A Model Driven Approach to the Development of Gamiﬁed Interactive

Clinical Practice Guidelines

Job N. Nyameino

1,3

, Fazle Rabbi

, Ben-Richard Ebbesvik

, Martin C. Were

3,4,5

and Yngve Lamo

Department of Informatics, University of Bergen, Bergen, Norway

Department of Computing, Mathematics, and Physics, Western Norway University of Applied Sciences, Bergen, Norway

Institute of Biomedical Informatics, Moi University, Eldoret, Kenya

Department of Biomedical Informatics, Vanderbilt University Medical Center, Nashville, TN, U.S.A.

Vanderbilt Institute for Global Health, Vanderbilt University Medical Center, Nashville, TN, U.S.A

Keywords:

Clinical Practice Guidelines, Model Driven Engineering, Gamiﬁcation.

Abstract:

Clinical practice guidelines (CPGs) play a fundamental role in modern medical practice since they summarize

the vast medical literature and provide distilled recommendations on care based on the current best evidence.

However, there are barriers to CPG utilization such as lack of awareness and lack of familiarity of the CPGs

by clinicians due to ineffective CPG dissemination and implementation. This calls for research into effective

and scalable CPG dissemination strategies that will improve CPG awareness and familiarity. We describe a

formal model-driven approach to design and implement a gamiﬁed e-learning system for clinical guidelines.

We employ gamiﬁcation to increase user motivation and engagement in the training of guideline content. Our

approach involves the use of models for different aspects of the system, an entity model for the clinical domain,

a workﬂow model for the clinical processes and a game model to manage the training sessions. A game engine

instantiates a training session by coupling the workﬂow and entity models to automatically generate questions

based on the data in the model instances. Our proposed approach is ﬂexible and adaptive as it allows for easy

updates of the guidelines, integration with different device interfaces and representation of any guideline.

1 INTRODUCTION

Medical knowledge is increasing at an exponential

rate and it is difﬁcult for clinicians to keep up with

this quantity of knowledge production (Fervers et al.,

2010). The development and use of clinical prac-

tice guidelines (CPGs) is a promising solution to this

problem. CPGs are systematically developed state-

ments that assist practitioners and patients to make

decisions about appropriate health care for speciﬁc

circumstances (Lohr et al., 1992). Guidelines are

a comprehensive summary of the available evidence

about medical conditions and provide recommenda-

tions for the management of those conditions (Goud

et al., 2009). A well-developed guideline reduces

variations in care, improves diagnostic accuracy, pro-

motes effective therapy and discourages ineffective

therapies all which contribute to improved quality of

care (Shiffman et al., 2004). The mere availability

of guidelines does not necessarily mean that the rec-

ommendations will be used in actual care. Indeed,

there has been a reported gap between recommended

care according to the evidence base and actual prac-

tice leading to preventable errors in practice (Donald-

son et al., 2000; Baker, 2001). This gap can be at-

tributed to several barriers to guideline dissemination

and implementation which include: internal barriers

(lack of awareness, lack of familiarity, lack of agree-

ment with the guideline content, and the inability to

overcome the inertia of previous practice) and exter-

nal barriers (i.e., patient, environmental, and guide-

line related factors such as ease of use and complexity

of the guideline) (Cabana et al., 1999).

The nature of guideline development means that

published guidelines are well-researched, compre-

hensive documents that can be prohibitively volumi-

nous. For example, the National Heart, Lung, and

Blood Institute (NHLBI) 2007 Guidelines for the Di-

agnosis and Management of Asthma full report is 440

pages long (NHLBI, 2007) while the National Insti-

tute for Health and Care Excellence (NICE) guide-

lines for the diagnosis monitoring and management of

chronic asthma (2017) report is 39 pages long (NICE,

2017). Such large texts are impractical for use at the

Nyameino, J., Rabbi, F., Ebbesvik, B., Were, M. and Lamo, Y.

A Model Driven Approach to the Development of Gamiﬁed Interactive Clinical Practice Guidelines.

DOI: 10.5220/0007736401470158

In Proceedings of the 14th International Conference on Evaluation of Novel Approaches to Software Engineering (ENASE 2019), pages 147-158

ISBN: 978-989-758-375-9

147

point of care. Additionally, poor guideline presenta-

tion has been identiﬁed as a factor in the lack of physi-

cian familiarity as some of the guidelines have been

described as being tedious, repetitive, confusing, and

unclear (Cabana et al., 2000).

To mitigate some of the barriers to knowledge

acquisition of guideline content, new dissemination

strategies aimed at improving awareness and famil-

iarity of guideline content are required. Active guide-

line dissemination strategies have been found to be

more effective than passive strategies at improving

the application of evidence based recommendations

in patient care (Grimshaw et al., 2012). In par-

ticular, educational interventions (e.g. distribution

of printed guidelines, educational meetings and out-

reaches) strengthen the effect of clinical educational

material. Further, the more intensely the information

is provided through these interventions, the greater its

effect on the recipients (Marriott et al., 2000). Re-

search into active strategies for clinical guideline dis-

semination are timely and relevant as they will poten-

tially help to plug the gap between recommended and

actual clinical practice.

One potentially useful active educational inter-

vention is in the distribution of gamiﬁed guidelines.

Gamiﬁcation is the use of game design elements in

non-game contexts (Deterding et al., 2011b; Deter-

ding et al., 2011a). It uses game based mechanics,

aesthetics and thinking to engage people, motivate

action, promote learning and solve problems (Kapp,

2012). The concept of Gamiﬁcation is relatively new

and has been used to describe the use of game-based

concepts and techniques, with the goal of increasing

the motivation and engagement of the participants and

improving the results.

