REPRESENT

ATION AND EFFICIENT MANAGEMENT

OF MULTI-VERSION CLINICAL GUIDELINES

Fabio Grandi

Dip. di Elettronica, Informatica e Sistemistica, Alma Mater Studiorum, Universit

a di Bologna

Viale Risorgimento 2, I-40136, Bologna, Italy

Keywords:

Clinical guidelines, Document retrieval, Temporal database, Ontologies, Personalization, Versioning, XML.

Abstract:

While the world wide web user is suffering from the disease caused by information overload, for which

personalization is one of the treatments which works, a physician accessing web-based clinical guideline

repositories is not immune from contagion. This seems a good reason to prescribe a personalization treatment

also to the professional user of a computerized clinical guideline library. To this end, we apply to clinical

guidelines solutions we previously developed for norm texts in the legal domain, and show how multi-version

representation capabilities and personalization query facilities can be added to their management.

1 INTRODUCTION

Clinical guidelines are deﬁnitions of “best practices”

encoding and standardizing clinical procedures for a

given disease. The advantages of adopting computer-

based guidelines as a support for improving the

work of physicians and optimizing hospital activi-

ties have been acknowledged by many authors and

several computer systems have been developed (see

e.g. (Fridsma, 2001; Gordon and Christensen, 1995)).

Clinical guidelines are subject to continuous develop-

ment and revision by committees of expert physicians

and health authorities and, thus, multiple versions co-

exist as a consequence of the clinical and healthcare

activity.

In this paper, we propose to apply to the manage-

ment of clinical guidelines some techniques we pre-

viously developed for norm documents in the legal

domain (Grandi et al., 2005; Grandi et al., 2009b),

which present strong similarities. Hence, we will in-

troduce solutions to model and to provide personal-

ized access to multi-version clinical guidelines, which

can be stored both in textual and in executable for-

mat in an XML repository. The XML language has

already been proposed by many authors and adopted

in several research projects (e.g. (Dubey and Chueh,

2000; Shiffman et al., 2000; Buchtela et al., 2008)) as

a suitable means to encode clinical guidelines. Hence,

our approach can be considered as a compatible ex-

tension of such proposals, to which we aim at adding

multi-version representation capabilities and person-

alization query facilities.

PATIENT PROFILE

CONTEXT INFO

TEMPORAL PERSPECTIVE

MULTI-VERSION

GUIDELINE REPOSITORY

PERSONALIZATION ENGINE

PERSONALIZED

GUIDELINE VERSION

Figure

1: Personalized access to multi-version guidelines.

To this end, we will describe how a multi-version

XML data model and the prototype system we devel-

oped for e-Government applications can be applied to

the representation and management of multi-version

clinical guidelines. In this way, multiple temporal

perspectives, patient proﬁle and context information

can be used by an automated personalization service

to build a guideline version tailored to a speciﬁc use

case (see Fig. 1).

The paper is organized as follows. In Section 2,

temporal and semantic versioning of clinical guide-

lines is introduced with reference to advanced appli-

cation requirements. Section 3 is devoted to the de-

scription of a multidimensional XML data model sup-

porting temporal and semantic versioning of guide-

lines. In Section 4, a prototype system efﬁciently

implementing the personalization engine sketched in

Grandi F. (2010).

REPRESENTATION AND EFFICIENT MANAGEMENT OF MULTI-VERSION CLINICAL GUIDELINES.

In Proceedings of the Third International Conference on Health Informatics, pages 54-61

DOI: 10.5220/0002727400540061

 SciTePress

Fig. 1 is brieﬂy described. Conclusions will ﬁnally be

found in Section 5.

2 MULTI-VERSION GUIDELINES

The fast evolution of medical knowledge and the dy-

namics involved in clinical practice imply the coex-

istence of multiple temporal versions of the clini-

cal guideline documents stored in a repository, since

guidelines are continually subject to amendments and

modiﬁcations. In fact, it is crucial to reconstruct, bor-

rowing the term from the legal ﬁeld, the consolidated

version of a guideline as produced by the application

of all the modiﬁcations it underwent so far, that is the

form in which it currently belongs to the state-of-the-

art of clinical practice and, thus, must be applied to

patients today. However, also past versions are still

important, not only for historical reasons: for exam-

ple, a physician might be called upon to justify his/her

actions for a given patient P at a time T on the basis

of the clinical guideline versions which were valid at

time T and applicable to the pathology of patient P. In

other words, temporal concerns are important in the

medical domain as they are in the legal domain and,

thus, a guideline management system should be able

to retrieve or reconstruct on demand any temporal ver-

sion of a given clinical guideline to meet advanced

application requirements.

