A HYBRID METHODOLOGY FOR CONSUMER-ORIENTED

HEALTHCARE KNOWLEDGE ACQUISITION

Elena Cardillo, Andrei Tamilin and Luciano Serafini

FBK-IRST, via Sommarive 18, 38100 Povo (TN) , Italy

Keywords: Medical Terminology, Medical Vocabulary Acquisition, Knowledge Acquisition, Consumer Healthcare.

Abstract: In spite of the improvements in Healthcare Informatics in answering consumer needs, it is still difficult for

laypersons who do not have a good level of healthcare literacy, to find, understand, and act on health

information. This is due to the communication gap which still persists between specialized medical

terminology used by healthcare professionals and “lay” medical terminology used by healthcare consumers.

So there is a need to create consumer-friendly terminologies reflecting the different ways consumers and

patients express and think about health topics. An additional need is to map these terminologies with

existing clinically-oriented terminologies. Following this direction, this work suggests a hybrid

methodology to acquire consumer health terminology for creating a Consumer-oriented Medical Vocabulary

for Italian that mitigates this gap. This resource could be used in Personal Health Records to provide

translation, search, and classification services, helping users to improve access to their healthcare data. In

order to evaluate this methodology we mapped “lay” terms with standard specialized terminologies to find

overlaps. Results showed that our methodology of acquisition provided many “lay” terms that can be

considered good synonyms for medical concepts.

1 INTRODUCTION

With the advent of the Social Web and Healthcare

Informatics technologies, we can recognize that a

linguistic and semantic discrepancy still exists

between specialized medical terminology used by

healthcare providers or professionals, and the so

called “lay” medical terminology used by healthcare

consumers. The medical communication gap became

more evident when consumers started to play an

active role in healthcare information access. In fact

they have become more responsible for their

personal healthcare, exploring information sources

on their own, consulting decision-support healthcare

sites on the web, and using patient-oriented

healthcare systems, which allow them to directly

read and interpret clinical notes or test results and to

fill in their Personal Health Record. During this

disintermediated interaction consumers can use only

their own knowledge, experience and preferences,

and this can often generate a wrong inference of the

meaning of a term, or the mis-association of a term

with its context (Zeng and Tse, 2006).

To help consumers fill this gap, the challenge is

to sort out the different ways consumers

communicate within distinct discourse groups and

map the common, shared expressions and contexts

to the more constrained, specialized language of

healthcare professionals. Though much effort has

been spent on the creation of medical resources such

as terminologies or classification systems, used

above all to help physicians in filling in Electronic

Health Records, there is little work based on the use

of consumer-oriented medical terminology, and in

addition most existing studies have been done only

for English.

A consumer-oriented medical terminology can

be defined as a “collection of forms used in health-

oriented communication for a particular task or

need by a substantial percentage of consumers from

a specific discourse group and the relationship of

the forms to professional concepts” (Zeng and Tse,

2006). Such terminology can be mainly used for

three bridging roles between consumers and health

applications: a) Information Retrieval, to facilitate

automated mapping of consumer-entered queries to

technical terms, producing better search results; b)

Cardillo E., Tamilin A. and Seraﬁni L. (2009).

A HYBRID METHODOLOGY FOR CONSUMER-ORIENTED HEALTHCARE KNOWLEDGE ACQUISITION.

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development, pages 64-71

DOI: 10.5220/0002303900640071

 SciTePress

Translation of Medical Records, supplementing

medical jargon terms with consumers-

understandable names to help patients interpretation;

c) Health Care Applications, to help integrating

different medical terminologies providing automated

mapping of consumer expressions to technical

concepts (e.g. querying for “short of breath” and

receiving information also for the concepts

“dyspnea”).

Given this scenario, the present work proposes a

hybrid methodology for the acquisition of consumer-

oriented medical knowledge and “lay” terminology

expressing particular medical concepts, such as

symptoms and diseases, for consequent creation of a

Consumer-oriented Medical Vocabulary for Italian.

We are particularly interested in performing analysis

of the clinical mapping between this consumer-

oriented terminology and the more technical one

used in standards Medical Classification Systems or

Nomenclatures. This resource could be integrated

with existing lexical and semantic medical

resources, and used in healthcare systems, like

Personal Health Records, to help consumers during

the process of querying and accessing healthcare

information, so as to bridge the communication gap.