The implementation of guideline summaries as in-

teractive, gamiﬁed ﬂowcharts on a mobile platform

will potentially mitigate the problems of guideline

complexity and presentation that plague the effective

dissemination of guideline content. In this paper we

present a formal model driven approach to gamiﬁca-

tion of clinical practical guidelines. To illustrate the

approach, we present three models, an entity model of

the clinical encounter domain, a workﬂow model for

the clinical processes and a game model all of which

will be integrated to create our gamiﬁed system. We

also describe a prototype mobile-based guideline app

that incorporates these models to present a gamiﬁed

interactive guideline training tool.

The rest of the paper is organized as follows: In

section 2 we give an introduction to the Diagram

Predicate Framework (DPF) and show how it can be

used for modelling CPGs workﬂows and entity mod-

els for the clinical domain. Moreover, we illustrate

how the guideline workﬂow information is synchro-

nized with the domain information. In section 3 we

introduce our approach to gamiﬁcation of workﬂows

and discuss in further detail the use of the different

models in our design. In section 4 we describe the

implementation of our approach in the development

of the prototype system. Finally in section 5 we com-

pare our approach to other works before we conclude

the paper and envision further work in section 6.

2 BACKGROUND

In this work we use a formal diagrammatic approach

to model driven software engineering (MDE), called

Diagram Predicate Framework (DPF). MDE is a sys-

tem development paradigm that promotes the use of

models as the primary artefacts that drives the whole

development process. In MDE models are speciﬁed

using a modelling language whose syntax and seman-

tics are deﬁned by a metamodel (Rodrigues da Silva,

2015). This allows for the development of domain-

speciﬁc modelling languages (DSLs) using notations

and abstractions that are unique to a given domain.

The use of DSLs allows for the development of more

expressive models and ease of use by domain experts.

In this section we provide an overview of how di-

agrammatic models can be created using DPF. We

chose to use DPF as it can be used to create cus-

tom domain speciﬁc modelling languages. Further-

more, we’ll present a metamodel for representing a

CPG workﬂow, a simple entity model for the medical

domain, a model for the game engine and ﬁnally an

integrated multi-metamodel that incorporates the en-

tity and CPG workﬂow models. UML is another al-

ternative approach for the modeling of various kinds

such as UML entity model, sequence model. How-

ever, DPF allows us to do multilevel metamodeling

and also visualizes constraints in the models.

2.1 Diagram Predicate Framework

(DPF)

DPF formalizes software development activities such

as metamodelling (Rutle et al., 2009) and model

transformations (Rutle et al., 2012) based on category

theory (Barr and Wells, 1990) and graph transforma-

tions (L

owe, 1993). By applying DPF we can formal-

ize clinical guidelines and clinical domain models at

different abstraction levels in form of diagrammatic

speciﬁcations. The diagrammatic nature of DPF also

facilitates visual representations of guidelines that can

be presented at different level of abstraction. A model

in DPF is represented by a diagrammatic speciﬁcation

ENASE 2019 - 14th International Conference on Evaluation of Novel Approaches to Software Engineering

148

Treatment

TreatmentName

treatmentName

Age

Weight

Patient

hasAge

hasWeight

History&Examination

Finding

hasFinding

Diagnosis

DiagnosisName

Severity

hasSeverity

diagnosisName

receives

hasTreatment

hasDiagnosis

conductedOn

PatientName

hasName

Symptom

hasSymptom

Plan

hasPlan

[pre-cond]

Gender

hasGender

Figure 1: A simpliﬁed entity model of the clinical encounter domain.

S = (S ,C

: Σ) which consists of a graph S and a set

of constraints C

speciﬁed by a predicate signature

Σ.

The predicate signature is composed of a collec-

tion of predicates, each having a name and an arity

(shape graph). A constraint consists of a predicate

from the signature together with a binding to the sub-

graph of the model’s underlying graph which is af-

fected by the constraint. In order to apply DPF for the

modeling of a game that operates over clinical prac-

tice guideline we need to formalize the concepts of a

guideline using DPF and also model the gamiﬁcation

concepts with DPF. In the following subsection we

present how DPF can be used to model different as-

pects of guidelines and representing the concepts for

gamiﬁcation.

2.2 Entity Modelling

We will now present an entity model of the clinical

domain. To do this we use a metamodel contain-

ing Concepts, Attributes and References. An actual

model typed over this metamodel is shown in Fig-

ure 1. We have concepts and corresponding rela-

tions for domain entities such as Patients, Diagnosis,

Treatments etc (see Figure 1). The model consists

of a constraint modeled with a predicate named ‘pre-

condition’ and visualized with symbol [pre − cond].

The constraint speciﬁes that all the treatment in-

stances must have a reference to a diagnosis instance.

2.3 Workﬂow Modelling

Clinical practice guidelines often consist of a ﬂow of

information. Workﬂow models may be used to repre-

sent the ﬂow of a guideline. In Figure 2 below we see

an example metamodel (M

) for behavioural models,

where Tasks can be connected by Flow edges. On the

next abstraction level (M

) we see a generic treatment

model that is typed by the ﬂow model. The treat-

ment model has three tasks Assessment and Diagno-

sis, Treatment and Evaluation. Finally, at (M

) we

see an instance of the treatment workﬂow of a severe

asthma diagnosis.

