TOWARDS A SEMANTIC SYSTEM FOR MANAGING CLINICAL

PROCESSES

Ermelinda Oro

DEIS, University of Calabria, Via P. Bucci 41/C, Rende (CS), Italy

Massimo Ruffolo

ICAR-CNR, University of Calabria, Via P. Bucci 41/C, Rende (CS), Italy

Keywords:

Knowledge Representation and Reasoning, Knowledge Management, Ontology, Workﬂow, Data Mining,

Workﬂow Mining, Decision Support System, Health Care Information System, Clinical Process, Medical

Knowledge.

Abstract:

Managing costs and risks is an high priority theme for health care professionals and providers. A promising

approach for reducing costs and risks, and enhancing patient safety, is the deﬁnition of process-oriented clini-

cal information systems. In the area of health care information systems, a number of systems and approaches

to medical knowledge and clinical processes representation and management are available. But no systems

that provide integrated approaches to both declarative and procedural medical knowledge are currently avail-

able. In this work a clinical process management system aimed at supporting a semantic process-centered

vision of health care practices is described. The system is founded on an ontology-based clinical knowledge

representation framework that allows representing and managing, in a uniﬁed way, both medical knowledge

and clinical processes. The system provides functionalities for: (i) designing clinical processes by exploit-

ing already existing and ad-hoc medical ontologies and guideline base; (ii) executing clinical processes and

monitoring their evolution by adopting alerting techniques that aid to prevent risks and errors; (iii) analyzing

clinical processes by semantic querying and data mining techniques for making available decision support

features able to contain risks and to enhance cost control and patient safety.

1 INTRODUCTION

Across the world the issue of patient safety, medical

errors preventionand healthcare risk management is a

very challenging and widely studied research and de-

velopment topic. It stimulates a growing interest in

the computer science researchers community.

A promising approach for reducing errors and

risks, and enhancing patient safety, is the deﬁnition

of process-oriented clinical information systems. In

fact, healthcare services and practices are character-

ized by complex clinical processes in which high risk

activities take place. A clinical process can be seen

as a workﬂow where clinical (e.g. treatments, drugs

administration, guidelines execution, medical exam-

inations) and general (e.g. patient enrolment, medi-

cal record instantiation, risk evaluation) activities and

events occur. Clinical processes and their activities

are, also, characterized by speciﬁc and sophisticated

medical knowledge. Systems that provide integrated

functionalities for representing and managing medi-

cal knowledge and for designing, executing (taking

into account risks rules and conditions) and analyzing

clinical processes, can change clinical practices and

help diffusion of a process and quality awareness in

healthcare organizations.

Currently, as described in Section 2, in the ﬁeld

of healthcare information systems, a number of ap-

proaches to medical knowledge and guidelines repre-

sentation and management have been proposed. Ex-

isting systems and approaches suffer of the following

shortcomings: (i) they have a lack of mechanisms for

errors and risks handling and prevention; (ii) they do

not use the same formalism for representing and man-

aging both medical knowledge and clinical processes,

hence, they are not able to exploit in a fully uniﬁed

way declarative and procedural knowledge during ex-

ecution and monitoring of clinical process; (iii) they

do not allow to organize clinical processes and their

element as an ontology; (iv) they do not allow to mod-

ify and customize represented knowledge and to exe-

cute clinical process in a ﬂexible and agile way.

180

Oro E. and Ruffolo M. (2009).

TOWARDS A SEMANTIC SYSTEM FOR MANAGING CLINICAL PROCESSES.

In Proceedings of the 11th International Conference on Enterprise Information Systems - Artiﬁcial Intelligence and Decision Support Systems, pages

180-187

DOI: 10.5220/0001994401800187

 SciTePress

This work describes the prototypical implemen-

tation of a Clinical Process Management System

(CPMS) that aims at supporting a semantic process-

centered vision of health care practices. The system

is founded on an Ontology-based Clinical Knowledge

Representation Framework (OCKRF) that allows to

express in a combined way medical ontologies, clini-

cal processes and errors and risks rules. More in de-

tail, the OCKRF allows the CPMS to provide meth-

ods for: (i) creating ontologies of clinical processes

that can be queried and explored in a semantic fash-

ion; (ii) expressing errors and risks rules (by means

of reasoning tasks) that can be used (during processes

execution) for monitoring processes; (iii) executing

clinical processes and acquiring clinical process in-

stances by means of either workﬂow enactment (pre-

deﬁned process schemas are automatically executed)

or workﬂow composition (activity to execute are cho-

sen step-by-step by humans); (iv) monitoring clini-

cal processes during the execution by running reason-

ing tasks; (v) analyzing acquired clinical process in-

stances, by means of querying and inference capabili-

ties, in order to recognize errors and risks for patients.

