Development of a Gestational Diabetes Computer Interpretable

Guideline using Semantic Web Technologies

Garazi Artola

, Jordi Torres

, Nekane Larburu

1,2

, Roberto Álvarez

1,2

and Naiara Muro

1,2,3,4,5

Vicomtech Research Centre, Mikeletegi Pasalekua 57, 20009, San Sebastian, Spain

Biodonostia Health Research Institute, P. Doctor Begiristain s/n, 20014, San Sebastian, Spain

UPMC, Univ. Paris 06, Paris, France

INSERM, Université Paris 13, Paris, France

LIMICS, Sorbonne Paris Cité, UMR S 1142, Paris, France

Keywords: Computer Interpretable Guideline, Semantic Web Technologies, Ontology, Gestational Diabetes, Decision

Support System.

Abstract: The benefits of following Clinical Practice Guidelines (CPGs) in the daily practice of medicine have been

widely studied, being a powerful method for standardization and improvement of medical care quality.

However, applying these guidelines to promote evidence-based and up-to-date clinical practice is a known

challenge due to the lack of digitalization of clinical guidelines. In order to overcome this issue, the use of

Clinical Decision Support Systems (CDSS) has been promoted in clinical centres. Nevertheless, CPGs must

be formalized in a computer interpretable way to be implemented within CDSS. Moreover, these systems are

usually developed and implemented using local setups, and hence local terminologies, which causes lack of

semantic interoperability. In this context, the implementation of Semantic Web Technologies (SWTs) to

formalize the concepts used in guidelines promotes the interoperability and standardization of those systems.

In this paper, an architecture that allows the formalization of CPGs into Computer Interpretable Guidelines

(CIGs) supported by an ontology in the gestational diabetes domain is presented. This CIG has been

implemented within a CDSS and a mobile application has been developed for guiding patients based on up-

to-date evidence based clinical guidelines.

1 INTRODUCTION

Clinical practice is based on the latest and most

reliable clinical evidence to provide the best

healthcare to patients. During the last years, studies

have shown the benefits of following Clinical

Practice Guidelines (CPGs) in the daily practice of

medicine (Grimshaw and Russell, 1993), being a

powerful method for standardization and

improvement of the medical care quality. According

to the Institute of Medicine’s (IOM) definition, CPGs

are “systematically developed statements to assist

practitioner and patient decisions about appropriate

health care for specific clinical circumstances”

(Institute of Medicine, 1990).

However, applying these guidelines to promote

the best evidence-based and most up-to-date clinical

practice is a known challenge. There is a lack of

digitalization of the guidelines, which makes it

difficult to maintain them updated in a dynamic way

and implement them in computerized systems.

In order to overcome these issues, Clinical

Decision Support Systems (CDSS) that formalize

guidelines in a computer interpretable way (i.e. as

Computer Interpretable Guidelines or CIGs) are

promoted. In this context, the application of Semantic

Web Technologies (SWTs) to formalize the

guidelines’ concepts could be a key to promote the

interoperability and standardization of the clinical

knowledge, by giving the opportunity to pursue a

reuse of ontologies.

In this paper, an architecture that will allow the

formalization of CPGs into CIGs, supported by an

ontology to assure a semantic validity of all the

formalized information is presented. As a use case,

the implementation of a gestational diabetes CIG is

described. Moreover, a mobile based application is

presented as the front-end of the CDSS for a patient-

oriented guidance.

This paper is organized as follows. Section 2

describes the state of the art done about the different

concepts needed in this work. Section 3 introduces the

104

Artola, G., Torres, J., Larburu, N., Álvarez, R. and Muro, N.

Development of a Gestational Diabetes Computer Interpretable Guideline using Semantic Web Technologies.

DOI: 10.5220/0008068001040114

In Proceedings of the 11th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2019), pages 104-114

ISBN: 978-989-758-382-7

methodology used during the development of the

work. Section 4 explains a practical scenario, and

finally, Section 5 presents our conclusions and future

guidelines.

2 STATE OF THE ART

In this chapter, the state of the art in Semantic Web

Technologies (SWTs), Clinical Practice Guidelines

(CPGs) for Gestational Diabetes Mellitus (GDM)

domain, Computer Interpretable Guidelines (CIGs),

and Clinical Decision Support Systems (CDSS) is

described.

2.1 Semantic Web Technologies

(SWTs)

The technological breakthrough in biomedical

engineering and health informatics is producing a

huge amount of data coming from different sources,

which causes a limited interoperability of healthcare

systems (Kolias et al., 2014). The Semantic Web

is a

Web of data that wants to enable computers to

interpret and process information on the World Wide

Web. SWTs provide the tools to process the data in a

more effective way, create the framework for

interoperability between systems and integrate data

from various sources.

In this context, many researchers have made use

of these technologies to cope with problems related to

semantic interoperability of ontologies or clinical

datasets. As an example, the work presented by El-

Sappagh et al. (El-Sappagh et al., 2018) introduced

the Diabetes Mellitus Treatment Ontology (DMTO)

as a basis for shared-semantics, domain-specific,

standard, machine-readable, and interoperable

knowledge relevant to type 2 diabetes mellitus

(T2DM) treatment. However, to the best of our

knowledge, there is no available gestational diabetes-

centered ontology in the main open source ontology

repositories, such as BioPortal

As can be seen, ontologies are becoming an

important tool in the field of semantics for

interoperability. In this sense, there are different

ontology languages available for representing

information on the semantic web, such as RDF

Schema or Web Ontology Language (OWL). RDF

Schema allows to build a simple hierarchy of

concepts and properties, while OWL has the ability to

https://www.w3.org/standards/semanticweb/

http://bioportal.bioontology.org/

http://www.snomed.org/snomed-ct/five-step-briefing

specify far more about the properties and classes,

adding semantics to the schema (e.g. defining two

concepts as equivalent or inferring implicit facts).