2.4 Game Modelling

At their core, games are goal-oriented activities with

reward and progress tracking mechanisms. The de-

sign of gamiﬁed e-learning systems should be under-

taken in the view of these core concepts. In our sys-

tem, the training will be done through a series of ques-

tions based on the guideline content. The game en-

gine in our model automatically generates questions

from the entity and workﬂow models to instantiate a

training module. The questions are categorized ac-

cording to the learner’s skill level (beginner, interme-

diate, advanced) and each question has a reward in the

form of points. A game model should also specify a

learner proﬁle that tracks the learner’s activities.

2.5 Integrating Models

The training model is built by the integration of the

entity and workﬂow models based on the principles

introduced by Rabbi et al (Rabbi et al., 2014a). The

states of the training module T M are deﬁned by a set

of elements that include a pair of workﬂow instance

W I and an entity instance EI: T M

=< EI

,W I

where i is a natural number. This integration of mod-

els is shown in Figure 3 and the concept is discussed

in more details in section 3. In Figure 3, we show a

A Model Driven Approach to the Development of Gamiﬁed Interactive Clinical Practice Guidelines

149

Task

flow

Assessment and Diagnosis

Treatment

Evaluation

Severe

Asthma

Re-evaluate

After 20 min

Oxygen

Salbutamol

Prednisolone

Cough +

Wheeze +

Cyanosis

Adjust treatment

Reassess patient to update diagnosis

Re-treat

Figure 2: The workﬂow model with its metamodel.

section of the entity model with values from a given

scenario where based on the History & Examination

ﬁndings, a Diagnosis of Severe Asthma is made and

its Treatment speciﬁed. The ﬂow of how this process

should happen is shown in the workﬂow model.

3 PROPOSED METHOD

3.1 Gamiﬁcation Elements

The core concepts of games that should inform the

design of gamiﬁed e-learning systems are goal ori-

ented activities with reward mechanisms and progress

tracking (Strme

cki et al., 2015). In the training of

guideline content, the main goal is for the trainees to

learn how to treat different aspects of a disease as de-

scribed in the guideline. The reward mechanisms and

progress tracking aid in increasing the users engage-

ment and motivation (Bernik et al., 2018).

3.2 CPG Modelling

In our approach, we separate two aspects from a clini-

cal practice guideline (CPG). Medical conditions and

clinical encounters of patients is one aspect which we

model in an entity model; recommended clinical pro-

cesses is another aspect which we model in a work-

ﬂow diagram. A ﬂow in a guideline often consist of

medical conditions such as ‘start giving oxygen if a

child is convulsing for more than 5 minutes’. Typ-

ically a modeling approach speciﬁes all the recom-

mended processes in a workﬂow diagram. An in-

stance of the diagram would then specify a concrete

scenario representing the care processes executed for

a patient.

In our approach, we skip modeling the general

recommendations in a workﬂow diagram and directly

specify concrete scenarios. For example, in our ap-

proach we model a scenario where a 2 year old boy

who is convulsing for 7 minutes is admitted to a hos-

pital and we start treatment by giving oxygen. We

follow this approach as it allows us to model concrete

scenario with less effort and we do not need to spend

time on encoding the whole guideline. Since the pur-

pose of modeling the guideline scenario is to generate

questions using our approach, it is sufﬁcient for us

to model the scenario representing the recommended

clinical processes according to the guidelines. To rep-

resent such scenarios we integrate a CPG workﬂow

model with an entity model which is encoded as a

DPF model. Typically, a CPG consist of a large num-

ber of pages with information from the clinical do-

main. There exists some approaches that allows us to

design a visual model of a CPG.

In (Rabbi et al., 2014a; Rabbi et al., 2014b), the

authors presented an approach where different aspects

of a system were coordinated by means of multiple

metamodels. The approach is based on the foundation

of DPF.

In the multi-metamodeling approach, a workﬂow

model is integrated with an entity model by means of

metamodel coordination. A workﬂow metamodel is

used to design the ﬂow of a system and an entity meta-

model used to design the entities and relationship of a

domain. A workﬂow model can be used to represent

an abstraction of a CPG but we need to incorporate the

ENASE 2019 - 14th International Conference on Evaluation of Novel Approaches to Software Engineering

150

:Treatment

Oxygen: TreatmentName

Salbutamol: TreatmentName

Prednisolone: TreatmentName

treatmentName

4 years: Age

16 Kg: Weight

:Patient

:hasAge

:hasWeight

:History&Examination

Cough: Finding

Wheeze: Finding

Cyanosis: Finding

:hasFinding

:Diagnosis

Asthma: DiagnosisName

Severe: Severity

:hasSeverity

:diagnosisName

:receives

:hasTreatment

:hasDiagnosis

:conductedOn

A 4yr old boy weighting

16 kg presents with a

history of cough, wheeze

and on examination he is

found to have central

Cyanosis

Diagnosis: Severe

ashma

Treatment with oxygen,

Salbutamol and

prednisolone

Reassess after 20

minutes and treat

according to new findings

Workflow Model, WI

Entity Model, EI

Male: Gender

:hasGender

Figure 3: Integrated entity and workﬂow models.

detailed domain knowledge in our modelling. In this

paper we exploit the use of the multi-metamodeling

approach to represent the domain knowledge of a clin-

ical guideline and the clinical process and apply them

to execute a training session. The idea of using the

CPG workﬂow model is to control the ﬂow of the

game such that the user is interacting with the right

gaming element at the right time.

