Exploring the design of mHealth systems for health behavior change using mobile biosensors

Abstract

A person's health behavior plays a vital role in mitigating their risk of disease and promoting positive health outcomes. In recent years, mHealth systems have emerged to offer novel approaches for encouraging and supporting users in health behavior change. A promising technology in this regard are mobile biosensors, that is, sensors that enable the collection of physiological data (e.g., heart rate, respiration, skin conductance) and that are intended to be worn, carried, or accessed during normal daily activities. Designers of mHealth systems have started to use the health information that can be gained from physiological data for the delivery of behavior change interventions. However, research providing guidance on how mHealth systems can be designed to utilize mobile biosensors for health behavior change is scant. In order to address this research gap, we conducted an exploratory study. Following a hybrid approach that combines deductive and inductive reasoning, we integrated a body of fragmented literature and conducted 30 semi-structured interviews with mHealth stakeholders. Arising from this study, a theoretical framework and six general design guidelines were developed, shedding light on the theoretical pathways for how the mHealth interface can facilitate behavior change and providing practical design considerations.

Full text

C
ommunications of the
A
I
S
ssociation for
nformation
ystems
Accepted Manuscript
Accepted Manuscript
Exploring the Design of mHealth Systems for Health Behavior Change using
Mobile Biosensors
Tyler J. Noorbergen
School of Electrical Engineering and Computing
The University of Newcastle
tyler.noorbergen@uon.edu.au
Marc T. P. Adam
School of Electrical Engineering and Computing
The University of Newcastle
John R. Attia
School of Medicine and Public Health
The University of Newcastle
David J. Cornforth
School of Electrical Engineering and Computing
The University of Newcastle
Mario Minichiello
School of Creative Industries
The University of Newcastle
Please cite this article as: Noorbergen, Tyler J.; Adam, Marc T. P.; Attia, John, R.; Cornforth, David J.; Minichiello,
Mario: Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors,
Communications of the Association for Information Systems (forthcoming), In Press.
This is a PDF file of an unedited manuscript that has been accepted for publication in the Communications of the
Association for Information Systems. We are providing this early version of the manuscript to allow for expedited
dissemination to interested readers. The manuscript will undergo copyediting, typesetting, and review of the resulting
proof before it is published in its final form. Please note that during the production process errors may be discovered,
which could affect the content. All legal disclaimers that apply to the Communications of the Association for Information
Systems pertain. For a definitive version of this work, please check for its appearance online at
http://aisel.aisnet.org/cais/.

C
ommunications of the
A
I
S
ssociation for
nformation
ystems
Research Paper ISSN: 1529-3181
Accepted Manuscript
Exploring the Design of mHealth Systems for Health
Behavior Change using Mobile Biosensors
Tyler J. Noorbergen
School of Electrical Engineering and Computing
The University of Newcastle
tyler.noorbergen@uon.edu.au
Marc T. P. Adam
School of Electrical Engineering and Computing
The University of Newcastle
John R. Attia
School of Medicine and Public Health
The University of Newcastle
David J. Cornforth
School of Electrical Engineering and Computing
The University of Newcastle
Mario Minichiello
School of Creative Industries
The University of Newcastle
Abstract:
A person’s health behavior plays a vital role in mitigating their risk of disease and promoting positive health outcomes.
In recent years, mHealth systems have emerged to offer novel approaches for encouraging and supporting users in
health behavior change. A promising technology in this regard are mobile biosensors, that is, sensors that enable the
collection of physiological data (e.g., heart rate, respiration, skin conductance) and that are intended to be worn, carried,
or accessed during normal daily activities. Designers of mHealth systems have started to use the health information
that can be gained from physiological data for the delivery of behavior change interventions. However, research
providing guidance on how mHealth systems can be designed to utilize mobile biosensors for health behavior change
is scant. In order to address this research gap, we conducted an exploratory study. Following a hybrid approach that
combines deductive and inductive reasoning, we integrated a body of fragmented literature and conducted 30 semi-
structured interviews with mHealth stakeholders. Arising from this study, a theoretical framework and six general design
guidelines were developed, shedding light on the theoretical pathways for how the mHealth interface can facilitate
behavior change and providing practical design considerations.
Keywords: Behavior Change, mHealth Systems, Mobile Biosensors.

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
1 Introduction
According to the WHO, health promotion refers to “the process of enabling people to increase control over,
and improve, their health” (WHO, 2016). In addition to systemic factors (e.g., availability and pricing of food)
and access factors (e.g., ability to pay for food), an individual’s own choices, known as health behavior, is
recognized to play a vital role in determining their risk of disease and promoting positive health outcomes.
1
Overall, it is estimated that the health burden of diseases related to lifestyle behaviors (e.g., cardiovascular
disease, diabetes) will amount to US$47 trillion over the next two decades (Bloom et al., 2011), a large
portion of which could be prevented by changing people’s health behavior. For instance, it has been shown
that more than 80% of cardiovascular events could be prevented if people engaged in health behavior to
modify the four main lifestyle risk factors, namely alcohol overconsumption, inadequate nutrition, physical
inactivity, and smoking (Urrea et al., 2015). Similarly, engaging in a healthier diet and increasing physical
activity can substantially reduce the incidence of diabetes (Hamman et al., 2006). However, despite the
staggering loss in economic welfare and the associated detrimental impact on people’s quality of life,
achieving sustained and lasting change in people’s health behavior remains a societal challenge.
Over the past decade, mobile health systems (or mHealth systems) have emerged as a promising
technology to increase people’s control over their health and facilitate health behavior change (O’Reilly &
Spruijt-Metz, 2013). Enabled by advances in mobile devices and ubiquitous computing, mHealth systems
refer to mobile technology used to enhance access to health services (Wowak et al., 2016). With respect to
health behavior change in particular, mHealth systems offer novel modes for the delivery of technology-
mediated interventions that support users in modifying their behavior for improved health outcomes (Direito
et al., 2017). Thereby, a behavior change intervention (BCI) can be defined as a “coordinated [set] of
activities designed to change specified behavior patterns” (Michie et al., 2011, p. 1).
2
For instance, an
education intervention may support a user to engage in a healthier diet by providing educational material
on the health benefits of increased vegetable consumption through advice in the mHealth interface
(Mummah et al., 2016). With the wide proliferation and ubiquity of mobile technology in society, mHealth
systems enable the delivery of BCIs in a practical and cost-effective way that can reach a large number of
individuals and that may be tailored to the individual user (Direito et al., 2017).
One of the recent key developments for mHealth systems design is the increasing availability of mobile
biosensors, that is, sensors that enable the collection of physiological data (e.g., blood pressure, heart rate,
respiration, skin conductance) and that are intended to be worn, carried, or accessed by the user during
normal daily activities (Kumar et al., 2013; Urrea et al., 2015). Combined with contextual information (e.g.,
location and self-report data), the data obtained from mobile biosensors provide valuable insights into a
person’s health status and their lifestyle choices (e.g., risk for cardiovascular disease and diabetes; Ballinger
et al., 2018). For instance, mobile heart rate sensors can provide insights into how a person’s self-reported
smoking habits affect their resting heart rate, where high resting heart rates have been linked to an increased
risk of cardiovascular disease (Palatini et al., 2006; Papathanasiou et al., 2013). Similarly, biosensors for
measuring heart rate, respiration, and skin conductance provide insights into a person’s (physiological)
stress levels, even before individuals consciously perceive stress (Riedl, 2013). Designers of mHealth
systems have started to utilize this source of health information in the delivery of BCIs. For instance, Xiong
et al. (2013) used mobile biosensors to deliver a training intervention for building capability to perceive and
control physiological stress responses. By creating a feedback loop between a user’s behavior and the
physiological changes resulting from that behavior, the system enables users to train to control their
physiology with paced breathing exercises, while receiving real-time biofeedback on their heart rate and
respiration. However, despite the widely-acknowledged potential of mobile biosensors, there has been
limited research that provides guidance for how mHealth systems can be designed to utilize mobile
biosensors for health behavior change (Free et al., 2013; Kumar et al., 2013; Payne et al., 2015).
In this paper, we address this gap by conducting an exploratory study to inform the design of mHealth
systems that utilize mobile biosensors for facilitating health behavior change. In particular, we follow a hybrid
1
It is important to highlight that the terms health promotion and health behavior refer to people in general rather than patients. Hence,
promoting a person’s health is not constrained to the treatment of a particular disease. Instead, acknowledging people as the main
health resource, health behavior refers to the mitigation of risk factors and the pursuit of positive health outcomes (WHO, 1986).
2
As described by Michie et al. (2015), an effective BCI commonly builds on one or more specific behavior change techniques (i.e.,
observable and replicable components for changing behavior). The authors identified 93 different techniques (e.g., self-monitoring of
behavior, information about health consequences, feedback on behavior) that are commonly employed in the delivery of BCIs.

Communications of the Association for Information Systems
Accepted Manuscript
approach that combines deductive and inductive reasoning. Firstly, we integrate a body of fragmented
literature (deduction) for the development of a theoretical framework (Gregory & Muntermann, 2011). This
body of literature, while fragmented, allows us to derive specific propositions that bring to light the theoretical
pathways for how mHealth systems may facilitate health behavior change by utilizing mobile biosensors. In
doing so, the framework may support systems development by providing researchers and practitioners with
a shared frame of reference that allows them to systematically map out how the elements of their mHealth
interface can target individual components of behavior and the types of BCIs through which this can be
achieved. Secondly, building on this theoretical groundwork and the stakeholder groups that we identified
in the mHealth literature, we conduct a series of exploratory interviews (induction) with representatives from
the identified stakeholder groups (health practitioners, health insurance providers, health behavior
scientists, IT professionals, designers, policy makers, and users), leading to the development of six general
guidelines for the design of such systems. The guidelines add to the mHealth knowledge base by providing
system designers with a starting point of practical design considerations that take into account multiple
stakeholder perspectives. In this vein, we address the following overarching research question:
RQ: How can mHealth systems be designed to utilize mobile biosensors for health behavior
change?
The remainder of this paper is organized as follows. In Section 2, we provide an overview of previous
research on the design of mHealth systems for behavior change and the challenges that arise in that context.
Section 3 presents the research methodology for the employed hybrid approach. Section 4 presents the
results of our deductive theorizing and introduces an integrative theoretical framework for mHealth systems
in the context of health behavior change. Based on a thematic analysis of the interviews, Section 5 derives
six general design guidelines for the design of mHealth systems utilizing mobile biosensors. Section 6
provides a general discussion of our findings, discusses the limitations of the study, and outlines
opportunities for future research. Section 7 provides a concluding note.
2 Related Work and Background
2.1 Related Work on Designing mHealth Systems for Behavior Change
Driven by the ubiquity and increasing capabilities of mobile user devices in recent years (Danaher et al.,
2015; O’Reilly & Spruijt-Metz, 2013), mHealth systems have become a growing area for IS research and
practice. mHealth can be defined as “medical and public health practice supported by mobile devices, such
as mobile phones, patient monitoring devices, personal digital assistants (PDAs), and other wireless
devices” (WHO, 2011, p. 6). According to recent estimates, the number of mHealth apps on the major online
stores related to mHealth exceeds 250,000 (R2G, 2016). The two primary application domains that have
emerged for mHealth systems over the past decade are: (1) disease management and (2) health behavior
change. Firstly, for disease management the focus builds on patient-centered care (Stewart, 2001), that is,
empowering patients to manage their medical conditions more effectively and more independently (e.g.,
helping diabetics control their blood sugar; Kitsiou et al., 2017). Secondly, for health behavior change the
focus is on prevention and facilitating better health choices to prevent disease, that is, supporting and
encouraging users to engage in health behaviors for the promotion of positive health outcomes (e.g.,
improved diet, smoking cessation). In this paper, we focus specifically on the latter category.
Scholars have recognized that the design of a mHealth system plays an important role in its effectiveness
for bringing about behavior change. Several factors for the effective design of such systems have been
identified. Firstly, scholars have argued that the design of a mHealth system for behavior change needs to
be guided by a theoretical framework rooted in the BCI literature (Free et al., 2013; Hingle & Patrick, 2016;
Oinas-Kukkonen & Harjumaa, 2009). For instance, Hingle and Patrick (2016) argued that a profound
understanding of BCIs is critical when making recommendations to users in regards to changing their
behavior and that this ideally should be accomplished through the use of an established intervention
framework. Similarly, Labrique et al. (2013) argued that the lack of a common framework creates difficulties
in identifying, cataloging, and synthesizing evidence for the design of mHealth systems. Hence, guiding the
design with a theoretical framework also facilitates the evaluation of such systems. However, as noted by
Davey et al. (2014, p. 181), at this stage “most m-health studies are not guided by any conceptual
framework, neither the research questions are instigated by existing theories.” Hence, there is a need for
research that explores how mHealth system design for health behavior change can be guided by a
theoretical framework that is rooted in the BCI literature.

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
Secondly, recent reviews of existing mHealth systems (Samdal et al., 2017) and web-based eHealth
systems (van Genugten et al., 2016) have shown that systems are more effective in bringing about behavior
change if their design implements a higher number of behavior change techniques. However, at this stage,
the number of techniques employed in the delivery of BCIs is small (Conroy et al. 2014, Direito et al. 2017).
For instance, in a review of the 200 most popular mHealth apps (free and paid apps on Apple iTunes and
Google Play), Conroy et al. (2014) found that apps for physical activity on average only implement four
techniques in the delivery of BCIs (see also Direito et al. 2017). Similarly, a review of systems for alcohol
reduction has shown that the reviewed systems implemented less than four behavior change techniques on
average (Crane et al., 2015). Hence, it is vital for system designers to consider how they can implement a
larger number of techniques through the different pathways of BCIs. As described in the first point, the
implementation of these techniques should be guided by a framework grounded in the BCI literature (Garnett
et al., 2016; Hingle & Patrick, 2016; Vandelanotte et al., 2016). In particular, guiding the design with an
established BCI framework enables system designers to implement a higher number of behavior change
techniques because they can systematically consider a range of different potential pathways for
implementing BCIs in their artifact (Michie et al., 2015).
Thirdly, scholars have argued that system designers should consider the potential of mobile biosensors in
the delivery of BCIs. More broadly, the design strategy of utilizing biosensors as built-in functions of
information systems (vom Brocke et al., 2013, p. 3) allows designers to develop “systems that recognize
the physiological state of the user and that adapt, based on that information, in real time” (Riedl et al., 2014,
p. i, see Lux et al., 2018 for a review). In doing so, mobile biosensors can facilitate a feedback loop between
a user’s health behavior and their physiological state (e.g., biofeedback, Xiong et al. 2013, Uddin et al. 2016;
just-in-time interventions, Gutierrez et al., 2015). For instance, Adam et al. (2017) conducted a series of
interviews to explore how employing biosignals may lead to the development of stress-sensitive enterprise
systems that support users in the management and reduction of stress through interventions at the individual
(e.g., biofeedback to increase stress awareness) and organizational levels (e.g., organize break schedules
by understanding stress patterns). Based on systematic reviews of academic literature on the effectiveness
of mHealth systems, Free et al. (2013) and Schoeppe et al. (2016) conclude that overall there is (1) limited
evidence for the effectiveness of mHealth systems (with the exception of SMS) and (2) a need to further
explore how technologies such as mobile biosensors and video can be utilized in the delivery of BCIs.
3
Hence, in the present paper, we specifically focus on the case of mobile biosensors as a promising
technology for bringing about behavior change.
Fourthly, the involvement of mHealth stakeholders plays a critical role for the design process (Facchinetti
et al., 2012; Lobelo et al., 2016; Petersen et al., 2015). For instance, Eckman et al. (2016) argued that
enabling a collaboration between all stakeholders and making sure that all voices are heard is critical for
successful mHealth design. Several scholars have suggested co-design methods as a potential mechanism
for stakeholder collaboration. For instance, Marzano et al. (2015, p.947) argued that the challenge of
mHealth systems design “is a multidisciplinary one and is likely to be best met through a careful process of
co-design.” Similarly, Burke et al. (2015) argued that many of the pitfalls in current mHealth approaches can
be addressed through a process of interdisciplinary collaboration that encompasses the inclusion of end
users in all phases. However, despite suggestions for the use of co-design in mHealth, research that uses
co-design in the context of mobile biosensors and health behavior change is sparse.
4
In the present paper,
we engage with different mHealth stakeholder categories to conduct a set of exploratory interviews with the
goal of developing a set of general guidelines to support system designers in mHealth systems design for
behavior change. Hence, our paper does not carry out the co-design of an actual mHealth system artifact,
but instead engages with stakeholders to explore design considerations based on multiple perspectives.
2.2 Involvement of Stakeholders and Remote Systems
While scholars have recommended to better involve stakeholders in the mHealth design process (e.g.,
Eckman et al., 2016; Lobelo et al., 2016), to our knowledge there has been limited research that defines the
3
Reviewing 26 randomized control trials, Free et al. (2013) found that, other than SMS-based interventions (e.g., for smoking
cessation), mHealth-based BCIs had limited effects on health outcomes. Specifically, Free et al. (2013 p.3) refer to primary outcomes
(i.e., objective measures of health or health service delivery or use) as well as secondary outcomes as (i.e., self-reported outcomes).
Similarly, Schoeppe et al. (2016) found limited effectiveness in a review of 30 studies with a focus on diet and physical activity.
4
Donetto et al. (2015) investigate the application of co-design in the healthcare context, but do not discuss mHealth specifically. There
have also been several instances of mHealth system design in the context of mental health (Bardram et al., 2013; Ben-Zeev et al.,
2015; Thieme et al., 2016), however none of these studies specifically look at the potential of mobile biosensors.

