ONTOLOGY BASED SEMANTIC REPRESENTATION OF THE

REPORTS AND RESULTS IN A HOSPITAL INFORMATION

SYSTEM

B. Prados-Suárez

Software Engineering Department. University of Granada, Spain

E. González Revuelta

Computer Science Department. Torrecárdenas Hospital, Almería, Spain

C. Peña Yáñez, G. Carmona Martínez

Computer Science Department. San Cecilio Hospital, Granada, Spain

C. Molina Fernández

Software Engineering Department. University of Jaén, Spain

Keywords: Health Ontologies, Hospital knowledge representation, Semantic organization of results, Hospital

Information Systems, Semantic retrieval of hospital reports.

Abstract: The main purpose of this paper is to contextualize the access to the huge amount of results and reports that a

Hospital Information System (HIS) can reach to have. Our target is to integrate a semantic layer with the

HIS, so that the user can employ this layer to access just the precise information needed, under his working

context. Here we propose a new navigation system based on the semantic characteristics of the data

acceded, their complementary characteristics, properties and relations, providing so the HIS with a new tool

to solve an evident problem for the users.

1 INTRODUCTION

Nowadays a characteristic of Hospital Information

Systems (HIS) is the huge amount of data they

manage, generate and store, as well as their wide

variety and typology. Among these data reports or

outputs can be found, which are the direct result of

concrete queries, as well as results, that are more

elaborated data obtained from statistical processes or

a basic exploitation of the information in the DB,

using the tools offered by the applications of the

concrete area or department Prados M., Peña M.C.

2003. Examples of reports are the list of patients

waiting for a given surgery or the list of medicines

prescribed by each medical speciality; while the

average duration of the hospital stay or the

efficiency indicators in each medical speciality are

examples of results. The set of all these documents

and data is what we call Universe of Result Reports

(URR).

A serious inconvenience is that the access to

these data must be done by a complicated browse

though the menus of different applications, which

gives rise to severe problems like:

• There is so much information available that the

user can not find what is looking for.

• There are a number of reports and outputs, that

are wrongly controlled, and these are sometimes

useless.

• There is no proper organization of the

information, making quite difficult the retrieval.

• There is also a great amount of redundant

results that, in addition, can be acceded from

different applications.

300

Prados-Suárez B., González Revuelta E., Peña Yáñez C., Carmona Martínez G. and Molina Fernández C. (2008).

ONTOLOGY BASED SEMANTIC REPRESENTATION OF THE REPORTS AND RESULTS IN A HOSPITAL INFORMATION SYSTEM.

In Proceedings of the Tenth International Conference on Enterprise Information Systems - ISAS, pages 300-305

DOI: 10.5220/0001711003000305

 SciTePress

• The user doesn’t know the structure of the

results, and hence what can he obtain from them

and from where (which application) must do it.

• The user obtains something that is not what

he/she really needs, or even can not find it.

• The models used to exploit the information

changes along the time, as well as the criteria,

purposes and requirements.

To solve these problems it would be necessary to

define logical structures to organize the URR,

designing it under semantic criteria. In this way new

navigation systems can be designed to allow the user

to focus or widen the information queries Lambrix

P., Habbouche M. and Perez M. (2003).

This is the target we face in this paper: The

analysis and design of a system that enables the user

to obtain accurate and appropriate information, as

well as to create fine and useful results user

interface, with as much knowledge as possible.

Our proposal to face this task is to establish a

conceptualization of the URR, by means of

Ontologies.

2 BACKGROUND

It is well known that Ontologies are conceptualizations

with a variety of utilities, among which we remark two:

on one hand, they are quite helpful in decision

making tasks like diagnostic decision in Medicine

Lambrix P., Habbouche M. and Perez M. (2003); on

the other hand, they are useful in Document

Retrieval Systems and Semantic Information

Retrieval Systems, where information is acceded by

its meaning and not just from a given terminological

description Berners-Lee, T., Hendler, J.and Lassila,

O., (2001).

