ENABLING INTEROPERABILITY THROUGH AN ONTOLOGY

APPROACH IN THE HETEROGENEOUS DOMAINS OF

COMPLEX CHRONIC CONDITIONS

Tara Sampalli

, Michael Shepherd

and Jack Duffy

Capital Health, Fall River, Nova Scotia, Canada

Faculty of Computer Science, Dalhousie University, Halifax, Nova Scotia, Canada

Keywords: Ontologies for complex chronic conditions.

Abstract: Complex and chronic health conditions have domain knowledge that is multidisciplinary, inconsistent, non-

standardized and poorly categorized making them heterogeneous. Consequently, challenges for collabora-

tive care management are widely prevalent due to lack of interoperability. Ontologies have come to the

forefront as effective method to improve interoperability in a domain due to their ability to enable higher

levels of specification. The primary objective of this study was to develop, test and evaluate a model and a

methodology for creating ontologies in heterogeneous domains of complex conditions, an area where there

is great paucity for research. The methodology in this research applied a two-staged approach for enabling

interoperability in the heterogeneous domain of two complex chronic health conditions, namely, multiple

chemical sensitivity and chronic pain. Four hundred and eight and three hundred forty five multidisciplinary

concepts were specified in the profile ontologies for multiple chemical sensitivity and chronic pain. A test-

ing and an evaluation process conducted in this research demonstrated that a high percentage of the multid-

isciplinary clinicians (>80%) agreed on the overall usefulness of the ontologies in improving the collabora-

tive environment. The results from the research are promising in terms of the potential applications of on-

tologies in heterogeneous knowledge domains.

1 INTRODUCTION

Chronic conditions such as multiple chemical sensi-

tivity (MCS), chronic fatigue syndrome (CFS) and

fibromyalgia (FM) are a significant burden to the

healthcare system. In a report by Statistics Canada,

at least 5% of Canadians have symptoms that cannot

be medically explained (Verhaak et al., 2006). Mul-

tidisciplinary care teams have come to the forefront

as an effective management strategy for these condi-

tions (Dysvik et al., 2005); (Fox et al., 2008). Stu-

dies have shown the consequences of poor commu-

nication among multidisciplinary care providers

resulting in poor care experiences for patients and

errors in care management such as repetitive or

redundant medical tests, misdiagnoses, delayed care

and inaccurate treatment plans (Pace et al., 2004);

(Schoen et al., 2008); (Kennedy, 2008). The domain

knowledge for these conditions is obscure with lack

of consensus among experts, non-standardized and

multidisciplinary. There is a clear need to enable

interoperability and better communication among

multidisciplinary care providers to improve care

environment for patients (Fox et al., 2008). An on-

tology approach to capturing domain knowledge is

explored as a possible modus operandi to organizing

knowledge in a heterogeneous knowledge base.

2 LITERATURE REVIEW

Ontologies have gained importance in recent years

as a knowledge management platform in many areas

including health care (Dominigue et al., 2001); (Ba-

neyx et al., 2005); (Mostefai et al., 2006). Ontolo-

gies are preferred to conventional classifications due

to the higher level of expressiveness that is possible

in describing concepts and their relationships

(Dominigue et al., 2001). Ontologies have been

typically developed in stable knowledge domains

(Lin et al., 2006); (Larson and Martone, 2009).

Challenges in developing ontologies in heterogene-

Sampalli T., Shepherd M. and Duffy J..

ENABLING INTEROPERABILITY THROUGH AN ONTOLOGY APPROACH IN THE HETEROGENEOUS DOMAINS OF COMPLEX CHRONIC

CONDITIONS.

DOI: 10.5220/0003744700460052

In Proceedings of the International Conference on Health Informatics (HEALTHINF-2012), pages 46-52

ISBN: 978-989-8425-88-1

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

ous domains have been discussed in the literature. In

this paper, heterogeneous domains are defined as

domains that are poorly categorized, multidiscipli-

nary, non-standardized and inconsistent.

