Software Design Principles for Digital Behavior Change Interventions

Lessons Learned from the MOPO Study

Lauri Tuovinen

, Riikka Ahola

2,3

, Maarit Kangas

2,3

, Raija Korpelainen

3,4,5

, Pekka Siirtola

Tim Luoto

, Riitta Pyky

2,3,4,5

, Juha R

oning

and Timo J

ams

2,3

Department of Computer Science and Engineering, University of Oulu, Oulu, Finland

Research Unit of Medical Imaging, Physics and Technology, University of Oulu, Oulu, Finland

Medical Research Center Oulu, Oulu University Hospital and University of Oulu, Oulu, Finland

Centre for Life Course Health Research, University of Oulu, Oulu, Finland

Department of Sports and Exercise Medicine, Oulu Deaconess Institute, Oulu, Finland

Department of Cultural Anthropology, University of Oulu, Oulu, Finland

Keywords:

Health Intervention, Behavior Change, Web Application, Software Design, Data Collection.

Abstract:

Using the Internet as a delivery channel has become a popular approach to conducting health promotion in-

terventions, and the evidence indicates that such interventions can be effective. In this paper we propose a set

of design principles and a generic architectural model based on experiences accumulated while developing a

Web-based application for a physical activation intervention. The proposed principles address the develop-

ment of an intervention application as an abstract entity, a platform for gathering data for the needs of three

principal stakeholder groups. The principles are derived from the purposes for which the data is gathered and

the constraints that may limit the availability of desired data; by observing these principles, developers of inter-

vention applications can identify the design trade-offs they need to make to ensure that all stakeholder needs

are adequately fulﬁlled. An evolutionary development process is proposed as a way of gradually working

toward an application that induces the desired effect on the behavior of the users.

1 INTRODUCTION

Computer software, and especially Web applications,

can be convenient and effective tools for the execution

of intervention studies that promote health behavior

change. However, the task of translating the objec-

tives of the intervention into software requirements is

far from trivial. The standard software engineering

practice of eliciting requirements from the customer

is not appropriate in this context, because there is no

customer in the conventional sense.

In the MOPO study

(Ahola et al., 2013), we have

developed and deployed a gamiﬁed Web portal in-

tended to persuade young men to improve their health

behavior, especially their physical activity. In this pa-

per we review the design challenges we needed to

overcome and the approach by which we chose to

address them. The paper focuses on problems that

http://www.tuunaamopo.ﬁ/sivu/ﬁ/mopo-

study in english/

we perceive as universal to all Internet-based behav-

ior change interventions, and on generalizing our ex-

periences as guidelines for developers of intervention

applications. The principal contribution of the paper

is a conceptual framework for designing the core ar-

chitecture of an intervention application and identify-

ing factors to be taken into account in the functional

speciﬁcation of the application.

In particular, we observe that when considered in

isolation of intervention objectives, a software appli-

cation developed for a health promotion intervention

is essentially a platform for collecting data from vari-

ous sources and delivering it to various stakeholders.

With all the details speciﬁc to a given intervention

thus abstracted away, specifying such an application

is reduced to identifying the information needs of the

stakeholders and the sources from which the required

information can be acquired. On the other hand, it is

equally important to identify the constraints that may

prevent the fulﬁlment of some of those needs. If there

Tuovinen, L., Ahola, R., Kangas, M., Korpelainen, R., Siirtola, P., Luoto, T., Pyky, R., Röning, J. and Jämsä, T.

Software Design Principles for Digital Behavior Change Interventions - Lessons Learned from the MOPO Study.

DOI: 10.5220/0005656101750182

In Proceedings of the 9th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2016) - Volume 5: HEALTHINF, pages 175-182

ISBN: 978-989-758-170-0

175

is some factor limiting the amount of data that it is

feasible to collect, then it may be necessary to bal-

ance what a stakeholder would ideally expect against

what is achievable in practice. Therefore, in order to

discover useful design guidelines, we will need to ex-

amine both the expectations and the limitations.

Regarding the scope of the paper, our intent is not

to claim that the abstraction of intervention applica-

tions as data collection platforms yields a compre-

hensive view, or that the list of stakeholders identi-

ﬁed in the paper is exhaustive. What we do argue is

that the abstraction provides one useful view that can

and should be supplemented with views addressing

other aspects and stakeholders of intervention soft-

ware. The result is not a complete methodology or

a universal architecture, but rather an exploration of

what can be learned about the design of intervention

software by adopting this particular perspective.