The CPMS adopts a semantics approach for rep-

resenting and managing both static and dynamic as-

pects of medical knowledge. So it enables better de-

sign, execution, control and management of clinical

processes and related errors and risks rules. CPMS

delivers health care professionals semantic decision

support functionalities able to contain risks (due to

medical errors and adverse events) in order to enhance

patient safety.

The remainder of this paper is organized as fol-

lows. Section 2 describes related work in the ﬁeld of

healthcare information systems. Section 3 sketches

the OCKRF which the CPMS is founded on. Section

4 depicts system features and architecture by using an

application example. Finally, Section 4 concludes the

paper and sketches future work directions.

2 RELATED WORK

In the recent past, a strong research effort has been

taken to provide standard representations of declara-

tive and procedural medical knowledge. In the fol-

lowing, available approaches and systems for medi-

cal ontologies and clinical process representation and

management are described.

Medical knowledge representation area provides

one of the most rich collection of domain ontolo-

gies available worldwide. A very famous and widely

adopted thesaurus is Mesh the Medical Subject Head-

ings classiﬁcation (MESH). It provides a controlled

vocabulary in the ﬁelds of medicine, nursing, den-

tistry, veterinary medicine, etc. MeSH is used to in-

dex, catalogue and retrieve the world’s medical liter-

ature contained in PubMed. Another classiﬁcation,

that has become the international standard diagnos-

tic classiﬁcation for all medical activities and health

management purposes, is ICD10-CM (ICD; WHO)

the International Classiﬁcation of Diseases Clinical

Modiﬁcation, arrived to its 10th Revision. The most

comprehensive medical terminology developed to

date is SNOMED-CT (SNOMED), the Systematized

Nomenclature of Medicine Clinical Terms, based on

a semantic network containing a controlled vocabu-

lary. Electronic transmission and storing of medi-

cal knowledge is facilitated by LOINC, the Logical

Observation Identiﬁers Names and Codes (LOINC),

that consists in a set of codes and names describ-

ing terms related to clinical laboratory results, test

results and other clinical observations. Machine-

readable nomenclature for medical procedures and

services performed by physicians are descried in CPT,

the Current Procedural Terminology (CPT), a regis-

tered trademark of the American Medical Associa-

tion. A comprehensive meta-thesaurus of biomedi-

cal terminology is the NCI-EVS (NCI-EVS) cancer

ontology. Some medical ontologies are, also, due

to European medical organizations. For example,

CCAM the Classiﬁcation Commune des Actes Med-

icaux (CCAM), is a French coding system of clinical

procedures that consists in a multi-hierarchical classi-

ﬁcation of medical terms related to physician and den-

tal surgeon procedures. A classiﬁcation of the termi-

nology related to surgical operations and procedures

that may be carried out on a patient is OPCS4, the Of-

ﬁce of Population Censuses and Surveys Classiﬁca-

tion of Surgical Operations and Procedures 4th Revi-

sion (OPCS-4), developed in UK by NHS. The most

famous and used ontology in the ﬁeld of healthcare

information systems is UMLS, the Uniﬁed Medical

Language System (UMLS), that consists in a meta-

thesaurus and a semantic network with lexical appli-

cations. UMLS includes a large number of national

and internationalvocabulariesand classiﬁcations (like

SNOMED, ICD-10-CM, and MeSH) and provides a

mapping structure between them. This amount of

ontologies constitutes machine-processable medical

knowledge that can be used for creating semantically-

aware health care information systems.