2 BACKGROUND

2.1 Consumer-oriented Medical

Terminologies

Over the last two decades research on Medical

Terminologies has become a popular topic and the

standardization efforts have established a number of

terminologies and classification systems such as

UMLS Metathesaurus

, SNOMED International

ICD-10 (International Classification of Diseases)

and ICPC-2 (International Classification of Primary

Care)

, as well as conversion mappings between

them to help medical professionals in managing and

codifying their patients health care data. They

concern, in fact, “the meaning, expression, and use

of concepts in statements in the medical records or

other clinical information systems” (Rector, 1999).

Despite the wide use of these terminologies, as we

have already mentioned, the vocabulary problem

continues to plague not only health professionals and



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

http://www.ihtsdo.org/snomed-ct/

http://www.who.int/classifications/icd/en/

http://www.fmrc.org.au/icpc2/

their information systems, but also consumers and in

particular laypersons, who are the most damaged by

the increased communication gap.

To respond consumer needs to support personal

healthcare decision-making, during the last few

years, many researchers have labored over the

creation of lexical resources that reflect the way

consumers/patients express and think about health

topics. One of the largest initiatives in this direction

is the Consumer Health Vocabulary Initiative

, by Q.

Zeng and colleagues at Harvard Medical School,

resulted in the creation of the Open Access

Collaborative Consumer Health Vocabulary (OAC

CHV) for English. It includes lay medical terms and

synonyms connected to their corresponding

technical concepts in the UMLS Metathesaurus.

They combined corpus-based text analysis with a

human review approach, including the identification

of consumer forms for “standard” health-related

concepts. Also (Soergel et al., 2004) tried to create

such a vocabulary to identify consumer medical

terms and expressions used by lay people and health

mediators. In fact they associated a Mediator

Medical Vocabulary with the consumer-oriented

one, and mapped them to a Professional Medical

Vocabulary. Even in this case the standard

terminology used for mapping was UMLS.

These and other similar studies examined large

numbers of consumer utterances (i.e., hundreds of

thousands of tokens) and consistently found that

between 20% and 50% of consumer health

expressions were not represented by professional

health vocabularies. Furthermore, a subset of these

unrepresented expressions underwent human

review. In most of these cases they performed

automatic term extraction from written texts, such as

healthcare consumer queries on medical web sites,

postings and medical publications. An overview of

all these studies can be found in (Keselman et al.,

2008).

It is important to stress that there are only few

examples of the applications as far as these initiative

are concerned. For example, in (Kim et al., 2007)

we find an attempt to face syntactic and semantic

issues in the effort to improve PHR readability,

using the CHV to map content in Electronic Health

Records (EHR) and PHR; (Zeng et al., 2007)

designed and implemented a prototype text

translator to make EHR and PHR more

comprehensible to consumers and patients. On the



http://www.consumerhealthvocab.org

A HYBRID METHODOLOGY FOR CONSUMER-ORIENTED HEALTHCARE KNOWLEDGE ACQUISITION

other hand (Rosembloom et al., 2007) developed a

clinical interface terminology, a systematic

collection of healthcare-related phrases to support

clinicians' entries of patient-related information into

computer programs such as clinical “note capture”

and decision support tools, facilitating display of

computer-stored patient information to clinician-

users as simple human-readable texts.

Concerning multilingual consumer-oriented

health vocabularies, we can mention the initiative of

the European Commission Multilingual Glossary of

Popular and Technical Medical Terms

, in nine

European languages, but it is a limited medical

vocabulary for medicinal product package inserts

accessible to consumers. In fact, it consists of a list

of 1,400 technical terms frequently encountered in

inserts, with corresponding consumer terms in all the

languages of the European Community. Greater

overlap between technical and lay terms was

observed for Romance languages and Greek than for

Germanic languages (except English), and some

technical terms had no lay equivalent.

2.2 Knowledge Acquisition in

Healthcare

Knowledge Acquisition process aims at identifying

and capturing knowledge assets and terminology to

populate a knowledge repository for a specific

domain. Central areas of this task are: terminology

work, relevant for the special subject field, including

terminography; content analysis of documents;

extraction of knowledge from various sources. A

major part of Knowledge Acquisition is capturing

knowledge from experts, a task that is made cost-

effective and efficient by using knowledge models

and special elicitation techniques. These techniques

should be used in different phases of the process,

since each of them permits the capture of a specific

typology of knowledge and the achievement of

specific aims. The most common techniques are

Interviews, direct observation of expert

performances to extract procedural knowledge,

mostly connected to manual skills, such as Think

Aloud Problem Solving, Self-report, and Shadowing.

