A SEMANTIC WEB TECHNOLOGIES-BASED SYSTEM

FOR CONTROLLING THE CORRECTNESS OF MEDICAL

PROCEDURES IN POLISH NATIONAL HEALTH FUND

Jan Andreasik

Zamo

c University of Management and Administration, Zamo

c, Poland

Institute of Biomedical Informatics, University of Information Technology and Management in Rzesz

ow, Rzesz

ow, Poland

Andrzej Ciebiera, Sławomir Umpirowicz

Institute of Biomedical Informatics, University of Information Technology and Management in Rzesz

ow, Rzesz

ow, Poland

Keywords:

Decision support system, Semantic web, Ontology, RDF, RDFS, OWL, SPARQL, SPIN.

Abstract:

The correctness veriﬁcation of the medical procedures described by enormous number of data stored in large,

relational databases is very difﬁcult. The database queries executed on redundant, incoherent, and contextless

data may cause erroneous results. The Semantic Web technology has been considered to solve the problem.

In this paper we propose the Semantic Web technologies-based system architecture for controlling medical

procedures. We performed the set of inference tasks that conﬁrmed the usefulness of the Semantic Web

technology. The experience with Semantic Web modelling can be used for implementation of solutions to the

contextual data integration and analysis.

1 INTRODUCTION

In the recent years, there has been intensive develop-

ment of the semantic Web technologies towards the

semantic analysis of data stored in large repositories.

In many institutions and enterprises, large distributed

databases based on relational database technologies

such as ORACLE, Microsoft SQL Server and oth-

ers, operate and are still extended. The data explo-

ration process based on this model causes a series of

issues that are related to preparation of reports. More-

over, the query technologies of distributed databases

does not meet the user’s requirements. The databases’

users are not familiarized with the database query

languages (SQL) because they usually are members

of the administrative staff, marketing specialists in

ﬁnance departments, and inspectors in various con-

trolling ofﬁces. These specialists want to formulate

queries in the natural language of the speciﬁc domain.

They also want to obtain reports with data or infor-

mation more concise than reports created on the basis

of relationships between data. They want to extract

the knowledge concerning certain issues creating the

knowledge based system (KBS). Semantic Web (Heb-

eler et al., 2009) gives them such possibilities.

In this paper we present the architecture of sys-

tem developed for the branch of the Polish National

Health Fund (NHF) in Podkarpackie voivodship. It

deﬁnes a list of issues including search processes of

the NHF distributed databases. The main issue is re-

lated to the control of medical procedures. The NHF

inspectors want to obtain the reports that respond to

the speciﬁc and complex queries.

This kind of systems is under the development.

Several systems, based on hospital repositories, and

used for patients monitoring can be mentioned here.

The HIWO (Hospital Ward Intelligent Ontology) sys-

tem was presented by P. Katari, R. Juric, S. Pau-

robally, K. Madani (Kataria et al., 2008). HIWO on-

tology is shown in the tool TopBraid whereas conver-

sion of data from the Oracle Express Edition 10g rela-

tional database to the RDF format is made with using

the D2RQ tool, which uses the Jena API libraries. An-

other approach was proposed by P. LePendu, D. Dou,

G.A. Frishkoff, J. Rong (LePendu et al., 2008), who

formed the ontological base for the analysis of elec-

troencephalographs. The data were collected in the

MySQL RDBMS, rules were written in the SWRL

331

Andreasik J., Ciebiera A. and Umpirowicz S..

A SEMANTIC WEB TECHNOLOGIES-BASED SYSTEM FOR CONTROLLING THE CORRECTNESS OF MEDICAL PROCEDURES IN POLISH

NATIONAL HEALTH FUND.

DOI: 10.5220/0003642103310336

In Proceedings of the International Conference on Knowledge Management and Information Sharing (KMIS-2011), pages 331-336

ISBN: 978-989-8425-81-2

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

Figure 1: System architecture for controlling the correctness of medical procedures.

