RFID-based Semantic-enhanced Ubiquitous Decision

Support System for Healthcare

Michele Ruta, Floriano Scioscia, Tommaso Di Noia and Eugenio Di Sciascio

Politecnico di Bari

via Re David 200, I-70125

Bari, ITALY

Abstract. We present an innovative Decision Support System for healthcare ap-

plications, based on a semantic enhancement of RFID standard protocol. Seman-

tically annotated descriptions of both drugs and patient’s case history stored in

proper RFID tags are used to help doctors in providing the correct therapy. The

proposed system allows to discover possible inconsistencies in a therapy suggest-

ing alternative treatments.

1 Introduction

RFID is an automatic identiﬁcation technology, relying on storing and remotely re-

trieving information located on a tag exploiting proper interrogators, namely readers.

The miniaturization of electronic components and circuits nowadays allows an RFID

transponder to be applied-to or incorporated-into objects, animals, or persons for iden-

tiﬁcation and tracking purposes. Some tags can be interrogated at distance and also

by-passing possible physical barriers and obstacles. They usually contain a unique code

which is read by the interrogator and can be used to identify the associated object via a

networked database on a server. Nevertheless, transponders with larger internal memory

open new interesting possibilities and enable further applications. Notice that current

Radio Frequency (RF) identiﬁcation methods only enable elementary recognition ap-

plications which exploit queries over a database for retrieving object features and prop-

erties. If tagged objects, animals or persons expose to a reader not simply a numeric

identiﬁer but a compressed semantic annotation, they may describe themselves without

referring to a centralized database, also allowing the possibility to update those de-

scriptions in real time and when needed considering information not completely known

when the tag has been attached/implanted. This is particularly useful in case: (i) a de-

pendable or networked link toward the ﬁxed information server is unavailable; (ii) the

information related to the object/subject has to be always and straightaway available;

(iii) an advanced description of object/subject characteristics and capabilities is needed

in order to enable complex inference procedures over data stored within the tag.

All the above features have an undoubted interest in the healthcare sector. E-health-

care information systems include applications for tele-medicine, tele-health, and tele-

home care services. RFID technology now has signiﬁcant impact on healthcare systems,

with speciﬁc reference to tracking and management of patients and drugs within hospi-

tals. Beneﬁts of RF identiﬁcation in those cases include error prevention in identifying

Ruta M., Scioscia F., Di Noia T. and Di Sciascio E. (2009).

RFID-based Semantic-enhanced Ubiquitous Decision Support System for Healthcare.

In Proceedings of the 3rd International Workshop on RFID Technology - Concepts, Applications, Challenges , pages 79-88

DOI: 10.5220/0002203500790088

 SciTePress

staff and regulation of accesses in various divisions for patients and doctors. Although

these applications are noteworthy, a more advanced exploitation of RFID could further

enhance the impact of this technology in e-healthcare. In this paper we present a novel

Decision Support System (DSS) for innovative medical applications based on a seman-

tic enhancement of RFID standard protocol. Thanks to semantic annotation of both

drugs to be administered and patient’s case history, the proposed system helps physi-

cians in conﬁrming and then choosing the best therapy based on the medical record of

the patient.

By referring to semantic metadata stored within RFID tags attached to drugs pack-

aging and patient’s RFID wristband, a matchmaking can be performed to discover

possible inconsistencies in a therapy, also suggesting further treatment options to the

physician. We borrowed ideas and technologies devised for the Semantic Web initia-

tive. We set our stage in an e-healthcare context, where RFID tags are dipped into an

enhanced Bluetooth framework. In a previous work both the RFID EPCglobal data ex-

change protocol

and the Bluetooth Service Discovery Protocol

have been modiﬁed to

enable support for advanced inference services, while maintaining legacy applications

[1]. Here we introduce novel semantic-based value-added services for decision support

in healthcare.

The remaining of the paper is structured as follows. In the next section relevant

related work referred to the exploitation of RFID technology in the medical ﬁeld is

surveyed. Section 3 outlines the framework, explaining the discovery and matchmaking

algorithms devised for the purposes outlined above. Section 4 illustrates the system

architecture and the approach in an example scenario; ﬁnally conclusion and future

work close the paper.