Other techniques, such as Card Sorting, Repertory

Grid, and Twenty Questions, are useful for

understanding how experts conceptualise knowledge

related to their own domain of reference (Milton,

2007).



http://users.ugent.be/~rvdstich/eugloss/welcome.html

In a task of Knowledge Acquisition, it is

important to identify two main components:

knowledge types, referred to knowledge orientation

and domain; and modalities, referred to the

representation medium in which knowledge exists.

In Knowledge Acquisition for the Healthcare

domain, according to (Abidi, 2008), many different

types of knowledge, which directly contribute to

clinical decision-making and care planning, can be

identified: Patient, Practitioner, Medical, Resource,

Process, Organizational, Relationship and, finally,

Measurement Knowledge. In the present work we

will only deal with Medical Knowledge and Patient

Knowledge. These knowledge types are represented

by different knowledge modalities. The most

common ones are: Tacit Knowledge of practitioner,

Explicit Knowledge, Clinical experiences,

Collaborative Problem-solving discussions, Social

Knowledge, etc. In our work we focus on Explicit

Knowledge, Clinical Experience and Social

Knowledge. In particular the last modality can be

viewed in terms of a community of practice and their

communication patterns, interest and expertise of

individual members.

3 METHODOLOGY

In this study we focused on the acquisition of

consumer-oriented knowledge about a specific

subset of healthcare domain that includes Anatomy,

Symptomatology and Pathology. For that task we

have chosen to use a hybrid methodology for the

identification of “lay” terms, words, and expressions

used by Italian speakers to indicate

Symptoms/Signs, Diseases and Anatomical

Concepts. Three different target groups were

considered for the application of our approach: First

Aid patients subjected to a Triage Process; a

community of Researchers and PhD students with a

good level of healthcare literacy, and finally a group

of elderly people with a low level of healthcare

literacy. The proposed methodology consists of the

following steps:

1. Familiarization with the domain and

exploitation of existing common lexical

resources (Glossaries, Thesauri, Medical

Encyclopedias, etc);

2. Choice and application of three different

Elicitation Techniques to the mentioned

groups of people:

a. Collaborative Wiki-based Acquisition;

b. Nurse-assisted Acquisition;

KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development

c. Interactive Acquisition combining

traditional elicitation techniques (Focus

Groups, Concepts Sorting and Games);

3. Automatic Term Extraction and analysis of

acquired knowledge by means of a Text

Processing tool;

4. Clinical review of extracted terms and manual

mapping to a standard medical terminology,

performed by physicians;

5. Evaluation of results in order to find candidate

terms to be included in the Consumer-oriented

Medical Vocabulary.

3.1 Wiki-based Acquisition

The first method for acquiring consumer-oriented

medical knowledge is based on the use of a

Semantic Media Wiki system

, an easy to use

collaborative tool, allowing users to create and link,

in a structured and collaborative manner, wiki pages

on a certain domain of knowledge. Using our online

eHealthWiki

system, users created wiki pages for

describing symptoms and diseases, using “lay”

terminology, specifying in particular the

corresponding anatomical categorization, the

definition and possible synonyms.

The system has been evaluated over a sample of

32 people: researchers, PhD students and

administrative staff of our research institute (18

females, 14 males, between 25 and 56 years old). In

one month, we collected 225 wiki pages, 106 for

symptoms and 119 for diseases, and a total of 139

synonyms for the inserted terms. It was very

interesting to test here also the understanding of the

collaborative nature of the Wiki for the specific task,

which gave users the possibility not only to insert

medical terms by creating wiki pages, but also to

update, cancel or correct the inserted information,

and above all to modify wiki pages added by other

users, in order to reach a convergence on the

common sense of medical terminology. In our case,

users were reluctant to modify concepts added by

others, even in the case of evident mistakes in

definitions or categorization (only 7 people out of 32

provided changes to wiki pages). Some examples of

categorization mistakes that had not been modified

are “Singhiozzo” (Hiccup), and “Mal di Testa”

(Headache), both categorized as Diseases instead of

Symptoms. In some cases, when users were in doubt

about the right categorization of a concept, they

inserted it in both the categories, e.g., “Ustione”



http://semantic-mediawiki.org/wiki/Semantic_MediaWiki

http://ehealthwiki.fbk.eu

(Burning). This test highlighted the fact that users

had problems in categorizing medical terms - mainly

due to their clinic ambiguity - and also the erroneous

use of these terms by them daily.

