-compliance in a silent mode.
From the time of intervention point of view:

HC 4017 P AT M AN FINAL P AGE 38
1.
the GL reacts in real time to the users actions (a user action is assumed to update the patient record), and it suggests, in real time as well, the next action(s), or highlight noncompliances. This modality could be suitable when GLs are adopted for ensuring a correct data collection, for example during history taking. For more complex procedures, i.e. tasks distributed in time and in different sites, this modality implies for the user the possibility of interacting very frequently with the computer, thus assuming the existence of a very efficient and distributed information system;
2.
the user accesses the GL only in specific, possibly pre-defined, time slices. For example, at the end of their shift, operators write down the data about tasks performed during the shift, and the GL either approves or gives some advises. Of course, in this case, the control is not in real time on every single task.
In addition to assist operators during their routine, the GL may be used for educational purposes. In this case, the user simulates patients, by filling the patient record with clinical findings, diagnostic tests, therapeutic interventions, etc., and he/she will obtain advises according to the chosen modality for each point of view.
For experimenting these ideas we implemented two GLs, the first one for the management of the stroke in an intensive care unit, and the second one for the prevention of bedsores in a general medicine ward.
(b)
(a)
Fig. 1 – The guideline for the stroke management. (a) The main page is shown in the central window. Shadowed blocks are expandable. The expansion of “acute phase” is shown in the right window.
(b) the advice produced by the guideline when the physician attempts to store data about a task that is not scheduled at that time
Figure 1 shows the main page of the GL for the stroke management, derived from the
American Heart Association. This GL is implemented with the silent modality; noncompliances are allowed, provided that the user justifies them, as shown in Fig. 1b; the GL reacts in real-time, as long as the patient record is updated. When interacting with the patient record, the user may ask for looking at the status of the tasks for the actual patient, i.e. which tasks have been performed and which are the possible next ones. Table I refers to the second
GL that we implemented. It is a slightly modified version of a GL published by the AHCPR
(Agengy for the Health Care Policy Research, USA). In this case, the GL is implemented with the explicit modality; non-compliances are allowed, with motivations (see the last column in
Tab. I); concerning time, the GL prints out the workplan (shown in Tab. I) at the beginning of a nurse shift, according to the data stored in the patient record. For practical reasons, nurses cannot interact in real rime with the computer, thus they write down on the paper workplan if suggested tasks have been performed or not (bold items). At the end of their shift,

HC 4017 P AT M AN FINAL P AGE 39 they fill the electronic patient record, in such a way that the GL is able to elaborate the workplan for the next shift.
Tab. I - NURSE WORKPLAN -wednesday, december 16 th
, 1998
Room Patient
21
22
24
Verdino Rosa
Nerastro Vito
Celestino Alba
Tasks to be performed
1) Ulcer evaluation
2) Anti-decubitus matress
3) Duoderm cream
1) Incontinence pad
1) Skin examination
Ulcer site
Elbow
Shoulder
Task done Notes/Motivation
YES
NO
YES
No more matress available
YES
NO Patient transferred for examinations
The same GL is used for educational purposes. In this modality, the user interacts with the patient record by entering simulated data and the GL engine immediately highlights, through the graphical editor, the tasks to be performed.