Today there is a wide variety of applications of

Ontologies within specialized fields, like “Health

Systems”, especially for medical terminology

analysis and the extraction of its semantic

characteristics Angelova, Galia, (2000). Within this

framework, is especially significant their use on

“query expansion”, to detect the kernel of a user’s

query and extend it with variants, finding the

documents with most relevant information

Strzalkowski, Tomek; Lin, Fang; Wang, Jin;and

Pérez-Carballo, José (1999). In addition, Ontologies

have been used to unify the medical language in

multilingual frameworks Scope Web site (2002)

and for semantic categorization of medical diagnosis

terminology Prados Reyes, M. Peña Yañez C., Vila

Miranda A. and Prados-Suarez B., (2006).

However, Ontologies themselves can also be directly

useful for the user as a tool to integrate data and

applications, offering the possibility to know

conceptualization itself, and use it to improve the

information access. This is the framework where we

present our proposal.

3 SUPPORT MATERIAL

To develop our proposal we have the HIS of the

Hospital San Cecilio, in Granada. This system

covers the traditional areas of every HIS, and has an

extensive catalogue of results and reports, whose

contents are structured according to their

membership to HIS applications.

This catalogue (our URR), can be assumed as a

“catalogue document”, where each document is a

report that, as every document, can be identified by a

set of formal and semantic characteristics. The

contents in the URR catalogue have different types,

(from statistical reports to reports relating attributes

from the data model), but they always have a sample

selection.

The representation of these data and their

characteristics gives raise to a Knowledge Base,

designed according to an Ontology, that will be the

support to develop the Semantic Retrieval System of

the universe URR.

4 HIS STRUCTURE

The actually implemented HIS covers the Economical-

administrative and Logistical- assistance fields,

managing the Operational Systems of the Hospital

Organization, and hence its activity.

Functionally it is structured in several levels, as

indicated in Figure 1.

5 METHODOLOGY

To develop the system proposed in this paper we

have followed the next steps:

1- Determine the Type of Universe to Study: Most

of the elements in our URR have small formal

differences, like the presentation order of the results,

or the grouping criteria. Hence, under a documentary

point of view they are different, while from the

semantic point of view they have the same value. It

lead us to represent the formal differences as

ONTOLOGY BASED SEMANTIC REPRESENTATION OF THE REPORTS AND RESULTS IN A HOSPITAL

INFORMATION SYSTEM

301

properties in the Ontology, or inside the executed

procedure, being more relevant the semantic value.

2- Definition of the Types of URR Formal

Characteristics and the Semantic Valuation

Criteria: The main purpose of this stage is to find

which components are interesting and allow the user

to access the URR, focusing or widening the queries.

The elements found in this stage will have a

correspondence in the Ontology design.

3- Design and Development of the Ontology: Here

we have structured the components according to

semantic characteristics, hierarchies of classes and

relations between them, according to the universe

URR. Hence we focus exclusively on the

Operational System, avoiding the inclusion of

classes, relations, or properties that don’t exist in the

operational system, though they would have logical

sense.

4- Ontology Representation: In this last stage we

have implemented the Ontology using the tool

Protégé, following a representation through

conceptual scheme.

Figure 1: Functional structure levels of HIS.

6 THE UNIVERSE TO STUDY

Analyzing the URR we have found the following

possible contents:

 Attributes from the data model tables.

 Attributes generated in the executed

procedure, as partial or total results.

 Statistical calculations of different types.

Under a functional point of view, these reports

belong to a concrete application of the HIS, and are

generated by procedures included on that application

as modules or tasks.

6.1 Formal Analysis of the URR

Formally a document from the URR has the

following characteristics:

Type: This property can have one of the next values:

“query relation report”, “operative recount” or

“statistical report”.