In this section we explain a training module which

consist of one or more CPG models and one or

more entity models represented as DPF speciﬁca-

tions. The states of the training module T M are

deﬁned by a set of elements that include a pair of

CPG workﬂow instance and a DPF entity instance

that represents the entities within a domain and re-

lationships between them. Figure 4 illustrates an

example of two states T M

and T M

of a training

module. The state T M

consists of a set of ele-

ments that include a pair of workﬂow instances and

DPF instances: {< WI

, EI

>, < W I

, EI

>, .. <

W I

, EI

} where W I

,W I

, ..W I

are workﬂow in-

stances and EI

, EI

, ..EI

are DPF entity instances.

Figure 4 shows a training session ﬂow which con-

sists of a sequence of states of training module i.e.,

Training

Flow1

:=< T M

, T M

, ....T M

>. Figure 3

shows an instance of a training session. In Figure 3

the game engine instantiates a training session by gen-

erating questions based on the DPF entity model and

CPG workﬂow model. For example, it could initially

generate a scenario based on the patient details and

history and examination ﬁndings and ask what the di-

agnosis is. If answered correctly, it will move on to

the next task and ask about the treatment. A training

session is composed of a sequence of training mod-

ules and is evolved from the initial state of a training

ﬂow and progresses based on the answer provided by

the user.

Figure 4: States of training module.

Training module

(State, TM )

Training module

(State, TM )

Training module

(State, TM )

Training module

(State, TM )

Transformation of training module’s state due to correct answer(s)

Figure 5: Progression of the states of training module.

In our approach a training session is evolved from

the initial state of a training ﬂow and progresses based

on the answer provided by the user. Figure 5 illus-

trates the idea of the progression of the states of train-

ing session. Depending on the answer given by the

user, a game engine consults with the training ﬂow

and evolves the state of the training session. We

use two DPF predicates < Enabled >, < Disabled >

to represent the current status of the training mod-

ules. A training module T M

when annotated with

the < Enabled > predicate indicates that the training

module is currently active and is being considered for

training.

The answers are collected from the user in two

different ways. We can ask the user to answer some

questions about the domain ontology i.e., entities and

their relationships; also the question can be based on

A Model Driven Approach to the Development of Gamiﬁed Interactive Clinical Practice Guidelines

151

Question

QuestionCategory

Game

requiredMinSkill

numberOfQuestions

passingCondition

Skill

title

dependency

Student

skill

PersonalInformation

questionCategory

title

questionCategory

questionFormat

Answer (ref to WI/EI)

reward

LearnerProfile

student

learningHistory

Figure 6: A conceptual model for the game elements.

the workﬂow instance. We utilize another DPF model

to formulate the questions that can be asked to the

user. Figure 6 shows the DPF model representing the

concepts of an e-learning game. The model represents

the game elements which include information about

how the game engine should control the game. In our

approach a game instance is associated with a train-

ing module. Using this DPF model we can specify

the number of questions to be asked, passing condi-

tion and can include questions under some question

category. We use references to the CPG instances and

entity instances for specifying correct and wrong an-

swers.

A general concern about this approach is the vali-

dation of the training ﬂow. While constructing a train-

ing ﬂow one might make mistake in two ways: (a)

wrong composition of instances of CPG model and

entity model; (b) wrong ﬂow of CPG instances. To

reduce the number of errors we apply inconsistency

checking as described in the next section.

4 SYSTEM DESCRIPTION

We propose to use a generic system based on the

idea of multilevel-metamodeling and their coordina-

tion. Figure 7 shows an overview of the system.

The responsibility of the ‘Game Engine’ is to control

the training ﬂow, maintain the status of the trainee,

produce dialogues or control the visualization of the

screen. The user should be able to interact with the

game engine via ‘Google Assistant’ or ‘Mobile appli-

cation’.

We plan to support different types of devices for

the training to facilitate training considering various

learning style of the trainee.

Question Flow Manager: The question ﬂow man-

ager selects the questions to be asked depending on

the level of difﬁculty of a training session. It main-

tains the order of questions to be shown to the user.

For example, user-A has skill-1 and chose to go

through the beginning session. While randomly se-

lecting questions that falls under the difﬁculty of ‘Be-

ginner’, it also looks into the questions that has been

used before for user-A. It puts more emphasize on the

questions that the user has been struggling with.

Conversation Manager: The conversation man-

ager keeps track of the conversation and manages the

context of the conversation. For example, if there are

three questions to be asked that is related to a child

who is 2 years old, then the conversation manager

produces a context for three questions and starts the

conversation saying “A 2 year old child comes to the

emergency department with <some condition>, an-

swer to the following questions:”. Afterwards it asks

the ﬁrst question, followed by the 2nd and 3rd ques-

tions.

User Management: The user management module

keeps track of the trainees skill, progress and effort.

The user management module is also used to produce

visualization showing the performance of a popula-

tion. If a group of trainee is particularly struggling

with a set of questions or question category then the

user management module will produce a report and

the trainer will be able to monitor it.

4.1 Adaptiveness

There are two ways our system facilitates adaptive-

ness:

• Customizing the gamiﬁcation process by means

of model-driven-engineering approach.

• Capturing the learning behaviour of trainee and

changing the model accordingly.

