AN AGENT BASED INFRASTRUCURE FOR FACILITATING

EVIDENCE BASED MEDICINE

Jennifer Sampson

Department of Computer and Information Science, Norwegian University of Science and Technology,Trondheim, Norway.

ywords: Multi-agent system, Evidence-based medicine

Abstract: Evidence-based medicine relies heavily on the timely dissemination of ‘best evidence’ to a wide audience of

health practitioners (Atkins and Louw, 2000). However, finding, assimilating and using this information

resource effectively can be difficult. In this paper we describe an infrastructure for facilitating evidence-

based health care using Agora - a multi-agent system. This paper discusses a new application for AGORA,

and also describes issues for disseminating such medical knowledge via an adaptive, intelligent, distributed,

mobile information service. We describe how an agent based approach can deliver clinical cases and

diagnosis information to clinicians at point of care tailored to her/his needs. This research in progress is

particularly important for the facilitating flow of information in health care.

1 INTRODUCTION

We are developing a multi-agent system to assist

with diagnosis decision making in health care using

mobile devices. This paper discusses an application

to AGORA, a multi-agent system, for disseminating

such medical knowledge via an adaptive, intelligent,

distributed, mobile information service. One of the

aims of the project is to provide a novel solution for

information service provision and selection for users

of mobile devices. The application of mobile

technology in this domain will involve the use of

software assistant agents to proactively

communicate availability of relevant clinical cases

to clinicians. The software agent will also

communicate with other software agents to make

informed decisions based on specific patient medical

results and documented medical research. This

research in progress is particularly important for the

facilitating flow of information in health care.

The overall research goal is to provide an

‘intelligent’ solution to medical services provision

through the use of agent assistants. The project is to

use agent technology to proactively communicate

availability of relevant clinical cases to clinicians

and provide personalised information tailored to

meet the clinician’s needs. We are particularly

interested in promoting a way to disseminate clinical

case reporting for diagnosis.

2 MOTIVATION

Previous research in the domain of evidence based

medicine and multi-agent systems (MAS) have

focused on information agents and appropriate

information retrieval methods (Abasola and Gómez

2000). We propose extending this work by using

agent technology to proactively communicate

availability of relevant clinical cases to clinicians

and provide personalised information tailored to

meet the clinician’s needs.

We are using an intelligent agents approach for

disseminating medical research knowledge via an

adaptive, intelligent, distributed, mobile information

service. By adaptive we mean the ability of the

service to adapt to the health professionals, service

providers and communication context. By intelligent

information services we understand ability of service

to employ planning, reasoning and knowledge

processing in order to satisfy the health professionals

requirements. An important intelligent feature of the

service is its pro-activity - ability to predict users’

needs and to take initiative. The pro-activity of the

service should be based on exploiting and (when it is

not available) building the health professionals

and/or service provider models by using machine

learning and inductive inference techniques. By

distributed information service we mean that

components of the service system can be distributed

493

Sampson J. (2004).

AN AGENT BASED INFRASTRUCURE FOR FACILITATING EVIDENCE BASED MEDICINE.

In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 493-496

DOI: 10.5220/0002657804930496

 SciTePress

across the network and that service can be composed

from these components. By mobile information

service we assume that both heath professionals and

service providers can employ mobile devices such as

mobile phones and PDAs for requesting/provision of

an information service. Currently we are developing

a prototype solution using Agora a multi-agent

system infrastructure. Future research will be

focused on the use of mobile devices to support the

dissemination of the service. A multi-agent system

infrastructure may include different levels and

components, however currently we are focused on

the cooperative work component in the

infrastructure which we consider to be at MAS

middleware level. An additional complicating factor

is the use of mobile devices for communicating

medical information to clinicians.

Due to the increasing number of mobile portable

devices in use opportunities for the development of a

wide range of mobile information services exists.

Matskin and Tveit (2003) comment that the

development of such technology as WAP (and other

mobile technology) makes it possible to provide

users of mobile phones with access to the Internet

and services which earlier were available only via

PCs connected to the Internet. The application of

mobile technology in the healthcare domain will

involve the use of software assistant agents to

proactively communicate availability of relevant

clinical cases, evidence-based research and patient

data to clinicians. The software agent will

communicate with other software agents to make

informed decisions based on specific patient medical

results and documented medical research. However

Matskin and Tveit (2003) note that mobile devices

have severe restrictions that may complicate

practical use of information services. These

restrictions are largely due to the limitations of

wireless data networks compare to wired networks

(less bandwidth, more latency, less connection

stability, less predictability and less standardized

protocols) and to the limitations of mobile handsets

compared to personal computers (small screen size,

complicated text input, limited memory, slow CPU

and more constrained energy supply).