One of the widely used ontology editors is Protégé

(Musen, 2015), which is fully compatible with the

latest OWL and RDF specifications. In Protégé,

terminologies are represented using classes, slots, and

facets (Noy and McGuinness, 2001), and it also

allows the use of annotation properties for adding

labels to the ontology classes and to link each concept

with its definition in validated and available standard

terminologies (e.g. SNOMED CT

, LOINC

, NCI

Thesaurus

, CIE-10-ES

). These permit the

representation of the biomedical concepts with stable

and unique codes, guaranteeing the interoperability of

the implemented knowledge.

2.2 Clinical Practice Guidelines

(CPGs)

CPGs are a set of criteria developed in a systematic

way to help professionals and patients in the decision-

making process, providing the latest evidence-based

diagnostic or therapeutic options when dealing with a

health problem or a specific clinical condition (Kredo

et al., 2016). Over the past years, these CPGs have

been widely used as part of CDSS by formalizing

them as CIGs. Such CIG-based CDSS have

demonstrated to be able to increase the chance of

impacting clinician behaviour compared to using only

narrative guidelines, as they provide updated patient

specific clinical data and advises at the point of care

(Latoszek-Berendsen et al., 2010).

Realizing that SWTs presented an increased

awareness when trying to cope with semantic

interoperability problems in ontologies, some other

researchers studied their usage to represent

computerized CPGs. For example, Hu et al. (Hu et al.,

2015) and Huang et al. (Huang et al., 2014) discuss

several use cases of semantic representation of

evidence-based medical guidelines, showing that they

are potentially useful for medical applications.

Due to our research interest, a state of the art in

Gestational Diabetes Mellitus (GDM) CPGs was

done. GDM is the most common metabolic disorder

of pregnancy, defined as a glucose intolerance

developed in the second or third trimester of

pregnancy (American Diabetes Association, 2016). It

confers an increased risk and complications during

pregnancy for both mother and child, including

https://loinc.org/

https://ncithesaurus-stage.nci.nih.gov/ncitbrowser/

https://eciemaps.mscbs.gob.es/

Development of a Gestational Diabetes Computer Interpretable Guideline using Semantic Web Technologies

105

cesarean delivery, shoulder dystocia, macrosomia,

and neonatal hypoglycemia (The HAPO Study

Cooperative Research Group, 2008). Furthermore,

women with GDM have a substantially increased risk

to develop type 2 diabetes and cardiovascular

diseases after pregnancy (Bellamy et al., 2009;

Sullivan et al., 2012). Therefore, strategies addressed

to optimize management of GDM including effective

prevention, and proper diagnosis and treatment are

mandatory (Chiefari et al., 2017).

There are several guidelines based on best

available and updated evidence for the GDM

management, such as the ones developed by the

National Institute for Health and Care Excellence

(NICE) or the World Health Organization (WHO).

These guidelines allow patient-centered decision

support by following several criteria: (i) the

measurement of clinical variables by the patient itself,

(ii) the readability and ease of follow-up of the given

recommendations, and (iii) the guideline ability to

deal with the guidance of the three pregnancy stages

(before, during, and after pregnancy).

After analysing the different guidelines for GDM

management, the Queensland Clinical Guideline

(Queensland Clinical Guidelines, 2015) was selected

to be used as the backbone guideline, extended with

the knowledge of several other guidelines in the

domain.

But, as it is known, following multiple CPGs in

parallel could result in statements that may interact

with each other, such as giving conflicting

recommendations. Furthermore, the impetus to

deliver customised care based on patient-specific

information, results in the need to be able to offer

guidelines in an integrated manner. In order to deal

with these problems, a harmonized patient-centered

CPG was created and formatted as CIG, enabling the

development of a CIG-driven CDSS.

2.3 Computer Interpretable

Guidelines (CIGs)

Computer Interpretable Guidelines (CIGs) are formal

representations of CPGs that can be executed to

provide guideline-based decision support. One of the

several well-known approaches for formally

representing CIGs are the “Task-Network Models”

(TNMs). These models structure the dependencies

among actions as hierarchical networks that when

fulfilled in a satisfactory way provide a

recommendation.

There exist several proposals to cope with

different clinical modelling challenges (Peleg et al.,

2003), such as GLIF (Patel et al., 1998), PROforma

(Sutton and Fox, 2003), Asbru (Seyfang et al., 2002)

or EON (Tu and Musen, 2001).

These types of CIG formalisms have been used in

many projects over the past years. For instance,

PROforma representation was used by Isern et al.

(Isern et al., 2012) for their proposal of ontology-

driven execution of CPGs. Eccher et al. (Eccher et al.,

2014) implemented and evaluated an Asbru-based

DSS for adjuvant treatment in breast cancer. In Peleg

et al. (Peleg et al., 2014) web-based interactive

clinical algorithms were developed based on GLIF

formalism for the sequencing of tasks to analyse

patients with particular clinical conditions.