3.2 Nurses-assisted Acquisition

The second acquisition technique involved the

nurses of a First Aid Unit (in a Hospital of the

Province of Trento)

, as figures of mediation for the

acquisition of terminology about patient symptoms

and complaints. Nurses here help patients to express

their problems using the classical subjective

examination performed during the Triage Process.

The Triage activity has the aim of prioritizing

patients according to the severity of their condition,

on the basis of examination costing few minutes.

This acquisition method involved 10 nurses, around

60 patients per day and a total of 2.000 Triage

Records registered in one month. During this period

nurses acquired the principal problems (symptoms

and complaint) expressed by their patients using

“lay” terminology and inserted them in the Triage

Record together with the corresponding medical

concepts usually used for codifying patient data. For

example, the lay expression “Ho i crampi alla

pancia” (I have a stomach ache) was inserted in the

Triage Record together with the corrisponding

medical concept “Addominalgia” (Abdominal pain).

3.3 Focus-group Acquisition

The last method used in our study consisted in

merging three different traditional elicitation

techniques: Focus Group, Concepts Sorting, and the

more experimental Board Games, in order to allow

interaction and sharing situations to improve the

process of acquisition. We have applied this

techniques to a community of 32 elderly persons in a

Seniors Club in the Province of Trento, between 65

and 83 years old.

During the process of acquisition we divided

participants into 4 groups, of 8 people each,

assigning to each one a specific body part category

(for instance, head and neck, abdomen and back,

arms, hands and chest, pelvic area, legs and feet).

Each group was asked to write on little cards all

known symptoms and diseases related to the

assigned area, starting from personal experience, and

comparing their idea with other members of the



Medicina d’Urgenza e Pronto Soccorso del Presidio Ospedaliero

di Cles (Trento): http://www.apss.tn.it/Public/ddw.aspx?n=26808

A HYBRID METHODOLOGY FOR CONSUMER-ORIENTED HEALTHCARE KNOWLEDGE ACQUISITION

group to find a common definition for the written

terms. Time allowed for card writing was about 20

minutes. About 160 medical terms were collected.

Then, all these terms were analyzed together,

creating discussions, exchanging opinions on term

definitions, synonyms, and recording preferences

and common sense. At the end of the discussion

about each medical concept, all participants gave

preferences for choosing the right body system

categorization of that concept. For this

categorization we provided a panel with 14 different

problem areas and body systems (digestive,

neurological, musculoskeletal, lymphatic, endocrine,

etc.). This allowed us not only to collect lay

terminology used by elderly people, but also to

understand how they define and categorize medical

concepts, in order to compare these results with

these obtained with the other two techniques

mentioned above.

4 TERM EXTRACTION

Three sets of collected data, including the

transcription of the Focus Group activity with

elderly persons, were further processed and

analyzed, to detect candidate consumer-oriented

terms, with the tool Text-2-Knowledge (T2K),

developed at the Institute of Computational

Linguistics of Pisa

. This tool allowed us to

automatically extract terminology from the data sets

and to perform typical text processing techniques

(normalization, pos tagging, chunking, etc.),

calculating, in addition, statistics such as term

frequency on the extracted data. The computational

system adopted by the tool includes a specific plug-

in for the analysis of Italian. It provides, as final

output, a term-based vocabulary whose added value

is represented by the terms’ semantic and conceptual

information regarding the vocabulary itself. These

terms, which can be either single or multi-word

terms, are organized in a hierarchical

hyponym/hyperonym relation depending on the

internal linguistic structure of the terms (Bartolini et

al., 2005); that is, by sharing the same lexical head.

In spite of the advantages of the automatic

extraction process, allowing for extraction of many

compound terms, a good amount of terms, certainly

representative of consumer medical terminology,

were not automatically extracted, since, due to the

quantitative limits of the corpus dimensions, their



http://www.ilc.cnr.it

occurrence was inferior with respect to the

predefined threshold value. Consequently, we

performed an additional manual extraction to take

into account such rare terms, usually mentioned by a

single participant. Statistical results about the three

different data sets are further discussed in Section 6.