We envisage necessary requirements for the

services are simplicity and expressiveness of the

services. It is possible that some of the limitations

will be relaxed in the future by development of

hardware or telecommunication networks

technology but at the moment we shall consider

these impediments when implementing mobile

services. Services should be personalized - they

should take into account changing healthcare

professional’s preferences. Thirdly, services should

be able to adopt both to changes in the healthcare

professionals preferences and in the context of

communication (Matskin and Tveit, 2003). In

addition we will assume that as much as possible

work for service customization and provision should

be done off-line without direct participation of the

user of the mobile device.

Each clinician should be able to present his or her

case profile by telling the EBM assistant specific

patient case information and define specific areas of

expertise. The confidentiality of the patient will be

adhered to through the means of agents who own the

sensitive patient data, this information will not be

released to other agents and is what we describe as

secret knowledge. The application is to facilitate

health care practice by providing relevant feedback

information regarding diagnosis and treatment

information based on similar cases via a mobile

service. This technology may be useful for doctors

treating patients where it is not suitable for patients

to attend a clinic, in this situation a remote check,

communication or diagnosis is essential. We foresee

that it should be possible to create certain default

profiles for specific types of health care professional

and case types.

Figure 1 depicts our high level

concept from an individual perspective.

3 THE AGORA CONCEPT

We are using AGORA as a prototype environment

for the development of an agent-based system for

facilitating evidence-based health care. Matskin et

al. (2001) have developed and successfully tested

the AGORA system in other domains such as e-

commerce. Our research is to modify the

architecture for use in the health care domain, in

particular for facilitating the dissemination of

clinical cases to clinicians. While AGORA has been

successfully used for virtual enterprises (Rao and

Petersen 2003) we recognise that it is necessary to

EBM internet

resources

EBM agent

assistant

Clinician

with WAP

phone/PDA

Clinician

with WAP

phone/PDA

Electronic

health

record

Medline

database

Cochrane

database

EBM

agent

assistant

Figure 1: Agent-based solution for evidence-based

healthcare

ICEIS 2004 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS

494

modify the architecture and underlying knowledge

base design to be appropriate for use in the

healthcare domain.

The term ‘Agora’ originating in Greece, is the

public forum and market place, where people

gathered to discuss politics, news, exchange views

etc. Similarly, agent Agoras are places where the

agents arrive to communicate, negotiate, coordinate,

collaborate and trade information and services

(Matskin et. al. 2001). We have used the Agora

metaphor in the development of a multi-agent

system which consists of a set of interconnected

Agoras and agents. Default agents are attached to an

Agora while the other agents are registered at an

Agora (Matskin et. al. 2001). The Agora system

allows where possible agents to register and present

activities. A matchmaker inside the agora compares

the activities, and if two or more activities match, a

negotiation agent will be started. This negotiation

agent will manage the negotiation between these

agents. The idea is that agents interested in some

topic can find an Agora that operates in the specific

domain, and then receive services and information

through the Agora. The registered agents suggest

that agents that want to "participate" in an Agora

have to be registered first. The agent will then

advertise themselves, and the Agora will help this

agent to find other agents and Agoras that may help

the agent to carry out tasks. Default agents are

agents that are bound to the Agora to perform some

services for the Agora. These agents may be

performing services like negotiation, coordination,

registration, or for example offering access to a

database. It is possible for agents to be registered at

several Agoras and that there could be interrelations

between Agoras (Matskin et. al. 2001).

Agora has been developed using Java, and

JATLite is used as a tool for communication

(Matskin et. al. 2001). Prior research has found that

it possible to run more than 70 agents concurrently

on the JATLite router. KQML is used as the basic

external communication language and protocol for

exchanging information and knowledge. A separate

communication format Message Wrapper has been

defined inside the agents and Agoras. Between the

internal format and the external KQML language

there is a translator. We are now using this translator

to use FIPA instead of KQML, by building a new

FIPA translator.

3.1 Using AGORA for facilitating

evidence-based heath care

The main advantage with the Agora approach is that

the agents have rather simple implementations

because of the gathering of specially designed

services and functionality inside the Agoras. In such

an application agents will represent the clinicians,

the case events and the evidence based medicine

service providers. In order to implement a virtual

health care Agora we first identify participants of the

cooperative work. We are interested in the roles

specific agents will undertake.