In this work, a simplified version of TNM was

implemented based on inference rules (i.e. IF-THEN

type rules), which is explained in Section 3.2.

2.4 Clinical Decision Support

Systems (CDSS)

Realizing the potential of using CIGs and ontologies,

most of the approaches in this field over the past years

focus on guideline development and implementation

for decision support. For instance, the work described

in (Galopin et al., 2015) proposes an ontological

reasoning method based on semantic web techniques

to bring more flexibility to CDSS and offer the ability

to deal with patients suffering from multiple

pathologies by including several modelled CPGs.

Another approach was done in (Riaño et al., 2012),

where the contents of an ontology for the care of

chronically ill patients were adapted to create

individual intervention plans describing health-care

general treatments and also to use it as the knowledge

base of a decision support tool.

CDSS aim aiding clinicians in their decision-

making process by providing the needed tools to

analyse clinical data with latest evidence in the

shortest time (Garg et al., 2005). However, they can

also support patients in the management of different

diseases. Advances in mobile communication for

health care (m-Health) allow the design and

development of patient-centric models to improve

patient’s self-management capabilities.

Recently, many studies have reported that

computer-assisted expert systems, such as CDSS,

might help diabetes practitioners and patients to make

reliable diagnoses and management decisions (Balas

et al., 2004; García-Sáez et al., 2014; Sun and

Costello, 2018; Wilkinson et al., 2013). On the other

hand, lifestyle plays an essential role in controlling

diabetes, in both the prevention and management of

the disease. Many reports in clinical research

(Mottola, 2007; Padayachee and Coombes, 2015;

KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development

106

Silva-Zolezzi et al., 2017) support the theory that

healthy eating and regular exercise are beneficial in

both preventing GDM and improving pregnancy

outcomes in women with GDM.

Taking this into account, among the different

CDSS designed in recent years, the ones related to the

management of GDM were analysed. For instance, in

the work of Caballero-Ruiz et al. (Caballero-Ruiz et

al., 2017) a web-based telemedicine platform for its

use as CDSS for the management of GDM was

designed to remotely evaluate patients, allowing them

to upload their data at home. In addition, mobile

applications allow using automatic processing tools

to provide real-time advice based on monitoring data.

Peleg et al. (Peleg et al., 2017) presented a

personalized and patient-centric CDSS for the

monitoring and evaluation of atrial fibrillation and

gestational diabetes. A newer study (Miremberg et

al., 2018) demonstrated the positive effect of a

smartphone-based daily feedback system among

women with GDM for improving patient compliance

to treatment and a better control of glycaemic levels.

With the aim of going beyond the current state of

the art of CPG-based CDSS, the objective of this

project was to design a patient-centered mobile CDSS

for the management of GDM with the novelty of

integrating SWTs to the system.

3 METHODOLOGY

In this chapter, the methodology followed for the

development of a semantically validated mobile

CDSS for giving guidance to women in the

management of Gestational Diabetes Mellitus

(GDM) is described.

3.1 GDM Ontology

In this section, an ontology formalizing all concepts

and knowledge coming from the different CPGs

related to GDM prevention, management and

treatment is presented. For this work, the Protégé

editor was used to define the different gestational

diabetes related clinical concepts and the

relationships among them. This ontology was built

using NCI Thesaurus terminology to assure a

semantic interoperability of the knowledge. The big

amount of biomedical concepts that NCI Thesaurus

contains and the fact that it is open access makes it an

appropriate terminology to be used for the GDM

ontology. Besides, the Unified Medical Language

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

System (UMLS)

code for each of the terminologies

was also specified in the ontology. UMLS integrates

the most notable vocabularies (e.g. SNOMED CT,

ICD-10, etc.) in its repository. The result was then

validated with a reasoner (Section 3.1.2) for ensuring

a consistent ontology.

3.1.1 Ontology Formalization

All the conditions and rules expressed in the CIG

contain variables and properties that are organized in

an ontology with their specific names. This ontology

is composed by a total number of 166 classes, which

are separated in two main groups (Figure 1). The first

group is under the class named DMG360Concept that

comprises all the necessary variables’ names for

creating the CIG. The second group, named

DMG360Value, compiles the classes that define the

possible values of the classes from the first group.

Figure 1: The two main groups of classes in the ontology:

DMG360Concept and DMG360Value.

As can be seen in Figure 1, the group of concepts

has four subclasses, and each of them comprises other

subclasses. The first concept in the list (i.e.

DiseaseOrSymptom) is composed by two subclasses:

the Disease class, which contains two diseases

definition (DiabetesMellitus and

PolycysticOvarianSyndrome); and the Symptom

class, with a list of 15 different symptoms as

subclasses. The second concept (i.e. Medication)

includes three subclasses corresponding to three

different medications: Antipsychotics,

Corticosteroids, and Insulin. The third one (i.e.

PatientInformation) is the concept with the biggest

number of subclasses, containing different

information about the patient (e.g. information about

maternal age, diabetes family history, ethnicity,

physical activity, body mass index (BMI), etc.). The

last concept (i.e. Recommendations) is related to the

recommendations that are given to the user, which

Development of a Gestational Diabetes Computer Interpretable Guideline using Semantic Web Technologies

107

can vary depending on the stage of pregnancy, as

explained in Chapter 3.2. In total, seven different

types of recommendations are specified.