Communications of the Association for Information Systems
Accepted Manuscript
necessary mHealth stakeholders and their degree of involvement in this process. Based on our review of
the literature, Figure 1 provides an overview of stakeholders and remote systems and how they interact in
the context of mHealth systems and health promotion. Overall, we identify seven different stakeholder
groups that may contribute important information to the design, these being: designers (D), health behavior
scientists (HBS), health insurance providers (HIP), health practitioners (HP), IT professionals (ITP), policy
makers (PM), and users (U) (Facchinetti et al., 2012; Lobelo et al., 2016; Petersen et al., 2015; Vandelanotte
et al., 2016). Thereby, the term direct stakeholders refers to stakeholders involved in the design and usage
of the system, whereas indirect stakeholders are involved in the design of the system, but do not use the
system directly themselves. For instance, HPs are direct stakeholders as they could use a mHealth system
to monitor users through a remote system, whereas HBSes and PMs have important influencing and
supporting roles. Specifically, the expertise of HBSes is vital for the design of a mHealth system as this
ensures that it builds on an established BCI framework (Lobelo et al., 2016; Petersen et al., 2015). PMs and
HIPs, on the other hand, can provide financial support for mHealth systems and make policy decisions
based on the data obtained from them (Facchinetti et al., 2012).
Further, the identification of different stakeholder categories highlights the need to understand how other
systems can be integrated into the approach. Specifically, we identify a set of remote systems which arise
in the context of mHealth systems. These remote systems are separate from the user device and allow for
more complex data analysis and involvement of other stakeholders (e.g., HPs). The first remote system is
the data aggregation and analysis service which aggregates and analyzes user data to allow for more
detailed feedback such as social comparisons between users. The health care provider IS provides
information relevant to HPs and allows them to send feedback to users. Lastly, the health insurance provider
IS provides information to HIPs which can help them become more involved in health promotion. Overall,
remote systems are vitally important for mHealth if there is to be any involvement of other stakeholders with
the system, and the importance of involving these stakeholders has been heavily emphasized in the
literature (Burke et al., 2015; Hingle & Patrick, 2016; Lobelo et al., 2016). In our interviews, we have recruited
representatives from all seven stakeholder groups shown in Figure 1.
Figure 1. Overview of Stakeholders and Remote Systems
3 Research Methodology
To address our research question, we decided to conduct an exploratory study and follow a hybrid approach
that combines deductive and inductive reasoning (Gregory & Muntermann, 2011). By combining deduction
and induction, this approach allows us to build on the advances in the established behavior change
literature, providing a theoretical grounding and focus for our research, and to explore a broad range of
design considerations based on multiple stakeholder perspectives in the mHealth space. Our research
methodology is summarized in Figure 2 (see Arnitz et al., 2017 for a similar conceptualization).
Biosensor
User
IT
Professionals
Direct Stakeholders
Feedback
Indirect Stakeholders
Influencing/Supporting
HEALTH PROMOTION
Data Aggregation &
Analysis Service
Remote Systems
Health
Care
Provider
IS
Health
Insurance
Provider
IS
Mobile Biosensor
Measurements
mHealth
Interface
User Device
mHealth System
Health
Behavior
Scientists
Designers
Policy
Makers
Health
Practitioners
Health
Insurance
Providers

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
3.1 Deduction: Development of an Integrative Theoretical Framework
Recent research on the efficacy of mHealth systems has argued that the design of such systems needs to
be appropriately underpinned by a framework grounded in the BCI literature (e.g., Free et al., 2013;
Morrissey et al., 2016; see also Section 2.1). In our deductive theorizing, we hence build on the extant
literature to investigate the theoretical pathways for how mHealth systems can utilize mobile biosensors to
facilitate health behavior change, and from this develop a set of propositions through an integrative
theoretical framework as suggested in Baumeister and Leary (1997). In order to address technological as
well as behavioral aspects in the design of mHealth systems, our synthesis of the literature covers research
in the disciplines of computer science, health, information systems, and psychology. As described by
Gregory and Muntermann (2011), deductive theorizing in the design of information systems builds on a
critical analysis and integration of the literature to develop propositions about the design of artifacts and the
way these artifacts provide utility to its users. It hence provides a theoretical underpinning for the design of
artifacts that is grounded in the literature.
In the first stage of our deductive theorizing, our study was informed by recent reviews on the efficacy of
mHealth systems (e.g., Direito et al., 2016; Free et al., 2013; Payne et al., 2015), BCIs (e.g., Fogg, 2009;
Michie et al., 2011; Weinmann et al., 2016), and the integration of biosignals into information systems (e.g.,
Riedl et al., 2014; vom Brocke et al., 2013). Building on the reading of academic literature, the five authors
of this article (with backgrounds in design, information technology, and public health) conducted several
interdisciplinary workshops with the goal of creating a list of key literature, stakeholders, and remote systems
for the subsequent stages of this study. During these workshops, which each lasted between one and two
hours, the authors engaged in group discussions on existing approaches, findings, and frameworks in the
literature that could lie the foundation for the development of an theoretical framework. In order to actively
seek expertise outside the boundaries of the author team, the authors invited six domain experts to three of
these workshops. The domain experts were selected to include a mix of expertise in BCIs, population health,
and user experience, and had between 5 and 30 years of research experience (avg. 16 years; one industry
practitioner, two postdoctoral researchers, three professors). In line with the exploratory nature of this
research, the domain experts were actively encouraged to bring in their expertise and suggest key research
streams, stakeholders, and remote systems for the context of our study. Thereby, it was explicit that it was
our goal to work across disciplinary boundaries and integrate a body of fragmented literature. This
consultation with domain experts enabled us to assemble a list of key literature in the respective areas (i.e.,
BCI frameworks, biosensor-enabled mHealth systems, co-design, and mHealth stakeholders) and create
an overview of stakeholders and remote systems to be investigated further (see Figure 1).
In the second stage of our deductive theorizing, the first and second authors developed an initial draft of the
framework by critically analyzing and integrating the academic literature identified in the first stage. Based
on a set of propositions derived from the literature, the framework conceptualizes the different pathways for
how mobile biosensors may be utilized to facilitate behavior change in a mHealth context. In the third stage,
the five members of the author team then iteratively refined the framework in three subsequent workshops.
During these workshops, which each lasted between one and two hours, the authors engaged in group
discussions on the components of the framework and the formulation of the propositions. Each author
individually prepared for those workshops by working through key literature on BCI interventions and already
existing biosensor-enabled mHealth approaches. This led to a clearer distinction of the different components
of human behavior, a separation of human physiology and mobile biosensor measurements, and a mapping
between specific BCI categories and the components of human behavior. The formulation of the
propositions in the theoretical framework is provided in Section 4.
3.2 Induction: Development of Design Guidelines
In our inductive theorizing, we conduct semi-structured interviews to develop a set of general design
guidelines for how mHealth systems can utilize mobile biosensors for behavior change. As described by
Gregory and Muntermann (2011), inductive theorizing in the design of information systems enables
researchers to integrate domain knowledge by considering multiple viewpoints and perspectives based on
real-world experience. The goal of our inductive theorizing is the development of general guidelines that
can aid the development of mHealth systems utilizing mobile biosensors for health behavior change.
Interview Design. The interview design was guided by the overview of stakeholders and remote systems
(Figure 1) and the results of our deductive theorizing. We decided to conduct semi-structured interviews,
because this enabled us to utilize the structure of the theoretical framework that we developed in our
deductive theorizing as a shared frame of reference with the interview participants and explore how the

Communications of the Association for Information Systems
Accepted Manuscript
theoretical pathways captured in the framework could be addressed by appropriate design. The interview
design was developed in several workshops involving all five authors of this article and comprises three
parts (see Appendix A). The first part of the interview builds on the overview of stakeholders and remote
systems. It explores participants’ understanding of mHealth systems in the context of health behavior
change and how it affects their own stakeholder domain. The second part refers to the theoretical pathways
for how mHealth systems may utilize mobile biosensors for behavior change, exploring how such pathways
can be realized from the perspective of the stakeholder, and whether there were any theoretical pathways
that we are missing. The third part refers to the development of general design guidelines based on the
experience and expertise of the stakeholders. In order to keep the interviews focused, we decided to refer
to mobile heart rate measurements as an example technology of mobile biosensors because heart rate
sensors provide important insights into a person’s health status (Acharya et al., 2007) and have become
increasingly accessible for daily use (e.g., Apple Watch, Samsung Gear). In particular, heart rate is
influenced by lifestyle behaviors such as smoking (Papathanasiou et al., 2013) and physical activity (Carter
et al., 2003), and heart rate measurements have been shown to be powerful markers for health, specifically
as a risk factor for cardiovascular disease, diabetes, and all-cause death (Fox et al., 2007; Palatini & Julius,
1997). For instance, a high resting heart rate has been linked to an increased risk of cardiovascular disease
and all-cause death (Fox et al., 2007; Palatini et al., 2006; Palatini & Julius, 1997), while a low resting heart
rate has been shown to be protective against cardiovascular disease (Palatini, 2009).
Early interviews were less structured to obtain a better grasp on the subject matter of the discipline and its
role in mHealth. However, as the study progressed the interview questions became more focused (Easterby-
Smith et al., 2002). All interviews included graphical representations depicting the integrative theoretical
framework that we developed in our deductive research, and the overview of stakeholders and remote
systems. As the study progressed, the current version of the design guidelines was shown to participants
in order for them to evaluate the guidelines and suggest refinements. Further, emphasizing the exploratory
nature of our interviews, it was made explicit that interview participants were asked to identify stakeholders,
theoretical pathways, and design aspects that they felt are currently missing in our work. The questions
were refined over time as more data were collected. Interviews were audio recorded so that responses
could be later analyzed to support the iteration and refinement of the design guidelines.
Sample. The sample chosen was based around the seven stakeholder categories identified in Section 2.2
(see Table 1; 30 interviews in total, one interview per participant). Participants were sourced by contacting
the directors of a medical research institute and of a local health district, asking for domain experts as
specified in the stakeholder categories. Further, users were sourced from the general population via face-
to-face contact and email. None of the participants were involved in this research in any capacity. The study
was approved by the ethics committee at the University of Newcastle, Australia (H-2016-0221) and informed
consent was obtained from all participants. Interviews occurred on campus or a location of the participant’s
choosing. Alternatively, interviews were also conducted via Skype. The duration of the interviews lasted for
an hour on average, however these varied in focus and length as the study progressed.
Table 1. Interview Table
Stakeholder Category
Number of Interviews
Designers (D)
5
Health Behavior Scientists (HBS)
6
Health Insurance Providers (HIP)
3
Health Practitioners (HP)
5
IT Professionals (ITP)
3
Policy Makers (PM)
2
Users (U)
6
Ʃ
30
Data Analysis and Development of Design Guidelines. The interviews were transcribed by the first
author, after which each transcript was validated by at least one of the other co-authors and sent to the
interview participants to check for potential corrections or omissions. Afterwards, the second author (an
experienced scholar in research on human-computer interaction), who was not involved in conducting the
interviews, used open and axial coding (Strauss & Corbin, 1990) to analyze the transcripts. This involved
carefully reading and color-coding the transcripts in order to identify an initial list of themes for the

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
development of design guidelines (91 codes linked to 14 themes). The themes were identified by critically
analyzing the codes assigned to the interview statements against the backdrop of the theoretical framework
and by identifying similar design concepts presented across the interviews. Afterwards, in five workshops
involving all five authors of this article, these 14 themes became the basis of the design guidelines which
were then iteratively refined. The authors prepared for the workshops by reading the interview transcripts
and carefully checking the analysis of the second author. The workshops, which each lasted between one
and two hours, were then used to arrive at consensus around the themes and design guidelines through a
process of selective coding (Strauss & Corbin, 1990). The goal of the selective coding was to examine the
initial list of themes and the linkages between them in order to identify general guidelines that best reflect
the expressed design considerations. This process was continued until the author team reached unanimous
agreement that the guidelines represented a coherent picture of the observations. Finally, each participant
was sent one email with the current version of the design guidelines asking for their feedback. Any feedback
received was used to further refine the design guidelines in the workshops.
Figure 2. Summary of the Research Methodology
Who? [When?] Research Task
DEDUCTION
Critical analysis and integration of
academic literature
List of key literature, stakeholders
and remote systems
Outcome
INDUCTION
Reading of all transcripts and color-
coding for initial list of themes.
Initial draft of integrative
theoretical framework
Intervention
Intervention
Feedback Loop
Intervention
Capability
Motivation
Opportunity
Health Behavior
Mobile Biosensors
P1
mHealth
Interface
Michie et al. (2011) COM-B System
P2
P3
P4
Tape-recorded and
transcribed interviews
Abstracted insights and
highlighted quotes
Design guidelines
Interdisciplinary workshops and
reading of academic literature
Workshops to revise framework and
develop interview design
Workshops to discuss abstracted insights,
highlighted quotes, and feedback
Interviews (N=30)
and transcription
[Nov 2015 –Mar 2016]
[Apr 2016 –May 2016]
[June 2016 –July 2016]
[Oct 2016 –Jul 2017]
[Aug 2016 –May 2017]
[Jan 2017 –Jul 2018]
Intervention
Intervention
Feedback Loop
Intervention Capability
(physical and psychological)
Motivation
(reflective and automatic)
Opportunity
(physical and social)
Health Behavior
(e.g., nutrition)
Mobile Biosensor
Measurements
(e.g., heart rate
measurements)
P1
Michie et al. (2011) COM-B System
P4
Physiology
(e.g., heart activity)
DESIGNOF THE MHEALTH SYSTEM USERPERCEPTION, BEHAVIOR, AND PHYSIOLOGY
Education
and Training
Coercion,
Incentivisation,
Modelling, and
Persuasion
Environmental
Restructuring,
Enablement, and
Restriction
P2
P3
mHealth Interface
User Device
Revised framework;
interview design
Authors 1-5
Authors 1 & 2
Authors 1-5
Authors 1-5
Author 2
Author 1
(interviews & transcription)
6 Health Behavior Scientists,
3 Health Insurance Providers,
5 Health Practitioners, 2 Policy Makers,
5 Designers, 3 IT Professionals, 6 Users
Authors 2-5
(validation of transcripts)
Consultation with six domain experts in
behavior change interventions, population
health, and user experience design.
Feedback from the 30 interview
participants on the current version of the
design guidelines (via one follow-up email
for each participant after their interview).
Current version of the design guidelines

Communications of the Association for Information Systems
Accepted Manuscript
4 An Integrative Theoretical Framework
In the following, we develop a framework to capture the theoretical pathways for how mHealth systems can
utilize biosensors to support behavior change. The development of our framework was informed by a review
of existing BCI frameworks, which led to the selection of the Behavior Change Wheel by Michie et al. (2011)
as the main underlying building block. We chose to build on the work of Michie et al. (2011) for several
reasons. Firstly, the Behavior Change Wheel by Michie et al. (2011) is based on an extensive review of
existing BCI frameworks and hence provides a comprehensive synthesis of the behavior change literature.
Secondly, this BCI framework has been used expansively in the health promotion literature and is therefore
familiar to stakeholders in health research, which is also the focus of our study (e.g., dietary interventions,
Robinson et al., 2013; cardiovascular disease risk management, Bonner et al., 2013). Thirdly, due to its
simplicity and accessibility the framework can be applied to a wide range of contexts, which is essential for
addressing behavior change challenges that require cross-disciplinary collaboration such as in our study.
4.1 The Behavior Change Wheel and the COM-B System of Behavior
The Behavior Change Wheel by Michie et al. (2011) is a BCI framework which allows various users to select
and design interventions and policies through an analysis of the nature of behavior, the components that
must be changed for initiating a behavioral change, and the interventions and policies necessary for
changing those components. The framework is organized into three layers going from outside to inside,
these being: (1) policies, (2) interventions, and (3) components of behavior. In this paper, we focus on the
latter two layers as the development of political interventions lies outside the scope of mHealth system
design. The Behavior Change Wheel catalogs nine different BCI categories (education, persuasion,
incentivization, coercion, training, enablement, modelling, environmental restructuring, and restriction) and
illustrates how these are linked to the components that make up behavior. These BCI categories can
influence one or more components within the inner most layer of the Behavior Change Wheel which Michie
et al. (2011) referred to as the COM-B system (see Figure 3).
Figure 3. BCI Categories and the COM-B System in the Inner Two Layers of the Behavior Change Wheel
The COM-B system comprises continually interacting components that generate behavior, these being:
capability, opportunity, and motivation (Michie et al., 2011). Capability is defined as an “individual’s
psychological and physical capacity to engage in the activity concerned” (Michie et al., 2011, p. 4). For
example, psychological capability involves having the necessary knowledge to achieve a behavioral target,
whereas physical capability involves being physically able to achieve a behavioral target. Opportunity relates
to “all the factors that lie outside the individual that make the behavior possible or prompt it” (Michie et al.,
2011, p. 4). Further, it can be subdivided into physical opportunity (opportunities in the physical environment;
e.g., having access to healthy foods) and social opportunity (opportunities in the social environment; e.g.,
language and concepts). Finally, motivation refers to “brain processes that energize and direct behavior”
(Michie et al., 2011, p. 4). Motivation can be broken into reflective motivation (conscious reflective
processes; e.g., planning and evaluation) and automatic motivation (affective processes; e.g., emotions and
impulses).
5
The single and double-sided arrows in the right part of Figure 3 conceptualize how a change in
5
Distinguishing between reflective and automatic motivation is important for mHealth system designers as it enables them to
systematically explore different pathways for addressing user motivation. As described by Michie et al., (2014), interventions targeting
Capability
(physical and psychological)
Motivation
(reflective and automatic)
Opportunity
(physical and social)
Health Behavior
(e.g., nutrition)
Michie et al. (2011) COM-B System
BEHAVIOR
Physical
Psychological
BCI CATEGORIES
Michie et al. (2011) Behavior Change Wheel
INTERVENTION DESIGN
Components of behavior
Intervention
Intervention
Intervention
…
…
Selection of
interventions