Section 2 of this paper presents an overview of

the MOPO study and the Web application developed

for it, along with a review of related work. Section

3 examines the expectations of different stakeholders

and how they affect the design of intervention appli-

cations. Section 4 considers major trade-offs that may

need to be made in order to design a feasible solution.

Section 5 discusses the ﬁndings and some directions

for future work, and Section 6 concludes the paper.

2 BACKGROUND

The medical research community has been aware

of the potential of the Internet as a communica-

tions channel in public health promotion interventions

since the late 1990s (Cassell et al., 1998). A vast

number of Internet-based interventions have been car-

ried out since then, targeting a variety of unhealthy

behavioral patterns such as physical inactivity [e.g.,

(Napolitano et al., 2003; Van Den Berg et al., 2006)],

smoking [e.g., (Graham et al., 2007; Saul et al.,

2007)], heavy drinking [e.g., (Linke et al., 2007;

Cunningham et al., 2009)], and high fat intake [e.g.,

(Frenn et al., 2005; Oenema et al., 2008)]. Several

reviews and meta-analyses have also been conducted,

indicating that Internet-based interventions have sig-

niﬁcant effects and providing recommendations for

future research and practice [e.g., (Bennett and Glas-

gow, 2009; Tate et al., 2009; Webb et al., 2010;

Brouwer et al., 2011)].

In the MOPO study, the principal target was to

persuade conscription-aged Finnish men to increase

their physical activity by means of a gamiﬁed Web

application. The effects of the application were stud-

ied in a 6-month randomized controlled trial (Ahola

et al., 2013). The study participants were recruited

at the annual conscription call-ups of the Finnish De-

fence Forces, which every male citizen of Finland is

required by law to attend in the year of his 18th birth-

day. Following two pilot studies carried out in 2011

and 2012 (Jauho et al., 2015), a full-scale interven-

tion was launched in 2013, with 248 participants in

the intervention group and 244 in the control group.

The interventions were carried out before the par-

ticipants began their national service, so that the study

tracked their activity as they went about their regular

daily lives. To measure their initial physical ﬁtness, a

series of ﬁtness tests were performed for each partic-

ipant and the results were recorded. At the end of the

study period, the same measurements were performed

again to see if there had been any signiﬁcant change

in the ﬁtness of the participants. To collect physical

activity data from the study participants, each partici-

pant was loaned a Polar Active activity monitor (Kin-

nunen et al., 2012) and asked to wear it for the du-

ration of the study; blinded devices, displaying only

date and time, were issued to the control group.

The application developed for the study is a Web

portal that provides the users with information on top-

ics related to health, well-being and ﬁtness, generates

visualizations and verbal feedback on their physical

activity, allows them to play a location-based digi-

tal game that rewards activity and mobility, and pro-

vides them with links to other relevant services and

resources. To collect the activity data recorded by

the study participants into a database, the participants

were instructed to periodically upload the data logged

by the activity monitor to the intervention server.

Manual logging of exercises was also possible, to ac-

count for situations where the wearable monitor was

unavailable or did not generate useful data. In addi-

tion to the activity data, information on how the ap-

plication was used and perceived was collected in the

study. The application itself was designed to log a

number of events and statistics, and the study partic-

ipants were also requested to voice their opinions on

what they liked and disliked about the application.

As the number of Internet-based health promotion

interventions such as the MOPO study continues to

increase, we argue that there is a need for a develop-

ment methodology that addresses the challenges spe-

ciﬁc to intervention applications. There is some pre-

vious work in this area that should be noted. Some

authors have created primarily descriptive conceptu-

alizations and critical reviews of various aspects of

the ﬁeld (Strecher, 2007; Barak et al., 2009; Danaher

and Seeley, 2009; Klasnja and Pratt, 2009), while oth-

ers have presented case reports of speciﬁc tools and

systems (Koskinen and Salminen, 2007; Alah

aiv

HEALTHINF 2016 - 9th International Conference on Health Informatics

176

et al., 2013). Most pertinent to our work are process

models, frameworks and design principles for the de-

velopment of persuasive systems for health behavior

change (Skinner et al., 2006; LaMendola and Krysik,

2008; Mohr et al., 2014; Nguyen et al., 2015), along

with more generic design methodologies considered

suitable for persuasive technology, such as participa-

tory design (Kensing and Blomberg, 1998).