Furthermore, two different properties are defined

within the ontology to relate the different classes and

their values: a data property and an object property.

Both are used to specify the range of values that can

be taken by these classes, thus they are both defined

using the noun hasRange. These ranges can handle

either common data types (i.e. integer, Boolean…)

when linked by a data property, or other possible

values expressed as classes in the DMG360Value

group when linked by an object property. For

example, the CurrentPhysicalActivityLevel class is

restricted to have a value corresponding to any

subclass of the ScalingValue class (i.e. High, Low, or

Moderate).

In addition, the clinical concepts’ names,

definitions and codes were extracted from NCI

Thesaurus repository and defined within the model

using annotation properties for its semantic

standardization. Five different annotation properties

were defined: (i) NCI_label, for giving the label of the

class as stated in the NCI Thesaurus, (ii)

NCI_definition, which contains the definition of the

concept by the NCI, (iii) NCI_code, with the unique

code of the term, (iv) UMLS_CUI, containing the

corresponding UMLS code, and (v) NCI_version,

specifying the version of the NCI Thesaurus

repository used.

For the ontology design, OWL language was

used. Having all the knowledge mapped, the model

was exported in RDF language for the integration

with the CDSS using the Jena API.

3.1.2 Ontology Validation

To validate the defined relationships between the

variables and their possible values in the ontology, a

tool called Reasoner was used. The reasoners offered

by Protégé (i.e. FaCT ++, HermiT or Pellet) are

programs that evaluate the consistency of an ontology

by identifying relationships between classes. In this

project, the FaCT++ reasoner was used.

After the triggering of this reasoner without

obtaining any unsatisfactory relationship between the

classes, the hierarchy of the designed GDM ontology

was validated.

Moreover, the SPARQL Query tool in Protégé

was used for the validation of the requests that could

be done to the ontology in the process of its

integration with the CDSS. For that, the SPARQL

query language

was applied, a language that can be

used to express queries across diverse data sources,

whether the data is stored natively as RDF or viewed

as RDF via middleware. Simple queries such as

requesting data/object properties of a class, finding

subclasses of a class, or showing the list of all classes

in the ontology were tested.

3.2 GDM CIG

As stated in the state of the art, CPGs must be

formalized in a computer interpretable way to provide

guideline-based decision support and allow the

evaluation of a patient in a computerized way. In this

work, several CPGs with the clinical statements that

describe the procedures to be followed in each

clinical setup for the GDM management were studied

and a more complete guideline that contained all the

needed information for monitoring patients before,

during and after pregnancy was created. This

extended CPG was then formalized as IF-THEN rules

in a computerized way (i.e. as a CIG) containing the

concepts defined in the ontology.

Figure 2: Components of a Rule object used in this work.

https://www.w3.org/TR/sparql11-query/

KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development

108

To create the rules for this project, conditions

(i.e. clinical statements to be accomplished) and

their respective consequences (i.e. the

recommendations to be given to the patient) were

identified and extracted from the selected CPGs in

the GDM domain. Once this knowledge was

formalized in our extended CPG, a translation into

computer interpretable language was done. For that,

statements were translated into IF-THEN kind rules,

where the conditional part is preceded by the IF

expression and the consequent part by the THEN

expression.

Each of the statements describing the criteria to

be followed for guiding patients in this setup were

defined as Rule objects, which encompasses (i) a

conditional part composed by one or more

conditions linked by (ii) a binary operator (i.e. AND,

or OR) and (iii) a consequent part containing the

recommendation for the patient, as can be seen in

Figure 2. Each of the conditions is based on a

Condition object, which stores (i) the name of the

clinical variable to be evaluated (i.e. concept defined

in the ontology), (ii) the mathematical operator (i.e.

>, ≥, =, <, ≤) and (iii) the value or the threshold of

the clinical variable to be evaluated. When the

conditions of a rule are matching a patient’s clinical

information, a recommendation is provided to the

patient.

Depending on the patient clinical stage, different

types of interventions for the management of GDM

were identified in the rule formalization process. For

the stage before pregnancy, information about

nutrition, physical activity and GDM risk factors

were formalized. For the pregnancy period, guides

related to nutrition, weight, physical activity, risk

factors, and glucose control were covered. And

finally, in the case of the post-pregnancy stage,

nutrition, weight control, physical activity,

breastfeeding, glucose control, and postpartum

depression recommendations were included. The

schema describing all this information is represented

in Figure 3.

When formalizing this knowledge in the

extended CPG and in its computerized version as

CIG, the pipeline presented in Figure 3 was

followed. The extended CIG contains 96 different

rules distributed in three stages: 13 for the

pre-pregnancy stage, 45 for the GDM management

during pregnancy, and 16 for post-pregnancy

stage. In addition, 22 characterization rules were

also formalized, whose action is the change of

variable values instead of providing a

recommendation. These rules were not introduced

into the CDSS as part of the rule base to be executed

by the engine, but they were implemented in the

mobile app itself.

3.3 Integration of the Ontology within

the CDSS

Once the CIG was formalized with all computer

interpretable rules and the GDM ontology was

validated, the next step was to integrate both into the

CDSS. For this objective, an authoring tool

previously designed (Muro et al., 2019) was used.

This tool enables an intuitive Graphical User

Interface (GUI) for including new clinical knowledge

Figure 3: The three stages of pregnancy and the different recommendations giving in each of them.