Dominigue et al. (2001) identify the key re-

quirement for an ontology approach to knowledge

management as a community’s perspectives being

stable on an issue with “well defined roles”, “speci-

fied criteria” and “codified procedures”. Challenges

related to developing ontologies when there is a lack

of consensus in a community are discussed in the

subsequent paragraphs.

A study by Larson and Martone (2009), the chal-

lenges of formalizing knowledge for neuroscience

were explored. The authors claimed that formalizing

knowledge about poorly understood biological sys-

tems presents many obstacles to the development of

ontologies. This study highlighted the importance of

developing a layer of standardization prior to at-

tempting higher level specification such as the crea-

tion of ontologies in the domain.

In a study by Lin et al. (2006), the challenges of

a mental health group of professionals working with

emerging knowledge was discussed. This study

describes the challenges and importance of building

knowledge through ontologies in heterogeneous

situations. This study presented the preliminary

challenges that exist in the knowledge capture for a

domain that has obscure definitions, lack of consen-

sus, unstructured data, inconsistent use of vocabu-

lary and assessment scales. A significant challenge

encountered in this work was to bring structure to

knowledge that continues to be generated in an ad

hoc manner.

In a study by Qin and Paling (2001), the impor-

tance of developing ontologies in heterogeneous

domain was examined. The research describes the

creation of an ontology from a well defined and well

used controlled vocabulary in order to provide a

higher level of semantics to the concepts in the vo-

cabulary. Digital objects, such as those in the Gate-

way to Educational Materials (GEM ontology) en-

compass multiple dimensions of characteristics

which often play important roles for users in search

of precise information in an efficient manner. The

authors suggest that a conventional cataloguing code

will be inadequate to describe these details in a les-

son plan, as many of these elements do not even

exist in the vocabulary. In this study, the authors

developed an ontology with the intention of adding

another layer of semantic operability to the termi-

nologies found in controlled vocabularies.

The review of literature has highlighted the fol-

lowing areas of research needs in the domain of

complex chronic conditions and the application of

ontologies in heterogeneous domains. Studies have

discussed the challenges of collaboration among

clinicians in the domain of complex chronic condi-

tions and the need to develop standardization and

interoperability in this domain. The potential of

ontologies is well recognized in the literature as

effective platform to enable interoperability and

standardization in a domain and finally researchers

have recognized the challenges of developing on-

tologies in heterogeneous domains. Overall, the

review has indicated that there is no research to-date

that has developed ontologies in the heterogeneous

domain of complex chronic conditions.

In this paper, we have discussed the development

of ontologies in the heterogeneous domains of two

complex chronic health conditions, multiple chemi-

cal sensitivity (Bartha, 1999) and chronic pain (Peng

et al., 2008). Both conditions require a multidisci-

plinary care management scheme and have all the

characteristics of heterogeneity defined in this paper.

3 METHODOLOGY

Model and methodology proposed in this research to

develop a layer of pragmatic interoperability in the

heterogeneous domain included a two-staged ap-

proach.

The first stage of the two-staged approach in-

cluded the creation of a standardized and controlled

clinical vocabulary. SNOMED CT® (2008), a

widely used reference terminology was used to stan-

dardize the concepts and terminologies found in the

patient charts. A pragmatic approach (Carlile, 2002)

was applied in the development of the controlled

vocabulary as the domain knowledge is heterogene-

ous.

The development of the controlled vocabulary

involved the developing of standardized and con-

trolled clinical vocabularies at the levels of syntac-

tic, semantic and pragmatic interoperability. The

method for creating the controlled vocabulary was

driven by the purpose of generating the goal and

usage of the vocabulary: chart audit and interviews

with experts to identify key concepts in the domain

of the complex condition (syntactic), standardization

of the vocabulary (semantic), and testing and eval-

uation of the vocabulary by the users (pragmatic).