Proprietary Application: This characteristic references

the functional module from the HIS logical design

that manages the information and procedures related

to the concrete target document. This characteristic

follows the logical levels structure of the HIS,

shown in bold letter in Figure 1 (System –

Subsystem – Functional Scope – Application –

Module – Procedure). Its values are, as an example,

“absenteeism” Module, from the “Human Resources

Management” Application, as well as those in

capital letter in Figure 1.

Computing Procedure that Generates (executes)

the Document: This attribute is referred to the

physical unit from the physical design that generates

the target document.

Domain of the Showed Attributes: This is the set

of data from the Operational System that are showed

in the document.

Generated Data: References the total results or

statistical values computed on the execution process.

Query Descriptors: Are those that allow the user to

filter the query for a particular document. Usually

they are very limited set like date, functional unit,

diagnosis, etc., and can be structured according to

their meaning.

6.2 Semantic Analysis of the URR

The semantic analysis of the URR offers a

perspective about its meaning from a given point of

view of interest, like the “business management”

perspective from the point of view of the

“administrative management”.

This task acquires maximum interest if we

consider that we have a logical structure of the

knowledge, showed in Figure 2, based on several

levels (contexts, scenarios, images, views), which

makes possible the semantic pertinency of a given

document to a concrete image or view Prados M.,

Peña M.C., Prados M.B. and Garrido J.M. (2004).

As an example, a report about the occupation of the

operating rooms is related (pertinent) to the

“Efficiency of operating rooms use” image, which

belongs to the “Surgical activity” scenario from the

“Medical Care” context.

In addition to the pertinency, there may be other

characteristics with semantic value like:

ICEIS 2008 - International Conference on Enterprise Information Systems

302

Orientation: That determines the type of user to

which the document is directed: Medical directors,

Management, Central Services…

Confidentiality level: With three possible values:

low, medium and high.

7 ONTOLOGY DESIGN

The aim of the Ontology design is to elaborate a

proposal according to the existing HIS, but opened

to the incorporation of new elements to the URR,

that could give raise to new “main classes” (defined

later) or instances.

The basic criteria that must be verified are the

hierarchization (ordering), modularization (concept

isolation), abstraction, clarity and coherency.

Considering it, we have defined the following

elements for the design:

z Domain: Set of reports and results (URR) that

can be obtained from the Hospital Information

System. Hence, the concepts or Domain

elements will be the Results, corresponding to

reports/outputs/lists issued from the variety of

information sources, on user demand.

z Classes: Each class corresponds to one of the

formal or semantic characteristics defined on

the URR. Here we define the concept of “main

class” to mean that the instances of that type of

class are elements from the URR, and we

establish as main class the semantic

categorization.

This way, as an example we say that an “extern

emergency assistance delay report” is an instance of

the class “Emergency efficiency”, which is a

subclass of “Emergency assistance activity”,

instance at the same time of “Assistance activity”

that is a subclass of the “Operational context”.

z Properties: Assigned to each element or class

of the URR domain, represent the set of

characteristics of a given element or set of

elements. There are two types of properties:

complex properties, defined as relations

between ontology classes; and explicit

properties, defined exclusively for a given class

or instance, without generating a new class.

z Instances: the elements in the URR are

assigned to instances of the semantic

categorization. They will be represented by the

naming of the physical unit of the HIS. There

will also be instances corresponding to the

individual terminal elements of the class

hierarchies, and will be defined in the Ontology

as “direct classes”.

Figure 1: Semantic Categorization scheme. Example:

Operational Context Scenarios.

7.1 Identification of Ontology Classes

The main conceptual classes defined in the Ontology

are:

Semantic Categorization: Represents the cognitive

focus approached by a URR document, and obeys to

user criteria formalized following the levels schema

in Figure 2. Its instances are the URR documents

expressed as a code according to the physical unit or

executable computing procedure.

Logical Model Functions: This class represents the

organization activities, computerized and structured

according to hierarchies of functional requirements.