Development of a Gestational Diabetes Computer Interpretable Guideline using Semantic Web Technologies

109

in a computerized way into the CDSS rule base. A

high-level project’s architecture representation is

shown in Figure 4.

Figure 4: High level representation of the architecture of the

project.

The development made within this project

extended the authoring tool by introducing a new

functionality: the syntactical and semantical

validation of the formalized knowledge through an

ontology. A research on different APIs (i.e. Protégé-

OWL API

, OWL API

, Apache Jena

, or RDF4J

)

for the integration of ontologies in this kind of

systems was done. After comparing them, Jena API

was selected for carrying out the integration of the

ontology with the Java based CDSS, because of its

ease of use and compatibility advantages with our

system.

In order to have an interaction between the GUI

and the CDSS, different web services were developed

as linkers to transmit data between them. These web

services were used by the authoring tool (i) to get the

list of variables (classes) for defining the evaluated

variable name within a Condition object, (ii) to get the

possible values of the specific variable selected in the

Condition object, (iii) to get the list of

recommendations for completing the consequent part

of the Rule object, and (iv) to post each generated rule

to the backend of the system. This backend is

composed by different modules based on Drools

, (i)

a rule engine and (ii) a rule file generator in Drools

Rule Language (.drl) extension.

To create the GDM CIG, each of the rules or

conditions were introduced manually in the authoring

tool using its GUI. In this process, four main blocks

are fulfilled. First, the name of the rule is defined.

Then, the conditions of the rule are introduced. Next,

https://protegewiki.stanford.edu/wiki/ProtegeOWL_API

https://github.com/owlcs/owlapi/wiki

https://jena.apache.org/documentation/ontology/

the recommendation for the introduced rule is

specified. And finally, the rule is sent to the backend.

In the second step for introducing the conditions

of each rule, the Condition objects are constructed

using the information obtained from querying the

integrated ontology in the authoring tool. For

requesting the filling out options, the previously

described web services are used. In this context, the

different types of ontology classes were used for

different purposes, as explained below.

The classes of the ontology that correspond to the

variables used for defining the conditional part of the

CIG are under the DMG360Concepts class. These

variables are requested and given as options for the

first parameter of a Condition object in the authoring

tool GUI.

Once the variable name of the Condition object is

specified, the respective condition operator is

selected. Then, in the third box for specifying the

value or threshold of the selected variable, the

authoring tool uses the web service for requesting its

possible values, and here is where the second type of

ontology classes interact. In the case of variables with

data properties, the backend of the authoring tool

defines the type of data that needs to be used by

changing the format of the box in the GUI. For the

case of the variables with object properties, a list

containing subclasses of the group DMG360Value is

given for being selected by the clinician.

Once the first condition is introduced, the binary

operator can be selected from the given options (i.e.

AND, or OR) and as many as required conditions can

be defined within the conditional part of a single rule

following the same procedure (see an example in

Figure 5).

Figure 5: Example of the definition of rule conditional

statements in the AT.

http://rdf4j.org/

https://www.drools.org/

KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development

110

The third step of the introduction of the rule

corresponds to the tab of the GUI for introducing

recommendations, which permits the user to select

the recommendation from a drop-down list. This list

is the result of the request made by the authoring tool

for the subclasses of the class RecommendationValue

from the DMG360Value group in the ontology. When

a recommendation is selected by the user using its

name or abbreviation, the text corresponding to that

recommendation is displayed above the selection box.

Each of the texts for the respective recommendation

class are specified in the ontology using annotation

properties. The authoring tool uses a web service for

obtaining the information from those annotations

when a recommendation name is selected from the

dropdown list.

Finally, the last step of the introduction of rules

uses the fourth web service for sending the rules to

the backend of the system and generating the .drl file,

as explained above. This last web service takes all the

values introduced by the clinician in the authoring

tool and sends them to the backend, where they are

processed and uploaded to the clinical database. As

the structure of the objects generated in the authoring

tool are the same as in the Rule object, a direct

mapping can be done. The generated Rule object is

then sent to the database, where it is stored, and then

retrieved for sending it to the backend and generating

the .drl file.

3.4 Patient-centered Mobile CDSS

Once the integration of the semantically validated

CIG was implemented within the CDSS as a .drl file,

a patient-centered mobile application was developed

to interact with it. The user profiles defined for this

application are (i) women that want to be pregnant

and are monitoring their health status for it, (ii)

women already pregnant, or (iii) women that have just

given birth, and they all want to receive

recommendations for managing the gestational

diabetes. In the next lines, some examples of the

different screens of the designed app and their

functionalities are described.

First, a logging system was developed with

registering or signing in options. This logging will

save the basic personal data (i.e. sex, ethnicity…) that

is supposed to be static during the monitoring period

to avoid introducing it each time the user logs into the

application.

Once the user is logged in, a main menu is

displayed providing the possible user profiles to start

the GDM management depending on their needs.

Each of the pregnancy stages will provide specific

recommendations, differing ones from the others as

each period has a different focus on what to evaluate

and how to treat the patient to reach her objective (see

Chapter 3.2): in pre-pregnancy period,

recommendations for nutrition, physical activity and

GDM risk factors are given. During pregnancy,

nutrition, weight, physical activity, risk factors, and

glucose control related guidance is given. And for

post-pregnancy period, nutrition, weight control,

physical activity, breastfeeding, glucose control, and

postpartum depression recommendations are given.