4.2 Methods

Our approach consists of modeling the entity which

represents the concepts related to clinical informa-

tion; the workﬂow which represents the ﬂow of a

guideline; and the structure of a game. We propose

ENASE 2019 - 14th International Conference on Evaluation of Novel Approaches to Software Engineering

152

Invocation

trigger

Conversation Action

1. invoke

fulfillment

Process

dialog input

2. user requests

Generate

dialog output

3. action responds

Game

Engine

Workflow

Model(s)

Entity

Model(s)

Game

Elements

Conversation

manager

Question

Flow

manager

User

management

DPF Models

Training managmenet

Figure 7: Overview of the proposed system architecture.

to use WebDPF tool for the modeling of these ele-

ments. The multilevel metamodeling supported by

the WebDPF tool allows us to design domain speciﬁc

modeling languages. The tool is also facilitated with

a constraint checker which allows us to check if the

models being constructed are valid or not. We per-

sist our models using JSON format. Figure 8 shows

an overview of the proposed system where the game

engine reads the models using a model parser. The

game engine executes the game by asking questions

to the trainee and also stores the answer in a relational

database. We have developed a report engine which

retrieves data from the relational database and visual-

izes learning analytics to the trainer. The trainer can

also interact with the report interface and can visu-

alize the data from different perspective. The trainer

can get an overview of the training modules and can

select individual trainees usage statistics. In our cur-

rent implementation we have not incorporated any

machine learning algorithms. Therefore, adapting a

training module according to the requirement of the

trainee are done manually by the knowledge engineer.

We used this modeling architecture to develop a

proof-of-concept game for the asthma guideline train-

ing. Figure 9 shows a sample conversation from the

asthma guideline training. While the participant is

using google assistant we use the google account for

registering the participant to our system. It is planned

to use OAuth 2.0 protocol for authenticating the user

from the mobile application to the participants Google

account. It will allow the user to switch from one de-

vice to another. While the participant is using the mo-

bile application they get more feature such as brows-

ing the guideline.

4.2.1 Implementation of the Mobile Application

The application is developed using React-Native and

JavaScript. React-Native is based on the React frame-

work, and is used to build mobile applications for An-

droid and iPhone. The motivation for using such a

framework is reuse of code when supporting both mo-

bile platforms as well as the web.

The game consists of a collection of quizzes,

where each quiz contains several questions. These

questions are based around a scenario, where the stu-

dent is presented with answer alternatives. Picking

an answer alternative will give the student points for

how close he was to the right action. The student is

presented with the answer key, an explanation, as well

as pointers to the evidence and the relevant guideline

for further study.

The quiz will conclude with a summary, giv-

ing feedback and statistics on students performance.

The quiz should have a passing condition to unlock

quizzes at a higher difﬁculty level. This is illustrated

in Figure 10.

4.2.2 Generating the Scenarios

To generate questions, we will write small scenarios

in the form of narrative templates where we use tags

to refer to variables in the entity model. The tag refers

to a path in the entity graph. The application will tra-

verse through the graph and return the value of the

given vertex.

A challenge with this method, is how to present

the data returned by the graph in a text. The value

from a measurement of the pulse-rate is just an in-

teger. An observation that the patient has a breath-

ing condition is a boolean, and an observation of the

patients level of consciousness is an enumerate of

A Model Driven Approach to the Development of Gamiﬁed Interactive Clinical Practice Guidelines

153

Model Parser

WebDPF

Persist

models in JSON

UI Flow

Dialog Flow

Game Engine

Training data

Report Engine

Learning analytics

Knowledge

Engineer

Trainee

Figure 8: Proposed method of the system.

Hello, welcome to the asthma guideline training program.

In order to continue, I need your permission to know your

name from google. Is that ok?

Welcome <<username>>. While the training is going on

you can always choose to go to the main menu or quit.

Please choose from the following options: See Progress;

Start Training; Quit.

What do you wish to do?

I would like to start training

I will ask 5 questions from the diagnosis and assessment

category. The questions have difficulty level ‘Easy’. Lets

start the training. A 2 year old boy comes to the

emergency department with a history of cough and

wheeze. Answer to the following questions.

First question, Central cyanosis would mean a diagnosis

of severe asthma. True or False

True

Correct answer! Second question…..

Figure 9: Sample ﬂow of conversation from the asthma

guideline training.

the AVPU (Alert, Verbal, Pain, Unresponsive) scale.

These values will have to be presented differently to

make a good sentence in the scenario. How we solved

this issue was by letting the vertex hold a string repre-

sentation of its value. This is illustrated in Figure 11.

<%Ben.name%> arrives at the emergency

department.

He <%Ben.hasConsciousness.value.name%>.

translates to

Ben arrives at the emergency

department.

He is not alert and not verbal,

but responds to pain.

5 RELATED WORK

In (Farkash et al., 2013) Farkash et al. presented a

model-driven approach to formalize clinical practice

guideline using natural rule language (NRL). Speci-

fying the constraints of a guideline with English-like

rule language reduces the gap of the representation

and processing of guidelines. The authors presented

a set of software components that support the repre-

sentation, interpretation of CPGs using NRL and also

can be applied directly to a patient’s EHR data for

analysis. Their approach is supported by a proof-of-

concept implementation for a simple essential hyper-

tension guideline directive. Our approach is different

with their approach as we use a graph based modeling

technique and the main contribution of our approach

is to support the training of a guideline by means of

gamiﬁcation.