Perhaps the single most relevant piece of previous

research is the BIT model (Mohr et al., 2014), which

addresses the software design of intervention appli-

cations among other considerations. However, the

model concentrates on a single perspective, namely

building a system capable of effecting health behav-

ior change, and this is generally true of the related

work cited above. Based on what we learned from the

MOPO study, we argue that the development of inter-

vention software should be approached from multiple

perspectives to account for the fact that designing an

application for a health intervention is an optimization

problem where maximizing the efﬁcacy of the inter-

vention is but one of several objectives. Another re-

cent and relevant study is (Nguyen et al., 2015), but its

topic is design of health information portals, whereas

our work concerns applications that are more interac-

tive and have more advanced application logic. On the

other hand, compared to more universal design prin-

ciples of persuasive technology or Web applications,

our work is more attuned to the problems and require-

ments speciﬁc to software for behavior change inter-

ventions. There is likely to be some overlap, but we

argue that intervention software is enough of a special

case to merit special treatment.

3 DESIGN PERSPECTIVES

Having established that the purpose of an interven-

tion application, when viewed on a high level of ab-

straction, is to collect, store and provide access to data

considered important for the success of the interven-

tion, we need to ask what exactly, then, is important.

The answer depends on which perspective the appli-

cation is viewed from; we discern three perspectives

that should be considered:

• From the perspective of the study participants, the

primary concern is operation of the application:

what its features are and how they are presented

to and used by the participants. The application

must be able to collect and store all data required

for it to deliver its functionality such that the ap-

plication is relevant and interesting to the users.

• From the perspective of the researchers, the pri-

mary concern is evaluation of the application:

how it performs with respect to the objectives set

for the intervention. The application must be able

to collect and store all data required for the re-

searchers to track relevant attributes of the study

participants and assess the performance of the ap-

plication objectively and reliably.

• From the perspective of the developers, the pri-

mary concern is modiﬁcation of the application:

how it could be improved to make it more effec-

tive as an intervention tool. The application must

be able to collect and store all data required for the

developers to clearly identify directions for fur-

ther development.

Although these are technically three distinct per-

spectives, in practice there is likely to be a substantial

amount of overlap. In the MOPO study, the physi-

cal activity data was used by both the researchers, to

enable them to see whether the interventions had any

effect, and the study participants, in the form of feed-

back intended to motivate them to achieve and retain a

healthy activity level. Similarly, the application usage

data was used by both the researchers, to enable them

to assess the extent to which the effects of the inter-

ventions were caused by the software, and the devel-

opers, to enable them to diagnose technical problems

and identify potential improvements.

Among the stakeholder groups identiﬁed above,

the researchers are in a key position when the func-

tionality of the application is speciﬁed. This is be-

cause all the requirements of the application are ulti-

mately dependent on the objectives of the study, and

it is up to the researchers to hypothesize on how to

achieve the objectives by means of software (opera-

tion), design a method for testing the hypothesis (eval-

uation), and draw conclusions on how to proceed af-

ter the study is ﬁnished (modiﬁcation). Translating

this knowledge into software speciﬁcations requires

close cooperation between the researchers and the de-

velopers, preferably with at least some members of

the development team having a dual role.

Probably the most elusive of the three design per-

spectives is the perspective of the study participants.

Deﬁning requirements to represent this perspective

involves a considerable amount of speculation on how

a given feature of the application would affect the be-

havior of the participants, and therefore a consider-

able amount of uncertainty. Consequently, this task

is substantially more difﬁcult than specifying features

required for the application to support evaluation and

interpretation of the study results. In the MOPO

study, representatives of the target demographic par-

ticipated in small-scale live testing of various devel-

opment versions of the intervention application, but

even with the feedback from these tests, it was difﬁ-

Software Design Principles for Digital Behavior Change Interventions - Lessons Learned from the MOPO Study

177

cult to predict how the participants of the intervention

would ultimately respond to the application.

The challenging nature of developing software ca-

pable of inducing behavior change emphasizes the

importance of the developer perspective. With so

much uncertainty concerning the impact of the soft-

ware, there is a high probability that the desired effect

will not be achieved on the ﬁrst attempt. This can be

responded to by adopting an evolutionary approach

where data collected during the study is used to make

decisions on how the software should be revised to

improve its impact on the behavior of the study par-

ticipants. The development process can thus be repre-

sented as a cycle of 4 Ds, as shown in Figure 1.