5 CLINICAL REVIEW

Term extracted by T2K were reviewed by two

physicians to find errors and incongruities in

categorization and synonymy. Many mistakes were

found in the first set (Wiki-based), where a wrong

categorization was assigned to 25 terms, and were

wrong synonyms were expressed for 8 terms. Many

mistakes were also found in the third set (Elderly

people), where wrong categorization were assigned

to 40 terms, e.g., “Giramento di Testa” or

“Vertigini” (Vertigo or Dizzines), in the

Cardiovascular System instead of the right

Neurological one. Concerning the second data set,

clinical review was performed during the process of

Triage by a nurse and a physician .

During the second part of our clinical review,

physicians have been asked to map a term/medical

concept pair by using a professional medical

terminology - in this study the International

Classification for Primary Care 2

Edition (ICPC2).

ICPC2 addresses fundamental parts of healthcare

process: it is used in particular by general

practitioners for encoding symptoms and diagnosis.

It has a biaxial structure that consider medical

concepts related to Symptoms, Diseases and

Diagnoses, and Medical Procedure, according to 17

Problem Areas/Body Systems. In previous work we

encode ICPC-2-E using a recently developed Web

Ontology Language (OWL)

(both for English and

Italian), that also provides the formalization of the

existing clinical mapping with the ICD10

classification system, as shown in (Cardillo et. al.,

2008). By means of this mapping we want to

reconstruct the meaning (concept)

inherent in the lay

usage of a term, and then to agree that

consonance

between lay and professional terms exists on the

basis of this deeper meaning, rather than the lexical

form. Five different types of relations are possible

between consumer terms and ICPC2 medical

concepts:

• Exact Mapping between the pairs; this

occurs when the term used by a lay person



http://www.w3.org/TR/owl-features/

KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development

can be found in ICPC2 terminology and

both terms correspond to the same concept.

E.g., the lay term “Febbre” (Fever) would

map to the ICPC2 term “Febbre”, and both

will be rooted to the same concept.

•

Related Mapping; it involves lay synonyms

and occurs when the lay term does not exist

in the professional vocabulary, but

corresponds to a professional term that

denotes the same (or closely related)

concept. E.g., lay term “Sangue dal Naso”

(Nosebleed) corresponds to “Epistassi”

(Epistaxis).

• Hyponymy Relation; this occurs when a lay

term can be considered as term of inclusion

of a ICPC2 concept. E.g., lay term

“Abbassamento della Voce” (Absence of

Voice) is included in the more general

ICPC2 concept “Sintomo o disturbo della

voce” (Voice Symptom/Complaint).

•

Hyperonymy Relation; in this case the lay

term is more general than one or more

ICPC2 concepts, so it can be considered as

its/their hyperonym. E.g., the term

“Bronchite” (Bronchitis) is broader than

“Bronchite Acuta/ Bronchiolite” (Acute

Bronchitis/ Bronchiolitis) e “Bronchite

Cronica” (Chronic Bronchitis) ICPC2

concepts.

• Not Mapped; it comprises those lay terms

that cannot be mapped to the professional

vocabulary. These can be legitimate health

terms whose omission reflects real gaps in

existing professional vocabularies; or they

can represent unique concepts reflecting lay

models of health and disease. E.g., the lay

term “Mal di mare” (Seasickness).

6 RESULTS EVALUATION

As previously mentioned, our methodology of

acquisition allowed us to acquire varied consumer-

terminology and to perform an interesting

terminological and conceptual analysis. Tables 1-4

provide term extraction and mapping evaluation in

terms of a statistical analysis. By means of the term

extraction process, we were able to extract a total of

692 medical terms from 225 Wiki pages, 375 of

which were not considered pertinent to our aim. We

performed mapping analysis on 587 terms as

summarized in Table 1.

Table 1: Wiki term collection.

Tot.

Terms

Exact

Map.

Iponym.

Relation

Iperonym.

Relation

Symptoms 306 26 50 40 9

Diseases 140 42 19 38 38

Anatomy 141 105 11 16 4

Other 375 / / / /

Tot. 962

Not Mapped. 186 Not Considered:375

We can observe that most of the exact mappings

with ICPC2 are related to anatomical concepts, and

which many synonyms in lay terminologies and

inclusion terms were found for symptoms. Table 2

shows the results related to the Triage acquisition

data.

Table 2: Nurses-assisted term collection.

Tot.

Terms

Exact

Map.