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
one component may indirectly influence another, and how the generated behavior can re-influence the
components in the COM-B system. For instance, an environmental restructuring intervention (e.g.,
increasing the availability of healthy food) which increases physical opportunity may also indirectly increase
motivation due to improving the convenience and access to performing the health behavior (e.g., eating
more healthily). Further, achieving this behavior may increase physical capability (e.g., weight loss) and
motivation (e.g., self-efficacy) which in turn can enable new behaviors to be performed.
4.2 Framework
In the following, we propose an application of the Behavior Change Wheel and the COM-B system in the
form of an integrative theoretical framework with four propositions (P1–4; see Figure 4).
Figure 4. An Integrative Theoretical Framework for Utilizing Mobile Biosensors in mHealth Systems for Health
Behavior Change (adapted from Michie et al., 2011)
On the right-hand side we conceptualize user perception, behavior, and physiology. This represents the
boundary of the user’s internal processes. Therefore, it is important to note that the perception of the user,
which involves their individual circumstances including their physical and social opportunities, influences
the effect an intervention delivered through the mHealth interface (P2, P3, and P4) will have on their COM-
B system configuration. Further, we extend the COM-B system in this framework by elaborating on the link
between health behavior and its accompanying change in physiology. For example, in the instance of the
health behaviors, physical activity and endurance training, both of these have been shown to reduce resting
heart rate (Carter et al., 2003; Woodward et al., 2014). We argue that the changes in physiology as a result
of enacted health behaviors can be captured by mobile biosensors, hence creating a feedback loop between
the user’s physiology and their perception, a link that is normally imperceptible to the user (P1).
4.3 Feedback Loop and Mobile Biosensor Measurements (P1)
The feedback loop in our framework encompasses the link between (1) a user’s physiology as a result of
their health behavior, (2) mobile biosensor measurements which allow the mHealth system to quantify
changes in the user’s physiology and use it as a system input, and (3) the interventions embedded in the
mHealth interface targeting a user’s health behavior through capability, opportunity, and motivation.
Importantly, a user’s health behavior has a direct influence on their physiology, regardless of whether the
user is provided with a feedback or not. These links between health behavior and physiology are well
documented in the established literature and physiological measurements have been shown to reveal early
indicators of health conditions (Fox et al., 2007; Palatini et al., 2006; Palatini & Julius, 1997). However,
users normally would not be able to discern how their health behavior affects their physiology and the long-
reflective motivation focus on instigating and supporting conscious processes that involve plans and evaluations (e.g., making a plan
to stop smoking after reflecting on the health benefits of smoking cessation). By contrast, interventions targeting automatic motivation
focus on affective responses and the reinforcement of routines and habits (e.g., reminders to reinforce the habit of reduced alcohol
consumption). By taking into account these different pathways, system designers may be more effective in addressing user motivation
as they can directly map out and consider how the design of each element of their mHealth interface may target one or even both
types of motivation (see also Section 4.6).
Intervention
Intervention
Feedback Loop
Intervention Capability
(physical and psychological)
Motivation
(reflective and automatic)
Opportunity
(physical and social)
Health Behavior
(e.g., nutrition)
Mobile Biosensor
Measurements
(e.g., heart rate
measurements)
P1
Michie et al. (2011) COM-B System
P4
Physiology
(e.g., heart activity)
DESIGN OF THE MHEALTH SYSTEM USER PERCEPTION, BEHAVIOR, AND PHYSIOLOGY
Education
and Training
Coercion,
Incentivisation,
Modelling, and
Persuasion
Environmental
Restructuring,
Enablement, and
Restriction
P2
P3
mHealth Interface
User Device

Communications of the Association for Information Systems
Accepted Manuscript
term consequences on their health only become apparent over years or even decades. By utilizing mobile
biosensors, mHealth systems have the capacity to close the loop between health behavior, physiology, and
user perception, providing saliency to underlying physiological processes that are normally imperceptible to
the user, and aiding decision making in a way that is motivating and timely.
6
In other words, while users
normally would not be able to see how their health behavior affects their physiology, the mHealth system
can make this link apparent and utilize it in the provision of interventions (e.g., showing positive physiological
consequences of enacted health behavior). These BCIs materialize through the mHealth interface where
they can influence capability, opportunity, and/or motivation, facilitated through the feedback loop that is
created based on mobile biosensor measurements.
Feedback has been shown to play an important role for bringing about behavior change. For instance, in
Control Theory by Carver and Scheier (1982), where behavior is seen as a goal-driven process, feedback
facilitates behavior change by revealing the discrepancy between current behavior and a behavioral goal.
This creates a feedback loop in which corrective adjustments are made to lower this discrepancy until the
behavioral goal is attained, or until the discrepancy is too large and causes disengagement from the goal
as a result of a lack of capability, opportunity, or motivation. In the context of information systems, persuasive
systems design also incorporates feedback as a method of attempting to change user attitudes or behavior
(Oinas-Kukkonen & Harjumaa, 2009). Fogg (2009) specifically looks at the link between persuasive design
and behavior change through the Fogg Behavior Model, which posits that in order for behavior change to
occur a person must be sufficiently motivated, have the ability to perform the target behavior, and be timely
triggered to perform the behavior. Feedback can be seen as one form of a trigger in this model. As for the
COM-B system by Michie et al. (2011), it is important to note that this framework does not explicitly include
feedback as it operates at a higher level of abstraction. However, in subsequent works, Michie et al. (2015)
developed a taxonomy of 93 behavior change techniques for the delivery of BCIs of which feedback
accounts for seven of these (e.g., biofeedback, feedback on behavior, feedback on outcome(s) of behavior).
Therefore, while not explicitly mentioned in the COM-B system in Michie et al. (2011), the concept of
feedback is implied through the relationships between the components.
We argue that feedback based on biosensor measurements is particularly important in facilitating behavior
change in the context of mHealth for several reasons. Firstly, it is currently difficult for people to monitor
changes in their physiology, and hence see how changes in their health behavior affect physiological
processes, because these processes are for the most part, imperceptible (Astor et al., 2013; Riedl et al.,
2014). However, with the increasing power, accuracy, and accessibility of mobile sensors in the collection
of physiological and contextual data relating to lifestyle behaviors (e.g., location, time of day), it is possible
to make these processes salient to the user through the use of feedback. This is important because changes
in physiology (e.g., decreased resting heart rate) usually precede changes that users can visually perceive
(e.g., weight loss). By making users aware of their physiological processes which are normally
imperceptible, it makes the relationship between specific behaviors and their resultant physiological
changes more salient, providing users with the opportunity to make more informed decisions. Secondly, the
provision of feedback can increase self-efficacy or the belief a person holds in their ability to influence events
that affect their lives (Bandura, 2010). This belief can change based on a person’s perception of how capable
they are in their own abilities. By providing feedback on a person’s current physiological state, the person
will be more psychologically capable due to having access to additional information which they would
otherwise not possess. Mobile biosensors can provide information about the connection between behaviors
performed and their subsequent effects on physiology. As a result, this access to information may lead to
an increase in a person’s perceived control over the outcome of their health, therefore leading to an increase
in self-efficacy, and possibly an increase in motivation to engage in related behaviors.
In sum, feedback plays an important role in behavior change as it can provide users with timely information
about their physiological processes in relation to lifestyle behaviors that would normally remain
imperceptible and, subsequently, increase users’ understanding of how their actions are leading to a change
in their bodily states. It hence makes the connection between health behavior and changes in physiological
process accessible to the user in a timely manner.
6
Miller (1978) states that feedback is important for instrumental learning (also referred to trial-and-error learning or operant
conditioning) in that feedback provides information about the successes and/or failures, providing an opportunity for users to adjust
their response. Without feedback, users are “like a blindfolded novice trying to learn to shoot baskets” (p. 291).

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
Proposition 1 (P1): Mobile biosensor measurements can be utilized to facilitate behavior change
interventions through the mHealth interface by creating a feedback loop
between users’ health behavior and their physiology.
However, the feedback loop described here only refers to the general pathway of mobile biosensor
measurements as a facilitator for BCIs. Hence, in order to leverage the potential of the established feedback
loop for health behavior change, mobile biosensor measurements need to be complemented with
appropriate mHealth interface design that addresses the components of the COM-B system. Building on
this feedback loop, the following subsections elaborate on the theoretical pathways for how the mHealth
interface can address capability (P2), opportunity (P3), and motivation (P4).
4.4 The Influence of the mHealth Interface on Capability (P2)
Capability refers to the psychological as well as the physical capacity to engage in an activity. Without the
capability to engage in the targeted activities, a change towards health behavior cannot be achieved. Michie
et al. (2011) identifies two interventions that can be used to increase capability: education and training. In
the following, we discuss how the mHealth interface can be utilized for increasing physical and psychological
capability through education and training interventions.
Michie et al. (2011) elaborate that physical capability can be increased through training interventions that
facilitate physical skill development. Numerous training interventions use activity sensors for improving
physical activity (Glynn et al., 2014), muscular fitness, movement skills, and weight-related behaviors (Smith
et al., 2014) by providing instructions on how to perform these behaviors through the mHealth interface.
Biofeedback training has also been used in the mHealth space to develop skills and improve users’ physical
capability to regulate their own physiological processes (Lux et al., 2018). For instance, Uddin et al. (2016)
developed a mobile training app called Beat, which uses an electrocardiographic sensor to provide real-
time biofeedback of heart rate variability. The application uses this biofeedback to build the user’s skill in
controlling their breathing rate to reduce stress and blood pressure. Similarly, Dillon et al. (2016) used
biofeedback for training based on mobile apps that utilize heart rate and skin conductance for stress
management. Hence, biosensors may be utilized to support users in training the regulation of their
physiological processes to improve their stress management capabilities.
On the other hand, Michie et al. (2011) explain that an increase in psychological capability can be
accomplished through education and training interventions that impart emotional, cognitive, and/or
behavioral skills. Common forms of education interventions in mHealth systems include self-monitoring and
performance feedback. Glynn et al. (2014) and Smith et al. (2014) employ education interventions by
providing users with performance feedback in relation to previously set health behavior goals (step count
and calories burned). Similarly, in the previously mentioned Beat app by Uddin et al. (2016), the authors
employ an education intervention in the form of a performance review which occurs at the end of the training
intervention. This review provides the user with feedback on their performance and visualizes the impact of
the breathing exercises and biofeedback on stress over time – a relationship between health behavior and
physiology that would normally be imperceptible for the user. Psychological capability can also be increased
through training interventions, for instance, through biofeedback based on heart rate or skin conductance
for improving the user’s emotion regulation capabilities (Astor et al., 2013; Peira et al., 2014).
Proposition 2 (P2): Mobile biosensor-based interventions that focus on education and training,
increase users’ psychological and physical capability to engage in health
behaviors.
4.5 The Influence of the mHealth Interface on Opportunity (P3)
Opportunity refers to the physical and social factors outside the individual that prompt behavior or make it
possible. Without the opportunity to engage in a particular activity, behavior change cannot occur. The BCIs
identified by Michie et al. (2011) to increase opportunity include: environmental restructuring, enablement,
and restriction. In the following, we discuss how the mHealth interface can utilize mobile biosensors to
facilitate interventions for increasing physical and social opportunity.
The mHealth interface can assist the user in changing the physical factors in their environment that prompt
behavior or make it possible. This can be accomplished through utilizing environmental restructuring,
enablement, and/or restriction interventions (Michie et al., 2011). Environmental restructuring interventions
aim to change the users’ physical or social context. One way the mHealth interface can increase physical

Communications of the Association for Information Systems
Accepted Manuscript
opportunity through environmental restructuring are just-in-time interventions, that is, interventions that
“deliver support at the moment and in the context that the person needs it most and is most likely to be
receptive” (Nahum-Shani et al., 2018, p.446). Researchers have begun to use mobile biosensors to facilitate
just-in-time interventions, thereby changing the user’s perception of their environment and, as a result of
this, increasing their opportunity to engage in health behaviors. For instance, Saleheen et al., (2015)
developed a mHealth system that uses respiration biosensors in combination with movement sensors
(accelerometers, gyroscopes) and contextual information (location based on GPS) to detect smoking
behaviors and trigger just-in-time interventions to stop smoking. Similarly, Gutierrez et al. (2015) developed
a mHealth system which detects alcohol intake for just-in-time interventions using heart rate and skin
temperature biosensors in combination with movement sensors (accelerometers, gyroscopes) and location
sensors (GPS). Another way the mHealth interface can increase physical opportunity is through enablement
which include interventions that increase means or reduce barriers beyond education or training (Michie et
al., 2011). For instance, based on biosensor measurements (e.g., a detected increase in resting heart rate
over time) the mHealth interface may facilitate the provision of individualized behavioral support (e.g., advise
on a change in routine), which increases the user’s means to engage in a targeted health behavior (e.g.,
decrease sodium intake). Lastly, the mHealth interface can increase physical opportunity through restriction
interventions which aim to reduce the opportunity to engage in adverse behaviors. In this sense, the systems
by Saleheen et al. (2015) and Gutierrez et al. (2015) could also be viewed as a restriction intervention as
their overall goal is to reduce the opportunity to engage in alcohol consumption or smoking.
By extending the social context of the user (e.g., facilitating access to communities), the mHealth interface
can also increase social opportunity in the users' cultural milieu. One way to achieve this through
environmental restructuring is to use mobile biosensors for facilitating social support (i.e., practical or
emotional help from friends, relatives, or colleagues). For example, Snyder et al. (2015) developed a mobile
biosensor-based system that facilitates social support for stress management. The system displays a user’s
current stress level (using skin conductance measurements) to people around the user, who can then
consider the user’s stress levels in their interactions with him/her. Further, Curmi et al. (2013) developed a
system that enables the opportunity for social support during physical activity by sharing the heart rates of
triathlon participants with members of their individual social networks in real time. Members of the social
network can express their social support by pressing a ‘Cheer’ button and the triathlon participant will receive
a direct feedback on this through their wearable device. Another way to increase social opportunity through
environmental restructuring is social comparison (i.e., comparison of a person’s own performance with that
of a peer). For instance, pointing out the percentile rank of users’ physiological stress levels (e.g., based on
heart rate and skin conductance) compared to their peers (e.g., same age and gender) creates a social
opportunity for them to improve their relative ranking (Lyons et al., 2014). Further, such social comparisons
may also be used for enablement. For instance, an enablement intervention may facilitate behavioral
support by enabling users to engage in online discussions with their peers (e.g., users who exhibit a similar
diet and resting heart rate) about practical approaches to attain a certain health goal (e.g., their individual
best practice for how they include additional servings of vegetables in their diet in order to lower their resting
heart rate), which in turn increases their means to engage in that behavior. Similarly, social opportunity can
also be addressed through restriction interventions, for instance, by using mobile biosensor measurements
to identify individuals or social groups who exhibit risk behaviors with adverse health effects (e.g., unhealthy
diet), and reduce the number of prompts that the user sees about such behaviors.
Proposition 3 (P3): Mobile biosensor-based interventions that focus on environmental
restructuring, enablement, and restriction, increase users’ physical and social
opportunity to engage in health behaviors.
4.6 The Influence of the mHealth Interface on Motivation (P4)
Motivation refers to reflective and automatic processes that energize and direct behavior (Michie et al.,
2011). Motivation is of central importance to behavior change because even if users have the capability and
the opportunity to carry out targeted activities, a change towards health behavior cannot be achieved without
a sufficient level of user motivation. The BCIs identified by Michie et al. (2011) to increase motivation include:
coercion, incentivization, modelling, and persuasion.
Reflective motivation focuses on instigating and supporting conscious processes that involve plans and
evaluations. BCIs in the Behavior Change Wheel which can be used to increase reflective motivation include
coercion, incentivization, and persuasion. Coercion interventions involve creating an expectation of
punishment or cost, while incentivization interventions create an expectation of reward (Michie et al., 2011).