Figure 2: Controlling system class diagram in the TBC editor.

language and queries were constructed in the fol-

lowing languages: SPARQL, OWL-QL, and SQL.

H. Chen (Chen et al., 2006) and co-workers devel-

oped a tool called Dartgrid for processing data from

the relational database using the SPARQL query lan-

guage. An application for the examination of Chi-

nese medicine procedures based on China Academy

of Traditional Chinese Medicine has been developed.

The other approach is to create systems, designed

in accordance with the principles of the SW tech-

nologies. Such systems are based on the medical

ontologies such as UMLS (NLM, 2011), SNOMED

(IHTSDO, 2011), and Galen (OpenGALEN, 2011).

The system for medical procedure registering with the

use of OWL was presented by A.L. Rector, R. Qamar,

T. Marley (Rector et al., 2009). The medical infor-

mation management system in a hospital, based on

the following languages: OWL-S, OWL, SWRL, was

presented by M.A. Casteleiro, J.J. Des Diz (Casteleiro

and Des Diz, 2008). They have developed an inter-

face between the UMLS thesaurus and the OWL lan-

guage editor called Protege-OWL. The model of pub-

lic health information network with use of the seman-

tic modeling was developed by Mirhaji P, Casscells

SW, Allemang D, Coyne R (Mirhaji et al., 2007).

2 SYSTEM ARCHITECTURE

At present, the usage of Semantic Web requires con-

nection to databases that supports the RDF model

storage or imports data into the RDF structures.

Source data for the system for controlling the correct-

ness of medical procedures is stored by the NHF in

their relational Oracle databases. Therefore, to make

connection to the data sources, the D2RQ (FUB,

2011) converter was used. We used the TopBraid

Composer (TBC) Free Edition (Quadrant, 2011) as

the SW modeling tool, which is the extended lan-

guage editor for RDF, RDFS and OWL, and more-

over, explorer of instances and a tool for performing

the SPARQL queries with a graphical user interface.

An additional advantage of TBC is the built-in rule

engine SPIN (SPARQL Inferencing Notation), that

KMIS 2011 - International Conference on Knowledge Management and Information Sharing

332

Figure 3: Organization of the class POZYCJE RECEPT

(prescriptions’ details) related to the class RECEPTY (pre-

scriptions).

Figure 4: Organization of the class OPAKOWANIA (drugs’

packages) related to the class RECEPTY (prescriptions).

makes the process of deﬁnition of additional condi-

tions and rules in the constructed queries of the sys-

tem easier.

In the future, when using Oracle 11g, the D2RQ

tools can be removed. The set of the relational

databases in NHF consists of over ten bases. For the

experiment purposes, three databases were chosen:

base of drugs, base of services and base of prescrip-

tions. Each base is characterized by a large number of

records (approximately 50 million records).

During the analysis of data based on the SQL

queries, carried out by NHF, we identiﬁed inconsis-

Figure 5: SPARQL query.

Figure 6: Recipes identiﬁers - the result of SPARQL query.

tencies in the source data. One of the objectives of

the project was also to conﬁrm the suitability of the

Semantic Web data model for database integration.

Source data for the controlling system were identiﬁed

as:

• Doctors’ base (BD),

• Patients’ base (BP),

A SEMANTIC WEB TECHNOLOGIES-BASED SYSTEM FOR CONTROLLING THE CORRECTNESS OF MEDICAL

PROCEDURES IN POLISH NATIONAL HEALTH FUND

333

Figure 7: SPIN rule deﬁned in the TBC.

Figure 8: Instances of the class Zdarzenia uprawniajace

(ZdUpr) deﬁned in the TBC.

• Prescriptions’ base (BR),

• Medicines’ base (BM),

• Diagnoses’ base (BDIAG),

• Services’ base (BS),

• Accounts’ base (BA),

• a dictionary of international drug names

(SLEK SL NAZWYM),

• the dictionary ICD 9 (International Classiﬁcation

of Medical Procedures),

• the dictionary ICD 10 (International Statistical

Classiﬁcation of Diseases and Related Health

Problems).

