ARCHETYPE-BASED SEMANTIC INTEROPERABILITY IN

HEALTHCARE

Alberto Marques, Ant

onio Correia, L

ucia Cerqueira

Centro Hospitar do T

amega e Sousa, Penaﬁel, Portugal

Jos

e Machado, Jos

e Neves

Universidade do Minho, Departamento de Inform

atica, Campus de Gualtar, Braga, Portugal

Keywords:

Electronic health record, Semantic interoperability, Quality-of-Information.

Abstract:

Advances in new Methodologies for Problem Solving and Information Technology enable a fundamental

redesign of health care processes based on the use and integration of data and/or knowledge at all levels, in a

healthcare environment. Indeed, new communication technologies may support a transition from institution

centric to patient-centric based applications, i.e., the health care system is faced with a series of challenges,

namely those concerning quality-of-information and the cost-effectiveness of such processes. The distribution

of cost-effective health care allowing the patient to take active part in the caring process, provision of evidence-

based care on all levels in the system and effective use and reuse of information are key issues for the health

care organization. The information and communication technology infrastructure should therefore reﬂect the

view of the health care system as a seamless system where information can ﬂow across organizational and

professional borders. Therefore, in this work it will be address key principles that must be at the center of

patient-centered technologies for disease management and prevention, namely those referred to above.

1 INTRODUCTION

The areas referred to above share the basic problem

of semantic interoperability, which simply means that

semantics is preserved in communication between

health care providers using different information sys-

tems, a circumstance which should be natural but has

proven to be very hard to achieve. Consequently,

demands of information handling within the health

care sector range from clinically valuable, patient-

speciﬁc information to a variety of aggregation lev-

els for follow-up and statistical and/or quantiﬁable

reporting. A number of protocols are for this pur-

pose put into use in domains such as diagnosis, health

dilemmas, interventions, and modus operandi. The

context is to assure quality of service at different lev-

els of abstraction, and to agree to process aggregation

according to different appearances or perspectives of

utilization.

However, almost all existing Electronic Health

Record (EHR) systems are built with an explicit do-

main model, a common approach in current EHR soft-

ware development practice. This means that the med-

ical domain knowledge present in these system results

in higher cost when new requirements in clinical prac-

tices occur. If a research program requires a large

amount of data to be collected from different clini-

cal sites using dissimilar EHR systems, this make it

much more difﬁcult to add a speciﬁc feature so the

collected data will be comparable. Indeed, the lack of

integration between the different EHR systems is not

only an obstacle for a more effective clinical practice

and research, but it is also a fact that may lead to a

suboptimal care for the individual patient, including

potential safety problems and an unnecessary waste

of resources. Therefore, the new systems should be

able to record data both from clinical routine docu-

mentation and experimental studies. The users should

not be forced to duplicate the data that already might

exist in others different information systems in use at

the organization.

On the other hand, the users should be able to

link the deﬁned data item to an external terminology,

which later can be utilized for possible data aggre-

gation and classiﬁcation purposes. More importantly,

such deﬁnitions should be shared and reused by other

305

Marques A., Correia A., Cerqueira L., Machado J. and Neves J. (2010).

ARCHETYPE-BASED SEMANTIC INTEROPERABILITY IN HEALTHCARE.

In Proceedings of the 2nd International Conference on Agents and Artiﬁcial Intelligence - Agents, pages 305-308

DOI: 10.5220/0002722303050308

 SciTePress

users than the original author for data recording, so

that the semantics and the quality of service of the

data collected from different information systems can

be maintained.

2 AGENT ORIENTED

PROGRAMMING

Although there is no universally accepted deﬁnition

of agent, in this work such an entity is to be under-

stood as a computing artefact, being used in hard-

ware or software devices, that exhibit the properties

of autonomy, reactivity, pro-activity and social be-

haviour. To develop such systems, a standard speciﬁ-

cation method is required, and it is believed that one

of the keyfactors for its wide acceptance is simplicity.

Indeed, the use of intelligent agents to simulate hu-

man decision making in the medical arena offers the

potential to set an appropriate software development

and analysis practice and design methodology that do

not distinguish between agent and human, until im-

plementation.

Agents in a health care facility conﬁgure applica-

tions or utilities that collect information about the as-

sets in the organization (Alves et al., 2005). Once that

information has been collected it can be posted di-

rectly to other entities (e.g. a physician), or a server,

saved in a ﬁle or emailed to someone to be handled

at a later date, or sent using HL7 (Health Level 7)

or web services in a Service Oriented Architecture

(SOA). Indeed, HL7 plays an essential role in extend-

ing the interoperability for the development of health

information exchange, in the standardization of XML

medical document structures and in the speciﬁcation

of robust vocabulary deﬁnitions for use in clinical

messages and documents (e.g., SNOMED CT),

en-

abling functional speciﬁcations for the EHR

3 AIDA

In order to fulﬁl this goal it was designed and de-

veloped an Agency for the Integration, Difusion, and

Arquive (AIDA), which allows to interoperate with

different information systems at the organization or

area levels. The agency conceptually consists of 9

http://www.hl7.org

http://www.ihtsdo.org/

AIDA-EHR is a portuguese EHR developed at Centro

Hospitalar do Porto and based in Problem Oriented Medical

Record methodology (Weed, 1969).