2 RFID for Healthcare

Hospital activities are characterized by complex workﬂows requiring the interaction

of several different actors and the coordination of multiple facilities [2]. Supply chain

management has suggested that the identiﬁcation capability of RFID can be leveraged

to improve healthcare services. Patients, equipment and staff can be tagged with RFID

transponders and tracked within a hospital by a network of RFID readers deployed in

key locations. Actions recognized via RFID can be logged automatically, thus avoid-

ing lengthy and error-prone manual data input by personnel. Research studies and pilot

projects have evaluated the impact of such infrastructures in ordinary hospital activities

[3,4] as well as in emergency conditions due to disasters [5] or epidemics [6]. The inte-

gration of RFID into a Hospital Information System (HIS) allows to: automate checks

for security (authorization enforcement) and safety (prevention of human error) during

critical processes such as patient admission, checkout and drug administration; reduce

response times in emergency situations; improve efﬁciency of resource allocation. Ul-

timately, these beneﬁts lead to higher conﬁdence and satisfaction of both patients and

personnel [2].

EPCglobal. http://www.epcglobalinc.org

Bluetooth. http://www.bluetooth.com

The integration of RFID with other pervasive computing technologies –such as

wireless protocols and sensor networks– is leading to further innovative applications

in the tele-medicine area, particularly for ubiquitous persistent monitoring of elderly

or disabled people as well as for patient follow-up during rehabilitation phase [7,8].

Context-awareness is the key aspect of such approaches to improve quality of health-

care services. Challenges and beneﬁts were clearly evidenced in a prototype of RFID-

enabled smart hospital bed [9], whose architecture resembles our solution. That sys-

tem, however, provided only basic identiﬁcation features and lacked more advanced

knowledge-based capabilities. Our proposal takes a step further in this direction, by

combining a pervasive and context-aware computing framework with decision support

features based on Knowledge Representation (KR). Decision support to clinical activity

is widely acknowledged as one of the most important beneﬁts of medical informatics

[10]. Research has also evidenced that artiﬁcial intelligence and rule-based systems can

be effective in helping clinicians to reduce errors in both diagnosis and treatment. Nev-

ertheless, the ﬁrst generation of Computerized Physician Order Entry (CPOE) systems

was mostly based on manual data entry and a fragmented collection of non-integrated

utilities. Experience taught that, in such cases, the improvements in overall quality of

patient care are not always clear, since the decrease in some kinds of errors is counter-

acted by slowdownsin operations and an increased frequencyof other types of mistakes.

Decision support is highly effective only when it is automatic and seamless [10].

Ontology-based knowledge modeling can ensure that only highly relevant informa-

tion about patient’s clinical conditions and appropriate treatments are supplied to physi-

cians. The use of lightweight wireless computing infrastructures and of widely-adopted

KR technologies can promote interoperability and integration of solutions designed for

hospital centres hosting tele-medicine applications. Finally, as pointed out in [7], ubiq-

uitous computing technologies allow the capture of health data at an unprecedented

scale: knowledge-based approaches can assist in the management, analysis and inter-

pretation of such data for research purposes and/or to improve clinical best practices.

3 Inference Services for Decision Support in Healthcare

In the approach we propose here, non-monotonic inferences presented in [11] are ex-

ploited to retrieve suitable treatments for a given disease taking into account the case

history of the patient. The system will calculate a score, based on the semantic com-

patibility between diseases affecting the patient and characteristics of available drugs,

so allowing to: (i) ﬁnd possible inconsistencies in a proposed therapy; (ii) arrange best

treatment options in relevance order; (iii) explain the matchmaking outcomes in both

cases.

In what follows, the Description Logics (DLs) setting we adopt are brieﬂy recalled

We refer to [11,13] for several examples and wider argumentation. From now on we

assume to model ontologies (Terminological Boxes T in DL-words), patient diseases

and drugs annotations in a language whose semantics can be mapped to the ALN

We assume the reader be familiar with basics of DLs formalisms and reasoning [12].

DL, for instance (a subset of) OWL-DL

or the more compact XML-based DIG [14]

language.

DL-based systems provide two basic reasoning services for T , namely (a) Satis-

ﬁability and (b) Subsumption in order to check (a) if a formula C is consistent w.r.t.

the ontology –T 6|= C ⊑ ⊥– or (b) if a formula C is more speciﬁc or equivalent to a

formula D –T |= C ⊑ D. It is possible to deﬁne at least ﬁve different match classes

based on subsumption and satisﬁability: exact, subsumption (full), plug-in, intersection

(potential) and disjoint (partial) match.