Not

Mapped

Symptoms 508 134 197 177

Diseases 325 86 94 145

Anatomy 275 120 95 60

Other 1281 / / /

Tot. 2389 Not Considered: 1281

We extracted a total of 2389 terms from 2.000

Triage records, but about half of these terms were

considered irrelevant for our evaluation, so mapping

was provided only for 1108 terms. Contrary to the

previous results, here is interesting to highlight the

high presence of lay terms used for expressing

symptoms with exact mappings to ICPC2, but also

many synonyms in lay terminology for ICPC2

symptoms and diseases. This is particularly related

to the context chosen for the acquisition, where

patients just ask for help about suspected symptoms

and complaints. Table 3 shows the results related to

the data acquisition from Elderly people.

Table 3: Focus Group /Game with Elderly Persons.

Tot.

Terms

Exact

Map.

Not

Mapped

Symptoms 79 35 44 /

Diseases 87 29 54 4

Anatomy 77 51 18 8

Other 68 / / /

Tot. 321 Not Considered: 78

A HYBRID METHODOLOGY FOR CONSUMER-ORIENTED HEALTHCARE KNOWLEDGE ACQUISITION

Concerning the last data set, 321 medical terms

were extracted by the transcription of the Focus

Group/Game activity. Here is interesting to note that

all the symptoms extracted had corresponding

medical concept in ICPC2 terminology.

Table 4 compares the three data set together and

shows that the most profitable methods for acquiring

consumer-oriented medical terminology was the one

assisted by nurses. But the limit of this method is

that it is time-consuming for nurses who have to

report all the patient “lay” health expressions. While

Wiki-based method, even if not exploited for the

collaborative feature, has demonstrated good

qualitative and quantitative results. Furthermore, are

interesting the results concerning mapping to ICPC2,

because 2/3 of the lay terms collected are covered by

ICPC2 terminology.

Table 4: Results Overview.

Sources

Tot.

Terms

Tot.

Mapped

Not

Mapped

eHealthWiki 962 398 186

Nurse-assisted 2389 726 382

Focus Groups 321 231 12

Tot. 3662 1355 580

To conclude our evaluation we have to highlight

that comparing the three sets of collected terms, the

overlap is only of 60 relevant consumer medical

terms. The total overlap with ICPC2 is about 508

medical concepts on a total of 706 ICPC2 concepts.

This means that all the other mapped terms can be

considered synonyms or quasi synonyms of the

ICPC2 concepts. The large number of not mapped

terms and the low overlap between the three sets of

extracted terms demonstrate that we collected a very

variegated range of medical terms, many compound

terms and expressions. These terms can be

representative of the corresponding technical terms

present in standard medical terminologies, and can

be used as candidate terms for the construction of

our Consumer-oriented Medical Vocabulary for

Italian.

7 CONCLUSIONS AND FUTURE

WORK

In this paper we have presented a hybrid

methodology for acquiring consumer-oriented

medical Knowledge and Terminology for Italian,

consisting of lay expressions and terms used to

indicate Symptoms, Diseases and Anatomical

Concepts. We applied three explorative elicitation

techniques to three different groups of people, and

we compared results on the basis of a term

extraction process, for statistical analyses, and on a

clinical mapping procedure, for finding overlaps

between extracted lay terms and specialized medical

concepts in the ICPC2 medical terminology. Our

methodology showed encouraging results, because it

allowed us to acquire many consumer-oriented

terms, to find a low overlap with medical concepts

and a high number of related mappings (most of the

time synonyms) to the ICPC2 terminology.

Taking each of these acquisition techniques

alone, they do not allow a good coverage of the

whole domain of pathology and symptomatology.

In the first case, in fact, most of the terms provided

by Researchers and PhD students are related to

Digestive and Musculoskeletal Systems, and Skin,

while in the case of the Triage activity patients

expressed most of the time symptoms related to

Musculoskeletal System (due to the geographic

context and period of the acquisition, i.e. mountain–

skiing area, end of winter), Respiratory System, and

Cardiovascular System. Furthermore, also in the

third acquisition, terms are mostly related to the

these Body Systems. Using a hybrid approach in

merging these techniques and involving a more

varied sample of people would improve the results,

both from the qualitative and the quantitative point

of view. Another limit could be seen in the process

of manual mapping performed by physicians. After

this pilot study we plan to implement a semi-

automatic procedure for mapping lay and specialized

terminology, which will be associated to the process

of automatic term extraction, and validated by the

review of physicians.