The chart audit and interviews with experts helped

generate the vocabulary. SNOMED CT® was used

as a reference terminology to standardize the terms

ENABLING INTEROPERABILITY THROUGH AN ONTOLOGY APPROACH IN THE HETEROGENEOUS

DOMAINS OF COMPLEX CHRONIC CONDITIONS

retrieved in the chart audit process. The re-coding of

patient profiles, evaluation and feedback from the

domain experts tested and evaluated the vocabulary.

A further step in the evaluation included feedback

from clinicians in the community.

The two-staged approach applied in this re-

search is shown in Figure 1.

Figure 1: The two-staged methodology applied in the

research.

The second stage of the two-staged approach

was the creation of an ontology in the heterogeneous

domain consisting of 3 phases: Development, testing

and evaluation. The development phase included the

experts in the domain specifying and organizing the

knowledge in the domain. This phase primarily drew

the knowledge from the controlled vocabulary. The

testing phase included the clinicians browsing the

profile ontology developed in this research to exam-

ine the concepts in the ontology, the relationships

between concepts, concept attributes and the indi-

viduals populated in the ontology. Following this

was an evaluation phase that included feedback from

the domain experts on the overall usefulness of the

ontology in patient care with emphasis on usefulness

from a health discipline perspective, from other

health disciplines and the multidisciplinary nature of

interactions captured in the ontology.

3.1 Development of the Ontology

The generic framework for the development of an

ontology was divided into three important phases: a

specification phase, a conceptualization phase, and

an implementation phase (Noy and McGuinness,

2000).

3.1.1 Specification Phase

Establishing the goal and scope was the first step in

developing the ontology. The goal of an ontology

determines the overall objective for developing an

ontology. The goal of the profile ontology devel-

oped in this study was to create a comprehensive

hierarchical controlled vocabulary and a representa-

tion of the multidisciplinary and multidimensional

relationships that exist among the concepts in the

controlled vocabulary. The scope of the ontology for

this research was maintained in the first layer of

organization, that is, in the domain of patient pro-

files. A patient profile or problem list generation is

typically the starting point of a care management

scheme for patients. The elements in this ontology

were maintained on the knowledge that existed in

this domain.

3.1.2 Conceptualization Phase

The conceptualization of the ontology commenced

with the controlled vocabulary that specified the

concepts that existed in the patient profile domain

for complex health condition. The key concepts

were explicitly related by establishing relationships

and attributes in the domain. Multidisciplinary inte-

ractions in the management of symptoms were spe-

cified in the ontology through relations and

attributes. Multidisciplinary classes were also

created in the profile ontology which showed the

involvement of various grouping of clinicians in the

management of a specific grouping of multidiscipli-

nary symptoms for patients. The knowledge in this

phase was derived from the patient charts and from

the domain experts. Instances from one-hundred

patient profiles were populated in the ontology.

3.1.3 Implementation Phase of the Ontology

Protégé 3.4.2 was used to implement the patient

profile ontology (Knublauch, 2004). The profile

ontology was exported into the Web Ontology Lan-

guage (OWL). A consistency check of the classes in

the ontology was conducted. Consistency checking

helped detect classes that cannot have instances.

The implementation phase also included the

evaluation of the ontology by domain experts for

accuracy, completeness and usefulness of the

knowledge represented in the ontology. The evalua-

HEALTHINF 2012 - International Conference on Health Informatics

tion phase included the clinicians browsing the on-

tology using an ontology browser. They browsed

various aspects of the ontology such as the classifi-

cation scheme, multidisciplinary relations between

concepts, instances, and standardization of concepts.

Google ontology browser was used by clinicians to

browse the ontology (Horridge et al., 2006). They

provided feedback on the usefulness of the ontology

through a survey questionnaire. Specifically, they

offered feedback on the overall usefulness of the

ontology, the relevance of the ontology in the con-

text of patient care and the value of shared knowl-

edge in the multidisciplinary domain. Individuals or

instances are used in the profile ontology to present

list of concrete concepts of relevance for each class.