Moreover, another kind of versioning, which we

will call semantic versioning, plays a fundamental

role, because clinical guidelines or some of their parts

have limited applicability with respect, for instance,

to the population of patients. In fact, a given guide-

line (e.g. involving treatment of heart diseases) may

contain different recommendations which are not uni-

formly applicable to the same classes of patients: one

general therapy may be non applicable to persons who

suffer from some metabolic disorders (e.g. diabetes

mellitus) or chronic diseases (e.g. kidney failure)

or present some addiction (e.g. cocaine); one ﬁrst-

choice drug may not be given to patients who are al-

ready under treatment with possibly interacting drugs

(e.g. anticoagulants), or show genetic or acquired hy-

persensitivity or intolerance to some substances (e.g.

patients with enzymatic defects or documented aller-

gies), and so on. Hence, when dealing with a speciﬁc

patient case, a physician may be interested in ﬁnd-

ing a personalized version of a clinical guideline, that

is a version tailored to the patient’s health state and

anamnesis, only containing recommendations which

are safely and effectively applicable to his/her per-

sonal case.

In addition to linking guidelines to classes of pa-

tients, semantic versioning can also involve more

generic applicability contexts (e.g. hospitals without

PET diagnostic equipment, or selected centers taking

part to a clinical trial), which might require the appli-

cation of a particular version of the general guideline,

which may also no longer be part of the consolidated

state-of-the-art guideline. For instance, consider ver-

sion v1 of a clinical guideline G which prescribes a

biopsy to conﬁrm a cancer diagnosis but has been su-

perseded by a new version v2 which introduces a PET

scan for the same cancer diagnosis, making in most

cases the biopsy unnecessary. However, in some hos-

pital H which is not equipped with a PET scanner, the

right version of G to be followed is v1, although no

longer considered valid by the medical community.

Therefore, the applicable version of the guideline for

context H is G(v1), with biopsy as a mandatory diag-

nostic means. This example also shows how temporal

and limited applicability aspects may also interplay in

the production and management of versions.

2.1 Temporal Versioning

As far as temporal versioning is concerned, several

independent time dimensions are involved in the rep-

resentation and management of clinical guidelines, in

particular when we consider an environment also sup-

porting the guideline authoring and approval process.

Relevant time dimensions include valid, event, avail-

ability, proposal and acceptance times (Combi and

Montanari, 2001; Terenziani et al., 2005). Even con-

sidering an environment where only approved guide-

lines are stored, and retrieved by ﬁnal users to be con-

sulted or followed, at least two time dimensions are

relevant:

Validity Time. It is the time the guideline is con-

sidered in force by the medical community and, thus,

is applied to patients. It has the same semantics of

valid time as in temporal databases (Jensen and et al.,

1998), since it represents the time the guideline actu-

ally belongs to the state-of-the-art of clinical practice.

Efﬁcacy Time. Borrowing the term from the legal

domain, it is the time the guideline can be applied to a

concrete case. It usually corresponds to validity, but it

might be the case that an obsolete, superseded guide-

line continues to be applicable to a limited number of

cases. While such cases exist, the guideline continues

its efﬁcacy though no longer considered in force.

Notice that validity and efﬁcacy time both have

the semantics of valid time but represent different and

independent valid time notions. Both are necessary

to correctly deal with cases as the one in the last de-

REPRESENTATION AND EFFICIENT MANAGEMENT OF MULTI-VERSION CLINICAL GUIDELINES

Figure 2: A sample ontology, where each class has a name

and is associated to a (pre-order,post-order) pair.

scribed example: the guideline version G(v1) for the

applicability context H can still be selected today as

its efﬁcacy includes current time, although its validity

does not. Furthermore, in addition to the time dimen-

sions which model the dynamics of guidelines in the

real world, transaction time (Jensen and et al., 1998)

plays an important role when automatic management

of information through computer systems is involved

and, thus, should never be neglected, since it allows to

execute retro- or pro-active modiﬁcations and to keep

track of their execution for audit purposes. For ex-

ample, it might be the case that a physician makes a

wrong decision in choosing a drug following the pro-

visions of a guideline retrieved from the system when

the returned consolidated version is actually out-of-

date; the decision is taken while a modiﬁed version

of the guideline (e.g. involving the adoption of some

more effective and less potentially dangerous drug) is

already available but has not been stored in the infor-

mation system yet. Hence, transaction time is needed

to ascertain a posteriori that the correct version was

stored retroactively and, thus, the physician acted in

good faith.