This information is managed by the app through a

menu, where for each stage different

recommendations on the above stated topics can be

provided.

Each time the user selects an option in the

different menus, the variable corresponding to that

specific topic is changed. For example, in the case of

the menu for the post-pregnancy stage, when the user

touches the image for obtaining recommendation

concerning the glucose control, the variable

GlucoseControlRecommendations is set to true.

If required, the application will retrieve more

clinical information by questionnaires (see Figure 6).

These questionnaires will ask for some questions that

have possible answers to be selected. These answers

are related to some clinical variables defined in the

GDM ontology (mostly Boolean or Object property

type variables), but sometimes it can require to the

user introducing numerical variables’ values (e.g. the

blood glucose level). Every time she answers a

question required in the app, this clinical information

will be stored as part of her profile to be evaluated by

the CDSS later.

Once all the questions for completing the needed

patient profile are answered, they are sent to the

CDSS in the backend, getting back the corresponding

recommendation(s) to be displayed in the

recommendation screen of the mobile application

(see Figure 7).

4 PRACTICAL GDM SCENARIO

In this chapter a practical scenario is presented to

show up how the whole architecture works. The

selected patient profile represents a pregnant woman

that would like to receive recommendations related to

her glucose control. For that, she will be using the

mobile application designed in this project. The

presented clinical case is a non-diabetic pregnant

woman that has suffered a decompensation in her

Blood Glucose Level (BGL) with a glucose value of

136 mg/dL before eating (pre-prandial BGL). In

Development of a Gestational Diabetes Computer Interpretable Guideline using Semantic Web Technologies

111

addition, her glucose levels showed higher values

than the maximum threshold more than twice during

the same week and in different intervals. Hence, her

clinical data would be as follows:

Table 2: Clinical data of the patient profile for the use case.

Variable name Value

DuringPregnancyPeriod TRUE

GlucoseControlRecommendations TRUE

GDM FALSE

Type1Diabetes FALSE

Type2Diabetes FALSE

NormalBGLResults FALSE

Insulin FALSE

TwoOrMoreDecomp1Week TRUE

InDifferentIntervals TRUE

ExceedMaximumValues TRUE

PrePrandialBGL 136

In the process for introducing these values in the

mobile application, first the diabetes related questions

are done to the user (i.e. the type of diabetes, and the

glucose level value are requested). Then, some other

questions are done to the user like (i) if she is taking

insulin, (ii) if she had more than one decompensation

of blood glucose levels in the same week, (iii) if they

were in different intervals, and (iv) if they exceeded

the maximum values. The screen of the app showing

these questions is visualized in Figure 6.

Figure 6: Questions of the questionnaire for glucose control

in the mobile app.

Once this data is gathered by the system, it is sent

to the backend to be evaluated by the CDSS. The

rule(s) fitting the values of the variables sent by the

user are triggered getting as result the

recommendation corresponding to the matching rule.

In this particular case, the recommendation obtained

was: “Notify your doctor to consider insulin

treatment” (see Figure 7).

Figure 7: Recommendation screen in the mobile app.

Once the user receives the recommendation, she

can follow it to solve her glucose level problem in this

case. Besides, she can also go back to the menu and

select other options for receiving recommendations

about more interventions for the management of

GDM.

5 CONCLUSIONS AND FUTURE

WORK

CPGs have been promoted as a powerful method for

standardization and improvement of medical care

quality and personalization of healthcare. However,

applying these guidelines to promote evidence-based

and up-to-date clinical practice is a known challenge

due to lack of digitalization of clinical guidelines. To

overcome these issues, the use of semantic web

technologies along with CDSS is proposed, in order

to avoid the lack of semantic interoperability and

promote the interoperability among systems in a

standardized way.

In this paper, an architecture that allows the

formalization of CPGs into Computer Interpretable

Guidelines (CIGs) supported by an ontology in the

gestational diabetes domain is presented. This CIG

has been implemented within a CDSS to give support

to the patients before, during, and after pregnancy. In

addition, a mobile application has been developed for

guiding patients based on up-to-date evidence based

KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development

112

clinical guidelines. This application is able to provide

users different recommendations based on their

clinical information.

As future work, applying tools for obtaining

information from patients’ electronic health records

could optimize the efficiency of the designed mobile

CDSS. In the same way, a more flexible way of

gathering the user’s clinical data will be implemented

(e.g. using wearables for obtaining patient data

without needing to introduce them manually).

Likewise, a way to facilitate the ontology generation

for clinicians will also be researched. Finally, it has

been also envisioned the future inclusion of feedback

tools within the mobile application in order to gather

the user appreciation of the system, as well as the

possibility to submit the system to the evaluation of

clinical specialists.

ACKNOWLEDGEMENTS

This work has been developed under the research

project DMG360 (2017-2018 / Exp. number ZE-

2017/00011, in collaboration with INIT Health, SL),

which has been funded by the Department of

Economic Development and Infrastructure of the

Basque Government under the Hazitek program.

REFERENCES

American Diabetes Association, 2016. Classification and

Diagnosis of Diabetes. Diabetes Care 39, S13–S22.

https://doi.org/10.2337/dc16-S005

Balas, E. A., Krishna, S., Kretschmer, R. A., Cheek, T. R.,

Lobach, D. F., Boren, S. A., 2004. Computerized

knowledge management in diabetes care. Med. Care

42, 610–621.