In (Kristensen et al., 2009) Kristensen et al.

presented a conceptual model for e-learning where

the learning materials are divided into atomic units

and organized in several graph based models such

as ‘Knowledge map’, ‘Learning map’ and ‘Student

map’. These conceptual models provide a better

structure for representing an e-learning environment

and an easy-to-use navigation interface for exist-

ing learning materials. We borrowed concepts from

this paper and adapted them for representing CPGs

and game elements by means of Diagram Predicate

Framework and multi-metamodelling approach.

A gamiﬁcation approach was presented in (Akl

et al., 2008) where the authors followed the format

of TV game shows in which two teams of residents

compete in answering questions that are based on the

recommendations of guidelines. However, their ap-

proach is lacking formalization and does not support

model based analysis. In our approach we emphasize

ENASE 2019 - 14th International Conference on Evaluation of Novel Approaches to Software Engineering

154

Figure 10: Flow of the mobile application.

Narrative

hasNarrative

Patient Symptom AVPU

hasSymptom

hasValue

Ben Consciousness

hasConsciousness

is not alert and

not verbal, but

responds to pain

Model

Instance

value name:narrative

Figure 11: Importing variables from graph into scenario.

on the modeling part and apply model driven engi-

neering technique to interface with various platforms.

In (Del Cura-Gonz

alez et al., 2016) the authors

conducted a study to assess the effectiveness of a

teaching strategy for the implementation of CPGs us-

ing educational games. They demonstrated the results

for an e-learning game EDUCAGUIA to improve

knowledge and skills related to clinical decision-

making by residents in family medicine. The game

consists of educational games with hypothetical clin-

ical scenarios in a virtual environment. To iden-

tify the effectiveness of teaching strategies through

e-learning, they proposed an average score compar-

ison of hypothetical scenario questionnaires between

the EDUCAGUIA intervention group and the control

group. Such evaluation is very important and it re-

ﬂects the usefulness of utilizing games in teaching

guidelines. We plan to conduct similar evaluation of

our gamiﬁcation approach with healthcare profession-

als in future.

(Aouadi et al., 2016) uses Technology-Enhanced

Learning standards to develop serious games which

can be used in technological/professional/academic

ﬁelds for learning. Their goal was to make a scenario-

building approach, built upon a model driven archi-

tecture. The game includes a health course with

demonstrative videos and evaluation quizzes with

each course having a passing condition. The game is

also demonstrated as a 3D game in a context of med-

ical training. In their approach Aouadi et al., used a

platform independent model for the development of

game components which was transformed into a plat-

form speciﬁc model by means of ATL transformation.

While their approach is very close to our proposed

method, they lack modularization and separation of

concerns. In our approach we do not only apply mul-

tilevel metamodelling but also the integration of dif-

ferent modeling hierarchies which allows us to con-

veniently articulate various aspects of an e-learning

system.

(Wyatt et al., 2013) presents OKWA (Okay with

Asthma) which is a game targeted on children.

They aim at educating children with asthma in self-

management skill. This includes information about

medications, how to avoid triggering the asthma,

monitoring, when and how to get help from others.

The game is an interactive animated movie-style nar-

rative, where the actions the child chooses will have

an effect on how the story develops. Our project dif-

fers from this one as our target group is adult health

care workers. We will also focus more on evalua-

tion of medical knowledge through tasks and quizzes

rather than just story telling.

(Shegog et al., 2006) is similar to the OKWA

project, as it targets self-management skills for chil-

dren with asthma during a role playing game. The

game uses the child’s asthma proﬁle, so the child can

see the responses to his/hers health information in the

game. To complete a scenario, the character in the

game needs to be symptom free.

A Model Driven Approach to the Development of Gamiﬁed Interactive Clinical Practice Guidelines

155

(Zolfo et al., 2010) describes an approach where

they use mobile phones as a personal learning envi-

ronment for health care workers in resource limited

environments. They put an emphasize on the impor-

tance of avoiding health care workers being absent

from the health station for training programs. They

use didactic learning material (3d animations, video,

presentations, sound) and evaluates learning through

multiple choice questions. They also use Skype and

Facebook to have clinical case discussions with a

network of experts. The project differs to ours, as

they have an emphasize on didactic distance-learning

while we are learning through actively solving prob-

lems and tasks through gaming elements.

(Bartel et al., 2017) aims for a generic gaming

platform for implementing gamiﬁed learning arrange-

ments in engineering education. Their approach to

implementation is based on the concept of domain-

speciﬁc modeling, which is descirbed as an instance

of model-driven software design. However both theirs

and our projects are in the ﬁeld of model driven de-

velopment, Bartels work aims at engineering educa-

tion, while we aim at education of health care work-

ers. Their project is a work in progress, and have lim-

ited results to analyze and compare to.

(Pesare et al., 2016) presents both Edugame and

Simulation of Clinical Cases. Edugame is aimed at

patients and caregivers, to manage the disease and

promote a healthy lifestyle to avoid critical situations

and hospitalization. The game is a role-playing game,

and the users mission is to answer correct on the prob-

lems posed on the character in the game. Simulation

of Clinical Cases is a single player simulation game.

The goal is to save the character in the game, by the

user suggesting the right therapy, action and/or exam-

ination to solve the condition the patient in the game

has. The game adapts to which role the clinician has,

so the game will be different for a nurse and a physi-

cian. The game provides scores according to if the

answer was correct, partially correct or wrong. Our

project has a larger focus on model driven develop-

ment and a data model to easier add new content and

other types of games.