Temporal versioning along multiple time dimen-

sions can be added to documents in an XML reposi-

tory by making temporal the XML encoding (Dyreson

and Grandi, 2009), that is introducing timestamps as

annotations in the XML document.

2.2 Semantic Versioning

Semantic applicability of multi-version resources can

be deﬁned with reference to domain ontologies.

Ontologies (Guarino, 1998; Gruber, 2009), which

are conceptualizations of a domain into a machine-

understandable format, have recently become quite

popular with the advent of the semantic web (Berners-

Lee et al., 2001), where the introduction of common

reference ontologies is necessary to allow information

and its interpretation to be shared by both human and

automatic agents.

Appropriate applicability of clinical guidelines to

individual patients can be deﬁned according to a con-

sensual taxonomy of diseases, like the ICD-10 en-

dorsed by the World Health Organization (ICD-10,

2009) or the MeSH Section C maintained by the US

National Library of Medicine (MeSH-C, 2009). For

instance, consider Fig. 2, which depicts a small por-

tion of a medical ontology representing a classiﬁca-

tion of principal heart diseases. Notice that, at this

stage of the research, we deal with “tree-like” ontolo-

gies deﬁned as class taxonomies induced by the IS-

A relationship. This will allows us to exploit during

query processing the pre-order and post-order prop-

erties of trees in order to enumerate the nodes and

check ancestor-descendant relationships between the

classes; such codes are displayed in the upper left cor-

ner of the ontology classes in the Figure, in the form:

(pre-order,post-order). For instance, the class “My-

ocardial ischemia” has pre-order “3” which is also its

identiﬁer, whereas its post-order is “6”. Before the

personalization engine can be used to build a guide-

line version tailored to a speciﬁc patient, the patient

must be classiﬁed with respect to the disease ontol-

ogy, on the basis of medical records by means of

a suitable reasoning service (Grandi et al., 2009b),

or through a proﬁle explicitly supplied by the physi-

cian. Moreover, additional semantic versioning coor-

dinates, referencing speciﬁc domain ontologies, can

also be considered to model context-dependent appli-

cability of guidelines.

Hence, in XML resource repositories, reference

to ontology concepts (e.g. using class identiﬁers like

those in Fig. 2) can be added to the resource represen-

tation and storage as a new versioning coordinate. In

this way, applicability annotations can be embedded

in the guideline documents to be used by automatic

personalization tools. Obviously, also the annotation

of clinical guidelines which deﬁnes their semantic

versioning must be effected by medical domain ex-

perts, as part of the guideline drafting and approval

process itself. Whenever an ontology deﬁnition is

changed, temporal versions of the ontology also must

be maintained, as the temporal perspectives for nav-

igating the ontology and for searching the guideline

repository must be same for consistency reasons. The

ontology temporal versioning techniques introduced

in (Grandi and Scalas, 2009) can be used to this pur-

pose.

One of the global effects of versioning is an in-

crease in the number or size of the documents to

be stored, also depending on the fact that different

versions of the same document are stored as sepa-

rate XML ﬁles or are arranged into a single multi-

HEALTHINF 2010 - International Conference on Health Informatics

version XML ﬁle, owing to a uniform encoding of

variant parts within the document structure. The lat-

ter solution, which is our choice, is often unavoid-

able in order to keep the growth of the storage space

under control, especially when different versions of

the same document may differ by a few nodes only.