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
One way the mHealth interface can utilize coercion or incentivization is by providing the user with information
of current and projected health benefits or ramifications based on their current physiological data and
behavior (e.g., future self; Rho et al., 2017). In particular, the health information extracted from physiological
data allows to project a user’s expected level of well-being (e.g., low stress levels) and risk of disease (e.g.,
cardiovascular disease, diabetes). In this sense, the prospect of disease can be interpreted as an expected
punishment or cost while the prospect of well-being and good health serves as an expected reward. By
providing this information to the user, the mHealth system can instigate reflective processes of planning and
evaluation (e.g., goal-setting for reducing stress levels as measured by skin conductance) in order to change
behaviors leading to these health outcomes (e.g., engaging in stress management). For example, Murray
et al. (2013) discussed how an avatar may mirror the health information of a user based on biosensors (e.g.,
stress) and support the user in devising a plan to change their health behavior. Building on this notion, the
commercial Oakwood Medical Avatar (1) uses an avatar that projects users’ current and future health by
extracting information from biosensors (e.g., blood pressure, muscle activity) and contextual data (e.g.,
sleep patterns, weight) and, based on this, (2) supports the user in planning health behaviors such as diet,
physical activity, and smoking cessation (Medical Avatar LLC, 2018). On the other hand, persuasion
interventions aim to trigger affective responses and stimulate action (Michie et al., 2011). Persuasion
interventions are similar to coercion and incentivization interventions, except that they focus more on how
a message is communicated. For instance, future health consequences based on the user’s current
behavior which are derived from mobile biosensors (e.g., smoking and drinking habits detected from
respiration, heart rate and skin temperature; Gutierrez et al., 2015; Saleheen et al., 2015) could be made
even more salient through imagery of those consequences (e.g., visually showing weight gain from drinking)
or facilitating a discussion with a health professional to devise a plan for action.
Automatic motivation involves processes which include “emotional reactions, desires (wants and needs),
impulses, inhibitions, drive states, and reflex responses” (e.g., reminders to reinforce the habit of reduced
alcohol consumption; Michie et al., 2014, p. 63). BCIs which can be used to increase automatic motivation
include coercion, incentivization, modelling, and persuasion. Coercion interventions can address automatic
motivation, for instance, by providing well-timed reminders with information about the health consequences
of risk behaviors (e.g., reminders containing imagery of negative health outcomes). For instance, the system
could use biosensors to detect smoking (Saleheen et al., 2015) or alcohol consumption (Gutierrez et al.,
2015) and use reminders to reinforce the formation of healthy habits. Conversely, reminders may also be
used to increase automatic motivation through incentivization. For instance, S. S. Martin et al. (2015) used
an incentivization intervention in the form of smart texts which provided positive reinforcement messages to
users based on their daily activity goal. Similarly, positive reinforcement messages could be sent for every
day that no smoking or drinking consumption was detected based on biosensors. Modelling interventions,
defined as “provid[ing] an example for people to aspire to or imitate” (Michie et al., 2011, p. 7), can also be
used to increase automatic motivation. For instance, using social comparison based on mobile biosensors
(e.g., using respiratory sinus arrhythmia measurements (Xiong et al., 2013) to display a paced breathing
leaderboard for stress management) can allow friends or people within a similar demographic to become
an example to aspire to for the user based on their positive example. Lastly, persuasion can be used to
increase automatic motivation. Similar to reflective motivation, imagery which makes salient the
consequences or benefits of behavior (e.g., visually showing the consequences of having a high resting
heart rate) can influence automatic motivation by triggering emotional responses to such visual stimuli.
Proposition 4 (P4): Mobile biosensor-based interventions that focus on coercion, incentivization,
modelling, and persuasion increase users’ reflective and automatic motivation
to engage in health behaviors.
5 Design Guidelines
Based on the thematic analysis, this section derives six general design guidelines for designing mobile
biosensor-enabled mHealth systems for health behavior change (see Appendix B for an overview). As
described in Section 3.2, the interviews referred to mobile heart rate measurements as an example
technology of mobile biosensors.
5.1 Guideline 1: Mobile Biosensor Recordings
Guideline 1 refers to the specifics of how biosensor data can be measured in mHealth systems to better
understand the circumstances of the user and the collection of contextual data which supports this. The
measurements provide the basis for the feedback loop between lifestyle behavior and physiology. We

Communications of the Association for Information Systems
Accepted Manuscript
extensively discussed this topic with participants, and the first consideration which emerged is that the
measuring device must be physically tuned to the person. This perspective was explicitly supported by HP1
and HP5. Other aspects of measurement that were discussed included duration, frequency, time of day,
and position. In regards to duration, HP1 recommended that for the example technology of mobile heart
rate measurements “[…] five minutes is fairly acceptable from current guidelines”. This is consistent with
Task Force (1996, p. 364) which recommend that “[…] 5 min recordings of a stationary system are preferred
unless the nature of the study dictates another design.” The advice from HPs in regards to time of day and
position indicated that consistency of these aspects (e.g., measurement start always around 8 am, always
sitting position) is the more important than the selection of a specific condition (e.g., 8 am vs 3 pm; sitting
position vs standing position). Further, HP5 added that in addition to these specific 5 min recordings, it
would be important to generally keep continuous recordings using a rolling time window of the last 48 hours
to be able to access this data in case the user exhibits a “funny turn”
7
. Generally, there was strong support
from users for the use of biosensor measurements in mHealth with the main reasons focusing on improved
visibility and improved understanding of physiological data in the context of health.
In order to better understand biosensor measurements, HPs emphasized that collecting contextual data is
critical in the interpretation of the physiological data. A major theme that emerged was the degree to which
physiological and contextual data should be measured passively or manually entered by the user. Types of
passive data discussed included steps, contextual location, blood pressure, and pulse. HP2 and HP4
supported the use of steps as feedback due to it being a discrete and cheap measure that can complement
physiological measurements for determining movement (resting vs moving heart rate). Further, the use of
contextual location from GPS was identified as useful data as it could assist in understanding the
circumstances leading up to health events. However, there were some users (e.g., U6) which disliked the
use of contextual location for privacy reasons. Manually entered data that was discussed included corrective
factors for physiological data (e.g. age, sex, medical / family history), utilization of health care resources
(e.g. hospital visits), and mood. Collecting corrective factors are necessary for improving the accuracy of
physiological measurements. For instance, β-blocker and rate limiting medications pharmacologically lower
heart rate, hence impairing the usefulness of physiological measurements as a measure of feedback
(Palatini, 2009). However, HPs emphasized that this was an issue more for chronic disease management
rather than prevention, with HP2 stating: “[…] resting heart rate is a measure of fitness as long as people
are not confounded with rate limiting drugs”. The majority of participants favored the use of passive
collection due to reasons of convenience and accuracy. For instance, U4 stated in regards to manual entry
that they would “[…] [try] to overdo it which would provide inaccurate results.” In sum, to better understand
physiological measurements a mixture of passive and manually entered contextual data should be collected,
with passive measurement being the predominantly utilized method.
5.2 Guideline 2: Affective Visual Assets
Guideline 2 refers to the use of affective visual assets to convey the health information extracted from
physiological data in an intuitive and meaningful way in order to increase users’ capability and motivation.
This addresses the challenge that physiological data are not intuitive to understand for the user (capability)
and meaningful to change their behavior (motivation).
Throughout the interviews it emerged that users find it difficult to understand the information embedded in
physiological data and how it relates to their health goals. For instance, HBS1 emphasized that “[…] as
someone from the general population, [heart rate] is not something that you know what it means in terms of
what’s good and what’s bad.” U2 echoed this sentiment, stating “Most people aren't aware of their health.
A lot of people don't go to the doctor unless they've got a health issue, so most people would have no idea
what their blood pressure or heart rate is, or what it means.” Hence, in order to support users’ psychological
capability to engage in healthy behaviors, the mHealth interface needs to convey this information in an
intuitive way. In the interviews, participants expressed that appropriately designed visual assets could help
bridge the gap between biosensor measurements and user understanding by providing a visual
representation of the relevant physiological measure. Recent examples in the literature support this notion.
For example, Tan et al. (2014) used visual assets resembling human elements (heart changing size, sweat
droplets) to convey a user’s stress levels extracted from mobile heart rate and skin conductance
measurements in an intuitive way. Similarly, scholars have used nature-inspired visualizations as analogies
7
Storing continuous biosensor recordings over a rolling time window of 48 hours allows to access health information in case of
exceptional circumstances. For instance, having access during and/or before a user is experiencing a ‘funny turn’ provides important
information to health practitioners to better understand the circumstances of the event (e.g., atrial fibrillation, tachycardia).

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
to represent a user’s health status. For instance, Al Osman et al. (2016) and Feijs et al. (2013) used visual
assets representing trees and flowers that change their appearance based on heart rate and respiration
measurements. Hence, by reducing the complexity of the underlying physiological data and by building on
analogies, visual assets can aid in making these measurements more intuitive to understand for users and
hence increase their psychological capability to engage in targeted behaviors.
Further, HBSes emphasized that even if visual assets convey the health information in an intuitive way, this
will not motivate users to engage in health behavior change unless this change becomes meaningful to
them. For instance, reducing a stress level measure extracted from respiration or skin conductance data
may in and by itself not be sufficiently motivating to change behavior. Hence, in order to not only address
capability but also motivation, visual assets should be designed in a way that makes it meaningful to change
their behavior. In particular, HBSes argued that visual assets should also be designed to be affective in the
sense that they can create an emotional bond with the user. For instance, HBS3 suggested that using
affective visual assets could make interventions more personal as it “[…] might feel like you're actually
talking to a person rather than just getting a brochure telling you to not smoke.” During our interviews, three
particular types of visual assets emerged for this goal: 1) mirrored-self avatars, 2) persuasive avatars, and
3) embodied agents. The mirrored-self avatar concept employed by Behm-Morawitz (2013) showed that the
appearance of an avatar representing a user can influence the real-world behavior of that user.
8
For
instance, a mirrored-self avatar could change its appearance (e.g. weight, age, skin) based on the user’s
physiological data (e.g., stress level, smoking and alcohol intake). Hence, by mirroring the health status of
the user, a mirrored-self avatar may establish a meaningful and at the same time intuitive link between the
user’s physiology and their health behavior. HBS1 supported the mirrored-self concept, particularly for
increasing self-efficacy, stating “[…] I think it does tie into the self-efficacy because when people are doing
well their avatar can reflect that […] it’s reinforcing in terms of the capability […] because that's one of the
issues is people not believing that they can do it.” By contrast, a persuasive avatar is an avatar that is
representative of another person, typically an authority figure (Hanus & Fox, 2015) such as virtual doctors
(Fujita et al., 2010) and virtual coaches (Buttussi et al., 2006). For example, feedback on biosensor data
(e.g., stress levels based on respiration and skin conductance) could be provided through a persuasive
avatar representing a HP (increasing trust and credibility in the interpretation of physiological data, see
Guideline 4). While participants generally supported the concept of a persuasive avatar, HBS4 added that
the level of persuasiveness needs to be carefully considered against the backdrop of the “person’s
relationship with authority figures”. Similar concerns were raised by several users. Finally, embodied agents
emerged as a third type of visual asset in the interviews. For instance, virtual pets which are a type of
embodied agent that employ a mix of anthropomorphic and non-human elements and are influenced by
user engagement (Kromand, 2007). Similarly, nature-inspired elements such as trees and flowers (e.g.,
reflecting physiological stress; Al Osman et al., 2016, Feijs et al., 2013) that we discussed earlier in the
context of capability can also be used to address motivation by means of an embodied agent that makes a
change in physiology more meaningful. In the interviews, virtual pets were seen as suitable for younger
users as these typically are situated in game settings.
While affective visual assets conveying the physiological state of the user was generally embraced by all
participants, particularly in terms of addressing capability, it became evident that the actual choice of which
visual assets is to be employed must be a personal choice in order to effectively address motivation. For
instance, in the context of avatars and embodied agents, there was a large variance of preferences among
the participants. Further, U1 stated that avatars would not work for them at all, but that other affective visual
assets would work, with the key being that the asset is individualized: “if somebody is an enthusiast in a
different area, it could be aimed at the things that they are enthusiastic about.“ One of the most popular
assets discussed was the mirrored-self avatar which received the most support, but also the most criticism.
The degree of realism was seen as both a strength and a weakness. While users preferred the directness
of the feedback on their physiological state from this avatar, with U2 stating “I think that people sometimes
need to be scared into doing something about their health […] I think it certainly would hit home to us more
than just reading numbers and things on a screen”, D1, D5, and ITP2 stated that it could be very confronting
and that if the avatar is too similar it could become eerie (see notion of “uncanny valley” by Mori, 1970);
therefore it may be preferable to deliberately pursue a design with moderate human likeness. In sum, while
overall using affective visual assets to convey the physiological state of the user was embraced by all
8
An avatar is defined as “a perceptible digital representation whose behaviors reflect those executed, typically in real time, by a specific
human being” (Bailenson & Blascovich, 2004, p. 64). Yee and Bailenson (2007) showed that users that had more physically attractive
avatars displayed increased confidence and kept shorter personal distances in virtual interactions compared to users operating less
physically attractive avatars. The authors explained this behavioral effect of the avatar on the user with the Proteus effect.

Communications of the Association for Information Systems
Accepted Manuscript
participants, the design of affective visual assets needs to adequately identify and reflect the motivational
factors of a particular target audience in order to effectively address motivation.
5.3 Guideline 3: Goal-setting Support
Guideline 3 refers to how mobile biosensor measurements can be used to provide users with effective goal-
setting support. Goal-setting is important for behavior change as it facilitates: (1) capability, by showing
users how they can achieve their goals, (2) opportunity, by showing users when they can achieve goals,
and (3) motivation, by showing feedback on progress towards goals and boosting self-efficacy (Michie et
al., 2015). From the interviews with HBSes and HPs it emerged that mobile biosensor measurements are
particularly useful for the provision of goal-setting support because these measurements can make visible
the short-term changes in physiology in response to healthy and unhealthy behaviors (hence, increasing
the saliency of the pathways between physiology and health behavior) and, thereby, facilitate a clearer
understanding of progress towards health goals. For instance, heart rate and skin temperature biosensors
can bring to light short-term physiological changes in response to alcohol intake which are normally
imperceptible to the user (Gutierrez et al., 2015), enabling users to set and monitor short-term goals around
their drinking behavior. In a different example, Paalasmaa et al. (2012) developed a system where users
could specify and monitor goals for their quality of sleep (measured by respiration and heart rate) as well as
link this health information to lifestyle behaviors (e.g., alcohol intake, exercise). As such, using mobile
biosensors in goal-setting allows to break down health goals (e.g., reduced risk for cardiovascular disease)
into small and achievable tasks that can be measured and reinforced with biosensors (e.g., reduced resting
heart rate). Taken together, this (1) makes lifestyle behavior change more approachable, (2) allows for more
opportunities to reinforce or correct behaviors and, (3) attempts to mitigate the problem that the long-term
consequences of unhealthy lifestyle behaviors only become apparent to the user after years or even
decades by shortening the time between when a behavior is performed and when feedback is received.
In the interviews, two user interface concepts emerged recurrently when discussing goal-setting support:
gamification and serious games.
9
D1 emphasized the importance of gamification for designing mHealth
systems, stating “[…] every app we build will be based around the gamification principle in some capacity.”
Combining gamification with mobile biosensors can facilitate goal-setting in numerous ways. Firstly,
gamification can break down unwieldy long-term health goals (e.g., improving stress management) into
incrementally achievable goals (e.g., first stabilizing and then reducing stress levels as measured by skin
conductance), and reinforce this through reminders which encourage habitual use and facilitate automatic
motivation (e.g., reminders reinforcing paced-breathing when detecting high stress levels). HBS5 stated that
“[health goals] have to be achievable [and] […] something that you inherently want to achieve.” Secondly,
gamification can mitigate the problem associated with the long-term consequences of unhealthy lifestyle
behaviors that only become apparent to the user after years or even decades (e.g., onset of cardiovascular
disease and diabetes) by providing short-term incentives based on game-like elements and provide
feedback on how changes in behavior affect the user’s physiology (e.g., changes in physiological stress
levels). However, there was broad agreement among participants that the game-like elements are only a
means to an end, and the goals as such need to be intrinsically focused. For instance, U5 stated: “[…] goals
should be intrinsic motivators – things from within.” An important factor in this context are self-comparisons
and social comparisons. Self-comparison was strongly supported by all participants as it provides feedback
which focuses solely on the user themselves and visualizes individual progress. Further, self-comparison
can increase self-efficacy as it can demonstrate a user’s individual progress over time (e.g., reduction in
smoking and drinking occasions as detected from respiration, heart rate, and skin temperature; Gutierrez
et al., 2015; Saleheen et al., 2015). On the other hand, social comparison allows for comparisons of the
user’s physiological state with other similar users (e.g., leaderboards for managing physiological stress
levels). Social comparison can be a motivating factor, however the sense of competition associated with it
may also be detrimental for some users. For instance, U5 stated: “I'm the type of person that believes you
should only be assessed based on yourself […] assessing against others can be a demotivator.”
9
Gamification, defined as “the use of game design elements in non-game contexts” (Deterding et al., 2011, p. 10), leverages design
and interface elements from entertainment games (e.g., achievements, badges, leaderboards, etc.) which provide additional motivation
for behavioral changes typically hard to motivate, such as lifestyle behaviors (Schoech et al., 2013). However, as with every user
interface concept, it became apparent that the actual choice and design of the employed user interface concept strongly needs to
consider the target audience, as gamification and serious games are likely to be more or less effective for particular groups and depend
on how these concepts are ultimately designed. For instance, HBS6 and U3 suggested that gamification may be effective with younger
users, but not as effective with the elderly. Similarly, U4 stated that “a serious game could be more of a fad”.