• Clinician agent - represents the medical

professional, at this point of time we only

consider practitioners of medicine who undertake

clinical work.

• Active case agent – represents a clinical case

currently under care

• Completed case agent – represents published

clinical case reports

• Case manager agent – a coordinator of cases and

clinicians in a health care environment.

• Diagnosis negotiation assistant agent –

matchmaking active case with other cases

• Cochrane library collaborator agent – manager of

the agents representing specific health domains.

Cooperative points for agents work in the health care

domain can be identified, as follows:

• Coordination of the health care activity

• Coordination of different clinician agents

• Coordination (information gathering) and

diagnosis support between a case manager agent

and the Cochrane library agents.

These cooperative points can be mapped into the

following set of agoras: a clinician agora, a case

agora, a diagnosis agora and a cochrane

collaboration agora.

Agora represents both registered and default

agents (refer Figure 2). Registered agents are either:

coordination agents, negotiation agents or

participant agents. The first two types of agents

manage coordination and negotiation activities to be

performed at the corresponding Agora. The

participant agents represent participants of the

cooperative activity. The default agents in Agora are

the Agora manager, service and matchmaking

agents. Each participant agent (for example case

agent) presents offers and requests for information.

Offered activity (OA) may be any activity the agent

can perform by itself, for example in the heath care

domain an offered activity could diagnosis

information offered by the diagnosis assistant agent.

Interrelations and connections among agents and

Agoras are depicted at Figure 2. Lines between

Agoras mean that agents from connected Agoras can

be registered on both of them. Each Agora is

shadowed as there is more than one Agora of each

type. The coordination agents manage corresponding

protocols.

AN AGENT-BASED INFRASTRUCTURE FOR FACILITATING EVIDENCE BASED MEDICINE

495

Wanted activity (WA) could be any requested

information from clinician agents regarding clinical

cases. Agora allows direct matching of OA and WA;

however we are currently overriding this default

matchmaking using semantic matchmaking and

ontological analysis. This is an important issue in

clinical case matching, we are employing fuzzy

matching techniques, where the matching procedure

is not so precise or complete (80%/50%). Another

important consideration is the composition of cases,

we are determining the efficacy of service

composition in this domain. Each of the agents in

the Agora platform has the following modules: goal

analyser, knowledge base, planner (in XML),

scheduler, action creator and communication

threads. We are currently working on defining the

ontology for the knowledge base module of the case

and diagnosis agents. The knowledge base module

maintains storage, retrieval and querying knowledge.

CONCLUSION

The platform described in this paper was previously

developed using the Agora concept for facilitating

cooperative work. Our research is towards using

Agora for facilitating the dissemination of medical

knowledge and for assisting diagnosis in a clinical

setting.

REFERENCES

Abasola, J. and Gómez, M. (2000). “MELISA. An

ontology-based agent for information retrieval in

medicine”, Proceedings of the First International

Workshop on the Semantic Web (SemWeb2000).

Atkins, C. and Louw, G., 2000. Reclaiming Knowledge: A

case for evidence-based Information Systems. In Proc.

8th European Conference on Information Systems.

Matskin, M., Divitini, M., Petersen, S.A., 1998. AGORA:

a Multi-Agent Support for Distributed Information

Technology Applications', In, NIK'98.

Matskin, M. Kirkeluten, O. J. Krossnes, S.B. Sæle, Ø.,

2001. Agora: An Infrastructure for Cooperative Work

Support in Multi-Agent Systems. Wagner, T. and Rana

O.F.(Eds.): Infrastructure for Agents, Springer-Verlag

Matskin, M. and Tveit, A., 2003. Software Agents for

Mobile Commerce Services Support. K. Siau (ed.).

Readings in Database Management. Idea Group.

Rao, J. and Petersen, S., 2003. Implementing Virtual

Enterprises Using AGORA Multi-agent System. In

Proceedings of CAiSE 2003 Forum.

Diagnosis

Agent N

Clinician

Agora

Coordinati

on Agent

Figure 2: Virtual Health Care AGORA

Case

Agora

Case Agent

Clinician

Agent N

Clinician

Agent 1

Coordination Agent

Case Agent 1

Cochrane

Collaboration

Agora

Cochrane

Agent 1

Cochrane

Agent N

Coordination Agent

Case

Manager

Agent

Diagnosis

Agent 1

Diagnosis

Agora

ICEIS 2004 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS

496