Personalization, which has shown to be a powerful

tool to cope with information overload on the inter-

net (Riecken, 2000), can also be particularly effective

when used in the management of large XML reposito-

ries of versioned documents (Grandi et al., 2009b). In

this case, the adoption of personalization techniques

can prevent in most cases users to have to go through

a huge amount of irrelevant information to ﬁnd out

the right version(s) of the one of interest and, thus,

might help to make their search faster and more accu-

rate. Hence, personalization based on semantic ver-

sioning may improve the quality of the interaction

with the user by further focusing the search on re-

ally relevant versions only, which is a desirable fea-

ture for clinical guideline management. For example,

one of the acknowledged most relevant obstacles in

the use and dissemination of guidelines (Cabana et al.,

1999) is the need for adapting them to constraints

in local settings (e.g. concerning available hospital

resources and practitioners’ skills). Management of

multi-version guidelines with context-based semantic

personalization might help to overcome this problem

(Fridsma et al., 1996; Terenziani et al., 2004). Other

use cases requiring a sort of location-based semantic

personalization can also be found: for instance, con-

sider a guideline involving the recommendation of a

new drug non yet registered in a given country, or in-

troducing a new protocol only available in selected

medical centers participating to an experimental pro-

gram: the actual contents of the guideline should be

changed according to the place where the guideline is

retrieved or executed.

3 AN XML DATA MODEL FOR

MULTI-VERSION GUIDELINES

In this Section, we introduce a multi-version XML

document model supporting multiple temporal and

semantic versioning coordinates. In doing this, we do

not refer to a speciﬁc document structure (e.g. deﬁned

via a DTD or XML Schema), but we rather introduce

a versioning annotation scheme which can be applied

to any generic XML resource. In particular, it can

be easily adapted to available proposals for the XML

encoding of clinical guidelines, including those de-

scribed in (Dubey and Chueh, 2000; Shiffman et al.,

2000; Buchtela et al., 2008).

RECOMMENDATIONS

1. IDENTIFICATION OF PATIENTS

WITH RISK OF UNSTABLE ANGINA

...

2. INITIAL EVALUATION AND MANAGEMENT

...

3. EARLY HOSPITAL CARE

3.1. Initial Treatment Strategy

...

3.2. Drug Therapy

3.2(v1). Anti-Ischemic and Analgesic Therapy

3.2(v1).1. Therapy with nitrates

...

3.2(v1).2. Therapy with beta-blockers

3.2(v1).2(v1). ...administration of drug D1...

3.2(v1).2(v2). ...administration of drug D2...

3.2(v1).2(v3). ...administration of drug D3...

...

3.2(v1).3. Therapy with ACE inhibitors

...

3.2(v2). Antiplatelet/Anticoagulant Therapy

...

4. CORONARY REVASCULARIZATION

...

5. LATE HOSPITAL CARE

...

Figure 3: The structure of a fragment of a sample multi-

version clinical guideline.

We start by formally deﬁning as version a piece

of text within a guideline document, with a common

temporal and semantic pertinence. Owing to the def-

inition, a version can be assigned a timestamp and an

applicability annotation to uniquely deﬁne its tempo-

ral and semantic pertinence. Obviously, different ver-

sions of the same object must differ in their temporal

and/or semantic pertinence.

For the sake of simplicity, but without loss of gen-

erality, we only consider in the examples which fol-

low one time dimension (i.e. validity) and one se-

mantic dimension (i.e. reference to classes in an on-

tology of diseases like the one in Fig. 2). Let us con-

sider as running example the clinical guideline frag-

ment in Fig. 3, involving recommendations for the

treatment of unstable angina patients. The ﬁgure dis-

plays the text organization, which has a three-level

section structure, where section 3.2. has two differ-

ent versions, namely 3.2(v1) and 3.2(v2), whereas

section 3.2(v1).2 has three different versions, namely

3.2(v1).2(v1), 3.2(v1).2(v2) and 3.2(v1).2(v3). The

multi-version XML encoding of such guideline frag-

ment is shown in Fig. 4.

In the XML encoding, we use the

ele-

ment to delimit the boundaries of a version within the

document. The

<valid>

and

elements

REPRESENTATION AND EFFICIENT MANAGEMENT OF MULTI-VERSION CLINICAL GUIDELINES

...

...

<title>Early Hospital Care</title>

...

<title>Drug Therapy</title>

<title>Anti-ischemic and Analgesic Therapy</title>

...

<title>Therapy with beta-blockers</title>

...

administration of drug D1

...

</version>

...

administration of drug D2

...

</version>

...

administration of drug D3

...

</version>

...

</section>

...

</version>

<title>Antiplatelet/Anticoagulant Therapy</title>

...

</version>

...

</section>

...

</version>

...

</section>

...

</version>

</recommendations>

...

Figure 4: An XML fragment showing the multi-version en-

coding of the guideline in Fig. 3.

are then used to assign the temporal and semantic per-

tinence, respectively, to the version which contains

them. Validity and applicability properties are inher-

ited by descendant nodes in the XML tree-structure

unless locally redeﬁned with a new version deﬁnition.