Achieving behavior change in the study partici-

pants is, of course, the primary goal in an interven-

tion study, but this does not mean that the other two

perspectives are of secondary importance: the im-

pact of the study cannot be reliably evaluated unless

the necessary data is available to the researchers, and

the evolutionary development process cannot advance

beyond the initial cycle unless the necessary data is

available to the developers. All three perspectives are

thus equally indispensable, and the needs of the corre-

sponding stakeholders are the fundamental forces that

shape the intervention application when considered as

a data collection platform. However, there are practi-

cal limits to what data, and how much data, it is feasi-

ble to collect in any given study; the factors that limit

data collection are discussed in the next section.

4 CONSTRAINTS AND

COMPROMISES

Analyzing the needs of the stakeholder groups yields

a speciﬁcation that states what data should ideally be

collected in order to maximize the likelihood that all

the identiﬁed needs can be fulﬁlled. In reality, how-

ever, it may be necessary to settle for a suboptimal

Figure 1: 4D model of the development process of interven-

tion applications. During each cycle of the process, a new

version of the application is developed and deployed, data

is collected for evaluation purposes, and decisions on what

to do in the next cycle are made based on the ﬁndings.

but acceptable solution. We discern three major fac-

tors that may force such trade-offs to be made:

• Cost: the resources available for data collection

and storage are limited by the budget of the study.

• Privacy: the study participants may not be willing

to disclose some data that would be desirable to

have for the purposes of the study.

• Convenience: the study participants may not be

willing to participate in data collection that re-

quires them to sacriﬁce their personal resources.

The cost of the resources required for data col-

lection and storage affects both which kinds of data

are available for collection and how much of them

can be collected during the study. If sensor devices

are to be used to collect data, then it is generally the

case that the more sophisticated the desired data is,

the more costly the required hardware will be. Data

storage, while less likely to prove prohibitively expen-

sive, nevertheless also needs some resources allocated

to it. Furthermore, both quantity and quality of data

may be constrained by personnel costs.

The privacy issue arises because some poten-

tial participants will object to being monitored, even

when given the assurance that their data will be used

only for the purposes of the study and kept strictly

separate from any information that might be used to

identify an individual participant. The more sensitive

the data that the participants are requested to give ac-

cess to, the more difﬁcult it will be to ﬁnd willing

participants, and the fewer participants there are, the

smaller the resulting body of data. Information con-

cerning the health of an individual is generally re-

garded as sensitive data, making privacy a potential

constraint in any study where some form of health in-

formation is collected from the participants.

Convenience is, in a sense, even more difﬁcult to

factor in than privacy, because its effect on data col-

lection cannot be known until when the study is al-

ready underway. Even a relatively small investment

of time or effort may prove enough of an inconve-

nience to cause some study participants to drop out of

the study, at which point it may be too late to replace

them. Using data collection methods that depend on

the participants actively providing data therefore runs

the risk of losing participants over the course of the

study such that it is not easy to predict how many will

stay committed. The usability of the main application

is also an important factor, since the participants are

likely to have a low tolerance to poor usability in an

application they use mainly because they are expected

to, not because they have a genuine use for it.

The MOPO study provides examples of compro-

mises made with respect to each of the constraints

HEALTHINF 2016 - 9th International Conference on Health Informatics

178

identiﬁed above. For instance, we wanted to keep

track of how certain key application usage statistics

changed over the course of the study. Instead of sim-

ply sampling these numbers at regular intervals, we

took periodic snapshots of the application database to

account for additional research ideas that might occur

during or after the collection of the dataset. However,

to reduce the need for storage resources, the most vo-

luminous component of the database — time series of

the study participants’ activity levels, sampled at 30-

second intervals — was left out of the snapshots. This

was judged an acceptable trade-off, given that includ-

ing this data in the snapshots would have multiplied

their size by two orders of magnitude.

An example of a trade-off between data collec-

tion and privacy is the collection of data represent-

ing the application usage behavior of the study partic-

ipants. From a technical point of view, it would have

been relatively simple to log the navigation history of

each participant individually, but there is a signiﬁcant

risk that this would have been viewed as overly in-

vasive. As a compromise, only statistical usage logs

were maintained, enabling the researchers and devel-

opers to see, for example, how much activity there

had been in each section of the application. An excep-

tion was made in the case of user logins; these were

logged individually, since this information was neces-

sary for the purpose of identifying and possibly con-

tacting study participants who had not used the appli-

cation, and also useful for troubleshooting purposes.