HC 4017 P AT M AN FINAL P AGE 40
Domingue, J. and Motta, E. (in-press). Planet-Onto: From News Publishing to
Integrated Knowledge Management Support . IEEE Intelligent Systems Special Issue on "Knowledge Management and Knowledge Distribution over the Internet" .
John Domingue, Enrico Motta
Knowledge Media Institute
The Open University
Walton Hall, MK7 6AA
Milton Keynes, UK
[ e.motta, j.b.domingue
]
@open.ac.uk
Abstract: Organizations need to keep members informed. Most organizations of a significant size do this via an internal publication. KMi Planet is a web-based news server which facilitates communication within the Knowledge Media Institute and allows the wider community to access lab-related items of interest. As the readership and the size of the news archive have grown, a number of knowledge management challenges have arisen: how to organize a speedy, low overhead publication process which can nevertheless yield high quality results; how to provide semantic search and knowledge retrieval facilities in an effective and sustainable way; how best to provide individualized presentations and news alerts; how best to emulate the behaviour of a newsroom team.
To address these questions we have developed an integrated suite of tools, which is called PlanetOnto. These tools support a speedy, but high quality publishing process, allow ontology-driven document formalization, and augment standard browsing and search facilities with deductive knowledge retrieval. In addition, the PlanetOnto architecture includes specialized agents, which provide personalized news feeds and alerts and can proactively identify potentially interesting news items.

HC 4017 P AT M AN FINAL P AGE 41
Motta, E., Buckingham Shum, S., and Domingue, J. (1999). Ontology-Driven
Document Enrichment: Principles and Case Studies. Proceedings of the 12th Banff
Knowledge Acquisition Workshop, Banff, Alberta, Canada, October 16-22, 1999 .
Available at: http://kmi.open.ac.uk/people/motta/papers/mottakaw99.pdf
:
Enrico Motta, Simon Buckingham Shum and John Domingue
Knowledge Media Institute
The Open University
Walton Hall, MK7 6AA
Milton Keynes, UK
[ e.motta, s.buckingham.shum, j.b.domingue
]
@open.ac.uk
Abstract: In this paper we present an approach to document enrichment, which consists of associating formal knowledge models to archives of documents, to provide intelligent knowledge retrieval and (possibly) additional knowledge services, beyond what is available using standard information retrieval and search facilities. The approach is ontology-driven, in the sense that the construction of the knowledge model is carried out in a top-down fashion, by populating a given ontology, rather than in a bottom-up fashion, by annotating a particular document. In the paper we give an overview of the approach and discuss its application to the domains of electronic news publishing, scholarly discourse and medical guidelines.

HC 4017 P AT M AN FINAL P AGE 42
Domingue J.. Tadzebao and WebOnto: Discussing, Browsing, and Editing Ontologies on the Web. Proceedings of the 11 th
Banff Knowledge Acquisition Workshop, Banff,
Alberta, Canada, April 18-23, 1998 .
John Domingue
Knowledge Media Institute
The Open University
Walton Hall, MK7 6AA
Milton Keynes, UK j.b.domingue@open.ac.uk
Abstract: In this paper I describe two systems which, in different ways, address the shortcomings of current approaches to enabling ontology construction and use via the
World-Wide Web. The first system Tadzebao, enables knowledge engineers to hold synchronous and asynchronous discussions about ontologies. Tadzebao addresses the fact that an integral part of communal design, dialogue, has largely been ignored by the community. The second system WebOnto uses a Java based client to alleviate the acknowledged problems of creating interfaces in ’vanilla HTML’.

HC 4017 P AT M AN FINAL P AGE 43
Domingue, J. B. and Motta, E. (1999) A Knowledge-Based News Server Supporting
Ontology-Driven Story Enrichment and Knowledge., Fensel, D. and Studer, R. (eds.),
Proceedings of the 11th European Workshop on Knowledge Acquistion, Modelling, and Management.. Dagstuhl Castle, Germany May 26 - 29, 1999 . Springer-Verlag, pp
103-120.
John Domingue, Enrico Motta
Knowledge Media Institute
The Open University
Walton Hall, MK7 6AA
Milton Keynes, UK
[ j.b.domingue, e.motta
]
@open.ac.uk
Abstract: We consider a knowledge management scenario in which members of an academic community collaboratively construct and share a common archive of news items. Given this scenario, a number of knowledge management challenges arise: how to organize a speedy, low overhead publication process which can nevertheless yield high quality results; how to provide semantic search and knowledge retrieval facilities in an effective and sustainable way; how best to provide individualized presentations and news alerts. To address these questions we have drawn on a number of technologies, including knowledge modelling, autonomous agents, software visualization, knowledge acquisition and distributed computing. In the paper we describe the resulting Planet-Onto architecture, which provides an integrated set of tools to support news publishing, ontology-driven document formalization, story identification and personalized news feeds and alerts.