Figure 1: The AIDA Architecture.

(nine) agent based subsystems: AIDA-RIS - Radio-

logical Information System; AIDA-MEIS - Medical

Exams Information System; AIDA-LIS - Laborato-

ries Information System; AIDA-ISM - Information

System for Monitoring (e.g., vital signals monitor-

ing); AIDA-PRM - Patient Relationship Management

(including communication using SMS); AIDA-OWM

- Organizartion and Work Management (Including

agenda, scheduling, planning and resource man-

agement; AIDA-EHR- Electronic Health Records;

AIDA-HL7 - Interoperation of systems; and AIDA-

SOA - Service Oriented Architecture; where the or-

ganization data (e.g. patient data) is stored and man-

aged (Figure 1). These nine subsystems are commu-

nicating using web services, in terms of HL7 or XML

based messages. The deﬁnition of attributes and tem-

plates can be shared among different healthcare units

having separate information systems. AIDA supports

Web based services (AIDA-SOA) to facilitate the di-

rect access to the information and communication fa-

cilities set by the humans, i.e. AIDA construction fol-

lows the acceptance of simplicity, the conference of

the achievement of common goals and the address-

ing of responsibilities. Indeed, the main goal is to

integrate, diffuse and archive large sets of informa-

tion from heterogeneous sources (departments, ser-

vices, units, computers, medical equipments). Un-

der these pressuppositions, a Healthcare Information

System (HIS) will be addressed in terms of (Figure

1). This system is also now a reality in some major

portuguese hospitals, being developed and conﬁgured

from generated forms of the EHR and sharing infor-

mation through AIDA.

This system is also now a reality in some major

ICAART 2010 - 2nd International Conference on Agents and Artificial Intelligence

306

portuguese hospitals, being developed and conﬁgured

from generated forms of the EHR and sharing infor-

mation through AIDA ((Rigor et al., 2008)). Another

interesting application of ubiquitous intelligent sys-

tems is described in ((Costa et al., 2007)). EHR and

AIDA are the starting point for the creation of Ambi-

ent Assisted Living practices on an ubiquitous com-

putational environment for improving the quality-of-

life of the elderly, being the core system for AIDA-CI

Clinical Inteliigence and Decison Support and AIDA-

IA - Ambient Intelligence.

4 THE COMPUTATIONAL

MODEL

With respect to the computational model, and in or-

der to fulﬁl all the pre-requisites associated to Agent

Oriented Programming (AOP) stated above, it were

considered extended logic programs with two kinds

of negation, classical negation, ¬, and default nega-

tion, not. Intuitively, not p is true whenever there is

no reason to believe p, whereas ¬p requires a proof of

the negated literal. An extended logic program (pro-

gram, for short) is a ﬁnite collection of rules and in-

tegrity constraints, standing for all their ground in-

stances, and is given in the form:

p ← p

∧ . . . ∧ p

∧ notq

∧ . . . ∧notq

; and

∧ . . . ∧ p

∧ notq

∧ . . . ∧notq

, (n, m ≥ 0)

where ? is a domain atom denoting falsity, the p

, and p are classical ground literals, i.e. either posi-

tive atoms or atoms preceded by the classical negation

sign ¬ (Neves et al., 2007). Every program is associ-

ated with a set of abducibles. Abducibles can be seen

as hypotheses that provide possible solutions or ex-

planations of given queries, being given here in the

form of exceptions to the extensions of the predicates

that make the program (Neves, 1984).

These extended logic programs or theories stand

for the agents (or programs) that populate the universe

of discourse. On the other hand, logic programming

enables an evolving agent to predict in advance its

possible future states and to make a preference. This

computational paradigm is particularly advantageous

since it can be used in program synthesis, employ-

ing the methodologies for problem solving that ben-

eﬁt from abducibles, in order to make and preserve

abductive hypotheses (Kakas et al., 1998)(Kowalski,

2006).

In order to accomplish such goal, i.e., to model

the universe of discourse in a changing environment,

the breeding and executable computer programs (or

agents) will be ordered in terms of the quality of ser-

vice that stems out of them, when subject to a pro-

Figure 2: A blended of the agents that make the Universe of

Discourse.

cess of conceptual blending (Turner and Fauconnier,

1995). In blending, the structure or extension of two

or more predicates is projected to a separate blended

space, which inherits a partial structure from the in-

puts, and has an emergent structure of its own. Mean-

ing is not compositional in the usual sense, and blend-

ing operates to produce understandings of composite

functions or predicates, the conceptual domain, i.e.,

a conceptual domain has a basic structure of enti-

ties and relations at a high level of generality (e.g.,

the conceptual domain for journey has roles for trav-

eler, path, origin, destination). In our work we will

follow the normal view of conceptual metaphor, i.e.,

metaphor will carry structure from one conceptual do-

main (the source) to another (the target) directly (Fig-

ure 2). In Figure 2 INPUT denote the agents (or pro-

grams) that are object of optimization and correlative

evolution.