Both subsumption and satisﬁability can be only used to check if there exists an ex-

act correspondence between two formulas. Hence they are not completely adequate in

scenarios like the healthcare ones, where simple yes/no answers are insufﬁcient because

exact matches are quite rare. In the proposed approach, considering a disease descrip-

tion S and a drug annotation D, solving the Concept Contraction Problem (CCP) and

the Concept Abduction Problem (CAP) [11] (see later on for further details) we are able

to provide a support in decision making for doctors determining a therapy.

Hereafter basics of algorithms to solve abduction and contraction problems are re-

ported:

– given a partial match between D and S, solving a CCP one can compute what has

to be given up G and kept K in D in order to have a potential match between

K (a contracted version of D) and S. Hence, the result of a CCP is a pair hG, Ki

representing respectively elements in D conﬂicting with S and the (best) contracted

D compatible with S;

– given a potential match between D and S, solving a CAP one can compute what

has to be hypothesized in S in order to have a full match with D (or its contracted

version K). Hence, the result of a CAP is a concept H representing in some way

what is underspeciﬁed in S in order to completely satisfy D. Please note that we

say underspeciﬁed instead of missing. This is because we are under an Open World

Assumption.

Of course, for both Concept Contraction and Concept Abduction we have to deﬁne

some minimality criteria on G (giveup as few things as possible) and on H (hypothesize

as few things as possible). Algorithms to solve CAPs and CCPs for ALN have been

proposed in [11] (not reported here for brevity) and they have been properly adapted

and exploited in our e-healthcare scenario based on RFID.

3.1 Matchmaking for Healthcare

In the application presented here, Concept Contraction and Concept Abduction in-

ference algorithms are used in a slightly different fashion w.r.t. current matchmaking

problems. Patient case history and drug annotations have distinct structures and are dif-

ferently described with each other

, so inference services outlined above have to be

properly used to reach the desired goals.

OWL Web Ontology Language. http://www.w3.org/TR/owl-features/

The illustrative example presented in Section 4 will clarify these aspects.

With reference to classical matchmaking –especially devised for e-commerce [13]–

where a demand is compared with a set of supplies, in the proposed approach we have

to compare the drug annotation stored within the packaging tag with the patient clinical

description in her RFID wristband. Notice that the semantic-based matchmaking is a

non-symmetric one and the ﬁnal purpose of our framework is to assess if a given drug

encounters patient’s diseases. This can be performed enriching the disease semantic an-

notation with the drugs classes suitable to cure the disease itself. In this way, Concept

Abduction allows to verify if a given treatment is suitable or not. For what concerns

contraindications, in the approach we propose treatments and disease descriptions are

modeled exploiting disjoint concepts in order to refer to interested organs and bodily

systems. Hence, if a given drug may present some undesired effects for a speciﬁc pa-

tient, the abduction check will fail due to the incompatibility between semantic descrip-

tions of drug and disease. So, thanks to Concept Contraction algorithm, the physician

can “see” the incompatibilities within a therapy annotation (i.e., the adverse indications

for the patient) which will make the part of the therapy to give up. The remaining K

component will be then used for a new abduction process.

Summarizing, the steps for therapy veriﬁcation are reported hereafter:

1. the system perform a Concept Abduction between therapy description D and pa-

tient’s case history S;

2. if D ⊓ S are satisﬁable w.r.t. T , the proposed therapy is veriﬁed by the system;

3. if D ⊓ S are incompatible, Concept Contraction algorithm allows to extract the

contraindications of the treatment for building a new compatible request to be sub-

mitted against the drugs in the hospital Knowledge Base (KB), in order to ﬁnd a

new pharmaceutical D *;

4. the system performs a new Concept Abduction between D * and S to verify the

new therapy.

Note that step 3 returns a list of further options in a relevance order. By means

of rankPartial and rankPotential algorithms [11], the system measures the seman-

tic distance (i.e., the compatibility level) between each treatment annotation and the

description of the patient case history. Results are arranged according to the semantic

correspondence with the disease.

4 Case Study

The proposed approach was tested in a case study for a hospital rheumatology unit.

A speciﬁc ontology was devised for connective tissue diseases, an important class of

autoimmune rheumatic diseases. Figure 1 shows a relevant excerpt of it, reported in

classical DL notation for the sake of readability.