In the approaches described in Section 2,

consumer-oriented health vocabularies were

developed by working only on big written corpora

(forum postings and queries to medical websites),

using machine learning algorithm and statistical

methods (naive Bayesian classifiers, C-value,

Logistic Regression etc.) to extract consumer-

oriented terminology. In comparison to these

approaches we gave more importance to qualitative

data, focusing on different methods for acquiring not

merely lay terminology but also knowledge directly

from consumers in different scenarios related to

General Practice. This allowed also to try to

KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development

understand how consumers make use of medical

terminology, how common expressions daily used in

health communication really match onto medical

concepts used by professionals.

To improve the results of the Knowledge

Acquisition process and to extract more variegated

consumer terminology, not related to the regional

context, one of the future tasks is to perform a

Knowledge Acquisition Process involving people in

a Social Network. This would allow to extend our

sample, including younger people. This task would

be very interesting for comparing results with what

resulted from the previous methodologies. Another

important improvement would be the analysis of

written texts such as forum postings of an Italian

medical website for asking questions to on-line

doctors

. Data extracted in this way could also be

used to validate the acquired verbal terminology, by

providing preferences between terms according to

frequency and familiarity score.

ACKNOWLEDGEMENTS

We would like to thank Antonio Maini and Maria

Taverniti, who provided us with useful support

respectively in the process of Knowledge

Acquisition and Term Extraction.

REFERENCES

Abidi, S.S.R., 2007. Healthcare Knowledge Management:

The Art of the Possible. In Proceedings of the

Knowledge Management for Health Care Procedures’

Conference, K4CARE 2007, Springer Berlin, pages 1-

20.

Bartolini,

R., Lenci, A., Marchi, S., Montemagni, S., and

Pirrelli, V., 2005. Text-2-knowledge: Acquisizione

semi-automatica di ontologie per l’indicizzazione

semantica di documenti, Technical Report fot the

PEKITA Project, ILC. Pisa p. 23;

Cardillo, E., Eccher, C., Tamilin, A., and Serafini, L.,

2008. Logical Analysis of Mappings between Medical

Classification Systems. In Proceedings of the 13

International Conference on Artificial Intelligence:

Methodology, Systems, and Applications,

AIMSA2008, Springer Berlin, pages 311-321

Keselman, A., Logan, R., Smith, C. A., Leroy, G., and

Zeng, Q., 2008. Developing Informatics Tools and

Strategies for Consumer-centered Health

Communication. Journal of the American Medical

Informatics Association, 14(4):473-483.



http://medicitalia.it

Kim, H., Zeng, Q., Goryachev, S., Keselman, A.,

Slaughter, L., and Smith, C.A., 2007. Text

Characteristics of Clinical Reports and Their

Implications for the Readability of Personal Health

Records. In Proceedings of the 12

World Congress

on Health (Medical) Informatics, MEDINFO2007,

IOS Press, pages 1117-1121

Milton, N. R., 2007. Knowledge Acquisition in Practice: A

Step-by-step Guide. London, Springer;

Rector, A. (1999). Clinical Terminology: Why is it so

hard?. Methods of Information in Medicine, 38(4):239-

252.

Rosembloom, T.S., Miller, R.A., Johnson, K.B., Elkin,

P.L., and Brown, H. S., 2006. Interface Terminologies:

Facilitating Direct Entry of Clinical Data into

Electronic Health Record Systems. Journal of

American Medical Informatics Association, 13(3):277-

287.

Smith, B. and Rosse, C., 2004. The Role of Foundational

Relations in the Alignment of Biomedical Ontologies.

In Proceedings of the 28

American Medical

Informatics Association's Annual Symposium,

AMIA2004.

Soergel, D., Tse, T., and Slaughter, L., 2004. Helping

Healthcare Consumers Understand: An “Interpretative

Layer” for Finding and Making Sense of Medical

Information. In Proceedings of the International

Medical Informatics Association’s Conference,

IMIA2004, pages 931-935.

Zeng, Q., Goryachev, S., Keselman, A., and Rosendale,

D., 2007. Making Text in Electronic Health Records

Comprehensible to Consumers: A Prototype

Translator. In Proceedings of the 31

American

Medical Informatics Association's Annual Symposium,

AMIA2007, pages 846-850.

Zeng Q. and Tse, T., 2006. Exploring and Developing

Consumer Health Vocabularies. Journal of the

American Medical Informatics Association, 13:24-29

A HYBRID METHODOLOGY FOR CONSUMER-ORIENTED HEALTHCARE KNOWLEDGE ACQUISITION