Therefore, there is no reason to repeat the valid or ap-

plies annotation when the pertinence is not changed

from the ancestor version in the XML tree-structure.

In general, redeﬁnition may involve only a subset of

the versioning dimensions, while the others dimen-

sions are inherited.

With reference also to Fig. 3, the XML frag-

ment in Fig. 4 shows, within the outermost

element, a hierarchical struc-

ture based on three levels of sections. The

element is composed of one

version, which deﬁnes its global semantic and tempo-

ral pertinence, that is applicable to class C3 in the on-

tology in Fig. 2 (patients with myocardial ischemia)

and valid from 1980 on. It is made of several ﬁrst-

level sections (see also Fig. 3), of which only section

3 is evidenced in the Figure. Such a section, made of

only one version to specify applicability to ontology

class C4 (patients with angina pectoris), deals with

Early Hospital Care. Its temporal pertinence is inher-

ited from the container element.

In general, by means of redeﬁnitions we can in-

troduce, for each part of a document, complex valid-

ity and applicability properties including extensions

or restrictions with respect to ancestors. For instance,

the applicability assignment to section 3 which we

just described is a restriction and the attribute

used to this end. Actually, the applicability assigned

to the version is the intersection of the

value and

of the value inherited by the ancestor version (in this

case C4∩C3, which equals C4 since it is a subclass

of C3). The same applies to second-level section 3.2

(entitled “Drug Therapy”), whose ﬁrst version (enti-

tled “Anti-ischemic and Analgesic Therapy”) applies

to class C5 (unstable angina), which is also a re-

striction, whereas the second version (entitled “An-

tiplatelet/Anticoagulant Therapy”) is also applicable

to class C7 (myocardial infarction), which is an exten-

sion indeed. Attribute

also

is used in this case, and

the applicability assigned to the version is the union

of the

also

value and of the value inherited by the

ancestor version (class C4∪C7). In other words, the

contents of section 3.2(v2) both apply to angina pec-

toris and myocardial infarction patients.

The third-level section 3.2(v1).2 entitled “Ther-

apy with Beta-blockers” is made of several versions,

each one dealing with the administration of a speciﬁc

drug and having its own temporal pertinence, whereas

the (inherited) applicability is the same (namely C5,

unstable angina). In order to derive the validities of

the three drugs shown in the Figure, we assume the

recommendations underwent the evolution which fol-

lows. Drug D1 was introduced in 1980 and then re-

placed by the drug D2 in 1991. However, the use of

drug D2 was suspended from 1999 to 2000, period

during which it had been under investigation since

suspected of causing adverse reactions. In 2004, due

to evidence of long-term adverse effects, D2 was def-

initely withdrawn. Drug D3 has been introduced in

1985. Hence, the resulting history of recommended

beta-blockers according to the guideline in Fig. 3

(which will in fact correspond to the answers to a

sequence of snapshot queries issued on the multi-

version document) is the following:

− from 1980 to 1984: drug D1

− from 1985 to 1990: drugs D1 and D3

− from 1991 to 1998: drugs D2 and D3

− from 1999 to 2000: drug D3

HEALTHINF 2010 - International Conference on Health Informatics

− from 2001 to 2003: drugs D2 and D3

− from 2004 on: drug D3

As for 3.2(v1).2(v2) in the Figure, versions can be as-

signed multiple intervals as validity: this corresponds

to adopt temporal elements (Gadia, 1988; Jensen and

et al., 1998), that is disjoint union of intervals, as

timestamps.

3.1 Operations

The multi-version XML data model can be equipped

with two basic operators for the management of

guideline authoring and maintenance: one devoted to

change the textual content of a guideline portion and

the other to allow modiﬁcations to the temporal and

semantic pertinence of a given version. The former

can be used for deletion of (a part of) the guideline

(abrogation), or the introduction of a new part of the

guideline (integration), or the replacement of (a part

of) the guideline (substitution). The latter can be used

to deal with the time/applicability extension or restric-

tion of (part of) the guideline. Such operators, in order

to preserve the well-formedness of the version struc-

ture and the inheritance semantics, can be deﬁned in

a similar way as the ones deﬁned for multi-temporal

norm documents in (Grandi et al., 2005).