Figure 9: Processing times depending of the database ac-

cess mode.

3 CONTROLLING SYSTEM

ONTOLOGY

Because the controlling system was designed to pro-

cess the existing source data, we applied organization

of classes within the scope of structures and proper-

ties based on the structure and Polish labels of source

data stored in the Oracle NHF database. This solution

makes the implementation of classes and their prop-

erties in OWL complicated, but facilitates the work of

NHF doctors (familiarity of existing structures) and

ensures immutability of organization of existing data

sources which is needed for other applications. To

conﬁrm the usefulness of the SW technology in the

analysis of the medical procedures we selected two

problems groups of medical procedures where drugs

are used:

(i) Tramadol and Pancreatinum,

(ii) Clopidogrel.

These medicines may be prescribed on prescriptions

in certain medical procedures (Proc) only with the

corresponding privilege code (KodUpr) and these

medicines are under supervised distribution. The dif-

ference between (i) and (ii) relays on the fact that the

prescription for (i) has no time validity. The medicine

is considered as a correct when it is prescribed at any

KMIS 2011 - International Conference on Knowledge Management and Information Sharing

334

time - before, during and after the event giving the

privileges (ZdUpr) for prescription. The prescription

for (ii) is the case when the medicine is prescribed

during the particular time depending on ZdUpr. An

ontology of the system for controlling the correctness

of medical procedures has been prepared in the TBC

editor and it is presented in Figure 2. The prepared

structure corresponds to structure organization cur-

rently used in the NHF data sources.

4 ANALYSIS OF MEDICAL

PROCEDURES USING SPARQL

AND SPIN IN THE

CONTROLLING SYSTEM

UNITS

Deﬁned inference problems for (i) and (ii) require dif-

ferent search paths. For case (i), searching for the

numbers of incorrect prescriptions (there is no ZdUpr

record) after the modeling OWL needs only to deﬁne

the query of the type SELECT in the SPARQL lan-

guage. For case (ii), a query in SPARQL requires

many additional conditions, as many as different pe-

riods that correspond to ZdUpr in the dictionary of

events. For this group of problems, a queries of the

Construct type in the SPARQL language we found

useful in the deﬁnition of a rule. All the conditions de-

ﬁned by the medical inspectors were taken into con-

sideration in the rule. A key SPIN features are the

following:

• possibility of calculation of property values based

on other properties,

• checking constraints and data validation,

• creating rule templates made under certain condi-

tions.

Figure 7 shows a rule deﬁned in SPIN. The struc-

ture of this rule allows creation of libraries of similar

queries which will be important facilities for the med-

ical inspectors.

For case (ii) in our analysis, we prepared the

class Zdarzenia uprawniajace (class for privileges

to prescribe drugs) with properties: Zdarzenie upr

(event identiﬁer), Zdarzenie upr kod (event code),

Zdarzenie upr liczba dni (the number of days just

after the date of event within the drug can be

applied), Zdarzenie upr nazwa m leku (international

drug name), each of data type properties, which allow

to refer to values of the ICD 9 and ICD 10 dictionar-

ies. The model of the controlling system supported by

TBC with embedded SPARQL and SPIN (Fuber and

Hepp, 2011) allows to create the optimal structures of

queries, and also allows further, ﬂexible development

of the system model with another structures, rules and

queries. Furthermore, TBC allows the development

of friendly interfaces for end users of the system, i.e.,

medical specialists.

5 PROCESSING TIME OF

SEMANTIC DATA

After the system correctness conﬁrmation the pro-

cessing time tests were carried out with the relational

database and the D2RQ converter and with the Oracle

11g database, which natively supports the storage of

RDF data.