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
Next, serious games, defined as “games used for purposes other than mere entertainment” (e.g., learning
or training; Wattanasoontorn et al., 2013, p. 231), were identified as another important concept for goal-
setting support based on mobile biosensors. By creating a tailored and engaging virtual environment,
serious games can provide users with the opportunity to improve their capabilities in a way that is motivating
and controlled. In the context of mobile biosensors and health behavior change, this can be used to provide
the user with an opportunity to increase their capability to engage in targeted health behaviors by directly
linking game elements to biosensor recordings. For instance, heart rate, respiration, and skin conductance
biosensors can be used in serious games to facilitate paced breathing exercises for relaxation (Dillon et al.,
2016; Xiong et al., 2013). HP1, HP2, and HBS2 supported the use of serious games, with HP1 stating: “[…]
why I like it is that you can in essence control the environment […] you’re controlling the input […] you’ve
got the sensor recording and you've got a known environment and you can put stressors in there”. In this
vein, serious games can utilize biosensors to adjust the game environment when aiming to achieve a
particular goal. For instance, Dillon et al. (2016) developed a mHealth serious game for stress reduction
where users’ skin conductance levels were used to progress the interface from a winter scene to a summer
scene. In sum, adjusting serious game elements based on physiological data can support goal-setting by
facilitating a tailored virtual environment which provides users with new opportunities to engage with health
goals and improve their capabilities to engage in targeted health behaviors. Because changes in physiology
are normally imperceptible to the user, users normally have limited opportunity for such training.
5.4 Guideline 4: External Support
Guideline 4 refers to facilitating a link between mobile biosensor recordings and external resources, which
we refer to as external support. Through discussion with participants we identify two primary types of
external support. The first type involves closing the gap between mHealth systems and external
stakeholders, particularly HPs. For instance, HP5 stated that mobile biosensor recordings could supplement
their understanding of a user’s physiological state as these measurements are more reflective of a user’s
typical day-to-day experience and are not influenced by phenomena such as white coat hypertension (i.e.,
elevated blood pressure in the presence of a HP; Martin & McGrath, 2014). By enabling the link to external
stakeholders, mHealth systems can address user capability by facilitating an information exchange
regarding the user’s physiological state between the user and their HPs, a source of data that is not
intuitively understood by most users and hence difficult for users to interpret by themselves. This notion is
similar to a design concept by Barakah & Ammad-uddin (2012), who proposed a virtual doctor platform
where HPs can remotely provide health advice to users based on their medical history and mobile biosensor
data (e.g., blood pressure, heart rate). Such an information exchange could be made more intuitive and
meaningful for users through the use of affective visual assets (e.g., a persuasive avatar of a doctor; see
Guideline 2). Additionally, this form of external support addresses opportunity because it allows for
characteristic patterns in physiological and contextual data to be detected and acted upon where they could
not before (e.g., abnormalities in blood pressure, detection of smoking and drinking occasions from
biosensors; Gutierrez et al., 2015; Saleheen et al., 2015). For instance, HP5 stated that if they had access
to physiological and contextual data from mHealth devices “[…] it will help me to determine what sort of
intervention the [user] needs, if at all. […] It may be on the wrist of the [user], but it's not available to me.”
Also, involving external stakeholders such as HPs into mHealth systems furthers the integration of mHealth
into clinical practice. Specifically, participants expressed that involving external stakeholders in the analysis
of their biosensor data can be motivating by building trust and credibility in mHealth systems. For example,
U5 supported this by saying: “[…] doctors are seen as [...] the ones that you can trust” and that “linking this
quite closely with the health professional would support the [mHealth system] and create credibility.”
The second type of external support identified involves utilizing physiological and contextual data to provide
the user with contextually relevant information from external resources that address psychological capability
through education and by creating awareness of opportunities in the physical environment. One of the
biggest benefits of mHealth is its ability to provide the user with just-in-time feedback (e.g., when detecting
physiological stress based on skin conductance and respiration biosensors) as mHealth devices are
immediately available in everyday situations (Danaher et al., 2015; Nahum-Shani et al., 2018). This ability
can be leveraged to point to external support material (e.g., video materials for paced breathing) and
opportunities in the physical environment (e.g., support hotlines, local meditation and healthy eating classes)
to reach users when this information is most relevant. In the interviews, HBSes and users emphasized that
links to relevant external support resources can support the information provided directly by the mHealth
system (e.g., when the user requires further assistance). For instance, U5 argued that this form of external
support may be useful as some people lack knowledge of resources available to them, stating: “a lot of

Communications of the Association for Information Systems
Accepted Manuscript
people don't know how to access places where resources are.” HBS3 added to this sentiment, stating: “It's
a really nice way to connect people up with those kind of services that they might not know exist otherwise.”
In this sense, users may be prompted to contact their HP when their physiological data (e.g., blood pressure,
heart rate) exhibit anomalies or provided with a list of relevant services (e.g., local HPs, healthy food options
based on location data) based on their location (GPS).
5.5 Guideline 5: Levels of Data Integration
In Guideline 5, we elaborate on four levels of data integration which attempt to connect the mobile biosensor
data collected by mHealth systems with external stakeholders. Considering the sensitivity of the health
information that can be extracted from physiological data, it is essential to adequately address any potential
security and privacy issues when integrating these four levels to ensure that all stakeholders can trust the
data and be assured of its provenance. The first level we refer to is Level A - Individual Feedback. This level
refers to the feedback the user sees through the mHealth interface (e.g., using skin conductance to show
changes in stress through a mirrored-self avatar). To facilitate this feedback, most current mHealth systems
primarily focus on the data stream between the user and the mHealth device, however this is not enough to
effectively address behavior change and neglects integration into clinical practice. In order to effectively
address the propositions, this necessitates that there is not only a data flow between the user and the
mHealth device, but that there are also other connections. Throughout the interviews we identified three
other levels that need to be considered.
The second level we identify is Level B - Data Aggregation and Analysis Service, which is a remote system
that analyzes mobile biosensor and contextual data from the mHealth device and sends it back in the form
of feedback (e.g., facilitating social comparison of health information extracted from physiological data). For
instance, ITP1 stated that the design needs to consider the computational complexity of analyzing and
aggregating biosensor data by “just [using] this mobile device to collect data and pass this data to a normal
computing system and use the most powerful technology to do the aggregation and analysis.” ITP1 further
explained that implementing a data aggregation and analysis service could allow mHealth systems to
perform more intelligently than current mHealth systems, as it could allow for applying more sophisticated
machine learning techniques, which is essential for suggesting targeted interventions that are tailored to the
individual situation of the user. For example, the inclusion of a data aggregation and analysis service has
the potential to facilitate statistical methods that not only allow for better personalization of interventions, but
also reduce follow-up time because of the ability to “track changes within the individual that predict outcomes
(e.g., heart arrhythmias) rather than waiting for the development of discrete, but rare, events (e.g., heart
attacks)” (Nilsen et al., 2012, p. 8). Additionally, as emphasized by PMs and HIPs, aggregated data (e.g.,
average physiological stress levels, severity of smoking and alcohol overconsumption as detected by
biosensors; Gutierrez et al., 2015; Saleheen et al., 2015) could serve as an important source of information
for policy creation and resource allocation (e.g., testing effectiveness of health promotion interventions for
particular user groups). For instance, PM1 elaborated that “if you can actually improve the clarity of their
vision as to the effectiveness of interventions, then the policy decision becomes much clearer to [PMs] in
terms of where should they be investing public dollars.” In sum, the data aggregation and analysis service
is not only important for providing more effective feedback to users. The health information that can be
extracted from mobile biosensors also provides insights about different user groups which could aid crafting
health promotion policy, and indirectly improve behavior change (e.g., better access to support for smoking
cessation and related resources in the environment).
The third level we describe is Level C - Integration with Health Practitioner Information System which is a
data link that allows HPs to monitor users’ physiology and send feedback back to the mHealth device.
Enabling this data link between users and HPs addresses an individual user’s (1) capability by facilitating
individualized feedback from HPs which can educate a user about their physiological state in relation to
their health goals (e.g., what their resting heart rate means; hence increasing psychological capability to
engage in health behavior), (2) opportunity, by enabling abnormal measurements to be detected and acted
upon proactively (e.g., by suggesting specific health behaviors based on the analyzed biosensor data), and
(3) motivation, by improving the integration of mHealth systems into clinical practice, thereby increasing
trust and credibility. Users generally supported the data link to a health practitioner IS. For instance, U2
particularly emphasized the usefulness of the remote aspect of the system, stating: “It would be amazing
and very helpful to doctors to not have to sit there and get half an hour's worth of conversation out of
somebody as to how they've been. They can just pretty much download that straight onto their system.” On
the other hand, U3 argued that some users may not be open to involving HPs and disliked the idea of having
their data visible to them. HPs strongly supported the data link as it further integrates mHealth into clinical

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
practice, allows for in-context measurements, and for users to be monitored remotely. HP5, for instance,
stated “If I had a [person] who had extra heart beats in 24 hours and I’m at a conference in Hong Kong,
[mHealth systems] should be linked to the [Health Practitioner IS] in some way so that it's captured and
when I see the patient again […] I can look at it.”
Finally, the fourth level we describe is Level D - Integration with Health Insurance Provider Information
System is a data link that gives HIPs access to user data. Regarding the potential role of HIPs, HIP1 stated:
“[…] health and healthier outcomes isn’t just about medical advice, but about lifestyle choices […] we can
potentially help people make better decisions about their health through getting information to them sooner,
and helping connect people who can help each other out.” However, currently the role of HIPs in
preventative health is limited: “[…] we get no data out of primary care until a person has a declared
significant medical condition that’s had to result in a hospitalization” (HP2). Utilizing mobile biosensors has
the potential to provide important data for preventative health such as utilization of health care resources,
factors leading up to health events, and risk status. HIP may use this data to provide individual users with
direct incentives (e.g., bonus points, financial support) to motivate them and create new opportunities for
health behavior change (e.g., support gym membership, discounts for healthy foods).
10
HIP1 strongly
supported this function, stating: “definitely wearables, and definitely phones, and definitely technology would
give us that data.” Overall, the involvement of HIPs in mHealth systems received a mixed response in the
interviews, with several participants expressing strong reservations. On the one hand, HPs were cautious
about the involvement of HIPs, with HP5 stating: “Whether users would trust you to transmit that data to the
insurance provider is a big issue. They might use it for their own commercial purpose […] I don’t know if we
are ready to involve them yet.” Conversely, users generally supported the involvement of HIPs, particularly
in the case where they are rewarded to increase motivation for health behavior change, with U4 stating: “I
like the idea that you can use the application and be rewarded for that by your insurer.” PM1 argued that
the most important factor is that there is mutual benefit for users to give up access to physiological data and
that this needs to be communicated through a narrative of why the data is being collected. In sum, Level D
could motivate users on an individual level through targeted incentives and create new opportunities for
them through improved health promotion at the population level. However, particularly against the backdrop
of the sensitivity of the health information that can be extracted from mobile biosensors, it is critical that the
involvement of HIPs is handled carefully to ensure that the needs of the user are kept paramount.
5.6 Guideline 6: Stakeholder Involvement
Guideline 6 refers to the involvement of stakeholders in all phases of design. Despite, and more explicitly
because of, the promising potential of utilizing mobile biosensor measurements in mHealth systems, it is
critical for system designers to acknowledge that the design of such systems for behavior change is a
challenging task, involving difficulties associated with the long-term consequences of unhealthy lifestyle,
imperceptibility of short-term changes in physiology, users’ understanding of physiological data, multiple
direct and indirect stakeholders with diverse background and interests, sensitivity of health information
extracted from biosensors, and a complex landscape of remote systems. Addressing these challenges, and
ensuring that the artifact design adequately reflects the intricacies of increasing individual users’ capability,
opportunity, and motivation to engage in targeted behaviors, requires the involvement of all relevant
stakeholders in all phases of design (Burke et al., 2015; Lobelo et al., 2016). This call for a co-design
approach, where stakeholders are actively involved from the early stages of design through to adoption,
was also consistently emphasized by participants from all stakeholder categories. As stated by HIP1, this
involvement is essential to make sure that the “technology seamlessly entwines its way in our lives”.
Co-design, also known as participatory design, is an approach to “facilitate users, researchers, designers
and others […] to cooperate creatively, so that they can jointly explore and envision ideas, make and discuss
sketches, and tinker with mock-ups or prototypes” (Steen, 2011, p. 52). One critical aspect of co-design is
that it involves the “user as a partner” rather than designing for the user as a subject (Sanders & Stappers,
2008, p. 5). Supporting this notion, D1 stated that “if you don't have key decision-makers involved in that
initial stage, you're doomed.” A second critical aspect of co-design is the shift in focus away from
technological aims to an emphasis on collaborative activities in contextual and generative design phases
10
For instance, the Australian health insurer Medibank engaged in a joint incentive scheme with Coles, a major chain of supermarkets.
In this scheme, Medibank customers receive additional rewards that can be redeemed in their grocery shopping at Coles if their activity
levels (as measured by a wearable fitness device) reach a certain goal and/or if they buy healthy foods (e.g., fresh fruit and vegetables)
at Coles (https://flybuys.medibank.com.au). A similar incentive scheme was introduced by the South African health insurance company
Discovery, where insurance clients were provided with discounts for healthy food purchases at certain supermarkets (An et al., 2013).

Communications of the Association for Information Systems
Accepted Manuscript
(Sanders & Stappers, 2014). As described by D1, “the development is completely… it's irrelevant.” Thereby,
particular emphasis is put on contextual research activities (e.g., using cultural probes and storytelling;
Gaver et al., 2004, Mitchell et al., 2015) which, as described by D2, help “to understand the situation that
[…] the problem exists in, and what other various kind of inputs or context surround that” (e.g., individual
motivators for health behavior change, contextual factors that may affect biosensor recordings).
6 Discussion and Conclusion
6.1 General Discussion
While there has been extensive research on how biosensors can provide insightful information about a
person’s health and lifestyle behaviors, researchers have only recently begun to use the information that
can be extracted from biosensors in the design of mHealth systems to deliver behavior change interventions
(Free et al., 2013). As such, there are no established guidelines for the design of such systems and scholars
have raised the concern that mHealth systems for behavior change are often not guided by a BCI framework
(Davey et al., 2014; Hingle & Patrick, 2016). Against this backdrop, the knowledge contribution of our study
constitutes an improvement (Gregor & Hevner, 2013), because it extends the knowledge base for a known
problem with low solution maturity (i.e., mHealth systems design) and high application domain maturity (i.e.,
health promotion). In particular, our research draws on the deep understanding of the societal problem of
health behavior change in the BCI literature and develops prescriptive knowledge for the design of
biosensors-enabled mHealth systems. In doing so, this research makes two core contributions.
The first contribution is the integrative theoretical framework which formulates the theoretical pathways for
how biosensor-enabled mHealth systems can bring about health behavior change. The framework
contributes to the prescriptive knowledge base by (1) providing researchers and practitioners with a shared
frame of reference for the implementation of a feedback loop between the user’s physiology and their
perception and (2) enabling system designers to systematically map out how the elements of their mHealth
interface can target individual components of behavior and the types of interventions through which this can
be achieved. Following a deductive theorizing approach, the propositions are grounded in the well-
established BCI framework by Michie et al. (2011), emphasizing that mHealth systems design needs to
simultaneously consider users’ capability, opportunity, and motivation. In particular, as elaborated in the
related work section, previous research argued that the design of mHealth systems for health behavior
change should be guided by a theoretical framework rooted in the BCI literature and that systems are more
effective the more behavior change techniques they implement (Garnett et al., 2016; Hingle & Patrick, 2016;
Vandelanotte et al., 2016). Many existing studies focus primarily on increasing users’ psychological
capability by providing them with additional information or providing that same information in a more intuitive
way (Michie et al., 2011). However, putting the focus only on increasing capability would fail to address
motivation, which is vitally important for lifestyle behavior change (Vandelanotte et al., 2016). Hence, the
framework allows more informed decisions as it enables system designers to consider a range of different
potential pathways and BCI categories for facilitating behavior change through the mHealth interface.
The second contribution is the set of six general guidelines for designing mobile biosensor-based BCIs,
which were developed in an inductive theorizing approach based on interviews with key mHealth
stakeholders. Constructing a mHealth system involves a multitude of design choices with links to a range
of different stakeholders and remote systems. In this sense, the guidelines contribute to the prescriptive
knowledge base by providing system designers with practical design considerations that take into account
multiple stakeholder perspectives. Hence, the guidelines can serve as groundwork for the development of
new solution artifacts for the delivery of technology-mediated interventions that support users in modifying
their behavior. Importantly, the guidelines are not constrained to a particular type of biosensor (e.g., heart
rate, skin conductance) or lifestyle behavior (e.g., nutrition, physical activity). Hence, while individual studies
usually focus on an individual solution artifact and a particular type of intervention (e.g., biofeedback training
for stress management; Xiong et al., 2013), our guidelines provide general considerations for a bigger class
of problems which system designers can then refine for their individual solution artifact. For instance, while
several studies have used avatars to convey the health information extracted from biosensors (e.g., Murray
et al., 2013), our study abstracts from the particular type of visual asset and instead emphasizes the need
for the asset to be intuitive and meaningful to the user. Hence, instead of directly adopting a solution that
worked in a different context or for a different cohort, this allows system designers to make explicit the actual
purpose of individual design choices (e.g., meaningfulness of biosensor data) and put the pursuit of this