The evidence-based medicine movement, that

aims at providing standardized clinical guidelines for

treating diseases (Sackett et al., 1996), has stim-

ulated the deﬁnition of a wide set of approaches

and languages for representing clinical processes. A

well known formalisms is GLIF, the Guideline Inter-

TOWARDS A SEMANTIC SYSTEM FOR MANAGING CLINICAL PROCESSES

181

change Format (GLIF). It is a speciﬁcation consist-

ing of an object-oriented model that allows to repre-

sent sharable computer-interpretable and executable

guidelines. In GLIF3 speciﬁcation is possible to re-

fer to patient data items deﬁned by a standard med-

ical vocabularies (such as UMLS), but no inference

mechanisms are provided. Proforma (Sutton and Fox,

2003) is essentially a ﬁrst-order logic formalism ex-

tended to support decision making and plan execu-

tion. Arden Syntax (Pryor and Hripcsak, 1993; Peleg

et al., 2001; HL7) allows to encode procedural med-

ical knowledge in a knowledge base that contains so

called Medical Logic Modules (MLMs). An MLM is

a hybrid between a production rule (i.e. an ”if-then”

rule) and a procedural formalism. It is less declarative

than GLIF and Proforma, its intrinsic procedural na-

ture hinders knowledge sharing. EON (Musen et al.,

1996) is a formalism in which a guideline model is

represented as a set of scenarios, action steps, deci-

sions, branches, synchronization nodes connected by

a ”followed-by” relation. EON allows to associate

conditional goals (e.g. if patient is diabetic, the tar-

get blood pressures are 135/80) with guidelines and

subguidelines. Encoding of EON guidelines is done

by Protg-2000 (Prot´eg´e) knowledge-engineering en-

vironment.

The group of formalisms, presented above, aim

at representing either static (ontologies) or procedural

(guidelines) medical knowledge. They pay less or no

attention to the combined speciﬁcation and manage-

ment of both static and procedural aspects of medi-

cal knowledge. Furthermore, they do not hold ad-hoc

mechanisms that allow errors and adverse events pre-

vention. In fact, no reasoning facilities able to ex-

ploit patient and disease status, prescribed cares and

drugs, current activities to execute, etc. are provided.

This limitations hinder clinical processes monitoring

aimed at discovering situations that could create risks

for patients during processes execution.

Two interesting healthcare information systems

that provide comprehensive framework for managing

clinical guidelines are DeGeL (Shahar et al., 2003)

and SEBASTIAN (Kawamoto and Lobach, 2005).

DeGeL is focused on providing automated support

for the speciﬁcation and implementation of clini-

cal guidelines in the treatment of patients, particu-

larly those with chronic conditions such as diabetes,

hypertension and depression. SEBASTIAN (Sys-

tem for Evidence-Based Advice through Simultane-

ous Transaction with an Intelligent Agent across a

Network) captures medical knowledge in XML doc-

uments known as Executable Knowledge Modules

(EKMs). An EKM encapsulates medical knowledge

in a machine-executable format that can be used to

generate patient-speciﬁc inferences useful for clinical

decision support (CDS). EKMs use a patient infor-

mation model based on the HL7 Reference Informa-

tion Model (HL7) , and medical concepts are prefer-

entially deﬁned using standard vocabularies included

in UMLS. These systems are mainly designed to sup-

port decisions during diagnosis or guidelines appli-

cation but do not support complex reasoning over

available knowledge for risk management scopes. A

further remarkable system is ASTI (S´eroussi et al.,

2001) because it is the only existing system that tack-

les the problem of errors prevention in prescriptions.

The ASTI project, in fact, has been focused on the

design of a guideline-based decision support system

to help general practitioners avoid prescription errors

and comply with best therapeutic practice, speciﬁ-

cally in the treatment of chronic diseases including

hypertension.

From the above discussion of related works

emerges that already existing approaches and sys-

tems: (i) have a lack of mechanisms for errors and

risks handling and prevention; (ii) do not use the same

formalism for representing and managing both medi-

cal knowledge and clinical processes, hence, they are

not able to exploit in a fully uniﬁed way declarative

and procedural knowledge during execution and mon-

itoring of clinical process; (iii) do not allow to orga-

nize clinical processes and their element as an ontol-

ogy in order to provide semantic querying and brows-

ing capabilities; (iv) do not allow to modify and cus-

tomize represented knowledge and to execute clinical

process in a ﬂexible and agile way. The system de-

scribed in the following section aims at overwhelm

these limitations.