Septris (Evans et al., 2015) is an online training

tool to help emergency clinicians to identify and dif-

ferentiate between the different forms of the sepsis

syndrome. Pick the right diagnostic tests and provide

optimal management of the syndrome. Diagnose and

treatment is a big part of our project as well, but it

will be made general enough to make games for sev-

eral different medical conditions.

6 CONCLUSION

In this work, we have presented a model-driven ap-

proach to the design and development of a gamiﬁed

system for learning clinical guideline content. We

also present a prototype mobile e-learning system that

utilized our design approach in its development. In

the near future we aim to test our system with clini-

cians to evaluate its usability, acceptability and effec-

tiveness.

ACKNOWLEDGEMENTS

This work was supported in part by the NORHED

program (Norad: Project QZA-0484). The content

is solely the responsibility of the authors and does not

necessarily represent the ofﬁcial views of the Norwe-

gian Agency for Development Cooperation.

REFERENCES

Akl, E. A., Mustafa, R., Slomka, T., Alawneh, A., Ab-

hishek, V., and Sch

unemann, H. J. (2008). An edu-

cational game for teaching clinical practice guidelines

to internal medicine residents: development, feasibil-

ity and acceptability. BMC Med Education, 8.

Aouadi, N., Pernelle, P., Ben Amar, C., Carron, T., and Tal-

bot, S. (2016). Models and mechanisms for imple-

menting playful scenarios. In 2016 IEEE/ACS 13th In-

ternational Conference of Computer Systems and Ap-

plications (AICCSA), pages 1–8. IEEE.

Baker, A. (2001). Crossing the quality chasm: a new health

system for the 21st century. BMJ: British Medical

Journal, 323(7322):1192.

Barr, M. and Wells, C. (1990). Category theory for comput-

ing science, volume 49. Prentice Hall New York.

Bartel, A., Hagel, G., and Wolff, C. (2017). Work in

progress: Towards a generic platform for implement-

ing gamiﬁed learning arrangements in engineering ed-

ucation. In 2017 IEEE Global Engineering Education

Conference (EDUCON), pages 1502–1505. IEEE.

Bernik, A., Buba

s, G., and Rado

sevi

c, D. (2018). Measure-

ment of the effects of e-learning courses gamiﬁcation

on motivation and satisfaction of students. In 41th In-

ternational Convention-Mipro 2018.

Cabana, M. D., Ebel, B. E., Cooper-Patrick, L., Powe,

N. R., Rubin, H. R., and Rand, C. S. (2000). Barriers

pediatricians face when using asthma practice guide-

lines. Archives of Pediatrics & Adolescent Medicine,

154(7):685–693.

Cabana, M. D., Rand, C. S., Powe, N. R., Wu, A. W.,

Wilson, M. H., Abboud, P.-A. C., and Rubin, H. R.

(1999). Why don’t physicians follow clinical practice

guidelines?: A framework for improvement. Jama,

282(15):1458–1465.

ENASE 2019 - 14th International Conference on Evaluation of Novel Approaches to Software Engineering

156

Del Cura-Gonz

alez, I., L

opez-Rodr

ıguez, J. A., Sanz-

Cuesta, T., Rodr

ıguez-Barrientos, R., Mart

ın-

Fern

andez, J., A.-C. G., Polentinos-Castro, E.,

R.-C. B., Escortell-Mayor, E., Rico-Bl

azquez,

M., Hern

andez-Santiago, V., Azcoaga-Lorenzo,

A., Ojeda-Ruiz, E., Gonz

alez-Gonz

alez, A. I.,

Avila Tomas, J. F., B.-C. J., Molero-Garc

ıa, J. M.,

Ferrer-Pe

na, R.and Tello-Bernab

e, M. E., and

Trujillo-Mart

ın, M. (2016). Effectiveness of a strat-

egy that uses educational games to implement clinical

practice guidelines among spanish residents of family

and community medicine (e-EDUCAGUIA project):

a clinical trial by clusters. Implementation Sci., 11:71.

Deterding, S., Dixon, D., Khaled, R., and Nacke, L.

(2011a). From game design elements to gamefulness:

deﬁning gamiﬁcation. In Proceedings of the 15th in-

ternational academic MindTrek conference: Envision-

ing future media environments, pages 9–15. ACM.

Deterding, S., Sicart, M., Nacke, L., O’Hara, K., and Dixon,

D. (2011b). Gamiﬁcation. using game-design ele-

ments in non-gaming contexts. In CHI’11 extended

abstracts on human factors in computing systems,

pages 2425–2428. ACM.

Donaldson, M. S., Corrigan, J. M., Kohn, L. T., et al.

(2000). To err is human: building a safer health sys-

tem, volume 6. National Academies Press.

Evans, K. H., Daines, W., Tsui, J., Strehlow, M., Maggio, P.,

and Shieh, L. (2015). Septris: a novel, mobile, online,

simulation game that improves sepsis recognition and

management. Academic medicine : journal of the As-

sociation of American Medical Colleges, 90(2):180–4.

Farkash, A., Timm, J. T. E., and Waks, Z. (2013). A model-

driven approach to clinical practice guidelines repre-

sentation and evaluation using standards. Studies in

health technology and informatics, 192:200—204.

Fervers, B., Carretier, J., and Bataillard, A. (2010). Clin-

ical practice guidelines. Journal of visceral surgery,

147(6):e341–e349.

Goud, R., de Keizer, N. F., ter Riet, G., Wyatt, J. C., Has-

man, A., Hellemans, I. M., and Peek, N. (2009). Ef-

fect of guideline based computerised decision support

on decision making of multidisciplinary teams: clus-

ter randomised trial in cardiac rehabilitation. Bmj,

338:b1440.