4 RESULTS

Two complex and chronic health conditions, name-

ly, multiple chemical sensitivity and chronic pain

were selected to test the viability of the proposed

methodology in heterogeneous knowledge systems.

One-hundred patient charts were reviewed, 9 do-

main experts and 36 community clinicians partici-

pated in the development of the MCS controlled

vocabulary. One-hundred patients, 8 domain experts

and 42 community clinicians participated in the

development of the chronic pain vocabulary. Seven

domain experts from MCS group and 6 from chronic

pain group participated in the development of the

ontologies.

4.1 Profile Ontologies for MCS and

Chronic Pain

The ontologies present a detailed taxonomic over-

view of the domain of complex health conditions.

The profile ontology for MCS contained 408 classes

describing the profile concepts for the condition of

MCS. At the basic level there are five relevant su-

per-classes under the primary areas of health focus

identified for the condition of MCS: Medical, Physi-

cal, Psychosocial, Rehabilitation and Nutrition (Fig-

ure 2). The profile ontology includes definitions of

over 70 properties, 46 data and 30 object properties.

The profile ontology for chronic pain contained

345 classes describing the profile concepts for the

condition of chronic pain. At the basic level there

are three relevant super-classes under the primary

areas of health focus identified for the condition of

chronic: Medical, Physical and Psychosocial. The

profile ontology includes definitions of over 80

properties, with 51 data and 38 object properties.

The profile ontologies contained explication of

all concepts included in the ontology such as the

multidisciplinary nature of patient profile, the man-

agement scheme and the various concepts under

each area of health focus. The properties in the on-

tologies introduce relations among concepts. A pa-

tient HasOrganization and the organization are

inversely linked to the class Patient by HasPatient.

The class Profile is linked to the class Management

Scheme by property hasCollaborativeManagement.

The class Psychosocial Profile is linked to the man-

agement scheme by property ManagementRequired

which has individual dietitian_referral or physi-

cian_referral.

Figure 2: View of MCS ontology in Protégé showing the

five super classes.

Standardized concepts are specified with their

SNOMED CT ID number (Concept Unique Identi-

fier) and with a list of synonyms. Class Fatigue has a

SNOMED CT concept ID of 84229001 with parent

concept being Energy and Stamina and synonyms

Weariness and Tiredness as shown in shown in

Figure 3. Examples of more intricate concepts that

benefit from standardization and consistency rele-

vant to this health condition include heightened

visual perception, heightened auditory perception,

emotional hypersensitivity, impairment of balance,

emotional regulation or emotional state finding and

hypervigilant behaviour. Similarly in the chronic

pain ontology, standardized concepts are specified

ENABLING INTEROPERABILITY THROUGH AN ONTOLOGY APPROACH IN THE HETEROGENEOUS

DOMAINS OF COMPLEX CHRONIC CONDITIONS

with their SNOMED CT ID number (Concept

Unique Identifier) and with a list of synonyms. Class

Lumbar spine - tender has a SNOMED CT® con-

cept ID of 298673002 with parent concept being

Finding of sensation of lumbar spine with finding

site as lumbar spine structure.

Figure 3: Query of a symptom “Fatigue” in the profile

ontology.

Multidisciplinary classes were created in the pro-

file ontologies which shows the involvement of

various grouping of clinicians in the management of

the multidisciplinary symptoms for patients with a

diagnosis of MCS. For instance, Multidisciplinary

class A involved management by a physician, a

nurse, a psychologist or a psychotherapist, a voca-

tional counselor and a physiotherapist or an occupa-

tional therapist. Multidisciplinary class L involved

management by a physician, a nurse and a dietitian.

The knowledge of these classes demonstrated to the

experts that despite the diagnosis, the management

schemes were driven by the presenting symptoms

which could greatly vary for patients. This was

evident in the knowledge that was retrieved from the

100 patient charts.