Clinical guideline repositories, like the US Na-

tional Guideline Clearinghouse (NGC, 2009) or the

UK National Library of Guidelines (NLG, 2009), are

usually managed by traditional information retrieval

systems where users are allowed to access their con-

tents by means of keyword-based queries expressing

the subjects they are interested in. Adopting a system

like the one described in (Grandi et al., 2009b) that we

developed for norm documents, users are offered the

possibility of expressing temporal and semantic spec-

iﬁcations for the reconstruction of a consistent version

of the retrieved guideline.

In particular, the queries can contain four types

of constraints: temporal, structural, textual and appli-

cability. Such constraints are completely orthogonal

and allow the users to perform very accurate searches

in the XML guideline repository. Let us focus ﬁrst on

the applicability constraint. Consider again the ontol-

ogy in Fig. 2 and guideline fragment in Fig. 4: for

the treatment of John Smith, an “infarctuated” patient

(i.e. belonging to class C7), the sample recommenda-

tions in Fig. 3 will be selected as pertinent, but only

the second version of Section 3.2 will be actually pre-

sented as applicable. Furthermore, the applicability

constraint can be combined with the other three ones

in order to fully support a multi-dimensional retrieval.

For instance, a physician (or an health insurance ofﬁ-

cer) could be interested in all the guidelines ...

• ... which have a section whose title (structural

constraint) contains the word anticoagulant (tex-

tual constraint), ...

• ... which were valid between 2007 and 2008 (tem-

poral constraint), ...

• ... and which are applicable to a patient suffering

from unstable angina (applicability constraint).

More precisely, the system is able to answer

queries having the XQuery syntax in Fig. 5, where

textConstr

tempConstr

, and

applConstr

are suit-

able functions allowing the speciﬁcation of the tex-

tual, temporal and applicability constraints, respec-

tively (the structural constraint is implicit in the XPath

expressions used in the XQuery statement).

4 IMPLEMENTATION

The personalization engine in Fig. 1, which is ca-

pable to execute queries like the one in Fig. 5, has

been implemented as a prototype Multi-version XML

Query Processor. The prototype code is written in

Java JDK 1.5 and employs ad-hoc data structures (re-

lying on embedded “light” DBMS libraries) and algo-

rithms which allow users to reconstruct on-the-ﬂy the

desired personalized version of the XML guideline,

by means of a multi-version extension of the holistic

twig join approach (Bruno et al., 2001). Guidelines

are stored in the XML repository using an indexing

scheme based on multi-version inverted indices, that

is an extension with timestamps and semantic anno-

tations of the indexing solution proposed in (Zhang

et al., 2001). In practice, the query processing al-

gorithm implements the temporal slice operator pro-

posed in (Mandreoli et al., 2006), to which the pro-

cessing of semantic constraints has been added, with-

out an appreciable overhead. In fact, thanks to the

properties of the adopted pre- and post-order encod-

ing of the ontology classes, applicability constraints

can be very efﬁciently tested during query processing

by means of simple comparisons. A detailed presen-

tation of the deployed data structures and holistic join

techniques, together with a related work discussion on

these topics, can be found in (Grandi et al., 2009a).

As a result, we obtain a high overall query pro-

cessing efﬁciency mated with low memory require-

ments. In order to evaluate the performance of the

prototype, a speciﬁc query benchmark was built and

several exploratory experiments were conducted to

test the personalization engine behavior under dif-

ferent workloads. The experiments have been ef-

fected on a Pentium 4 3Ghz Windows XP Profes-

sional workstation, equipped with 1GB RAM and a

REPRESENTATION AND EFFICIENT MANAGEMENT OF MULTI-VERSION CLINICAL GUIDELINES

FOR $a IN guidelines.xml

WHERE textConstr ($a//section/title/text(), ’anticoagulant’)

AND tempConstr (’vTime OVERLAPS PERIOD(’2007-01-01’,’2008-12-31’)’)

AND applConstr (’C5’)

RETURN $a

Figure 5: An XQuery-equivalent query executable on a clinical guideline personalization system.

160GB EIDE disk with NT ﬁle system (NTFS). Test

were performed on three XML document collections

of increasing size (namely 5,000, 10,000 and 20,000

guidelines, with a total size of 120MB, 240MB and

480MB, respectively). In all collections the guide-

lines were synthetically generated by means of a suit-

able tool, which is able to produce XML documents

compliant to our multi-version model under different

parameter conﬁgurations. For each collection, the av-

erage, minimum and maximum document size was

24KB, 2KB and 125KB, respectively. Experiments

were conducted by submitting queries of ﬁve differ-

ent types, mixing in various ways structural, textual,

temporal and applicability constraints.