Grimshaw, J. M., Sch

unemann, H. J., Burgers, J., Cruz,

A. A., Heffner, J., Metersky, M., and Cook, D. (2012).

Disseminating and Implementing Guidelines. Pro-

ceedings of the American Thoracic Society, 9(5):298–

303.

Kapp, K. M. (2012). The gamiﬁcation of learning and

instruction: game-based methods and strategies for

training and education. John Wiley & Sons.

Kristensen, T., Lamo, Y., Hinna, K. R. C., and Hole, G. O.

(2009). Dynamic content manager - A new concep-

tual model for e-learning. In Liu, W., Luo, X., Wang,

F. L., and Lei, J., editors, Web Information Systems

and Mining, International Conference, WISM 2009,

Shanghai, China, November 7-8, 2009. Proceedings,

volume 5854 of Lecture Notes in Computer Science,

pages 499–507. Springer.

Lohr, K. N., Field, M. J., et al. (1992). Guidelines for

clinical practice: from development to use. National

Academies Press.

owe, M. (1993). Algebraic approach to single-pushout

graph transformation. Theoretical Computer Science,

109(1-2):181–224.

Marriott, S., Palmer, C., and Lelliott, P. (2000). Dissem-

inating healthcare information: getting the message

across. BMJ Quality & Safety, 9(1):58–62.

NHLBI (2007). Expert panel report 3: guidelines for the

diagnosis and management of asthma. Number 97.

DIANE Publishing.

NICE (2017). Asthma: diagnosis, monitoring and chronic

asthma management. Nice Guideline 80.

Pesare, E., Roselli, T., Corriero, N., and Rossano, V. (2016).

Game-based learning and Gamiﬁcation to promote en-

gagement and motivation in medical learning con-

texts. Smart Learning Environments, 3(1):5.

Rabbi, F., Lamo, Y., and MacCaull, W. (2014a). Co-

ordination of multiple metamodels, with application

to healthcare systems. In The 5th International Con-

ference on Emerging Ubiquitous Systems and Perva-

sive Networks (EUSPN-2014)/ The 4th International

Conference on Current and Future Trends of Informa-

tion and Communication Technologies in Healthcare

(ICTH 2014)/ Afﬁliated Workshops, September 22-25,

2014, Halifax, Nova Scotia, Canada, volume 37 of

Procedia Computer Science, pages 473–480. Elsevier.

Rabbi, F., Lamo, Y., and MacCaull, W. (2014b). A ﬂexible

metamodelling approach for healthcare systems. In

Jaatun, E. A. A., Brooks, E., Berntsen, K. E., Gilstad,

H., and Jaatun, M. G., editors, Proceedings of the 2nd

European Workshop on Practical Aspects of Health

Informatics, Trondheim, Norway, May 19-20 , 2014.,

volume 1251 of CEUR Workshop Proceedings, pages

115–128. CEUR-WS.org.

Rodrigues da Silva, A. (2015). Model-driven engineering.

Comput. Lang. Syst. Struct., 43(C):139–155.

Rutle, A., Rossini, A., Lamo, Y., and Wolter, U. (2009). A

diagrammatic formalisation of mof-based modelling

languages. In Oriol, M. and Meyer, B., editors, Ob-

jects, Components, Models and Patterns, pages 37–

56, Berlin, Heidelberg. Springer Berlin Heidelberg.

Rutle, A., Rossini, A., Lamo, Y., and Wolter, U. (2012). A

formal approach to the speciﬁcation and transforma-

tion of constraints in mde. The Journal of Logic and

Algebraic Programming, 81(4):422–457.

Shegog, R., Bartholomew, L. K., Gold, R. S., Pierrel, E.,

Parcel, G. S., Sockrider, M. M., Czyzewski, D. I.,

Fernandez, M. E., Berlin, N. J., and Abramson, S.

(2006). Asthma Management Simulation for Chil-

dren: Translating Theory, Methods, and Strategies to

Effect Behavior Change. Simulation in Healthcare:

The Journal of the Society for Simulation in Health-

care, 1(3):151–159.

Shiffman, R. N., Michel, G., Essaihi, A., and Thornquist,

E. (2004). Bridging the Guideline Implementation

Gap: A Systematic, Document-Centered Approach to

Guideline Implementation. J Am Med Inform Assoc,

11(5):418–426.

A Model Driven Approach to the Development of Gamiﬁed Interactive Clinical Practice Guidelines

157

Strme

cki, D., Bernik, A., and Rado

sevi

c, D. (2015). Gami-

ﬁcation in e-learning: introducing gamiﬁed design el-

ements into e-learning systems. Journal of computer

science and technology, 11(12):1108–1117.

Wyatt, T. H., Li, X., Huang, Y., Farmer, R., Reed, D., and

Burkhart, P. V. (2013). Developing an interactive story

for children with asthma. The Nursing clinics of North

America, 48(2):271–85.

Zolfo, M., Iglesias, D., Kiyan, C., Echevarria, J., Fu-

cay, L., Llacsahuanga, E., de Waard, I., Su

arez, V.,

Llaque, W., and Lynen, L. (2010). Mobile learning

for HIV/AIDS healthcare worker training in resource-

limited settings. AIDS Research and Therapy, 7(1):35.

ENASE 2019 - 14th International Conference on Evaluation of Novel Approaches to Software Engineering

158