The convenience of data collection is the area

where it proved the most difﬁcult to ﬁnd a good trade-

off. Uploading activity data from Polar Active to the

application server required the study participants to

nicates with the monitor through Flowlink, a small

peripheral plugged into the USB port of the computer

running the client. To avoid overrunning the memory

buffer of Polar Active, each study participant needed

to repeat the upload process at least once every 21

days. In theory, this would not seem to be an exces-

sive inconvenience, but in practice, it was necessary

to send out reminders to the participants and to offer

them chances to win prizes by uploading their activ-

ity data. It was also found that any glitches in the

upload process, or in any other application function,

could have a substantial demoralizing effect, caus-

ing some participants to stop trying rather than con-

tact the developers for a solution (Luoto et al., 2014).

The precarious motivation of the participants poses a

considerable problem, since it would be important for

the developers to get their feedback on how to make

the application more captivating, but lacking the mo-

tive, the participants may not even use the application

enough to be able to give any meaningful feedback.

With regard to each constraining factor, it is im-

perative to make a deliberate decision on a trade-

off, to ensure that the outcome is both acceptable

and achievable. Overemphasizing either of these puts

the success of the study at risk: if too much weight

is placed on achievability, there is a risk of diluting

the objectives of the study, whereas if acceptability is

weighted too heavily, there is a risk that the study will

fail to collect enough data to support the objectives.

Either way, the further off balance the trade-off is, the

more difﬁcult it will be to achieve good results.

5 DISCUSSION

Table 1 shows some metrics based on login statis-

tics that can be used to evaluate the acceptance of our

application in the 2012 and 2013 interventions. The

most notable change is the increase in the percentage

of the intervention group who tried the application at

least once; there is no improvement when we look at

the ratio of the number of users who logged in mul-

tiple times to the number of all users. Thus it seems

that although the 2013 study had more success in per-

suading participants to start using the application, the

users lost interest in the application at a similar rate,

with only a relatively small number of them becoming

regulars. This does not mean that the studies did not

yield any positive results, but data from other sources

suggests that these were mostly caused by other mo-

tivating factors such as the activity monitor.

Some of the difﬁculty of motivating the partici-

pants of the MOPO study can probably be attributed

to the target population being particularly challeng-

ing to address effectively. The adverse health effects

of sedentary lifestyle generally do not occur until later

in life, so it is difﬁcult to get young people to feel that

Table 1: Some metrics computed from application login

statistics for the 2012 and 2013 interventions. Applica-

tion users are deﬁned as intervention group members who

logged in at least once, returning users as those who logged

in at least twice, and monthly users as those who logged in

at least four times between the beginning of the intervention

in September and the end of the year.

2012 2013

Intervention group (N) 141 248

Application users (N

) 66 160

Ratio of N

to N 0.47 0.65

Returning users (N

) 49 117

Ratio of N

to N

0.74 0.73

Monthly users (N

) 24 59

Ratio of N

to N

0.36 0.37

Software Design Principles for Digital Behavior Change Interventions - Lessons Learned from the MOPO Study

179

participating in the intervention is important to them

personally, if their only motivation is the abstract

knowledge that they have an elevated risk of suffering

from certain medical conditions in the future. Under

these circumstances, when executing a study based on

monitoring the participants while they act freely in an

uncontrolled environment, it may prove a consider-

able challenge to ensure that the participants do not

forget or neglect their commitment.

Related to this is our observation that it is not ad-

visable to rely on study participants to report prob-

lems encountered while using the application. People

who use an application because they need it for some-

thing have an incentive to report glitches, namely the

hope that they will be patched in a future update, but

this logic fails in the case of intervention applica-

tions, because the users may have no particular inter-

est in using the application. In this case their willing-

ness to report problems again depends on the depth of

their commitment, and an uncommitted study partic-

ipant, when encountering a problem that effectively

prevents them from using the application, may sim-

ply give up rather than seek help to solve the prob-

lem. Lesser usability issues, such as an unintuitive or

unattractive user interface, are similarly more discour-

aging if the user is not internally motivated to con-

tinue using the application.