3 THE FRAMEWORK

The Ontology-based Clinical Knowledge Represen-

tation Framework (OCKRF) allows to represent

machine-executable and ﬂexible model of declarative

(static) and procedural(dynamic) medical knowledge.

The framework is organized in tree layers as shown in

Figure 1.

The ﬁrst layer, called OCKRF meta-model is

founded on an ontology-based approach to medical

knowledge representation that allows to express in a

combined way medical ontologies and clinical pro-

cesses. The adopted approach is grounded on a meta-

model that merges expressive power of ontology and

workﬂow representation formalisms. More in detail,

the meta-model is expressed by means of a formal-

ism based on an object-oriented version of datalog

that holds typical ontology representation constructs

ICEIS 2009 - International Conference on Enterprise Information Systems

182

like: classes, relations, attributes, objects (instances).

Furthermore, querying and reasoning capabilities of

Datalog allow to query represented ontologies and ex-

ecute reasoning task over them in a semantic fash-

ion. The meta-model exploits a ﬂow-graph oriented

workﬂow modeling approach (Figures 2 and 3) in-

spired to the JPDL (JPDL). In a ﬂow-graph based

approach a workﬂow is represented by a labeled di-

rected graph whose nodes correspond to the activities

to be performed, and whose arcs describe the prece-

dences among them. The key idea, which the pre-

sented framework is based on, is that elements of the

workﬂow meta-model (i.e. processes, nodes, tasks,

events, transitions, actions, decisions) are expressed

as ontology classes. So, by using the adopted ap-

proach ontologies of clinical processes can be ob-

tained. So, by using the adopted approach: (i) clin-

ical processes and their elements can be organized as

an ontology (for instance, possible events can be de-

ﬁned as a taxonomy in which each class of event has a

speciﬁc set of attributes); (ii) workﬂow elements like

nodes, taskes, activities, events etc. are represented

in terms of ontology classes so types of their param-

eters can be other classes, hence clinical process can

be managed in a semantic fashion; (iii) decisions can

be deﬁned as reasoning tasks that involve not only the

execution state but also the knowledge represented in

the medical ontologies (for example, at each prescrip-

tion a reasoning task can check if the administered

drug is compatible with allergies and/or with the state

of the patient); (iv) special reasoning tasks aimed at

controlling possible risks and errors conditions can be

deﬁned for a process (reasoning tasks generate events

that are properly handled in the process).

The second layer is constituted by medical ontolo-

gies and clinical processes expressed by the formal-

ism provided in the ﬁrst layer. In particular: (i) med-

ical ontologies represent concepts related to different

medical domains (e.g. diseases, drugs and their inter-

actions, prescription criteria, medical examinations,

medical treatments, laboratory terms, anatomy, pa-

tients administration concepts, possible errors). Med-

ical ontologies can be obtained by importing already

existing ontologies and thesaurus or designed by hand

(for example, concepts regarding patients, wards,

hospital, could depend from the speciﬁc hospital or

ward). For instance, concepts related to breast neo-

plasm, needed in the process presented in the Sec-

tion 4.1, has been imported from the International

Classiﬁcation of Diseases (ICD10-CM), the Anatom-

ical Therapeutic Chemical (ATC) classiﬁcation sys-

tem, and the Medical Subject Headings (Mesh) Tree

Structures. (ii) clinical processes are both medical

guidelines and clinical practices depending from the

Figure 1: The Ontology-based Clinical Knowledge Repre-

sentation Framework.

speciﬁc ward. Clinical processes can by imported

from already available guideline bases or designed by

hand when they represent clinical practices followed

in a speciﬁc ward for caring a given disease. Medi-

cal ontologies and clinical processes are stored in the

system knowledge base.

Figure 2: The process meta-model.

The third layer consists of a set of medical ontolo-

gies and clinical process instances stored in a knowl-

edge base. Instances are generated during medical on-

tologies deﬁnition and/or clinical process execution.