4.2 Evaluation of the Ontologies

Google ontology browser (Horridge et al., 2006)

was used by clinicians to browse the ontology and

offer their evaluation. They viewed the individual

patient profiles, multidisciplinary information rele-

vant to their discipline and other disciplines, in-

depth query of symptoms and their management

scheme.

Query of a symptom such as Lumbar spine –

tender shows the number of patients with the symp-

toms and the super class of the concept in the ontol-

ogy. The instances in profiles show the multifaceted

nature of symptoms as substantiated under each area

of health focus that exist in the domain of a patient.

Pain symptom as presented in the patient charts

has been viewed in the patient charts by a psycho-

therapist, physician or physiotherapist from various

angles of importance such as pattern of pain, ana-

tomical site or in relation to the pain threshold. Fig-

ure 4 shows the view of a patient profile that shows

the multidisciplinary care involved in the manage-

ment of Pain symptom.

Figure 4: Multidisciplinary interactions in the categoriza-

tion of Pain.

The clinicians also viewed information on vari-

ous symptoms including the profiles under which a

symptom was categorized, the number of patients

that had a symptom and the standardization informa-

tion for the symptoms.

Figure 5 shows the evaluation of the ontologies

by the domain experts of both groups. The ontology

had > 80% of agreement from the experts on its

usefulness in direct patient care. The use of the mul-

tidisciplinary classes in the ontology brought a high-

er level of agreement from both groups of experts

(62%). The ontology had a consistently small to

moderate percentage of clinicians showing strong

agreement on its usefulness on all categories of the

questionnaire. The ontology also had a very small

percentage of disagreement on all categories of the

survey questionnaire.

Figure 5: Evaluation of the ontologies.

HEALTHINF 2012 - International Conference on Health Informatics

5 DISCUSSION AND

CONCLUSIONS

A novel methodology and model has been presented

in this research for the development of ontologies in

heterogeneous knowledge domains. The broad ob-

jective of the research was to enhance communica-

tion in the multidisciplinary care management of

chronic, complex and lesser known health condi-

tions. The ontology approach was selected to de-

velop consistency, standardization, organization and

interoperability of domain knowledge with the broad

goal of improving collaboration and communication

for multidisciplinary clinicians involved in the care

of patients with complex chronic conditions.

The development of the profile ontologies in this

study was divided into three phases: specification,

conceptualization and implementation (Noy and

McGuinness, 2000). The methodology includes

several key components or criteria that were identi-

fied in past research such as acknowledging the

heterogeneous nature of the domain knowledge

(Larson and Martone, 2009) involving clinicians

(experts and non-experts) in the process of devel-

opment and evaluation and exploring the potential of

the study by testing it in clinical workflow (Lin et

al., 2006). However there are several limitations to

this research such as the scope being limited to the

domain of patient profile information, a convenience

sample of participants, size of the sample, the fact

that the potential of the boundary objects in improv-

ing communication or collaboration among clini-

cians or the impact on patient care was not explored.

The results do indicate that this direction of research

has significant potential and requires further explo-

ration.

An ontology can reach a wider audience and has

been deliberately selected to explicate the knowl-

edge of lesser known and complex health condi-

tions. Ontologies provide a pragmatic interoperable

format for collaborative sharing of knowledge

across communities of practice. The ontology has

the potential to get richer as more users contribute

new knowledge and as more patient instances are

populated in the ontology. The overall agreement

shown by experts in this study is very promising for

the use of ontologies in the heterogeneous domains

of complex health conditions.

REFERENCES

Baneyx, A., Charlet, J., Jaulent, M., 2005. Building an

ontology of pulmonary diseases with natural language

processing tools using textual corpora. Studies In

Health Technology And Informatics.

Bartha, 1999. Multiple chemical sensitivity: A 1999 con-

sensus. Archives of Environmental Health, 54(3), 147.

Retrieved April 20, 2003, from InfoTRAC database.

Carlile, P. R., 2002. A pragmatic view of knowledge and

boundaries: Boundary Objects in new product devel-

opment. Organization Science, 13 (4): 442-455.