This uncertainty concerning the commitment of

the study participants makes it all the more impor-

tant to design the application to automatically gather

data for development purposes. Automatic data col-

lection cannot completely substitute human feedback,

especially on subjective issues such as how interest-

ing or persuasive the application is, but it can be suf-

ﬁcient for the developers to diagnose situations where

the application performs inadequately. Most impor-

tantly, automatic notiﬁcation, whenever possible, is

the fastest way to inform the developers when a fail-

ure has occurred.

We can summarize the design imperatives derived

from our experience as follows:

• Designers of intervention applications should

identify the data required to fulﬁll the needs of all

stakeholders, and design the application to collect

all the required data. In particular, the application

should collect:

– all the data required for the application to de-

liver its speciﬁed functionality to the users.

– all the data required for the successful execu-

tion of the scientiﬁc work of the researchers.

– all the data required to inform the future deci-

sions and actions of the developers.

• Designers of intervention applications should

identify the constraints that may prevent the ap-

plication from collecting desired data, and design

trade-offs to ensure that the outcome is satisfac-

tory to all stakeholders. In particular, the data col-

lection objectives should be set such that:

– achieving them is possible using the resources

available for the execution of the intervention.

– achieving them is possible without unduly lim-

iting the privacy of the study participants.

– achieving them is possible without unduly in-

conveniencing the study participants.

Although these principles are all technically de-

rived from a single study, the fact that there were two

pilots before the full-scale intervention means that we

had multiple opportunities to accumulate experience

on the design and execution of Internet-based health

promotion interventions. This was particularly use-

ful in that it enabled us to observe the importance

of collecting data to support the speciﬁcation of the

next version of the intervention application, both au-

tomatically and by eliciting input from members of

the target population. Thus we came to apply the evo-

lutionary development approach outlined in Section

3, and as a result of this, both the ﬁrst pilot study

and the second one led us to identify major software

modiﬁcations that should be made before the next it-

eration. They also improved our understanding of

what data should be collected; for example, the full

metabolic equivalent (MET) data from Polar Active

was not used in the ﬁrst pilot, and some additional

digested values calculated from the MET data during

upload were used only in the ﬁnal intervention.

Given the amount of development work done in

the MOPO study, along with the fact that we have

abstracted away all application-speciﬁc details, we

would argue that the design principles proposed here

are generalizable to other Internet-based interven-

tions. Furthermore, the architecture of the MOPO

software is also potentially generalizable to a wider

range of applications. If we apply the abstraction used

in this paper to the architecture, we can reduce it to a

data repository and a set of interfaces through which

the three stakeholder groups access the data:

• The data upload interface, by which whatever

data on the study participants is being monitored

in the intervention is sent to the application.

• The user interface, by which the the study partic-

ipants submit requests to the application and have

information presented to them.

• The administrator interface, by which the re-

searchers and developers have unlimited access to

the data stored by the application.

HEALTHINF 2016 - 9th International Conference on Health Informatics

180

To beneﬁt from this abstract architecture, it could

be implemented in the form of a software framework

that would provide a generic foundation on which ap-

plications for many different digital behavior change

interventions could be built. The architecture of an

intervention application would then look like Figure

2, where the data interface APIs are provided by the

framework and the interface modules, which imple-

ment the application-speciﬁc functionality, are coded

by the application developers, using data import and

export services provided by the APIs. The inter-

face APIs handle communication with the application

database, either directly or via the intermediate an-

alytics API, which provides services for generating

various digested presentations from the original data.

The data repository is divided into two conceptu-

ally distinct sections: static content, which only the

administrators can write to, and collected data, which

holds the data acquired over the course of the study

by monitoring the participants. The internal archi-

tecture of the repository should reﬂect the modularity

of the data interfaces; due to the evolutionary nature

of the development process, data collection require-

ments are likely to change and the core architecture

must be able to accommodate these changes. With

several different stakeholder needs to be fulﬁlled, and

several factors constraining the quality or quantity of

the data that can be acquired to fulﬁll them, it is es-

pecially important to ensure that the design of the

application does not introduce any additional con-

straints. Making the data interfaces and storage struc-

tures modular will eliminate or at least signiﬁcantly

reduce the risk of not being able to use an important

data source because the developers were not aware of

it when the application was originally speciﬁed.