TOWARDS A SEMANTIC SYSTEM FOR MANAGING CLINICAL PROCESSES

183

Figure 3: The nodes hierarchy.

Instances are stored in the knowledge base and made

available for querying and reasoning.

More in general, the OCKRF allows methods for:

(i) creating ontologies of clinical processes that can

be queried and explored in a semantic fashion; (ii) ex-

pressing errors and risks rules (by means of reason-

ing tasks) that can be used (during processes execu-

tion) for monitoring processes; (iii) monitoring clin-

ical processes during the execution by running rea-

soning tasks that enable to prevent errors and adverse

events that can cause risks for patients; (iv) to deﬁne

semantic Electronic Medical Records schemas (i.e.

Meta EMR); (v) analyzing acquired clinical process

instances, by means of querying and inference ca-

pabilities, in order to recognize errors and risks for

patients. For example, advanced browsing capabili-

ties of the medical knowledge base by using concept-

based queries that enables to retrieve clinical pro-

cesses and guidelines by using available medical on-

tology concepts.

The OCKRF is implemented by the DLP+ ontol-

ogy language (Ricca and Leone, 2007) that beside

complete and expressive ontology representation fea-

tures, holds also powerful ASP reasoning capabili-

ties (Eiter et al., 1997; Leone et al., 2006; Ricca and

Leone, 2007) over represented knowledge.

4 SYSTEM DESCRIPTION

In this section are described architecture (Figure 4)

and functionalities of the the Clinical Process Man-

agement System (CPMS) prototypical implementa-

tion. The prototype has been obtained by combin-

ing the JBPM engine (JPDL) and the DLV+ system

(Ricca and Leone, 2007). The prototype is designed

to follow a clinical processes life-cycle model based

on 3 phases: (i) representing (importing) medical

clinical processes and ontologies; (ii) executing and

monitoring clinical processes; (iii) acquiring, query-

ing and analyzing clinical process instances. Each

phase is implemented by an ad-hoc software module

as described in the following.

Figure 4: The CPMS architecture.

4.1 Clinical Processes Design

The Design module mainly exploits the DLV+ sys-

tem. It provides functionalities for deﬁning medical

ontologies and clinical processes by: (i) import fa-

cilities that allow to acquire already existing ontolo-

gies and guidelines; (ii) direct ”on-screen” drawing

and manual speciﬁcation functionalities based on an

ontology and process editor. The editor enables ag-

ile guidelines representation and browsing. It con-

tains a rules editor that allows to deﬁne ontology con-

straints and/or reasoning tasks used to control risks

and errors during processes execution. Process ele-

ments (e.g. nodes, tasks, decisions) that do not belong

to a speciﬁc process can be, also, represented. Ac-

quired and/or represented schemas and instances are

stored in a knowledge base and can be queried by us-

ing querying and meta-querying capabilities of DLV+

system.

In order to brieﬂy describe how the design module

works, in the following is presented, an application of

the system to a real case concerning a clinical process

for caring the breast neoplasm. The clinical process

(depicted in Figure 5) is referred to the practices car-

ried out in the oncological ward of an Italian hospi-

tal, for this reason it is not a general guideline but a

speciﬁc clinical process adopted in the domain of the

considered ward.

The clinical process is organized in the following

10 activities and sub-processes:

1. The task node acceptance models the patient en-

rollment. The patient arrives to the ward with an

already existing clinical diagnosis of a breast neo-

plasm. This node is manually executed by an on-

cologist that collect patient personal data that are

ICEIS 2009 - International Conference on Enterprise Information Systems

184

Figure 5: A clinical process for caring the breast neoplasm.

stored as new instances of the classes that describe

process and patient information.

2. The group node anamnesis represents a set of

anamnesis activities that can be executed without

a speciﬁc order. Activities in the group are: (i)

general anamnesis, in which physiological gen-

eral data (e.g. allergies, intolerances) are be-

ing collected; (ii) remote pathological anamnesis,

concerning past pathologies; (iii) recent patholog-

ical anamnesis, in which each data or result de-

rived from examinations concerning the current

pathology (or pathologies) are acquired.

3. The task node initial clinical evaluation allows to

acquire the result of an examination of the patient

by a local oncologist.