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
purpose at the heart of the design process. Figure 5 provides a graphical summary of the direct connections
between the six design guidelines and the theoretical framework, which are discussed in the following.
Figure 5. Mapping of the Design Guidelines to the Integrative Theoretical Framework
Measurement. By bringing to light important considerations regarding the measurement of physiological
data, Guideline 1 directly relates to the creation of a feedback loop between the user’s physiology and their
perception (P1). Further, as shown in Figure 5, Guideline 1 also has important links with other design
considerations. Firstly, in order to close the feedback loop, the collected physiological data need to be
conveyed to the user via the mHealth interface (Guidelines 2–4). Secondly, there is an opportunity to use
mobile biosensors not only for individual feedback, but also to facilitate the provision of external support
(Guideline 4) by leveraging data integration with remote systems (Guideline 5). Despite the increased
accessibility of sensor technology for consumers, the link to health care providers is currently underutilized.
HP5 emphasized this missing link stating “[…] all these apps can monitor [heart rate] continuously even
now, but it may not be available to your medical practitioner […] we need to close the loop.” Similarly, there
have been calls for better integration of mHealth systems into clinical practice (Clifton et al., 2013; Lobelo
et al., 2016) as there is the promise of being able to remotely monitor users and allow for early detection of
medical complications (Dobkin & Dorsch, 2011). Hence, by creating a link between a user’s physiological
measurements and remote systems, the feedback loop between a person’s health behavior and the
resulting physiological changes may be assisted by the provision of external support. With the support of
external resources, this may help the user to better understand how their lifestyle behavior affects their
physiology, and subsequently, their health.
Interface Design. The design of the mHealth interface is of critical importance for addressing a user’s
capability, motivation, and opportunity, as it mediates the information flow to the user. Building on the
foundations of the established feedback loop, Guidelines 2–4 directly relate to how propositions P2–P4 can
be implemented in BCIs in the mHealth interface (see Figure 5), which needs to convey the information
extracted from mobile biosensors for the user in an intuitive way (P2), create an emotional bond that makes
it meaningful for the user to engage in a targeted behavior (P3), and create opportunities for the user to do
so (P4). Importantly, these three guidelines are interrelated. For instance, by employing the concept of a
persuasive avatar, using affective visual assets (Guideline 2) can be an effective way for facilitating external
support (Guideline 4). One way of accomplishing this is by using persuasive avatars which receive input
from subject matter experts. Further, by leveraging the stereotype of a HP, persuasive avatars may be able
to activate similar responses from users (e.g., increased trust and credibility) through priming mechanisms
(i.e., situational cues and social stereotypes that activate concepts and behaviors; Bargh & Chartrand,
2000).
11
Similarly, affective visual assets (Guideline 2) can incorporate goal-setting (Guideline 3) in a way
that is engaging and motivating. For instance, showing a future projection of a user’s mirrored-self avatar
11
Overall, there was consensus that persuasive avatar may be particularly beneficial in the elderly cohort as there is the potential to
link persuasive avatars to health care providers which could potentially increase trust, credibility, and adherence of the feedback. In
the context of disease management, for instance, Javor et al. (2016) showed that Parkinson’s disease patients show higher initial trust
in avatar faces than in human faces. Therefore, a persuasive avatar of a health practitioner may elicit more trust in Parkinson’s disease
patients than a real health practitioner. However, as expressed by HBS4 and U6, there were also concerns about the use of persuasive
avatars, with U6 stating “I know a lot of people have issues with authority so it could be a downfall.” Therefore, the success of persuasive
avatars may depend on the perception of the user on authority figures.
Feedback Loop
Capability
(physical and psychological)
Motivation
(reflective and automatic)
Opportunity
(physical and social)
Health Behavior
(e.g., nutrition)
Mobile Biosensor
Measurements
(e.g., heart rate
measurements)
Michie et al. (2011) COM-B System
Physiology
(e.g., heart activity)
DESIGN OF THE MHEALTH SYSTEM USER PERCEPTION, BEHAVIOR, AND PHYSIOLOGY
Education
and Training
Coercion,
Incentivisation,
Modelling, and
Persuasion
Environmental
Restructuring,
Enablement, and
Restriction
mHealth Interface
User Device
Interface Design
Measurement
Stakeholders &
Remote Systems
GUIDELINE 2
Affective Visual
Assets
GUIDELINE 3
Goal Setting
Support
GUIDELINE 4
External
Support
GUIDELINE 1
Recordings
GUIDELINE 5
Levels of Data
Integration
GUIDELINE 6
Stakeholder
Involvement
Mobile Biosensor
DESIGN GUIDELINES
P1
P4
P2
P3
Intervention
Intervention
Intervention
Integration with
Remote Systems

Communications of the Association for Information Systems
Accepted Manuscript
based on their physiology could increase self-efficacy as it provides an incentive for healthy behaviors and
an opportunity to change their behavior before their projection becomes real (Rho et al., 2017). Users
strongly supported the use of projection, with U5 for instance, stating: “I would love something like that
[future self] because you actually know what the end results are going to be and you can see if you need to
tweak or do something for your end goal. It's good to set those goals and be able to visually see what you're
aiming for.” HBS6 argued that using a mirrored-self avatar for projecting the future self may “[…] slowly build
some sort of capability in [users] to make that sort of prediction.” However, HBS2 added that it is important
that projections that are directly associated with goals should be short-term oriented, stating “[…] you don't
want to be projecting something in a week, because a week is too long, especially if they're anticipating or
struggling now, that's far too long.” Therefore, goal-setting can be incorporated into affective visual assets
in order to better address capability, opportunity, and motivation.
Importantly, goal-setting also has important connections with external support (Guideline 4), as the goal-
setting process around physiological data involve consultation with external stakeholders (e.g., HPs). To
this end, HP3 and HP5 suggested that mHealth should involve a prescription-like process where the user
can set goals with a HP and refine these goals over time through routine checkups. By doing so, HP5
argued, mHealth would be better suited to address long term behavior change, stating: “It should be a long
term process. If you're giving somebody blood pressure medication, you're not in it for a short term. You're
for the rest of the [person’s] life. It needs to be similar.” Further, users brought up several other benefits of
involving HPs in the goal-setting process such as increased trust in the feedback and increased compliance
with advice. For example, U4 stated in regards to accountability and long-term use: “It puts a face behind
what the app is, so then they know if someone of that stature, that profession, has taken the time out to be
involved in this, then it's something they would consider not only using, but continue using.” Therefore, given
the difficulty to interpret physiological data in the context of the user and the long-term process and
implications of health behavior change, the goal-setting process (Guideline 3) should involve external
stakeholders (Guideline 4) in order to tailor user goals, increase trust, and compliance with advice.
Stakeholders & Remote Systems. Through the integration of remote systems (Guideline 5) it is not only
possible for HPs to monitor the user’s physiological and contextual data remotely, but also to send feedback
back to the user, for example, in the event that the user has a ‘funny turn’ and the doctor wants to arrange
an appointment with the user. Furthermore, what emerged clearly from the results is that many existing
mHealth systems focus heavily on the data link between the user and the user device (Level A), but neglect
the integration of the user device with remote systems (Levels B, C, and D) (Free et al., 2013). For instance,
Winters et al. (2017, p. 119) emphasizes this missing link by stating that there is a “lack of informed
engagement with health-sector stakeholders and key decision-makers on mHealth innovation, as well as a
distinct lack of integration with the formal health system.” The design guidelines presented in this study aim
to contribute towards a better understanding of the relationship between mHealth systems and remote
systems which can help to bridge the gap between mHealth and clinical practice. Creating this data link
between users and other stakeholders, and considering the privacy and security issues that this link entails,
can improve behavior change outcomes by enabling opportunities that would not exist otherwise (e.g.,
detection of abnormal measurements that a HP can act on), as well as improving the credibility of mHealth
systems by involving HPs which could motivate usage intentions and the sustainability of mHealth
interventions (external support; Guideline 4). After all, most people see their HP as a partner in their health
and trust them (Roy Morgan, 2017). In this vein, through its connections with the other guidelines, Guideline
5 has indirect links to the design of the mHealth interface and the mobile biosensor measurements.
Finally, previous research has emphasized the early inclusion of stakeholders in the design process (Burke
et al., 2015; Petersen et al., 2015). By involving stakeholders in all phases of the design process, Guideline
6 can be seen as a facilitator for Guidelines 1-5 in the design of BCIs. Hence, there are important indirect
links between Guideline 6 and the design of the mHealth system. For instance, co-design can facilitate the
creation of visual assets that are intuitive for the user in conveying health information derived from mobile
biosensors and meaningful by addressing the motivational factors of a particular target cohort (e.g., adapting
a self-avatar or a persuasive avatar; Guideline 2). By involving stakeholders early in the design process,
these motivational factors can already be determined in the contextual phase. For example, as described
by D1, by engaging users in co-design workshops, users can collaboratively explore how meaningful the
selection of a particular user interface concept is for their individual motivational pathways, allowing system
designers to “[…] find out what that selection means to [the user], as opposed to what you subjectively
interpret that information as.” Similarly, in order to address the problems associated with the long-term
consequences of poor lifestyle behavior, co-design can support the creation of compelling goal-setting
support (e.g., specification of self-comparison and/or social comparison; Guideline 3). Further, when

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
devising the information streams between the user device and remote systems, co-design can contribute to
ensuring that the mHealth system meets agreed-upon ethical standards of privacy and data protection when
handling information extracted from biosensor recordings and other data sources (levels of data integration;
Guideline 5). Thereby, it is vital to involve not only the user, but also other mHealth stakeholders in the
design process in order to address the problem in a way that meets the needs of the entire mHealth system.
6.2 Limitations
Despite the findings of this study, there are limitations that need to be addressed. Firstly, our study did not
engage in the design of an actual solution artifact, but instead explored general considerations for the design
of biosensor-enabled mHealth systems at a conceptual level. Hence, the design, implementation, and
evaluation of a specific mHealth system will require tailoring to the individual problem domain and take into
account the individual characteristics of the user cohort (e.g., differences in motivational factors for different
age groups). For instance, Guideline 2 emphasizes that mobile biosensor recordings are difficult to
understand and interpret for the user, and hence require the use of affective visual assets to convey the
embedded health information in an intuitive and meaningful way. However, we did not investigate which
particular visual assets are most effective for a particular user cohort (e.g., young adults) and health
behavior (e.g., healthy nutrition). Nevertheless, considering multiple stakeholder perspectives, the design
guidelines provide system designers with a set of general considerations to take into account when
implementing the theoretical pathways for bringing about health behavior change. Further, our framework
can provide researchers and practitioners with a shared frame of reference to map their approach to and
consider which other theoretical pathways are worth pursuing in their mHealth interface. For instance, using
the framework to guide the design may enable system designers who are focusing on implementing a just-
in-time intervention to address automatic motivation (e.g., by using mobile biosensors to detect unhealthy
behaviors) to complementarily pursue other BCIs (e.g., education, training) to address capability.
Secondly, while we developed design guidelines that are contextualized to the case of biosensor-enabled
mHealth systems for behavior change, the guidelines may at least partially also apply to the design of other
types of complex end-user information systems. For instance, affective visual assets (Guideline 2) and goal-
setting support (Guideline 3) are also important design considerations for end-user systems in education
(mLearning, Garcia-Cabot et al., 2015). Similarly, it is straightforward that external support (Guideline 4)
and data integration with remote systems (Guideline 5) are important for the design of mHealth systems in
disease management (e.g., managing diabetes; Kitsiou et al., 2017).
12
However, and notwithstanding that
some or even all of the design guidelines may also apply to other areas, it is important to note that we have
developed the guidelines in a process of inductive reasoning based on exploratory interviews for the specific
context of mobile biosensors and health promotion. In all stages of this process, the exploratory nature of
this research was explicit to interview participants and they were actively encouraged to raise any further
design considerations that they felt important to be included. Further, our study elaborates why and how the
developed design guidelines are important for the case of biosensor-enabled mHealth systems for health
behavior change (e.g., difficulty to interpret physiological data, imperceptibility of short-term changes in
physiology, sensitivity of health information extracted from biosensors).
Nevertheless, we acknowledge that not all guidelines may be applicable at all times and they may need to
be adjusted based on the requirements of the system in question. Hence, even though we believe that our
guidelines constitute a useful starting point for the development of biosensor-enabled mHealth systems, we
advise against the mandatory or rote use of the design guidelines. Also, evaluating and refining the
proposed guidelines to a particular problem domain (e.g., improved nutrition, reduced alcohol intake) and
user cohort (e.g., young or middle-aged adults) will require dedicated design science projects, involving
their own data collection for testing the theoretical propositions and for evaluating the effectiveness of their
solution artifact in bringing about behavior change.
6.3 Future Research
Findings from this study can be extended to future research in several areas. To evaluate the design
guidelines presented in this study, we suggest the development of a methodological framework for co-
12
In this context, it is important to note that the design of complex end-user systems in other contexts may additionally also require
other design considerations. Notably, it is critical for the design of mHealth systems for disease management (e.g., managing diabetes)
to take into account the specific characteristics and treatment requirements of the disease. Because disease management is outside
the scope of this work, our interviews did not take such factors into account.

Communications of the Association for Information Systems
Accepted Manuscript
design in a mHealth context, as currently to our knowledge, no such methodological framework exists that
captures the intricacies of mHealth systems, such as the necessity to include clinical trials at different stages
in the design process, particularly for integrating the mHealth system as a facilitator for health behavior
change into clinical practice. From here, a co-design study that focuses on a specific cohort can be
performed from which prototype applications can be created to evaluate and refine the design guidelines.
On another note, currently many mHealth studies have only established the efficacy of intervention effects
over a short time (Vandelanotte et al., 2016). Therefore, further studies need to be conducted in order to
better understand what BCIs and techniques are effective for sustainable behavior change in a mHealth
context over the long-term, addressing the problem that rigorous testing for the efficacy of mHealth systems
is being outpaced by adoption (Burke et al., 2015).
7 Concluding Note
Mobile biosensors hold great potential for the development of mHealth systems that support users in
mitigating their risk of disease and promoting positive health outcomes through health behavior change. We
hope this study can serve researchers and practitioners as a reference guide for the design of biosensor-
enabled mHealth systems by providing an overview of stakeholder perspectives that need to be considered,
conceptualizing the theoretical pathways for how the components of behavior can be addressed through
different BCIs in the mHealth interface, and providing general guidelines for the development of such
systems.
Acknowledgements
This research was supported by an Australian Government Research Training Program (RTP) Scholarship.

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
References
Acharya, U. R., Joseph, K. P., Kannathal, N., Lim, C. M., & Suri, J. S. (2006). Heart rate variability: A review.
Medical and Biological Engineering and Computing, 44(12), 1031–1051.
Adam, M. T. P., Gimpel, H., Maedche, A., & Riedl, R. (2017). Design blueprint for stress-sensitive adaptive
enterprise systems. Business & Information Systems Engineering, 59(4), 277–291.
Al Osman, H., Dong, H., & El Saddik, A. (2016). Ubiquitous biofeedback serious game for stress
management. IEEE Access, 4, 1274–1286.
An, R., Patel, D., Segal, D., & Sturm, R. (2013). Eating better for less: a national discount program for
healthy food purchases in South Africa. American Journal Of Health Behavior, 37(1), 56–61.
Arnitz, T., Hütter, A., & Riedl, R. (2017). Mutual trust between the chief information officer and chief
executive officer: Insights from an exploratory interview study. Journal of Information Technology
Theory and Application, 18(3), 66–99.
Astor, P. J., Adam, M. T. P., Jerčić, P., Schaaff, K., & Weinhardt, C. (2013). Integrating biosignals into
information systems: A NeuroIS tool for improving emotion regulation. Journal of Management
Information Systems, 30(3), 247–277.
Bailenson, J. N., & Blascovich, J. (2004). Avatars. In W. S. Bainbridge (Ed.), Encyclopedia of Human-
Computer Interaction (pp. 64–68). Great Barrington, MA: Burkshire Publishing Group.
Ballinger, B., Hsieh, J., Singh, A., Sohoni, N., Wang, J., Tison, G. H., … Pletcher, M. J. (2018). DeepHeart:
Semi-supervised sequence learning for cardiovascular risk prediction. In AAAI 2018 Proceedings (pp.
1–9). New Orleans, USA.
Bandura, A. (2010). Self-Efficacy. In I. B. Weiner & W. E. Craighead (Eds.), The Corsini Encyclopedia of
Psychology. Hoboken, USA: John Wiley & Sons.
Barakah, D. M., & Ammad-uddin, M. (2012). A survey of challenges and applications of wireless body area
network (WBAN) and role of a virtual Doctor server in existing architecture. In ISMS 2018 Proceedings
(pp. 214–219). Kota Kinabalu, Malaysia.
Bardram, J. E., Frost, M., Szántó, K., Faurholt-Jepsen, M., Vinberg, M., & Kessing, L. V. (2013). Designing
mobile health technology for bipolar disorder: a field trial of the monarca system. In CHI 2013
Proceedings (pp. 2627–2636). NY, USA.
Bargh, J. A., & Chartrand, T. L. (2000). The mind in the middle: a practical guide to priming and automaticity
research. In H. T. Reis & C. M. Judd (Eds.), Handbook of Research Methods in Social and Personality
Psychology (pp. 253–285). New York: Cambridge University Press.
Baumeister, R., & Leary, M. R. (1997). Writing narrative literature reviews. Review of General Psychology,
1(3), 311–320.
Behm-Morawitz, E. (2013). Mirrored selves: The influence of self-presence in a virtual world on health,
appearance, and well-being. Computers in Human Behavior, 29(1), 119–128.
Ben-Zeev, D., Schueller, S. M., Begale, M., Duffecy, J., Kane, J. M., & Mohr, D. C. (2015). Strategies for
mHealth research: Lessons from 3 mobile intervention studies. Administration and Policy in Mental
Health and Mental Health Services Research, 42(2), 157–167.
Bloom, D. E., Cafiero, E. T., Jané-Llopis, E., Abrahams-Gessel, S., Bloom, L. R., Fathima, S., … Weinstein,
C. (2011). The global economic burden of non-communicable diseases. Geneva: World Economic
Forum.
Bonner, C., Jansen, J., Mckinn, S., Irwig, L., Doust, J., Glasziou, P., … McCaffery, K. (2013). General
practitioners’ use of different cardiovascular risk assessment strategies: a qualitative study. Medical
Journal of Australia, 199(7), 485–489.
Burke, L. E., Ma, J., Azar, K. M. J., Bennett, G. G., Peterson, E. D., Zheng, Y., … Quinn, C. C. (2015).
Current science on consumer use of mobile health for cardiovascular disease prevention: A scientific
statement from the American Heart Association. Circulation, 132(12), 1157–1213.