Domingue, J., Motta, E., Shum, S. B., Vargas-Vera, M.

and Kalfoglou, Y., 2001. Supporting ontology driven

document enrichment within communities of practice.

In: 1st International Conference on Knowledge Cap-

ture.

Dysvik, E., Natvig, G. K., Eikeland, O. J., Brattberg, G.,

2005. Results of a multidisciplinary pain management

program: a 6- and 12-month follow-up study. Reha-

bilitation nursing: the official journal of the Associa-

tion of Rehabilitation Nurses. Vol. 30, No. 5, 198-206.

Fox, R., Sampalli, T., Fox., J., 2008. Measuring health

outcomes of a multidisciplinary care approach in indi-

viduals with chronic environmental conditions using

an abbreviated symptoms questionnaire. Journal of

Multidisciplinary Healthcare. December 2008.

Horridge, M., Drummond, N., Goodwin, J., Rector, A.,

Stevens, R. and Wang, H., 2006. OWL experiences

and directions workshop. Retrieved from

http://www.webont.org/owled/2006/acceptedLong/sub

mission_9.pdf

Kennedy, 2008. Clinical documentation improvement in

MS-DRGs as a strategy for compliance: facilities may

consider clinical documentation audits to look for cod-

ing errors. Journal of Healthcare Compliance.

Knublauch, H., 2004. "The Protégé OWL Plugin," in 7th

International Protégé Conference, Bethesda, MD.

Larson, S. D. and Martone, M. E., 2009. Ontologies for

neuroscience: What are they and what are they good

for? FrontNeurosci. May; 3(1):60

Lin, Y., Poschen, M., Procter, R., 2006. Ontology as a

social-technical process: a case study. ORA Confer-

ence/Workshop Paper.

Mostefai S, Bouras A and Batouche M., 2006. Effective

collaboration in product development via a common

sharable ontology. International journal of computa-

tional intelligence.

Noy, N. F., and Mcguinness, D., 2000. Ontology devel-

opment 101: A guide to creating your first ontology.

Stanford KSL Technical Report KSL-01-05.

Pace, W. D., Dickinson, L. M., Staton, E. W., 2004. Sea-

sonal variation in diagnoses and visits to family physi-

cians. Ann Fam Med, Vol.2, 411–7.

Peng, P., Stinson, J. N., Choiniere, M., Dion, D., Intrater,

H., LeFort, S., Lynch, M., Ong, M., Rashiq, S.,

Tkachuk, G., Veillette, Y., 2008. Role of health care

professionals in multidisciplinary pain treatment fa-

cilities in Canada. Pain Res. Manag. Nov-Dec: 13(6):

484-8.

Qin, J., and Paling., 2001. Converting a controlled vo-

cabulary into an ontology: the case of GEM.

ENABLING INTEROPERABILITY THROUGH AN ONTOLOGY APPROACH IN THE HETEROGENEOUS

DOMAINS OF COMPLEX CHRONIC CONDITIONS

Information Research, 2001; 6(2)

Schoen, C., Osborn, R., How, S. K. H., Doty, M. M. and

Peugh, J., 2008. In Chronic Condition: Experiences of

Patients with Complex Health Care Needs, in Eight

Countries, Health Affairs Web Exclusive, 1-16

SNOMED CT® @ IHSTDO, 2008. SNOMED CT Clini-

cal Terms Basics. Retrieved from

http://www.ihtsdo.org/fileadmin/user_upload/Docs_01

/Recourses/Introducing_SNOMED_CT/SNOMED_C

T_Basics_IHTSDO_Taping_Aug08.pdf

Verhaak, P. F., Meijer, S. A., Visser, A. P., Wolters, G.,

2006. Persistent presentation of medically unexplained

symptoms in general practice. Fam Pract. Vol. 23,

414–420.

HEALTHINF 2012 - International Conference on Health Informatics