The system behavior showed a good efﬁciency in

every context, providing a response time (including

query analysis, retrieval of the qualifying guideline

parts and reconstruction of the result) of a few sec-

onds for most of the queries. Moreover, the selec-

tivity of the query predicates does not impair per-

formances, even when large amounts of documents

containing some (typically small) relevant portions

have to be retrieved. The system is able to deliver

a fast and reliable performance in all cases, since it

practically avoids the retrieval of useless document

parts. For the same reasons, the main memory re-

quirements of the Multi-version XML Query Proces-

sor are quite limited, less than 5% with respect to an

approach like the one adopted in (Grandi et al., 2005),

where complete documents are retrieved with a tra-

ditional XML engine working on structural and tex-

tual constraints, and then temporal and applicability

constraints are applied using a DOM representation

to prune out non-qualifying XML nodes. Notice that

this property is very interesting for a system which is

likely to run in a highly concurrent multi-user envi-

ronment, since memory requirements are not crucial

for performance. The prototype system also showed

a good scalability behavior in every type of query set-

ting, as the computing time for the same query al-

ways grows linearly with the number of documents.

Full details on performance evaluation can be found

in (Grandi et al., 2009a; Grandi et al., 2009b).

5 CONCLUSIONS

In this paper, we applied to the representation and

management of clinical guidelines some techniques

we previously developed for norm documents in the

legal domain (Grandi et al., 2009a; Grandi et al.,

2009b). In particular, we introduced solutions to

model and to provide personalized access to multi-

version guidelines, supporting multiple temporal and

semantic versioning coordinates. The proposal in-

volves the deﬁnition of a multi-version XML data

model and the implementation of a prototype person-

alization engine.

Preliminary experimental work on query perfor-

mance, with repositories of syntectic XML docu-

ments, showed encouraging results. In particular, the

personalization engine proved to be very efﬁcient in a

large set of experimental situations and showed excel-

lent scale-up ﬁgures with varying load conﬁgurations.

We underline that the very same techniques we

presented for personalized access to multi-version

textual guideline documents can also be applied to the

enactment of workﬂows implementing multi-version

clinical guidelines, provided that workﬂows are spec-

iﬁed using an XML-based deﬁnition language, like

BPEL (WS-BPEL, 2009) or XPDL (XPDL, 2009),

which can be enriched as well with temporal and se-

mantic annotations in order to deﬁne versions (Grandi

et al., 2009b).

Future work will consider the improvement of the

approach to cope with more advanced application re-

quirements (e.g. relaxing of constraint of tree-like

ontologies) and the completion of the technological

infrastructure required to set up the personalization

platform with the design and implementation of aux-

iliary services (e.g. for automatic patient classiﬁca-

tion with respect to the disease ontology). Further

work will also include the assessment of our devel-

oped system in a concrete working environment, with

real users and in the presence of a repository of real

clinical guidelines.

HEALTHINF 2010 - International Conference on Health Informatics

REFERENCES

Berners-Lee, T., Hendler, J., and Lassila, O. (2001). The

semantic web. Scientiﬁc American, 284(5):34–43.

Bruno, N., Koudas, N., and Srivastava, D. (2001). Holistic

twig joins: Optimal xml pattern matching. In Proc. of

SIGMOD 2001, pages 310–321.

Buchtela, D., Pele

ska, J., Vesel

y, A., Zv

arov

a, J., and

Zvolsk

y, M. (2008). An xml-based format for guide-

line interchange and execution. In Proc. of MIE 2008,

pages 151–156.

Cabana, M., Rand, C., Powe, C., Wu, A., Wilson, M., Ab-

boud, P., and Rubin, H. (1999). Why don’t physicians

follow clinical practice guidelines? a framework for

improvement. Journal of American Medical Associa-

tion, 282(15):1458–1465.

Combi, C. and Montanari, A. (2001). Data models with

multiple temporal dimensions: Completing the pic-

ture. In Proc. of CAiSE 2001, pages 187–202.

Dubey, A. and Chueh, H. (2000). An xml-based format for

guideline interchange and execution. In Proc. of AMIA

2000, pages 205–209.