We construct a dynamic virtual world of complex

and interacting population of agents, entities that are

built as evolutionary logic programs or theories that

compete against one another in a rigorous selection

regime, in order to produce the optimal model to a

particular problem(i.e., the OUTPUT, Figure 2). In

other words, the agents or logical theories evolve in

order to model the universe of discourse, in which ﬁt-

ness is judged by one criterion alone, the quality of

service (Analide et al., 2008).

It is now possible to measure or quantify the qual-

ity of service, which is given in terms of the logic

program or theory, deﬁned at meta-level, according to

the logic program (or agent) given below:

{¬qos(Agent, Q

service

) ←

not qos(Agent, Q

service

)∧

not exception

qos

(Agent, Q

service

exception

(ag

, 0.5),

exception

(ag

, 0.33),

?(((pa(ag

, X)∨

ARCHETYPE-BASED SEMANTIC INTEROPERABILITY IN HEALTHCARE

307

Figure 3: Quality of service reported by the qos agent for

the Healthcare Information System.

pa(ag

, Y )∧

¬((pa(ag

, X)∧

pa(ag

, Y ))),

exception

(X, Y ) ← da(ag

, Y ),

da(ag

, da),

ca(ag

, 1),

. . .}

qosagent

where the integrity constraint or invariant for ag

stated above denotes an exclusive or, i.e., the qual-

ity of service associated with the ag

is tailored by

the exceptions referred to above for ag

(in this case

the value of 0.5). Therefore, the quality of service

reported by the qosagent for the Healthcare Informa-

tion System is given by the dashed area of the Figure

3, where pa, da, ca, ra and iea are respectively pred-

icates for evaluating the quality of service of proxy

agents, decision agents, computing agents, resource

agents and interaction and explanation agents.

5 CONCLUSIONS

This work presents ongoing research and some de-

velopments on improving semantic interoperabil-

ity of different information systems, using open

archetypes. It was introduced an archetype-based

agency-independent testing framework, the agency

AIDA, that can validate archetype implementations

and help ensure quality of service and interoperabil-

ity of singular information systems. Challenges for

integrating archetype and terminology was discussed

and a candidate open language for expressing termi-

nological value sets was presented. Finally, advanced

archetype-based data sharing using clinically mean-

ingful scenarios was demonstrated. The aim was not

only to view the exchanged data but also utilized the

archetype semantics of the data. The scenarios in-

cluded the use of local decision support rules on re-

ceived data, namely drug interactions and warnings.

REFERENCES

Alves, V., Machado, J., Abelha, A., and Neves, J. (2005).

Agent based decision support systems in medicine. In

WSEAS on Biology and Biomedicine, Issue 2, Volume

Analide, C., Abelha, A., Machado, J., and Neves, J. (2008).

An agent based approach to the selection dilemma in

cbr. In Badica, C., Mangioni, G., Carchiolo, V., and

Burdescu, D. D., editors, IDC, volume 162 of Studies

in Computational Intelligence, pages 35–44. Springer.

Costa, R., Neves, J., Novais, P., Machado, J., Lima, L., and

Alberto, C. (2007). Intelligent mixed reality for the

creation of ambient assisted living. In Neves, J., San-

tos, M., and Machado, J., editors, Progress in Artiﬁ-

cial Intelligence, volume 4874. LNAI, Springer.

Kakas, A., Kowalski, R., and Toni, F. (1998). The role of

abduction in logic programming. In Gabbay, D., Hog-

ger, C., and Robinson, J., editors, Handbook of logic

in Artiﬁcial Intelligence and Logic Programming, vol-

ume 5, pages 235–324. Oxford University Press.

Kowalski, R. (2006). The logical way to be artiﬁcially intel-

ligent. In Toni, F. and Torroni, P., editors, Proceedings

of CLIMA VI. LNAI, Springer Verlag.

Neves, J. (1984). A logic interpreter to handle time and

negation in logic databases. In Proceedings of ACM

1984 Annual Conference, San Francisco, USA.

Neves, J., Machado, J., Analide, C., Abelha, A., and Brito,

L. (2007). The halt condition in genetic programming.

In Neves, J., Santos, M. F., and Machado, J., editors,

EPIA Workshops, volume 4874 of Lecture Notes in

Computer Science, pages 160–169. Springer.

Rigor, H., Machado, J., Abelha, A., Neves, J., and Alberto,

C. (2008). A web-based system to reduce the nosoco-

mial infection impact in healtcare units. In Cordeiro,

J., Filipe, J., and Hammoudi, S., editors, WEBIST (1),

pages 264–268. INSTICC Press.

Turner, M. and Fauconnier, G. (1995). Conceptual inte-

gration and formal expression. In Johnson, M., edi-

tor, Journal of Metaphor and Symbolic Activity, vol-

ume 10.

Weed, L. (1969). Medical records, medical education, and

patient care. the problem-oriented record as a basic

tool.

ICAART 2010 - 2nd International Conference on Agents and Artificial Intelligence

308