Bellamy, L., Casas, J.-P., Hingorani, A.D., Williams, D.,

2009. Type 2 diabetes mellitus after gestational

diabetes: a systematic review and meta-analysis. Lancet

Lond. Engl. 373, 1773–1779. https://doi.org/10.1016/

S0140-6736(09)60731-5

Caballero-Ruiz, E., García-Sáez, G., Rigla, M., Villaplana,

M., Pons, B., Hernando, M. E., 2017. A web-based

clinical decision support system for gestational

diabetes: Automatic diet prescription and detection of

insulin needs. Int. J. Med. Inf. 102, 35–49.

https://doi.org/10.1016/j.ijmedinf.2017.02.014

Chiefari, E., Arcidiacono, B., Foti, D., Brunetti, A., 2017.

Gestational diabetes mellitus: an updated overview. J.

Endocrinol. Invest. 40, 899–909. https://doi.org/

10.1007/s40618-016-0607-5

Eccher, C., Seyfang, A., Ferro, A., 2014. Implementation

and evaluation of an Asbru-based decision support

system for adjuvant treatment in breast cancer. Comput.

Methods Programs Biomed. 117, 308–321.

https://doi.org/10.1016/j.cmpb.2014.06.021

El-Sappagh, S., Kwak, D., Ali, F., Kwak, K.-S., 2018.

DMTO: a realistic ontology for standard diabetes

mellitus treatment. J. Biomed. Semant. 9.

https://doi.org/10.1186/s13326-018-0176-y

Galopin, A., Bouaud, J., Pereira, S., Seroussi, B., 2015. An

Ontology-Based Clinical Decision Support System for

the Management of Patients with Multiple Chronic

Disorders. Stud. Health Technol. Inform. 216, 275–279.

García-Sáez, G., Rigla, M., Martínez-Sarriegui, I., Shalom,

E., Peleg, M., Broens, T., Pons, B., Caballero-Ruíz, E.,

Gómez, E.J., Hernando, M.E., 2014. Patient-oriented

Computerized Clinical Guidelines for Mobile Decision

Support in Gestational Diabetes. J. Diabetes Sci.

Technol. 8, 238–246. https://doi.org/10.1177/1932296

814526492

Garg, A. X., Adhikari, N. K. J., McDonald, H., Rosas-

Arellano, M. P., Devereaux, P. J., Beyene, J., Sam, J.,

Haynes, R. B., 2005. Effects of computerized clinical

decision support systems on practitioner performance

and patient outcomes: a systematic review. JAMA 293,

1223–1238. https://doi.org/10.1001/jama.293.10.1223

Grimshaw, J. M., Russell, I. T., 1993. Effect of clinical

guidelines on medical practice: a systematic review of

rigorous evaluations. Lancet Lond. Engl. 342, 1317–

1322.

Hu, Q., Huang, Z., Gu, J., 2015. Semantic representation of

evidence-based medical guidelines and its use cases.

Wuhan Univ. J. Nat. Sci. 20, 397–404.

https://doi.org/10.1007/s11859-015-1112-y

Huang, Z., ten Teije, A., van Harmelen, F., Aït-Mokhtar, S.,

2014. Semantic Representation of Evidence-Based

Clinical Guidelines, in: Miksch, S., Riaño, D., ten Teije,

A. (Eds.), Knowledge Representation for Health Care.

Springer International Publishing, Cham, pp. 78–94.

https://doi.org/10.1007/978-3-319-13281-5_6

Institute of Medicine, 1990. Clinical Practice Guidelines:

Directions for a New Program. National Academies

Press (US), Washington (DC).

Isern, D., Sánchez, D., Moreno, A., 2012. Ontology-driven

execution of clinical guidelines. Comput. Methods

Programs Biomed. 107, 122–139. https://doi.org/

10.1016/j.cmpb.2011.06.006

Kolias, V.D., Stoitsis, J., Golemati, S., Nikita, K.S., 2014.

Utilizing Semantic Web Technologies in Healthcare.

Springer Int. Publ. Switz. 16. https://doi.org/10.1007/

978-3-319-06844-2_2

Kredo, T., Bernhardsson, S., Machingaidze, S., Young, T.,

Louw, Q., Ochodo, E., Grimmer, K., 2016. Guide to

clinical practice guidelines: the current state of play. Int.

J. Qual. Health Care 28, 122–128. https://doi.org/

10.1093/intqhc/mzv115

Latoszek-Berendsen, A., Tange, H., Herik, H. J. van den,

Hasman, A., 2010. From Clinical Practice Guidelines to

Computer-interpretable Guidelines. Methods Inf. Med.

49, 550–570. https://doi.org/10.3414/ME10-01-0056

Miremberg, H., Ben-Ari, T., Betzer, T., Raphaeli, H.,

Gasnier, R., Barda, G., Bar, J., Weiner, E., 2018. The

impact of a daily smartphone-based feedback system

Development of a Gestational Diabetes Computer Interpretable Guideline using Semantic Web Technologies

113

among women with gestational diabetes on compliance,

glycemic control, satisfaction, and pregnancy outcome:

a randomized controlled trial. Am. J. Obstet. Gynecol.

218, 453.e1-453.e7. https://doi.org/10.1016/j.ajog.

2018.01.044

Mottola, M.F., 2007. The role of exercise in the prevention

and treatment of gestational diabetes mellitus. Curr.