Communications of the Association for Information Systems
Accepted Manuscript
Buttussi, F., Chittaro, L., & Nadalutti, D. (2006). Bringing mobile guides and fitness activities together: a
solution based on an embodied virtual trainer. In MobileHCI 2006 Proceedings (pp. 29–36). NY, USA.
Carter, J. B., Banister, E. W., & Blaber, A. P. (2003). Effect of endurance exercise on autonomic control of
heart rate. Sports Medicine, 33(1), 33–46.
Carver, C. S., & Scheier, M. F. (1982). Control theory: a useful conceptual framework for personality-social,
clinical, and health psychology. Psychological Bulletin, 92(1), 111–135.
Clifton, L., Clifton, D. A., Pimentel, M. A. F., Watkinson, P. J., & Tarassenko, L. (2013). Gaussian processes
for personalized e-health monitoring with wearable sensors. IEEE Transactions on Biomedical
Engineering, 60(1), 193–197.
Conroy, D. E., Yang, C.-H., & Maher, J. P. (2014). Behavior change techniques in top-ranked mobile apps
for physical activity. American Journal of Preventive Medicine, 46(6), 649–652.
Crane, D., Garnett, C., Brown, J., West, R., & Michie, S. (2015). Behavior change techniques in popular
alcohol reduction apps: Content analysis. Journal of Medical Internet Research, 17(5), 1–12.
Curmi, F., Ferrario, M. A., Southern, J., & Whittle, J. (2013). HeartLink: Open broadcast of live biometric
data to social networks. In SIGCHI 2013 Proceedings (pp. 1749–1758). Paris, France.
Danaher, B. G., Brendryen, H., Seeley, J. R., Tyler, M. S., & Woolley, T. (2015). From black box to toolbox:
Outlining device functionality, engagement activities, and the pervasive information architecture of
mHealth interventions. Internet Interventions, 2(1), 91–101.
Davey, S., Davey, A., & Singh, J. V. (2014). Mobile-health approach: A critical look on its capacity to
augment health system of developing countries. Indian Journal of Community Medicine, 39(3), 178–
182.
Deterding, S., Dixon, D., Khaled, R., & Nacke, L. (2011). From game design elements to gamefulness:
Defining “gamification.” In MindTrek 2011 Proceedings (pp. 9–15). Tampere, Finland.
Dillon, A., Kelly, M., Robertson, I. H., & Robertson, D. A. (2016). Smartphone applications utilizing
biofeedback can aid stress reduction. Frontiers in Psychology, 7, 1–7.
Direito, A., Carraça, E., Rawstorn, J., Whittaker, R., & Maddison, R. (2017). mHealth technologies to
influence physical activity and sedentary behaviors: Behavior change techniques, systematic review
and meta-analysis of randomized controlled trials. Annals of Behavioral Medicine, 51(2), 226–239.
Dobkin, B. H., & Dorsch, A. (2011). The promise of mHealth: daily activity monitoring and outcome
assessments by wearable sensors. Neurorehabilitation and Neural Repair, 25(9), 788–798.
Donetto, S., Pierri, P., Tsianakas, V., & Robert, G. (2015). Experience-based co-design and healthcare
improvement: Realizing participatory design in the public sector. The Design Journal, 18(2), 227–
248.
Easterby-Smith, M., Thorpe, R., & Lowe, A. (2002). Management Research: An Introduction (2nd ed.).
London, UK: Sage.
Eckman, M., Gorski, I., & Mehta, K. (2016). Leveraging design thinking to build sustainable mobile health
systems. Journal of Medical Engineering & Technology, 40(7–8), 422–430.
Facchinetti, T., Fernando, A., & Quoi, R. (2012). “This process is just beginning”: Connecting mobile medical
devices. Biomedical Instrumentation & Technology, 46(2), 19–25.
Feijs, L., Kierkels, J., van Schijndel, N. H., & van Lieshout, M. (2013). Design for relaxation during milk
expression using biofeedback. In DUXU 2013 Proceedings (pp. 494–503). Las Vegas, USA.
Fogg, B. J. (2009). A behavior model for persuasive design. In PERSUASIVE 2009 Proceedings. Claremont,
USA.
Fox, K., Borer, J. S., Camm, A. J., Danchin, N., Ferrari, R., Lopez Sendon, J. L., … Tendera, M. (2007).
Resting heart rate in cardiovascular disease. Journal of the American College of Cardiology, 50(9),
823–830.

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
Free, C., Phillips, G., Galli, L., Watson, L., Felix, L., Edwards, P., … Haines, A. (2013). The effectiveness of
mobile-health technology-based health behaviour change or disease management interventions for
health care consumers: A systematic review. PLoS Medicine, 10(1), 1–45.
Fujita, H., Hakura, J., & Kurematsu, M. (2010). Virtual doctor system (VDS): medical decision reasoning
based on physical and mental ontologies. In IEA-AIE 2010 Proceedings (pp. 419–428). Cordoba,
Spain.
Garcia-Cabot, A., De-Marcos, L., & Garcia-Lopez, E. (2015). An empirical study on m-learning adaptation:
Learning performance and learning contexts. Computers & Education, 82, 450–459.
Garnett, C., Crane, D., Michie, S., West, R., & Brown, J. (2016). Evaluating the effectiveness of a
smartphone app to reduce excessive alcohol consumption: Protocol for a factorial randomised control
trial. BMC Public Health, 16(536), 1–10.
Gaver, W. W., Boucher, A., Pennington, S., & Walker, B. (2004). Cultural probes and the value of
uncertainty. Interactions, 11(5), 53–56.
Glynn, L. G., Hayes, P. S., Casey, M., Glynn, F., Alvarez-Iglesias, A., Newell, J., … Murphy, A. W. (2014).
Effectiveness of a smartphone application to promote physical activity in primary care: The SMART
MOVE randomised controlled trial. British Journal of General Practice, 64(624), e384–e391.
Gregor, S., & Hevner, A. R. (2013). Positioning and presenting design science research for maximum
impact. MIS Quarterly, 37(2), 337–355.
Gregory, R. W., & Muntermann, J. (2011). Theorizing in design science research: Inductive versus deductive
approaches. In ICIS 2011 Proceedings. Shanghai, China.
Gutierrez, M. A., Fast, M. L., Ngu, A. H., & Gao, B. J. (2015). Real-time prediction of blood alcohol content
using smartwatch sensor data. In ICSH 2015 Proceedings (pp. 175–186). Phoenix, AZ, USA.
Hamman, R. F., Wing, R. R., Edelstein, S. L., Lachin, J. M., Bray, G. A., Delahanty, L., … Wylie-Rosett, J.
(2006). Effect of weight loss with lifestyle intervention on risk of diabetes. Diabetes Care, 29(9), 2102–
2107.
Hanus, M. D., & Fox, J. (2015). Persuasive avatars: The effects of customizing a virtual salesperson׳s
appearance on brand liking and purchase intentions. International Journal of Human-Computer
Studies, 84, 33–40.
Hingle, M., & Patrick, H. (2016). There are thousands of apps for that: Navigating mobile technology for
nutrition education and behavior. Journal of Nutrition Education and Behavior, 48(3), 213–218.
Javor, A., Ransmayr, G., Struhal, W., & Riedl, R. (2016). Parkinson patients’ initial trust in avatars: Theory
and evidence. PLoS ONE, 11(11), 1–21.
Kitsiou, S., Paré, G., Jaana, M., & Gerber, B. (2017). Effectiveness of mHealth interventions for patients
with diabetes: An overview of systematic reviews. PLOS ONE, 12(3), 1–16.
Kromand, D. (2007). Avatar categorization. In DiGRA 2007 Proceedings (pp. 400–406). Tokyo, Japan.
Kumar, S., Nilsen, W. J., Abernethy, A., Atienza, A., Patrick, K., Pavel, M., … Swendeman, D. (2013). Mobile
health technology evaluation: The mHealth evidence workshop. American Journal of Preventive
Medicine, 45(2), 228–236.
Labrique, A. B., Vasudevan, L., Kochi, E., Fabricant, R., & Mehl, G. (2013). mHealth innovations as health
system strengthening tools: 12 common applications and a visual framework. Global Health: Science
and Practice, 1(2), 160–171.
Lobelo, F., Kelli, H. M., Tejedor, S. C., Pratt, M., McConnell, M. V., Martin, S. S., & Welk, G. J. (2016). The
wild wild west: A framework to integrate mHealth software applications and wearables to support
physical activity assessment, counseling and interventions for cardiovascular disease risk reduction.
Progress in Cardiovascular Diseases, 58(6), 584–594.
Lux, E., Adam, M. T. P., Dorner, V., Helming, S., Knierim, M. T., & Weinhardt, C. (2018). Live biofeedback
as a user interface design element: A review of the literature. Communications of the Association for
Information Systems, 43(18), 257–296.

Communications of the Association for Information Systems
Accepted Manuscript
Martin, C. A., & McGrath, B. P. (2014). White-coat hypertension. Clinical and Experimental Pharmacology
and Physiology, 41(1), 22–29.
Martin, S. S., Feldman, D. I., Blumenthal, R. S., Jones, S. R., Post, W. S., McKibben, R. A., … Blaha, M. J.
(2015). mActive: a randomized clinical trial of an automated mHealth intervention for physical activity
promotion. Journal of the American Heart Association, 4(11), 1–9.
Marzano, L., Bardill, A., Fields, B., Herd, K., Veale, D., Grey, N., & Moran, P. (2015). The application of
mHealth to mental health: Opportunities and challenges. The Lancet Psychiatry, 2(10), 942–948.
Medical Avatar LLC. (2018). Medical Avatar. Retrieved October 2, 2018, from http://medicalavatar.com/
Michie, S., Atkins, L., & West, R. (2014). The Behaviour Change Wheel: A Guide to Designing Interventions.
London: Silverback Publishing.
Michie, S., van Stralen, M. M., & West, R. (2011). The behaviour change wheel: A new method for
characterising and designing behaviour change interventions. Implementation Science, 6(42), 1–13.
Michie, S., Wood, C. E., Johnston, M., Abraham, C., Francis, J. J., & Hardeman, W. (2015). Behaviour
change techniques: the development and evaluation of a taxonomic method for reporting and
describing behaviour change interventions. Health Technology Assessment, 19(99), 1–187.
Miller, N. E. (1978). Biofeedback and visceral learning. Annual Review of Psychology, 29(1), 373–404.
Mitchell, V., Ross, T., May, A., Sims, R., & Parker, C. (2015). Empirical investigation of the impact of using
co-design methods when generating proposals for sustainable travel solutions. CoDesign, 12(4),
205–220.
Mori, M. (1970). The uncanny valley. Energy, 7(4), 33–35.
Morrissey, E. C., Corbett, T. K., Walsh, J. C., & Molloy, G. J. (2016). Behavior change techniques in apps
for medication adherence: A content analysis. American Journal of Preventive Medicine, 50(5), e143–
e146.
Mummah, S. A., King, A. C., Gardner, C. D., & Sutton, S. (2016). Iterative development of Vegethon: a
theory-based mobile app intervention to increase vegetable consumption. International Journal of
Behavioral Nutrition and Physical Activity, 13(1), 1–12.
Murray, T., Hardy, D., Spruijt-Metz, D., Hekler, E. B., & Raij, A. (2013). Avatar interfaces for biobehavioral
feedback. In DUXU 2013 Proceedings (pp. 424–434). Las Vegas, NV, USA.
Nahum-Shani, I., Smith, S. N., Spring, B. J., Collins, L. M., Witkiewitz, K., Tewari, A., & Murphy, S. A. (2018).
Just-in-time adaptive interventions (JITAIs) in mobile health: key components and design principles
for ongoing health behavior support. Annals of Behavioral Medicine, 52(6), 446–462.
Nilsen, W., Kumar, S., Shar, A., Varoquiers, C., Wiley, T., Riley, W. T., … Atienza, A. A. (2012). Advancing
the science of mHealth. Journal of Health Communication, 17(sup1), 5–10.
O’Reilly, G. A., & Spruijt-Metz, D. (2013). Current mHealth technologies for physical activity assessment
and promotion. American Journal of Preventive Medicine, 45(4), 501–507.
Oinas-Kukkonen, H., & Harjumaa, M. (2009). Persuasive systems design: Key issues, process model, and
system features. Communications of the Association for Information Systems, 24(28), 485–500.
Paalasmaa, J., Waris, M., Toivonen, H., Leppäkorpi, L., & Partinen, M. (2012). Unobtrusive online
monitoring of sleep at home. In EMBC 2012 Proceedings (pp. 3784–3788). San Diego, CA.
Palatini, P. (2009). Recommendations on how to measure resting heart rate. Medicographia, 31(4), 414–
419.
Palatini, P., Benetos, A., Grassi, G., Julius, S., Kjeldsen, S. E., Mancia, G., … Zanchetti, A. (2006).
Identification and management of the hypertensive patient with elevated heart rate: statement of a
European Society of Hypertension Consensus Meeting. Journal of Hypertension, 24(4), 603–610.
Palatini, P., & Julius, S. (1997). Heart rate and the cardiovascular risk. Journal of Hypertension, 15(1), 3–
17.

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
Papathanasiou, G., Georgakopoulos, D., Papageorgiou, E., Zerva, E., Michalis, L., Kalfakakou, V., &
Evangelou, A. (2013). Effects of smoking on heart rate at rest and during exercise, and on heart rate
recovery, in young adults. Hellenic Journal of Cardiology, 54(3), 168–177.
Payne, H. E., Lister, C., West, J. H., & Bernhardt, J. M. (2015). Behavioral functionality of mobile apps in
health interventions: a systematic review of the literature. JMIR MHealth & UHealth, 3(1), 1–9.
Peira, N., Fredrikson, M., & Pourtois, G. (2014). Controlling the emotional heart: Heart rate biofeedback
improves cardiac control during emotional reactions. International Journal of Psychophysiology,
91(3), 225–231.
Petersen, C., Adams, S. A., & DeMuro, P. R. (2015). mHealth: Don’t forget all the stakeholders in the
business case. Medicine 2.0, 4(2), 1–4.
R2G. (2016). mHealth App Developer Economics 2016. Retrieved from
https://research2guidance.com/product/mhealth-app-developer-economics-2016/
Rho, S., Lee, I., Kim, H., Jung, J., Kim, H., Jun, B. G., & Lim, Y. (2017). FutureSelf: What happens when
we forecast self-trackers’ future health statuses? In DIS 2017 Proceedings (pp. 637–648). NY, USA.
Riedl, R. (2013). On the biology of technostress: Literature review and research agenda. DATA BASE for
Advances in Information Systems, 44(1), 18–55.
Riedl, R., Davis, F. D., & Hevner, A. R. (2014). Towards a NeuroIS research methodology: Intensifying the
discussion on methods, tools, and measurement. Journal of the Association for Information Systems,
15(10), i–xxxv.
Robinson, E., Higgs, S., Daley, A. J., Jolly, K., Lycett, D., Lewis, A., & Aveyard, P. (2013). Development and
feasibility testing of a smart phone based attentive eating intervention. BMC Public Health, 13, 639.
Roy Morgan. (2017). Roy Morgan Image of Professions Survey 2017. Retrieved from
http://www.roymorgan.com/findings/7244-roy-morgan-image-of-professions-may-2017-
201706051543
Saleheen, N., Ali, A. A., Hossain, S. M., Sarker, H., Chatterjee, S., Marlin, B., … Kumar, S. (2015).
puffMarker: A multi-sensor approach for pinpointing the timing of first lapse in smoking cessation. In
UbiComp 2015 Proceedings (pp. 999–1010). Osaka, Japan.
Samdal, G. B., Eide, G. E., Barth, T., Williams, G., & Meland, E. (2017). Effective behaviour change
techniques for physical activity and healthy eating in overweight and obese adults; systematic review
and meta-regression analyses. International Journal of Behavioral Nutrition and Physical Activity,
14(1), 14–42.
Sanders, E. B.-N., & Stappers, P. J. (2008). Co-creation and the new landscapes of design. CoDesign, 4(1),
5–18.
Sanders, E. B.-N., & Stappers, P. J. (2014). Probes, toolkits and prototypes: three approaches to making in
codesigning. CoDesign, 10(1), 5–14.
Schoech, D., Boyas, J. F., Black, B. M., & Elias-Lambert, N. (2013). Gamification for behavior change:
Lessons from developing a social, multiuser, web-tablet based prevention game for youths. Journal
of Technology in Human Services, 31(3), 197–217.
Schoeppe, S., Alley, S., Van Lippevelde, W., Bray, N. A., Williams, S. L., Duncan, M. J., & Vandelanotte, C.
(2016). Efficacy of interventions that use apps to improve diet, physical activity and sedentary
behaviour: A systematic review. International Journal of Behavioral Nutrition and Physical Activity,
13(127), 1–26.
Smith, J. J., Morgan, P. J., Plotnikoff, R. C., Dally, K. A., Salmon, J., Okely, A. D., … Lubans, D. R. (2014).
Smart-phone obesity prevention trial for adolescent boys in low-income communities: The ATLAS
RCT. Pediatrics, 134(3), e723–e731.
Snyder, J., Matthews, M., Chien, J., Chang, P. F., Sun, E., Abdullah, S., & Gay, G. (2015). MoodLight:
Exploring personal and social implications of ambient display of biosensor data. In CSCW 2015
Proceedings (pp. 143–153). Vancouver, Canada.
Steen, M. (2011). Tensions in human-centred design. CoDesign, 7(1), 45–60.