Conﬁguration of the server:

• Hardware: Intel Core 2 Duo E7600, 8 GB RAM.

• Software: Windows Server 2003 R2, Oracle

11.1.0.6.0, TopBraid Composer ME 3.2.0.

Processing times depending of the database access

mode shows the Table 1 and the Figure 9.

Table 1: Processing times depending of the database access

mode.

Number of RDF Time [h:m:s] Time [h:m:s]

triples D2RQ Oracle 11g

3 452 958 00:01:09 00:16:00

14 233 326 00:11:10 02:47:30

27 708 786 00:38:54 08:23:30

41 184 246 01:42:10 More than 24 hours

6 CONCLUSIONS

The result of our experiment is the achievement of a

high degree of data integration in order to obtain the

expected reports without any intervention to the exist-

ing distributed structure of the databases. The system

for controlling the correctness of medical procedures

model based on the TBC and D2RQ tools conﬁrmed

the usefulness of SW in the analysis of the medical

procedures described by the records stored in the re-

lational databases. The experience with controlling

system modelling can be used to:

• extension of the developed model by new data

structures, relations, inference rules, and the tools

for result data visualization,

• sharing knowledge structures stored in the RDF,

RDFS, OWL, and the SPIN / SPARQLMotion

rules in description of the medical procedures,

A SEMANTIC WEB TECHNOLOGIES-BASED SYSTEM FOR CONTROLLING THE CORRECTNESS OF MEDICAL

PROCEDURES IN POLISH NATIONAL HEALTH FUND

335

• implementation of the SW tools based on a similar

data model or analogical purposes of inference,

• implementation of solutions to the contextual data

analysis.

REFERENCES

Casteleiro, M. and Des Diz, J. (2008). Clinical practice

guidelines: A case study of combining owl-s, owl, and

swrl. Knowledge-Based Systems, 21:247–255.

Chen, H., Wang, Y., Wang, H., Mao, Y., Tang, J., Zhou, C.,

Yin, A., and Wu, Z. (2006). Towards a semantic web

of relational databases: a practical semantic toolkit

and an in-use case from traditional chinese medicine.

In Proc. of the Fifth International Semantic Web Con-

ference, pages 750–763.

FUB (2011). The d2rq platform-treating non-rdf databases

as virtual rdf graphs. http://www4.wiwiss.fu-

berlin.de/bizer/d2rq/.

Fuber, C. and Hepp, M. (2011). Using sparql and spin

for data quality. http://heppnetz.de/ﬁles/feurber-hepp-

sparql-spin-dqm.pdf.

IHTSDO (2011). Snomed ct.

http://www.ihtsdo.org/snomed-ct/.

Kataria, P., Juric, R., Paurobally, S., and Madani, K. (2008).

Implementation of ontology for inteligent hospital

wards. In Proc. of the 41st Hawaii International Con-

ference on System Sciences (HICSS 2008), pages 1–9.

IEEE.

LePendu, P., Dou, D., Fishkoff, G., and Rong, J. (2008).

Ontology database: A new method for semantic mod-

eling and an application to brainwave data. In Lu-

dacher, B. and Mamoulis, N., editors, Proc. of the

SSDBM 2008, pages 313–330, Berlin Heidelberg.

Springer-Verlag.

Mirhaji, P., Casscells, S., Allemang, D., and Coyne, R.

(2007). Improving the public health information net-

work through semantic modeling. IEEE Intelligent

Systems.

NLM (2011). Umls. http://www.nlm.nih.gov/research/

umls.

OpenGALEN (2011). http://www.opengalen.org/.

Quadrant, T. (2011). Topbraid composer.

http://www.topquadrant.com/products/TB Composer.

html#free.

Rector, A., Qamar, R., and Marley, T. (2009). Binding

ontologies and coding systems to electronic health

records and messages. Applied Ontology, 4:51–69.

KMIS 2011 - International Conference on Knowledge Management and Information Sharing

336