An approach to this class hierarchization is shown in

Figure 1. As an example, the control of the

“temporary working inability” is a member of

“Absenteeism control”, member at the same time of

“Human resources management”, member again of

“Administrative management”, that belongs to the

Economic-Administrative subsystem.

Receiver: Final destination of the document. We

have planned three types: “Local”, “Extern” and

“Publication”. The class “Local” generates a

hierarchic line following the organic structure of the

Hospital government. The “Extern” class represents

different institutions that are users of the hospital

information, while “Publication” refers different

ways to announce the information (web pages,

notice board…).

Chronology: Establishes the “periodical” (and

period) or “on demand” character.

Confidentiality Level: Unavoidably associated to

every hospital document.

Query Strategy: Represents the dimensions through

which the user can select or filter the content of a

ONTOLOGY BASED SEMANTIC REPRESENTATION OF THE REPORTS AND RESULTS IN A HOSPITAL

INFORMATION SYSTEM

303

document It is useful to inform the user about the

possibilities to focus the query.

Leaning to decision: Indicates the interest of the

document under a strategic point of view, like

periodic control, evolution or research.

Semantic Valoration of the Contents: It refers to

the set of resulting information in a URR document.

Concretely, it refers the physical data model, and the

structure and contents of the data tables themselves.

Hierarchies in this class are identified according to

three levels: “application”, “table type” (historical,

movements, master, and base) and “data type” (own

attribute, situation attribute, movement attribute).

7.2 Properties Definition

The aim of the properties is to enrich the conceptual

structure of the Ontology, to properly define URR

documents according to their formal and semantic

characteristics. We have two types of properties:

Complex Properties: Are an integral part of the

Ontology, and are defined as relations between

classes. As an example, a report of the “Attended

emergencies”, which is a member of “Emergency

frequentiation”, has the properties of being

confidential, of local use for Medical management,

can be selected chronologically, functionally or by

its characters, and show the data in the table

“Emergency movements”.

Explicit Properties: Represent qualities of one or

several classes, and are useful only when they are

assigned to a few classes since, otherwise they

would finally become a new class. This type of

properties is reserved for very specific

characteristics of a class or instance, like a document

which totals some concrete values according to a

specific calculation. These properties can be easily

added and removed to enable or disallow some

given characteristic.

7.3 Ontology Generation with

PROTÉGÉ

Once made the intellectual tasks to define the

Ontology, the use of the tool Protégé Natasha Noy

and Samson Tu, (2003) has made possible to easily

obtain the conceptual scheme, the properties and

characteristics, with the advantage of the possibility

of generate a model in XML and RDF. Since this

tool is quite useful for maintenance tasks it also

offers the possibility of facing the problem in an

increasing way.

8 USE OF THE ONTOLOGY IN

RETRIEVAL TASKS

As previously mentioned our aim is to enable the

access through the user’s concepts, to improve the

accessibility to the information without having to

know the technical or functional structure of the

system.

Since every document in the URR is marked by

the Ontology according to a set of formal or

semantic properties, it can be indexed by a set of

descriptors, as in any Documentary System Gil

Leyva, I. and Rodríguez Muñoz, J.V. (1996). With

it, we can offer the user different access ways:

Simple Documentary Query: The user formulates

the query equation from which the “query kernel” is

obtained, by “query expansion”, as a set of

descriptors of the concepts in the Ontology. From

this query, the reports indexed by this set are

obtained Scope Web site (2002).

Focusing by Semantic Categorization: It locates

the user in the cognitive field and selects the reports

pertinent to the target study. It is performed by

accessing the class “semantic categorization”, whose

navigation allows the centre the query and know the

properties related to each level or instance.

Ontology Navigation: The user gets into the

Ontology structure, knowing its classes, and from

them focus or widen the query field or analyze the

properties of a given class up to reach the instance

level that is the target of the selection.