The prototypical system simulates a “smart bed” tablet computer, equipped with a

touchscreen and an RFID reader (see Figure 2 for details). The device connects to the

HIS through semantic-enhanced Bluetooth Service Discovery Protocol, via a hotspot

placed in the ward within radio range of beds. Each resource (patient, staff member and

drug) is identiﬁed by means of an RFID tag with unique EPC code, unique identiﬁer of

Anatomy

- Immune System ⊑ Anatomic P art

- Adverse Immune System ⊑ Anatomic P art ⊓ ¬Immune System

- Circulatory System ⊑ Anatomic P art

- Adverse Circulatory System ⊑ Anatomic P art ⊓ ¬Circulatory System

- Skeletal System ⊑ Anatomic P art

- Muscular System ⊑ Anatomic P art

- Adverse Skeletal System ⊑ Anatomic P art ⊓ ¬Skeletal System

- Adverse M uscular System ⊑ Anatomic P art ⊓ ¬Muscular System

- Bone ⊑ Skeletal System

- Adverse Bone ⊑ ¬Bone ⊓ Adverse Skeletal System

- V isual System ⊑ Anatomic P art

- Adverse V isual System ⊑ Anatomic P art ⊓ ¬V isual System

- Eye ⊑ V isual System

- Adverse Eye ⊑ Adverse V isual System ⊓ ¬Eye

Diseases

- Disease ⊑ ∃aff ects

- Musculoskeletal System Disease ⊑ Disease ⊓ ∀aff ects.(Adverse Skeletal System ⊓

Adverse M uscular System)

- Immune System Disease ⊑ Disease ⊓ ∀affects.Adverse Immune System

- Autoimmune Disease ⊑ Immune System Disease

- Connective T issue Disease ⊑ Autoimmune Disease

- Systemic Lupus Erythematosus ⊑ Connective T issue Disease⊓ ∀af f ects.(Adverse Integumentary System⊓

Adverse Hematopoietic System ⊓ Adverse Joint ⊓ Adverse Kidney ⊓ Adverse N ervous System ⊓

Adverse Lung ⊓ Adverse M uscle ⊓ Adverse Gastrointestinal T ract ⊓ Adverse Circulatory System)

- Severe SLE ⊑ Systemic Lupus Erythematosus ⊓ ∀therapy.(N SAID ⊓ P lasmapheresis ⊓ Corticosteroid ⊓

Immunomodulator Immunosuppressant)

- M ild SLE ⊑ Systemic Lupus Erythematosus ⊓ ∀therapy.(N SAID ⊓ Corticosteroid ⊓

Immunomodulator Immunosuppressant)

Treatments

- Drug ⊑ T reatment

- P hysiotherapy ⊑ T reatment

- P lasmapheresis ⊑ T reatment

- N SAID ⊑ Drug

- Corticosteroid ⊑ Drug

- Immunomodulator Immunosuppressant ⊑ Drug

- Anti T N F alpha ⊑ I mmunomodulator Immunosuppressant

Fig.1. Excerpt of the ontology engineered for the case study.

Semantic

enhanced RFID

INFERE

Semantic

enhanced RFID

INFERE

RFID

Reader

WNIC

Drug’s RFID tag

Patient’s RFID tag

Fig.2. System architecture.

the reference ontology (OUUID), semantic-based annotation in compressed DIG format

and data-oriented resource attributes [1].

The key capabilities of the system (control of drug submission procedures and de-

cision support to the physician for therapy management) will be better explained by

means of a small example. The smart bed hosts a patient with a mild form of Systemic

Lupus Erithematosus (SLE) and a generic disease of the muscular and skeletal system.

This is expressed w.r.t. the reference ontology with the following ALN formula:

S: Mild SLE ⊓ M usculoskeletal System Disease

A rheumatologist approaches the bed to give cortisone to the patient. The RFID

reader detects the triple {patient, staffmember,drug} so that the display is activated:rel-

evant information extracted from tags is shown and the authorized operation is recorded

into the HIS. Figure 3 shows the output in our current prototype (the panel on the left

hand side allows to computer simulate tag reading events). If an unauthorized staff

member –e.g., a janitor or a physician from another ward– approached the bed with

a drug, then the system would provide a warning as shown in Figure 4 and record the

event into the HIS log.

Fig.3. Prototype display: access granted.

Fig.4. Unauthorized access is detected.