6 CONCLUSION

Web applications hold considerable potential as tools

for executing effective health interventions promot-

ing behavior change. In this paper we proposed a

set of design principles for such applications based on

the observation that an intervention application can be

viewed as an abstraction whose principal functional-

ity consists of gathering data from various sources and

presenting it to stakeholders in various formats. From

this observation, we derived three principles that pre-

scribe identiﬁcation of stakeholder needs, and another

three that prescribe identiﬁcation of constraints. Fur-

thermore, we concluded that an evolutionary devel-

opment approach is suitable for tackling the problem

of predicting how the users of the application will re-

spond to it. These ﬁndings are the result of hands-on

Figure 2: A conceptual model of the architecture of an in-

tervention software application developed using a software

framework. The interface APIs are generic components

provided by the underlying framework; the application-

speciﬁc modules use the APIs to implement the function-

ality of the application.

experience of developing and deploying three itera-

tions of a Web-based application designed to persuade

young men to change their physical activity behavior.

ACKNOWLEDGEMENTS

The MOPO study was supported by the Finnish Min-

istry of Education and Culture; Juho Vainio Founda-

tion; Centre for Military Medicine Finland; Northern

Ostrobothnia Hospital District; Centre for Economic

Development, Transport and the Environment of

North Ostrobothnia; European Social Fund; Finnish

Funding Agency for Technology and Innovation; and

European Regional Development Fund.

REFERENCES

Ahola, R., Pyky, R., J

ams

a, T., M

antysaari, M., Koskim

aki,

H., Ik

aheimo, T. M., Huotari, M.-L., R

oning, J.,

Heikkinen, H. I., and Korpelainen, R. (2013). Gami-

ﬁed physical activation of young men — a multidisci-

plinary population-based randomized controlled trial

(MOPO study). BMC Public Health, 13(32).

Alah

aiv

a, T., Oinas-Kukkonen, H., and Jokelainen, T.

(2013). Software architecture design for health BCSS:

Case Onnikka. In Proceedings of the 8th International

Conference on Persuasive Technology, pages 3–14.

Barak, A., Klein, B., and Proudfoot, J. G. (2009). Deﬁning

Software Design Principles for Digital Behavior Change Interventions - Lessons Learned from the MOPO Study

181

Internet-supported therapeutic interventions. Annals

of Behavioral Medicine, 38(1):4–17.

Bennett, G. G. and Glasgow, R. E. (2009). The delivery of

public health interventions via the Internet: Actualiz-

ing their potential. Annual Review of Public Health,

30:273–292.

Brouwer, W., Kroeze, W., Crutzen, R., de Nooijer, J., de

Vries, N. K., Brug, J., and Oenema, A. (2011). Which

intervention characteristics are related to more expo-

sure to Internet-delivered healthy lifestyle promotion

interventions? a systematic review. Journal of Medi-

cal Internet Research, 13(1):e2.

Cassell, M. M., Jackson, C., and Cheuvront, B. (1998).

Health communication on the Internet: An effective

channel for health behavior change? Journal of

Health Communication: International Perspectives,

3(1):71–79.

Cunningham, J. A., Wild, T. C., Cordingley, J., Van Mierlo,

T., and Humphreys, K. (2009). A randomized con-

trolled trial of an Internet-based intervention for alco-

hol abusers. Addiction, 104(12):2023–2032.

Danaher, B. G. and Seeley, J. R. (2009). Methodological

issues in research on Web-based behavioral interven-

tions. Annals of Behavioral Medicine, 38(1):28–39.

Frenn, M., Malin, S., Brown, R. L., Greer, Y., Fox, J., Greer,

J., and Smyczek, S. (2005). Changing the tide: An

Internet/video exercise and low-fat diet intervention

with middle-school students. Applied Nursing Re-

search, 18(1):13–21.

Graham, A. L., Cobb, N. K., Raymond, L., Sill, S., and

Young, J. (2007). Effectiveness of an Internet-based

worksite smoking cessation intervention at 12 months.

Journal of Occupational & Environmental Medicine,

49(8):821–828.

Jauho, A.-M., Pyky, R., Ahola, R., Kangas, M., Virtanen, P.,

Korpelainen, R., and J

ams

a, T. (2015). Effect of wrist-

worn activity monitor feedback on physical activity

behavior: A randomized controlled trial in Finnish

young men. Preventive Medicine Reports, 2:628–634.

Kensing, F. and Blomberg, J. (1998). Participatory design:

Issues and concerns. Computer Supported Coopera-

tive Work, 7(3–4):167–185.