Dyreson, C. and Grandi, F. (2009). Temporal xml. In

Ozsu,

M. and Liu, L., editors, Encyclopedia of Database

Systems. Springer-Verlag (in press).

Fridsma, D. (2001). Special issue on workﬂow management

and clinical guidelines. AIM Journal, 22(1):1–80.

Fridsma, D., Gennari, J., and Musen, M. (1996). Mak-

ing generic guidelines site-speciﬁc. In Proc. of AMIA

1996, pages 597–601.

Gadia, S. (1988). A homogeneous relational model and

query languages for temporal databases. ACM Trans.

on Database Systems, 13(3):418–448.

Gordon, C. and Christensen, J. (1995). Health Telematics

for Clinical Guidelines and Protocols. IOS Press.

Grandi, F., Mandreoli, F., and Martoglia, R. (2009a). Is-

sues in personalized access to multi-version xml doc-

uments. In Pardede, E., editor, Open and Novel Issues

in XML Database Applications, pages 199–230. IGI

Global.

Grandi, F., Mandreoli, F., Martoglia, R., Ronchetti, E.,

Scalas, M., and Tiberio, P. (2009b). Ontology-based

personalization of e-government services. In Mourlas,

C. and Germanakos, P., editors, Intelligent User Inter-

faces, pages 167–203. IGI Global.

Grandi, F., Mandreoli, F., and Tiberio, P. (2005). Tem-

poral modelling and management of normative docu-

ments in xml format. Data & Knowledge Engineering,

54(3):327–354.

Grandi, F. and Scalas, M. (2009). The valid ontology: A

simple owl temporal versioning framework. In Proc.

of SEMAPRO 2009, pages 98–102.

Gruber, T. (2009). Ontology. In

Ozsu, M. and Liu, L.,

editors, Encyclopedia of Database Systems. Springer-

Verlag (in press).

Guarino, N., editor (1998). Formal Ontology in Information

Systems. IOS Press.

ICD-10 (2009). International statistical classiﬁcation of dis-

eases and related health problems. World Health Or-

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

Jensen, C. and et al., C. D. (1998). The Consensus Glos-

sary of Temporal Database Concepts - February 1998

Version. In Etzion, O., Jajodia, S., and Sripada, S., ed-

itors, Temporal Databases — Research and Practice,

pages 367–405. Springer-Verlag.

Mandreoli, F., Martoglia, R., and Ronchetti, E. (2006). Sup-

porting temporal slicing in xml databases. In Proc. of

EDBT 2006, pages 295–312.

MeSH-C (2009). Medical subject headings - section

c: Diseases. US National Library of Medicine,

http://www.nlm.nih.gov/mesh/2009/mesh browser/

MeSHtree.C.html.

NGC (2009). National guideline clearinghouse. US

Agency for Healthcare Research and Quality,

http://www.guideline.gov.

NLG (2009). National library of guidelines. UK Na-

tional Institute for Health and Clinical Excellence,

http://www.library.nhs.uk/GUIDELINESFINDER/.

Riecken, D. (2000). Personalized views of personalization.

Communications of the ACM, 43(8):27–28.

Shiffman, R., Karras, B., Agrawal, A., Chen, R., Marenco,

L., and Nath, S. (2000). Gem a proposal for a more

comprehensive guideline document model using xml.

Journal of AMIA, 7(5):488–497.

Terenziani, P., Montani, S., Bottrighi, A., Molino, G., and

Torchio, M. (2005). Clinical guidelines adaptation:

Managing authoring and versioning issues. In Proc.

of AIME 2005, pages 151–155.

Terenziani, P., Montani, S., Bottrighi, A., Torchio, M.,

Molino, G., and Correndo, G. (2004). A context-

adaptable approach to clinical guidelines. In Proc. of

MEDINFO 2004, pages 169–173.

WS-BPEL (2009). The web services business pro-

cess execution language. WfMC Coalition,

http://www.wfmc.org/standards/docs.htm.

XPDL (2009). The xml process deﬁnition lan-

guage. OASIS Organization, http://www.oasis-

open.org/committees/tc home.php?wg abbrev=wsbpel.

Zhang, C., Naughton, J., DeWitt, D., Luo, Q., and Lohman,

G. (2001). On supporting containment queries in re-

lational database management systems. In Proc. of

SIGMOD 2001, pages 425–426.

REPRESENTATION AND EFFICIENT MANAGEMENT OF MULTI-VERSION CLINICAL GUIDELINES