Query Contextualization: Quite similar to a

documentary query, but by descriptors in the

knowledge base, returning to the user the context by

means of the hierarchical relations and the properties

defined in the Ontology, as well as the documents

pertinent to that context. Among them, the user will

select those interesting for him/her. As can be seen,

this is a semantic query procedure.

9 CONCLUSIONS:

IMPLEMENTABLE DATA

MODEL

The design proposed in this paper is a

conceptualization of the URR obtained from the

HIS. This design satisfies the needs posed by the

habitual users of the HIS, offering them an interface

that makes easier the access to the URR contents,

allowing the user to know a priori what can be

interesting for a concrete study, the characteristics

ICEIS 2008 - International Conference on Enterprise Information Systems

304

and limitations of the contents and other documents

that can be useful.

Hence, the proposed design has a mainly practical

character, and represent a practical application of the

Ontologies, restricted but useful to easily solve a

problem that can be generalized and extended to

other fields, especially in the hospital framework.

The structural richness of the Ontology is

conditioned by the characteristics and extent of the

HIS, but we think that it is secondary, since the

update, maintenance and extent of the Ontology and

its data model is easy to bring up to date.

The process to implement the model has been

cyclic and iterative: on an initial model and

implementation we have iteratively discover

relations or classes, those have modified the

conceptual model and have unchained the pertinent

modifications cascade.

ACKNOWLEDGEMENTS

This work was partially supported by the Junta de

Andalucía, under Project P07-TIC 03175.

REFERENCES

Prados M. and Peña M.C. (2003), Sistemas de

Información hospitalarios. Organización y gestión de

Proyectos. EASP (Escuela Andaluza de Salud

Pública), Granada, 1nd edition.

Lambrix P., Habbouche M. and Perez M. (2003).

Evaluation of ontology development tools for

bioinformatics. Bioinformatic. Vol 19, n 12, pag 1564-

1571.

Berners-Lee, T., Hendler, J.and Lassila, O., (2001) The

Semantic Web, Scientific American, 284(5):34–43

Angelova, Galia, (2000) Ontologies for Natural Language

Processing Applications. In: Simov, Viril; Kiryakov,

Atanas (eds.) Ontologies and Lexical Knowledge.

Sofia: OntoText Lab, Sirma AI EOOD, p. 1-15.

Strzalkowski, Tomek; Lin, Fang; Wang, Jin;and Pérez-

Carballo, José (1999) Evaluating Natural Language

Processing Techniques in Information Retrieval. In:

Strzalkowski, Tomek(ed.) (1999) Natural Language

Information Retrieval. Dordrecht: Kluwer Academic

Publishers, p. 113-145. ISBN 0-7923-5685-3.

Scope Web site (2002), http://tecn.upf.es/scope.

Prados Reyes, M. Peña Yañez C., Vila Miranda A. and

Prados-Suarez B., (2006) Generation and Use of one

Ontology for Intelligent Information Retrieval from

Electronic Record Histories. In Proceedings of the 8th

International Conference on Enterprise Information

Systems. pp 565– 571, Ed. INSTICC, Paphos, Chipre.

May 23-27.

Prados M., Peña M.C., Prados M.B. and Garrido J.M.

(2004), Gestión de conocimiento en el ámbito

hospitalario. EASP (Escuela Andaluza de Salud

Pública), Granada, 1nd edition.

Natasha Noy and Samson Tu, (2003) Developing Medical

Informatics Ontologies using Protégé, (Tutorial

materials from the AMIA 2003 Protégé tutorial). In

AMIA Symposium

Gil Leyva, I. and Rodríguez Muñoz, J.V. (1996)

Tendencias en los Sistemas de Indización Automática.

Estudio evolutivo. Revista Española de

Documentación Científica, vol 19, nr 3 p 273-291.

ONTOLOGY BASED SEMANTIC REPRESENTATION OF THE REPORTS AND RESULTS IN A HOSPITAL

INFORMATION SYSTEM

305