4. The decision node more clinical test requested

represents the decision of the physician about the

necessity to perform or not additional examina-

tion on the patient.

5. the group node other exams models possible ad-

ditional clinical tests. Each node of the group

correspond an exam. If requested these tests are

conducted to ﬁnd out general (or particular) con-

ditions of patient and disease not fully deducible

from the test results already available.

6. The task node therapeutic strategy deﬁnition

models guideline to follow for caring the given

neoplasm. At design time the physician picks a

guideline (among those available in the knowl-

edge base) that depends upon actual pathology

state as well as other collected patient data. The

semantic indexing strategy allows the retrieval of

guidelines by using concept-based queries. At

each guideline corresponds the prescription of re-

lated drugs. So, the selection of a guideline im-

plies computation of doses, which may depend on

patient’s biomedical parameters, such as body’s

weight or skin’s surface. Cross-checking doses at

execution time (by ad hoc reasoning tasks) is fun-

damental here, because if a wrong dose is given to

the patient the outcome could be lethal.

7. The task node informed agreement sign models

the agreement of the patient concerning under-

standing and acceptance of consequences (either

side effectsor beneﬁts) which may derivefrom the

chosen chemotherapy, and privacy agreements.

8. The sub-process node chemotherapy administra-

tion, models the guideline to execute for caring

the patient.

9. The decision node therapy ended models a control

about effect of the therapy and the possibility to

stop, continue or change cares.

10. The task node discharging models the discharging

of the patient from the ward and allow to acquire

ﬁnal clinical parameter values.

By using the rule editor, for each clinical process

(or its elements) a set of risk and error conditions can

be described in terms of ontology constraints and/or

reasoning tasks. So during both manual and auto-

matic executionof clinical processes, these conditions

can be executed in order to check possible risks and

errors that are going to happen. If a condition is veri-

ﬁed, the system generates an event that alerts the actor

that is executing the activity, so risks and errors can be

prevented.

4.2 Clinical Processes Execution

and Monitoring

The Execution & Monitoring module provides func-

tionalities for the assisted execution of clinical pro-

cesses and the acquisition of process instances. The

module is mainly constituted by the JBPM engine that

interact with DLV+ system.

Clinical process execution is performed in two

ways: (i) by using a workﬂow enactment strategy. In

this case, a process schema, designed and stored in

DLV+, is imported in JBPM and automatically ex-

ecuted involving actors that can be humans or ma-

chines (e.g. legacy systems supplying results of med-

ical examinations); (ii) by using a dynamic workﬂow

composition strategy. In this case, nodes to execute

are selected step by step by choosing the most ap-

propriate one in a given moment. Nodes are chosen

by using semantic querying capabilities of DLV+ sys-

tem and executed by JBPM. Queries allow to spec-

ify patient clinical data and each signiﬁcant informa-

tion available in medical ontologies (e.g. drug inter-

action, allergies, etc). So, queries exploit patient clin-

TOWARDS A SEMANTIC SYSTEM FOR MANAGING CLINICAL PROCESSES

185

ical data coming from anamnesis and medical exam-

ination, and each information available in the partic-

ular moment within ontologies and already executed

process activities. The execution of each activity is

performed either: (i) automatically by executing cus-

tom nodes that run java code designed to involveother

existing systems (e.g. batch acquisition of clinical re-

sults from a computer posted in a laboratory); or (ii)

manually by a web based graphical user interface that

shows forms to human process actors. During manual

execution physician and nurses can store values of ac-

tivity parameters by ﬁlling forms by hand. They can

insert the values by writing them in the related textual

ﬁelds. Special queries and reasoning tasks can be trig-

gered to user input in order to check the correctness of

entered values. The execution generates both process

and ontology instances that are stored in the DLV+

knowledge base. Since, process elements (e.g. nodes,

decisions, transitions, tasks) are represented by means

of ontology classes, process instances are in turn con-

stituted by ontology class instances. This way process

schemas and instances intrinsically constitute an on-

tology. Ad hoc extensions permit the interaction be-

tween JBPM and DLV+. So reasoning, querying and

meta-querying over schemas and available instances

are possible.