Cortisone is described in the KB as:

D: ∃af fects ⊓ ∀therapy.Corticosteroid⊓ ∀affects.Eye ⊓ Circulatory System⊓ Bone

so it has potential adverse effects towards eyes, bones and circulatory system. Our pa-

tient has no eye problems, but a skeletal disease, while SLE can affect the circulatory

system. Special care must be taken for this patient during treatment of SLE with corti-

sone. The system is capable to perform this inference automatically and issue a warning

to the doctor. Figure 5 shows the warning, with an alert and the description of the con-

ﬂicting characteristics in a box in the lower part of the display. Further decisions are

then left to the judgment of the human expert. Steps 1 and 2 of the sequence in Section

3.1 produce the following outcome: Give up: ∀affects.Circulatory System ⊓ Bone

Keep: ∀therapy.Corticosteroid ⊓ ∀af f ects.Eye

The physician can now query the system for other therapy options. The smart bed

computer sends via Bluetooth a request to the ward hotspot. For each drug in the hos-

pital KB, rankPotential is computed w.r.t. the patient’s description. If compatibility

arises, rankPartial is computed to extract and evaluate the incompatible part of the de-

mand, then rankPotential is computed again for the “Keep” part.

In order to improve ﬂexibility of decision support, the matchmaking framework ex-

plained in Section 3 is combined with context-speciﬁc variables by means of an overall

utility function. The following parameters are taken into account: (1) age of patient;

(2) estimated frequency of drug adverse effects; (3) severity of patient’s condition, ex-

pressed in a numeric scale from 0 to 4 according to guidelines of the International

Classiﬁcation of Functioning, Disability and Health framework issued by World Health

Organization [15]. The utility function has the following formula:

par

pot

max

pot

· tanh

age

· sever ity · tanh

adv frequency

The function was modeled as a distance measure, hence a lower value means a bet-

ter overall match. The ﬁrst factor allows different drugs to be ranked according to their

compatibility w.r.t. patient’s conditions: r

par

and r

pot

are the rankPartial and rankPo-

tential values between the drug and the patient, while max

pot

is the highest (worst)

rankPotential considering all drugs in the hospital KB (i.e., the less effective treatment).

The next components take patient’s age and severity into account: a younger patient or

with a lower impairment level will tolerate therapy better in general (the model is not

deﬁned for pediatric patients). The last factor models drug contraindications, where

adv f requen c y is the statistical frequency of the main adverse effects, expressed in

number of occurrences per 100 patients. Empirical evaluation has suggested values for

the two tunable weights α = 50 and β = 10 respectively.

Let us suppose the patient is 96 years old and has an overall impairment degree of

2. Cortisone has a 6% frequency of adverse effects. With respect to the patient, cortisone

has a rankPartial value of 4 and its compatible part has a rankPotential of 7, whereas

the maximum rankPotential among all drugs in the KB is 9. Hence the ﬁnal outcome

is: f

4+7

· tanh

· 3 · tanh

= 1.2575

Iteration of the procedure over each drug in the KB produces the ranking depicted in

Figure 6, which is shown to the physician. We notice that prednisone is quite similar to

cortisone (both are corticosteroids), having the same risks toward the particular patient;

however, its estimated frequency of adverse effects is lower (5% vs 6%), hence it is

preferable.

The physician selects an appropriate therapy and leaves. RFID reader detects the

event and the application screen is closed.

Even though expressiveness of the logic language is limited by the need to provide

acceptable reasoning performance, the example showed that careful modeling of the

domain can provide useful knowledge-based decision support. Our prototypical DSS

system helps the domain expert in an unobtrusive way, by automatically invoking infer-

ence procedures upon relevant fragments of knowledge extracted directly from RFIDs.

Fig.5. Physician is alerted of potential adverse effects.

Fig.6. Suggested drugs for the patient, in relevance order.

5 Conclusions and Future Work

We have presented a novel DSS for healthcare purposes based on a semantic enhance-

ment of RFID standard protocol. The proposed system exploits semantic annotations

of drugs to be administered as well as of patient’s case history, to help medical person-

nel in providing the correct therapy. Thanks to the semantic metadata accompanying

the description of both drugs (stored on RFID tags attached to packaging) and diseases

(saved on patient’s RFID wristband), it is possible to discover possible inconsistencies

in a therapy and to suggest alternative care options.