Kinnunen, H., Tanskanen, M., Kyr

ainen, H., and Westert-

erp, K. R. (2012). Wrist-worn accelerometers in as-

sessment of energy expenditure during intensive train-

ing. Physiological Measurement, 33(11):1841–1854.

Klasnja, P. and Pratt, W. (2009). Healthcare in the pocket:

Mapping the space of mobile-phone health interven-

tions. Journal of Biomedical Informatics, 45(1):184–

198.

Koskinen, E. and Salminen, J. (2007). A customizable mo-

bile tool for supporting health behavior interventions.

In Proceedings of the 29th Annual International Con-

ference of the IEEE Engineering in Medicine and Bi-

ology Society, pages 5907–5910.

LaMendola, W. and Krysik, J. (2008). Design imperatives

to enhance evidence-based interventions with persua-

sive technology: A case scenario in preventing child

maltreatment. Journal of Technology in Human Ser-

vices, 26(2–4):397–422.

Linke, S., Murray, E., Butler, C., and Wallace, P. (2007).

Internet-based interactive health intervention for the

promotion of sensible drinking: Patterns of use and

potential impact on members of the general public.

Journal of Medical Internet Research, 9(2):e10.

Luoto, T., Korpelainen, R., R

oning, J., Ahola, R., Enwald,

H., Hirvonen, N., Tuovinen, L., and Heikkinen, H. I.

(2014). Gamiﬁed persuasion: User experiences of an

online activation service. International Journal of So-

ciotechnology and Knowledge Development, 6(4).

Mohr, D. C., Schueller, S. M., Montague, E., Burns, M. N.,

and Rashidi, P. (2014). The behavioral interven-

tion technology model: An integrated conceptual and

technological framework for eHealth and mHealth in-

terventions. Journal of Medical Internet Research,

16(6):e146.

Napolitano, M. A., Fotheringham, M., Tate, D., Sciamanna,

C., Leslie, E., Owen, N., Bauman, A., and Marcus, B.

(2003). Evaluation of an Internet-based physical activ-

ity intervention: A preliminary investigation. Annals

of Behavioral Medicine, 25(2):92–99.

Nguyen, B. V., Burstein, F., and Fisher, J. (2015). Improv-

ing service of online health information provision:

A case of usage-driven design for health information

portals. Infomation Systems Frontiers, 17(3):493–511.

Oenema, A., Brug, J., Dijkstra, A., de Weert, I., and de

Vries, H. (2008). Efﬁcacy and use of an Internet-

delivered computer-tailored lifestyle intervention, tar-

geting saturated fat intake, physical activity and smok-

ing cessation: A randomized controlled trial. Annals

of Behavioral Medicine, 35(2):125–135.

Saul, J. E., Schillo, B. A., Evered, S., Luxenberg, M. G.,

Kavanaugh, A., Cobb, N., and An, L. C. (2007). Im-

pact of a statewide Internet-based tobacco cessation

intervention. Journal of Medical Internet Research,

9(3):e28.

Skinner, H. A., Maley, O., and Norman, C. D. (2006). De-

veloping Internet-based eHealth promotion programs:

The spiral technology action research (STAR) model.

Health Promotion Practice, 7(4):406–417.

Strecher, V. (2007). Internet methods for delivering behav-

ioral and health-related interventions (eHealth). An-

nual Review of Clinical Psychology, 3:53–76.

Tate, D. F., Finkelstein, E. A., Khavjou, O., and Gustafson,

A. (2009). Cost effectiveness of Internet interventions:

Review and recommendations. Annals of Behavioral

Medicine, 38(1):40–45.

Van Den Berg, M. H., Ronday, H. K., Peeters, A. J., Le

Cessie, S., Van Der Giesen, F. J., Breedveld, F. C., and

Vliet Vlieland, T. P. M. (2006). Using Internet tech-

nology to deliver a home-based physical activity inter-

vention for patients with rheumatoid arthritis: A ran-

domized controlled trial. Arthritis Care & Research,

55(6):935–945.

Webb, T. L., Joseph, J., Yardley, L., and Michie, S. (2010).

Using the Internet to promote health behavior change:

A systematic review and meta-analysis of the im-

pact of theoretical basis, use of behavior change tech-

niques, and mode of delivery on efﬁcacy. Journal of

Medical Internet Research, 12(1):e4.

HEALTHINF 2016 - 9th International Conference on Health Informatics

182