Clinical process monitoring is based on the auto-

matic execution and veriﬁcation, on live clinical pro-

cesses, of ontology constraints and reasoning tasks

modeled at design time. This way, the system can

generate events that inform the user about exceptions

and unusual or undesired behaviors. The availabil-

ity of domain medical ontologies concerning drugs’

interactions, side effects and contraindications could

reduce dramatically the probability of fatal mistakes.

Furthermore, the set of activities executed until a

given moment, the current state of a patient, etc. can

be retrieved and visualized by physician and nurse in

order to check the evolution of cares.

By considering the example introduced in the pre-

vious paragraph, at execution time, for each node of

the clinical process and for each ontology concept in-

volved in the process, instances are created. For ex-

ample results of the anamnesis and the check-up are

stored as instances of the noticed pathologies where

attributes are ﬁlled with the related observed values.

During guideline execution, due to drug inherent tox-

icity, before each dose administration a check about

drug administration is performed by analyzing pro-

cess evolution, patient conditions, and information

about the particular drug contained in the drug ontol-

ogy (e.g. maximum absolute ratings for certain drug

in a ”whole life” or other ratings relative to biomed-

ical parameters of the patient). Problems related to

drugs administration can be immediately ﬂagged as

dangerous, or even lethal, and then associated risk is

notiﬁed to the oncologist.

4.3 Clinical Processes Analytics

The Analytics module aims at allowing analysis of

the clinical processes instances after their acquisi-

tion. The execution of clinical processes, in fact,

makes available process instances organized as an on-

tology. This way a large amount of semantically en-

riched data becomes available for querying and re-

trieval. Analysis are possible by reports composed by

using semantic querying capabilities of DLV+ applied

to process instances contained in the system knowl-

edge base.

5 CONCLUSIONS AND FUTURE

WORK

This work describes the prototypical implementa-

tion of a semantic clinical process management sys-

tem founded on an ontology-based clinical knowl-

edge representation framework. The system allows to

jointly represent both medical knowledge (by means

of medical ontologies) and clinical processes (by

means of a ontology-based workﬂow representation

approach). The system allows: (i) creating ontolo-

gies of clinical processes that can be queried and ex-

plored in a semantic fashion; (ii) expressing errors

and risks rules (by means of reasoning tasks) that

can be used (during processes execution) for mon-

itoring processes; (iii) manual process execution in

which each clinical activity to perform in a given

moment is chosen by physician on the base of the

current conﬁguration of patient and disease parame-

ters; (iv) automatic execution by means of the enact-

ment of a designed process schema; (vi) automatic

monitoring of clinical processes execution by running

ad-hoc reasoning tasks that exploit knowledge repre-

sented in medical ontologies and clinical processes

to check possible error and risk causes. The execu-

tion of clinical processes allows to acquire process

and ontology instances that are stored in a knowledge

base. Acquired instances can be analyzed by means

of reports obtained by querying the system knowl-

edge base. Currently the system has been applied to

the clinical process depicted in Figure 5. The practi-

cal application shown that the system enables better

health care decision making capabilities that allows

health care professionals to improve risks and errors

prevention.

ICEIS 2009 - International Conference on Enterprise Information Systems

186

The main challenging future research and devel-

opment problems are the deﬁnition of an efﬁcient

query engine working on the conjunct representation

of workﬂows and ontologies and the deﬁnition of

further monitoring and analytical technics. In par-

ticular, existing process instances can be organized

in datasets that can be analyzed by means of data

and workﬂowmining techniquesaimed at discovering

patterns related to risks and adverse events. In partic-

ular, workﬂow mining techniques are able to classify

clinical process instances on the base of their behav-

ior and, possibly, to suggest new schemas (precess re-

engineering) able to reduce risks for patients and the

impact of errors. Furthermore, new import/exportfea-

tures will be implemented in order to make the system

compliant with already existing health care informa-

tion systems standards.

REFERENCES

CCAM. Classiﬁcation Commune des Actes M´edicaux

(CCAM). http://www.ccam.sante.fr/.

CPT. Current Procedural Terminology (CPT): A

Registered trademark of the American Med-

ical Association (AMA). http://www.ama-

assn.org/ama/pub/category/3113.html.