Communications of the Association for Information Systems
Accepted Manuscript
Stewart, M. (2001). Towards a global definition of patient centred care: The patient should be the judge of
patient centred care. British Medical Journal, 322(7284), 444–445.
Strauss, A., & Corbin, J. (1990). Basics of Qualitative Research. Newbury Park, CA: Sage.
Tan, C. S. S., Schöning, J., Luyten, K., & Coninx, K. (2014). Investigating the effects of using biofeedback
as visual stress indicator during video-mediated collaboration. In CHI 2014 Proceedings (pp. 71–80).
Toronto, Canada.
Task Force of the European Society of Cardiology the North American Society of Pacing Electrophysiology.
(1996). Heart rate variability: Standards of measurement, physiological interpretation, and clinical
use. Circulation, 17, 354–381.
Thieme, A., McCarthy, J., Johnson, P., Phillips, S., Wallace, J., Lindley, S., … Olivier, P. (2016). Challenges
for designing new technology for health and wellbeing in a complex mental healthcare context. In CHI
2016 Proceedings (pp. 2136–2149). NY, USA.
Uddin, A., Morita, P., Tallevi, K., Armour, K., Li, J., Nolan, R., & Cafazzo, J. (2016). Development of a
wearable cardiac monitoring system for behavioral neurocardiac training: a usability study. JMIR
MHealth UHealth, 4(2), 1–11.
Urrea, B., Misra, S., Plante, T. B., Kelli, H. M., Misra, S., Blaha, M. J., & Martin, S. S. (2015). Mobile health
initiatives to improve outcomes in primary prevention of cardiovascular disease. Current Treatment
Options in Cardiovascular Medicine, 17(59), 1–12.
van Genugten, L., Dusseldorp, E., Webb, T. L., & van Empelen, P. (2016). Which combinations of
techniques and modes of delivery in internet-based interventions effectively change health behavior?
A meta-analysis. Journal of Medical Internet Research, 18(6), 1–14.
Vandelanotte, C., Müller, A. M., Short, C. E., Hingle, M., Nathan, N., Williams, S. L., … Maher, C. A. (2016).
Past, present, and future of eHealth and mHealth research to improve physical activity and dietary
behaviors. Journal of Nutrition Education and Behavior, 48(3), 219–228.
vom Brocke, J., Riedl, R., & Léger, P.-M. (2013). Application strategies for neuroscience in information
systems design science research. Journal of Computer Information Systems, 53(3), 1–13.
Wattanasoontorn, V., Boada, I., García, R., & Sbert, M. (2013). Serious games for health. Entertainment
Computing, 4(4), 231–247.
Weinmann, M., Schneider, C., & Vom Brocke, J. (2016). Digital nudging. Business & Information Systems
Engineering, 58(6), 433–436.
Winters, N., Oliver, M., & Langer, L. (2017). Can mobile health training meet the challenge of ‘measuring
better’? Comparative Education, 53(1), 115–131.
Woodward, M., Webster, R., Murakami, Y., Barzi, F., Lam, T. H., Fang, X., … Rodgers, A. (2014). The
association between resting heart rate, cardiovascular disease and mortality: evidence from 112,680
men and women in 12 cohorts. European Journal of Preventive Cardiology, 21(6), 719–726.
World Health Organization. (1986). The Ottawa Charter for health promotion. Retrieved from
http://www.who.int/healthpromotion/conferences/previous/ottawa/en/
World Health Organization. (2011). mHealth: New horizons for health through mobile technologies.
Retrieved from http://www.who.int/goe/publications/goe_mhealth_web.pdf
World Health Organization. (2016). Health promotion. Retrieved from
http://www.who.int/topics/health_promotion/en/
Wowak, K. D., Adjerid, I., Angst, C. M., & Guzman, J. C. (2016). A tutorial on empirical ICT4D research in
developing countries: Processes, challenges, and lessons. Communications of the Association for
Information Systems, 38, 353–374.
Xiong, M., He, R., Ji, L., & Wu, J. (2013). A biofeedback system for mobile healthcare based on heart rate
variability. In ICSH 2013 Proceedings (pp. 84–95). Beijing, China.
Yee, N., & Bailenson, J. N. (2007). The Proteus effect: The effect of transformed self-representation on
behavior. Human Communication Research, 33(3), 271–290.

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
Appendix A: Interview Guide
In the following, we provide an example interview guide. Each interview comprised three parts, as
summarized in the following. While the questions provided below were used as a guide for the interviews,
these questions were accompanied by follow-up questions to the participant’s responses.
Part 1: General Understanding of mHealth for Health Behavior Change
Questions:
• What is your opinion on the current state of health promotion? Why do you think it is/isn't working?
[HBS, HP, P, U, HIP]
• What advantages do you think mHealth systems could have over traditional forms of health
promotion? [all]
• What are the biggest challenges for mHealth systems from a [stakeholder’s area] perspective? Why
do you think this is the case? Is there anything missing in the existing approaches? [all]
• What do you see as the role of [stakeholder’s area] in health promotion? [all]
• From an economic and policy perspective, should mHealth systems focus on rehabilitation/recovery
or prevention? Are there specific cohorts should be prioritized initially? [P, HP, HIP, HBS]
• What might the shifts in laws and governance look like as a result of mHealth systems? How can
we address these? [P]
• How can we better integrate mHealth systems into clinical practice? [HP]
Material:
• Figure: Overview of stakeholders and remote systems (see Figure 1)
Part 2: Theoretical Pathways for how mHealth can utilize Mobile Biosensors
Questions:
• How can we ensure that mHealth systems are tailored to the individual circumstances of the user?
(e.g. their baseline). What should the beginning of the health behavior change process look like in
a mHealth system context? Who should be involved? [all]
• What are some important considerations when attempting to change health behavior? [HBS, HP,
U]
• How important is it that you actually believe that you’re going to be successful when trying to change
your health behavior? [HBS, HP, U]
• What feedback can heart rate give us about someone’s health? How could this feedback be
represented in a mHealth system to support health behavior change in the context of a user’s
capability, opportunity, or motivation? [HP, HBS, U]
• How can resting heart rate and heart rate variability be measured? Would you be able to walk me
through the process? What are the factors that need to be considered in a mHealth system context?
[HP, ITP]
• How could we use mHealth systems to increase a user’s capability / opportunity / motivation in the
context of health behavior change? [HBS, HP, U]
• Is there any other data that can be collected from mobile biosensors that would be useful for
supporting health behavior change? How should this data be measured and how should it be
represented in mHealth systems to address user capability, opportunity, or motivation? [all]
• What do you think is the difference psychologically between starting a health behavior change and
maintaining a health behavior change? How should these difference approaches be represented in
a mHealth system? [HBS, HP, U]
• What do you think is important for bringing about lasting health behavior change in lifestyle? How
can this be applied in mHealth systems? [HP, HBS, HIP, P, U]
• Are there any theoretical pathways that we are missing? [all]

Communications of the Association for Information Systems
Accepted Manuscript
Material:
• Figure: Integrative theoretical framework (see Figure 4)
Part 3: Development of General Design Guidelines
Questions:
• What kinds of measurements could be collected that would be useful in a health promotion context?
[HP, U, HIP, HBS, ITP, P]
• Is there any other data (e.g. contextual data such as location, or corrective factors such as age and
sex) that should be collected to improve accuracy and/or provide deeper insights into physiological
measurements? How should these be implemented in mHealth systems? [HP, ITP, HIP, HBS, U,
P]
• What are the challenges or limitations of using mobile biosensors such as heart rate in a health
promotion context? How can we overcome/adjust for these? [HP, ITP]
• How frequently can/should heart rate be measured? Why? [HP, HIP, HBS, U, ITP]
• What position should the user be in when measuring heart rate? Why? [HP]
• What aspects of design do you think are important for creating an empathic connection with the
user? [all]
• How do you think the mHealth interface should look? What things must be there? How should
feedback be represented? Why? [all]
• What stakeholders should be involved in the goal-setting process? Why? [all]
• How should the approach/feedback of mHealth system change for long-term goals and short-term
goals? [all]
• How do you think mHealth systems could be utilized for supporting the setting and accomplishment
of goals? [all]
• What level of involvement do you think health practitioners and health insurers should have in the
use of mHealth systems? [all]
• What are the emerging social, privacy, and security issues to arise as a result of mHealth? How can
we address these? [all]
• How can the data collected through mHealth systems be used to support policy decision making?
What data is this, and what form should it take? [P]
• What are some potential pitfalls and challenges of designing mHealth systems [D]
• Would you be able to walk me through the design process? [D]
• Are there any particular design methods and techniques you think would be particularly useful for
designing mHealth systems? [D]
• Who are the stakeholders in mHealth? What role do you think these stakeholders should have in
the design and operationalization of mHealth systems? [all]
• Are there any other design considerations that we are missing? [all]
Material:
• Table: Current version of the design guidelines (see Appendix B)

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
Appendix B: Design Guidelines Overview
Table B1. Design Guidelines for mHealth Systems Utilizing Mobile Biosensors for Health Behavior Change
Design Guideline
Brief Description
Guideline 1: Employ dedicated mobile
biosensor recordings of adequate length in a
resting state combined with contextual data to
assess the overall health status of the user over
time while also keeping a continuous recording,
using a rolling time window, to retain critical
information in case of a ‘funny turn’.
Mobile biosensor recordings enable a feedback loop between the
user’s physiological state and their perception (P1). For instance,
heart rate is a measure that is easily obtained and that provides
critical information on a person’s health status. The recording
length of five minutes is an inevitable compromise between
convenience and completeness/accuracy. Further, the
measurement conditions (i.e., body position, frequency, time of
day) should be as consistent as possible. Contextual data include
other collected data such as age, location based on GPS, or
medication regime. Biosensor recordings are then interpreted in
the context of the user’s holistic health environment to provide
timely feedback that is situated in and tailored to the user’s
individual situation. In case of an adverse event or ‘funny turn’, for
example, arrhythmia in heart rate, additional data may be
accessed by collecting it continuously in a rolling time window
(e.g., the past 48 hours).
Guideline 2: Use affective visual assets to
convey the physiological state of the user and
make the health information embedded in the
biosensor recordings obtained from mobile
biosensors more intuitive and meaningful to the
user.
Biosensor recordings are difficult to understand for the user in
relation to their health goals and lifestyle behavior. This impairs
users’ capability and motivation to engage in healthy behaviors.
Appropriate visual assets can address capability (P2) by making
the health information embedded in biosensor data intuitive for
users to understand in the context of their health goals. Further,
affective visual assets that link their appearance to the
physiological state of the user (e.g., mirrored-self avatars,
persuasive avatars, virtual pets that adjust based on biosensor
measurements) can address motivation (P4) by creating an
emotional bond between the user and the visual asset that makes
a change in behavior meaningful to them.
Guideline 3: Provide effective goal-setting
support to help the user to develop achievable
health goals relating to their physiological state,
identify actions they can take to achieve those
goals, and increase their self-efficacy through
the feedback loop enabled by mobile biosensor
data.
The consequences of lifestyle behavior often only become salient
over a long period of time, while the short-term changes in
physiology are normally imperceptible to the user. Mobile
biosensors allow to identify short-term effects of lifestyle behaviors
(e.g., changes in physiological stress levels) and, hence, to break
down health goals into small and achievable tasks that are linked
to the users’ physiology. Through the user interface, mobile
biosensors can be used to provide goal-setting support that
facilitates capability (P2) by showing users how they can achieve
their goals (e.g., gamified targets for physiology, biosensor-
enabled serious games to train capabilities for behavior change),
opportunity (P3) by showing users when they can achieve goals
(e.g., triggering reminders by using biosensors to detect health
behaviors); and motivation (P4) by showing feedback on progress
in the user’s physiology and boosting self-efficacy (e.g., self-
comparison of physiological data over time).

Communications of the Association for Information Systems
Accepted Manuscript
Guideline 4: Provide external support to users
by allowing health practitioners to review the
biosensor recordings, to allow the health
practitioner to give feedback to the user, and to
provide the user with access to contextually-
relevant resources based on their physiological
state.
Assessing a user’s health status based on biosensor recordings is
a complex task that goes beyond what an individual user device
can achieve. It is essential for health practitioners to be involved
and to support the user by reviewing the biosensor recordings,
making adjustments to the user’s targeted activities and health
goals, and providing feedback to the user on their physiological
state. External support can extend to recommending a visit to a
health practitioner, or provide access to external sources of
information. Integrating external support into the user interface
addresses capability (P2) by facilitating information exchange and
opportunity (P3) by making individual adjustment to targeted
activities and health goals. The involvement of trusted health
practitioners can increase user trust and compliance, and can
enable monitoring of users’ physiological data to intervene if
necessary.
Guideline 5: Consider the four levels of data
integration in the collection, management, and
use of biosensor recordings, ensuring high
levels of privacy and security of sensitive health
data and enabling effective support for users
and the development of policy.
While the primary focus of mHealth systems commonly lies on the
data stream between the user and the mHealth device, it is critical
to consider additional data integration levels to ensure biosensor
data is used in an effective and responsible way. Data streams of
biosensor recordings to remote systems are critical for increasing
users’ psychological capability (P2) by facilitating information
exchange, for creating opportunities (P3) by including
stakeholders in the goal setting and data assessment process,
and for increasing motivation (P4) by facilitating social interactions
and identifying the motivational factors of the individual user. The
identified levels are: the immediate biosensor data gathering for
individual feedback (Level A), biosensor data aggregation and
analysis (Level B), and interfaces with information systems of
health care providers (Level C) and health insurance providers
(Level D).
Guideline 6: Ensure effective stakeholder
involvement in all design phases to account for
the complexities associated with using mobile
biosensor measurements for behavior change
and to reconcile the diverging backgrounds,
interests, and perspectives of all relevant
stakeholders.
While mobile biosensor recordings offer important insights into a
person’s health status, designing systems to utilize these
recordings for behavior change is a challenging task, involving
complexities associated with the long-term consequences of
unhealthy lifestyle behaviors, multiple direct and indirect
stakeholders with diverse backgrounds and interests, and a
complex landscape of remote systems. Actively involving
stakeholders already in the contextual and conceptual design
phases contributes to reconciling different stakeholder
perspectives early and ensuring that the mHealth artifact
adequately reflects the intricacies of increasing individual users’
capability (P2; e.g., convey biosensor information in a way that is
intuitive for the target cohort), opportunity (P3; e.g., ensure
seamless integration of biosensor data with the health care
provider IS), and motivation (P4; target meaningful motivational
factors for a particular cohort) to engage in targeted behaviors.

Exploring the Design of mHealth Systems for Health Behavior Change using Mobile Biosensors
Accepted Manuscript
About the Authors
Tyler Noorbergen is a PhD candidate in information systems at the University of Newcastle, Australia.
Tyler received a Bachelor of Information Technology and a Bachelor of Business from The University of
Newcastle, Australia in 2015, and a Bachelor of Information Technology (Honours) in 2016. His research
interests center around human-computer interaction, persuasive technologies, and behavior change.
Marc T. P. Adam is a Senior Lecturer in Computing and Information Technology at the University of
Newcastle, Australia. In his research, he investigates the interplay of cognitive and affective processes of
human users in human-computer interaction. He received an undergraduate degree in Computer Science
from the University of Applied Sciences Würzburg, Germany, and a PhD in Economics of Information
Systems from the Karlsruhe Institute of Technology, Germany. His research has been published in top
international outlets such as Business & Information Systems Engineering, Communications of the
Association for Information Systems, International Journal of Electronic Commerce, Journal of Management
Information Systems, Journal of the Association for Information Systems, Journal of Retailing, and others.
John R. Attia is Professor of Medicine and Clinical Epidemiology at the University of Newcastle, Australia,
and has expertise in population, clinical, molecular, and genetic epidemiology. He trained at McMaster
University in clinical medicine and obtained his fellowship with the Royal College of Physicians of Canada
and the Royal Australasian College of Physicians. During this time he was awarded the Outstanding
Housestaff award, the J.T. Walsh award for outstanding Internal Medicine resident, and Best Teacher in
Internal Medicine. He also obtained a BSc in Physiology (Faculty scholar at McGill University), a MSc in
Epidemiology (McMaster University), and a 5 year MRC scholarship to complete his PhD in Molecular
Genetics (University of Toronto). He is currently academic director of general medicine at John Hunter
Hospital responsible for the advanced training program, as well as director of the Clinical Research Design,
IT, and Statistical Support (CReDITSS) Unit at the Hunter Medical Research Institute (HMRI), a unit that
provides epidemiological and statistical methodological advice to clinical researchers.
David J. Cornforth is a Senior Lecturer in Computing and Information Technology at the University of
Newcastle, Australia. He received the BSc degree in Electrical and Electronic Engineering from Nottingham
Trent University, UK, in 1982, and the PhD degree in Computer Science from the University of Nottingham,
UK, in 1994. He has been an educator and researcher at Charles Sturt University, the University of New
South Wales, and now at the University of Newcastle, Australia. He has also been a research scientist at
the Commonwealth Scientific and Industrial Research Organisation (CSIRO), Newcastle, Australia. His
research interests are in health information systems, pattern recognition, artificial intelligence, multi-agent
simulation, and optimization.
Mario Minichiello is Professor in Design at the University of Newcastle, Australia. In his research, he is
focused on the role of design and visual communication in the areas of climate change, economic
betterment, and human behavior. He has over twenty years of experience in international high-level industry
and academic leadership; including the BBC, The Guardian, Ogilvy and Mather, with interaction design at
the natural history museum UK and as a user-centered designer, illustrator, and visualizer for a number
companies around the world. He is director smart design network and of the Hunter Centre for Creative
Industries and Technology, a research cluster that brings together representatives from industry, business,
government, and the university to develop research breakthroughs and new ways of thinking.
Copyright © 2019 by the Association for Information Systems. Permission to make digital or hard copies of
all or part of this work for personal or classroom use is granted without fee provided that copies are not
made or distributed for profit or commercial advantage and that copies bear this notice and full citation on
the first page. Copyright for components of this work owned by others than the Association for Information
Systems must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on
servers, or to redistribute to lists requires prior specific permission and/or fee. Request permission to publish
from: AIS Administrative Office, P.O. Box 2712 Atlanta, GA, 30301-2712 Attn: Reprints or via e-mail from
publications@aisnet.org.