HC 4017 P AT M AN FINAL P AGE 44
Buckingham Shum, S., Motta, E., & Domingue, J. (1999). Representing Scholarly
Claims in Internet Digital Libraries: A Knowledge Modelling Approach. In S.
Abiteboul and A.-M. Vercoustre (eds.), Proceedings of ECDL’99: Third European
Conference on Research and Advanced Technology for Digital Libraries. Paris,
France, September 22-24, 1999.
Springer-Verlag (Lecture Notes in Computer
Science). Available at: http://kmi.open.ac.uk/techreports/papers/kmi-tr-80.pdf
Simon Buckingham Shum, Enrico Motta and John Domingue
Knowledge Media Institute
The Open University
Walton Hall, MK7 6AA
Milton Keynes, UK
[ e.motta, s.buckingham.shum, j.b.domingue
]
@open.ac.uk
Abstract: This paper is concerned with tracking and interpreting scholarly documents in distributed research communities. We argue that current approaches to document description, and current technological infrastructures particularly over the World
Wide Web, provide poor support for these tasks. We describe the design of a digital library server which will enable authors to submit a summary of the contributions they claim their documents makes, and its relations to the literature. We describe a knowledge-based Web environment to support the emergence of such a communityconstructed semantic hypertext, and the services it could provide to assist the interpretation of an idea or document in the context of its literature. The discussion considers in detail how the approach addresses usability issues associated with knowledge structuring environments.

HC 4017 P AT M AN FINAL P AGE 45
S. Quaglini, M. Grandi, P. Baiardi, M.C. Mazzoleni, C. Fassino, G. Franchi, S.
Melino.
Acting an Ulcer Prevention Plan through Guideline-based Tools, Proceedings
AMIA 99 Annual Symposium, Washington, DC, November 6-10,1999 (Poster)
S. Quaglini*, M. Grandi o
, P. Baiardi o
, M.C. Mazzoleni o
,
C. Fassino*, G. Franchi o
, S. Melino o
*Dipartimento di Informatica e Sistemistica, University of Pavia, sil@ipvaimed2.unipv.it
o
IRCCS Fondazione S. Maugeri , Clinica del Lavoro e della Riabilitazione,
Pavia, Italy
Abstract This work describes how some problems related to a guideline diffusion and implementation have been tackled and solved for an application concerning the pressure ulcer prevention (the original guideline was delivered by the Agency for
Health Care Policy and Research). The developed system facilitates both trained nurses by producing daily work-plans and novices through an educational tool.

HC 4017 P AT M AN FINAL P AGE 46
Project Consortium: HC 4017 PatMan
Contact address for the Project:
Name: Virtual Centre for Health Informatics
Street: Fredrik Bajers Vej 7D2
City: 9220 Aalborg East
Country: Denmark
Telephone: +45 9635 8809
Fax: +45 9815 4008
E-mail: info@v-chi.dk
WWW: www.v-chi.dk
The Project PatMan Patient Workflow Management System has been supported by the
European Commission under the auspices of the TELEMATICS APPLICATIONS
Programme.
Further information on the TELEMATICS APPLICATIONS Programme:
You can obtain more information on the projects of the TELEMATICS
APPLICATIONS Programme from:
European Commission, DG XIII, C/1
TELEMATIC APPLICATIONS Programme Information desk
Rue de la Loi 200 - BU 29 - 04/05
B - 1049 Brussels
Fax: +32 2 295 23 54
E-mail: telematics@dg13.cec.be
Or on the TELEMATICS APPLICATIONS Programmes’ homepage: http://www.echo.lu/telematics