Eiter, T., Gottlob, G., and Mannila, H. (1997). Disjunctive

Datalog. ACM Transactions on Database Systems,

22(3):364–418.

GLIF. Guide Line Interchange Format (GLIF). http://

www.glif.org.

HL7. Health Level Seven (HL7). http://www.hl7.org/.

ICD. International Classiﬁcation of Diseases, Tenth

Revision, Clinical Modiﬁcation (ICD-10-CM).

http://www.cdc.gov/nchs/about/otheract/icd9/icd10cm.htm.

JPDL. jBPM Process Deﬁnition Language (jPDL). http://

www.jboss.org/jbossjbpm/jpdl/.

Kawamoto, K. and Lobach, D. F. (2005). Design, imple-

mentation, use, and preliminary evaluation of sebas-

tian, a standards-based web service for clinical deci-

sion support. In AMIA Symp., pages 380–4.

Leone, N., Pfeifer, G., Faber, W., Eiter, T., Gottlob, G.,

Perri, S., and Scarcello, F. (2006). The dlv system for

knowledge representation and reasoning. ACM Trans.

Comput. Logic, 7(3):499–562.

LOINC. Logical Observation Identiﬁers Names and Codes

(LOINC). http://loinc.org/.

MESH. Medical Subject headings (MeSH). http://

www.nlm.nih.gov/mesh/2008/index.html.

Musen, M. A., Tu, S. W., Das, A. K., and Shahar, Y. (1996).

Eon: A component-based approach to automation of

protocol-directed therapy. Journal of the American

Medical Informatics Association, 3(6):367–388.

NCI-EVS. NCI Enterprise Vocabulary Services (EVS).

http://ncicb.nci.nih.gov/NCICB/infrastructure/

cacore

overview/vocabulary.

OPCS-4. The Ofﬁce of Population Censuses and

Surveys Classiﬁcation of Surgical Opera-

tions and Procedures, 4th Revision (OPCS-4).

http://www.openclinical.org/medTermOPCS4.html.

Peleg, M., Ogunyemi, O., Tu, S., Boxwala, A. A., Zeng, Q.,

Greenes, R. A., and Shortliffe, E. H. (2001). Using

features of arden syntax with object-oriented medical

data models for guideline modeling. In in American

Medical Informatics Association (AMIA) Annual Sym-

posium, 2001, pages 523–527.

Prot´eg´e. Ontology Editor and Knowledge Acquisition Sys-

tem. http://protege.stanford.edu.

Pryor, T. A. and Hripcsak, G. (1993). The arden syntax for

medical logic modules. Journal of Clinical Monitor-

ing and Computing, 10(4):215–224.

Ricca, F. and Leone, N. (2007). Disjunctive logic program-

ming with types and objects: The dlv+ system. J. Ap-

plied Logic, 5(3):545–573.

Sackett, D. L., Rosenberg, W. M. C., Gray, M. J. A.,

Haynes, B. R., and Richardson, S. W. (1996). Ev-

idence based medicine: what it is and what it isn’t.

BMJ, 312(7023):71–72.

S´eroussi, B., Bouaud, J., Dr´eau, H., Falcoff, H., Riou, C.,

Joubert, M., Simon, C., Simon, G., and Venot, A.

(2001). Asti: A guideline-based drug-ordering system

for primary care. MEDINFO.

Shahar, Y., Young, O., Shalom, E., Mayafﬁt, A.,

Moskovitch, R., Hessing, A., and Galperin, M.

(2003). Degel: A hybrid, multiple-ontology frame-

work for speciﬁcation and retrieval of clinical guide-

lines. In AIME, pages 122–131.

SNOMED. Systematized Nomenclature of

Medicine-Clinical Terms (SNOMED CT).

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

Sutton, D. R. and Fox, J. (2003). The syntax and semantics

of the proforma guideline modeling language. JAMIA,

10(5):433–443.

UMLS. Uniﬁed medical Language System (UMLS). http://

www.nlm.nih.gov/research/umls/.

WHO. World Health Organization (WHO). http://

www.who.int/classiﬁcations/en/.

TOWARDS A SEMANTIC SYSTEM FOR MANAGING CLINICAL PROCESSES

187