Sports Med. Rep. 6, 381–386. https://doi.org/10.1007/

s11932-007-0056-1

Muro, N., Larburu, N., Torres, J., Kerexeta, J., Artola, G.,

Arrúe, M., Macía, I., Seroussi, B., 2019. Architecture

for a Multimodal and Domain-Independent Clinical

Decision Support System Software Development Kit.

Presented at the 41st Annual International Conference

of the IEEE Engineering in Medicine and Biology

Society (EMBC), Berlin, Germany.

Musen, M.A., 2015. The Protégé Project: A Look Back and

a Look Forward. AI Matters 1, 4–12. https://doi.org/

10.1145/2757001.2757003

Noy, N. F., McGuinness, D. L., 2001. Ontology

Development: A Guide to Creating Your First Ontology

25.

Padayachee, C., Coombes, J. S., 2015. Exercise guidelines

for gestational diabetes mellitus. World J. Diabetes 6,

1033–1044. https://doi.org/10.4239/wjd.v6.i8.1033

Patel, V. L., Allen, V. G., Arocha, J. F., Shortliffe, E. H.,

1998. Representing Clinical Guidelines in GLIF. J. Am.

Med. Inform. Assoc. JAMIA 5, 467–483.

Peleg, M., Fox, J., Patkar, V., Glasspool, D., Chronakis, I.,

South, M., Nassar, S., Gaglia, J. L., Gharib, H., Papini,

E., Paschke, R., Duick, D. S., Valcavi, R., Hegedüs, L.,

Garber, J. R., 2014. A Computer-Interpretable Version

of the AACE, AME, ETA Medical Guidelines for

Clinical Practice for the Diagnosis and Management of

Thyroid Nodules. Endocr. Pract. Off. J. Am. Coll.

Endocrinol. Am. Assoc. Clin. Endocrinol. 20, 352–359.

https://doi.org/10.4158/EP13271.OR

Peleg, M., Shahar, Y., Quaglini, S., Fux, A., García-Sáez,

G., Goldstein, A., Hernando, M.E., Klimov, D.,

Martínez-Sarriegui, I., Napolitano, C., Parimbelli, E.,

Rigla, M., Sacchi, L., Shalom, E., Soffer, P., 2017.

MobiGuide: a personalized and patient-centric

decision-support system and its evaluation in the atrial

fibrillation and gestational diabetes domains. User

Model. User-Adapt. Interact. 27, 159–213.

https://doi.org/10.1007/s11257-017-9190-5

Peleg, M., Tu, S., Bury, J., Ciccarese, P., Fox, J., Greenes,

R.A., Hall, R., Johnson, P.D., Jones, N., Kumar, A.,

Miksch, S., Quaglini, S., Seyfang, A., Shortliffe, E.H.,

Stefanelli, M., 2003. Comparing Computer-

interpretable Guideline Models: A Case-study

Approach. J. Am. Med. Inform. Assoc. 10, 52–68.

https://doi.org/10.1197/jamia.M1135

Queensland Clinical Guidelines, 2015. Guideline:

Gestational diabetes mellitus 38.

Riaño, D., Real, F., López-Vallverdú, J.A., Campana, F.,

Ercolani, S., Mecocci, P., Annicchiarico, R.,

Caltagirone, C., 2012. An ontology-based personaliza-

tion of health-care knowledge to support clinical

decisions for chronically ill patients. J. Biomed. Inform.

45, 429–446. https://doi.org/10.1016/j.jbi.2011.12.008

Seyfang, A., Miksch, S., Marcos, M., 2002. Combining

diagnosis and treatment using ASBRU. Int. J. Med. Inf.

68, 49–57.

Silva-Zolezzi, I., Samuel, T. M., Spieldenner, J., 2017.

Maternal nutrition: opportunities in the prevention of

gestational diabetes. Nutr. Rev. 75, 32–50.

https://doi.org/10.1093/nutrit/nuw033

Sullivan, S. D., Umans, J. G., Ratner, R., 2012. Gestational

Diabetes: Implications for Cardiovascular Health. Curr.

Diab. Rep. 12, 43–52. https://doi.org/10.1007/s11892-

011-0238-3

Sun, S., Costello, K. L., 2018. Designing Decision-Support

Technologies for Patient-Generated Data in Type 1

Diabetes. AMIA. Annu. Symp. Proc. 2017, 1645–1654.

Sutton, D. R., Fox, J., 2003. The Syntax and Semantics of

the PROforma Guideline Modeling Language. J. Am.

Med. Inform. Assoc. JAMIA 10, 433–443.

https://doi.org/10.1197/jamia.M1264

The HAPO Study Cooperative Research Group, 2008.

Hyperglycemia and Adverse Pregnancy Outcomes. N.

Engl. J. Med. 358, 1991–2002. https://doi.org/10.1056/

NEJMoa0707943

Tu, S.W., Musen, M.A., 2001. Modeling data and

knowledge in the EON guideline architecture. Stud.

Health Technol. Inform. 84, 280–284.

Wilkinson, M.J., Nathan, A.G., Huang, E.S., 2013.

Personalized Decision Support in Type 2 Diabetes

Mellitus: Current Evidence and Future Directions.

Curr. Diab. Rep. 13, 205–212. https://doi.org/10.1007/

s11892-012-0348-6

KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development

114