Future work includes: wider tests on the proposed methods; extension of the pro-

totype to support multiple hospital rooms and beds; improvement of the user interface,

possibly with guidance by medical personnel without speciﬁc computing knowledge.

Acknowledgements

The authors are grateful to Crescenzio Scioscia (M.D. at Rheumatology Unit, Depart-

ment of Internal Medicine and Public Health, University of Bari) for guidance in the

modeling of the problem domain and acknowledge partial support of Apulia Region

Strategic Project PS 121.

References

1. Di Noia, T., Di Sciascio, E., Donini, F.M., Ruta, M., Scioscia, F., Tinelli, E.: Semantic-based

bluetooth-rﬁd interaction for advanced resource discovery in pervasive contexts. Interna-

tional Journal on Semantic Web and Information Systems (IJSWIS) 4 (2008) 50–74

2. Sbrenni, S., Piazza, T., Farinella, E.: La tracciabilit`a del paziente in strutture ospedaliere.

Notiziario dell’Istituto Superiore di Sanit`a 20 (2007) 3–8

3. Sangwan, R., Qiu, R., Jessen, D.: Using RFID tags for tracking patients, charts and med-

ical equipment within an integrated health delivery network. In: Networking, Sensing and

Control, 2005. Proceedings. IEEE. (2005) 1070–1074

4. Holzinger, A., Schwaberger, K., Weitlaner, M.: Ubiquitous Computing for Hospital Appli-

cations: RFID-Applications to Enable Research in Real-Life Environments. In: 29th Annual

International Computer Software and Applications Conference, 2005 – COMPSAC 2005.

Volume 2. (2005)

5. Fry, E., Lenert, L.: MASCAL: RFID Tracking of Patients, Staff and Equipment to Enhance

Hospital Response to Mass Casualty Events. In: AMIA Annual Symposium Proceedings.

Volume 2005., American Medical Informatics Association (2005) 261

6. Wang, S., Chen, W., Ong, C., Liu, L., Chuang, Y.: RFID Application in Hospitals: A Case

Study on a Demonstration RFID Project in a Taiwan Hospital. In: Proceedings of the 39th

Annual Hawaii International Conference on System Sciences, 2006 – HICSS’06. Volume 8.

(2006)

7. Ho, L., Moh, M., Walker, Z., Hamada, T., Su, C.: A prototype on RFID and sensor networks

for elder healthcare: progress report. In: Proceedings of the 2005 ACM SIGCOMM work-

shop on Experimental approaches to wireless network design and analysis, ACM New York,

NY, USA (2005) 70–75

8. Jovanov, E., Milenkovic, A., Otto, C., de Groen, P.: A wireless body area network of intel-

ligent motion sensors for computer assisted physical rehabilitation. Journal of NeuroEngi-

neering and Rehabilitation 2 (2005) 1–10

9. Bardram, J.: Applications of context-aware computing in hospital work: examples and design

principles. In: Proceedings of the 2004 ACM symposium on Applied computing, ACM New

York, NY, USA (2004) 1574–1579

10. Handler, J., Feied, C., Coonan, K., Vozenilek, J., Gillam, M., Peacock, P., Sinert, R., Smith,

M.: Computerized Physician Order Entry and Online Decision Support. Academic Emer-

gency Medicine 11 (2004) 1135–1141

11. Di Noia, T., Di Sciascio, E., Donini, F.: Semantic matchmaking as non-monotonic reasoning:

A description logic approach. Journal of Artiﬁcial Intelligence Research (JAIR) 29 (2007)

269–307

12. Baader, F., Calvanese, D., Mc Guinness, D., Nardi, D., Patel-Schneider, P.: The Description

Logic Handbook. Cambridge University Press (2002)

13. Colucci, S., Di Noia, T., Pinto, A., Ragone, A., Ruta, M., Tinelli, E.: A non-monotonic

approach to semantic matchmaking and request reﬁnement in e-marketplaces. International

Journal of Electronic Commerce 12 (2007) 127–154

14. Bechhofer, S., M¨oller, R., Crowther, P.: The DIG Description Logic Interface. In: Proceed-

ings of the 16th International Workshop on Description Logics (DL’03). Volume 81 of CEUR

Workshop Proceedings., Rome, Italy (2003)

15. Organization, W.H.: International Classiﬁcation of Functioning, Disability and Health.

World Health Organization, Geneva (2001)