Content uploaded by Jan-Niklas Kramer
Author content
All content in this area was uploaded by Jan-Niklas Kramer on Feb 07, 2019
Content may be subject to copyright.
Protocol
Investigating Intervention Components and Exploring States of
Receptivity for a Smartphone App to Promote Physical Activity:
Protocol of a Microrandomized Trial
Jan-Niklas Kramer1, MSc; Florian Künzler2, MSc; Varun Mishra3, BTech; Bastien Presset4, MSc; David Kotz3,5, PhD;
Shawna Smith6,7, PhD; Urte Scholz8, PhD; Tobias Kowatsch1, PhD
1Center for Digital Health Interventions, Institute of Technology Management, University of St. Gallen, St. Gallen, Switzerland
2Center for Digital Health Interventions, Department of Management, Technology and Economics, Swiss Federal Institute of Technology, Zurich,
Switzerland
3Department of Computer Science, Dartmouth College, Hanover, NH, United States
4Institute of Sports Studies, University of Lausanne, Lausanne, Switzerland
5Center for Technology and Behavioral Health, Geisel School of Medicine at Dartmouth, Lebanon, NH, United States
6Institute for Social Research, University of Michigan - Ann Arbor, Ann Arbor, MI, United States
7Medical School, University of Michigan - Ann Arbor, Ann Arbor, MI, United States
8Department of Psychology, University of Zurich, Zurich, Switzerland
Corresponding Author:
Jan-Niklas Kramer, MSc
Center for Digital Health Interventions
Institute of Technology Management
University of St. Gallen
Dufourstrasse 40a
St. Gallen, 9000
Switzerland
Phone: 41 71224 ext 7249
Fax: 41 71224 7301
Email: jan-niklas.kramer@unisg.ch
Abstract
Background: Smartphones enable the implementation of just-in-time adaptive interventions (JITAIs) that tailor the delivery of
health interventions over time to user- and time-varying context characteristics. Ideally, JITAIs include effective intervention
components, and delivery tailoring is based on effective moderators of intervention effects. Using machine learning techniques
to infer each user’s context from smartphone sensor data is a promising approach to further enhance tailoring.
Objective: The primary objective of this study is to quantify main effects, interactions, and moderators of 3 intervention
components of a smartphone-based intervention for physical activity. The secondary objective is the exploration of participants’
states of receptivity, that is, situations in which participants are more likely to react to intervention notifications through collection
of smartphone sensor data.
Methods: In 2017, we developed the Assistant to Lift your Level of activitY(Ally), a chatbot-based mobile health intervention
for increasing physical activity that utilizes incentives, planning, and self-monitoring prompts to help participants meet personalized
step goals. We used a microrandomized trial design to meet the study objectives. Insurees of a large Swiss insurance company
were invited to use the Ally app over a 12-day baseline and a 6-week intervention period. Upon enrollment, participants were
randomly allocated to either a financial incentive, a charity incentive, or a no incentive condition. Over the course of the intervention
period, participants were repeatedly randomized on a daily basis to either receive prompts that support self-monitoring or not and
on a weekly basis to receive 1 of 2 planning interventions or no planning. Participants completed a Web-based questionnaire at
baseline and postintervention follow-up.
Results: Data collection was completed in January 2018. In total, 274 insurees (mean age 41.73 years; 57.7% [158/274] female)
enrolled in the study and installed the Ally app on their smartphones. Main reasons for declining participation were having an
incompatible smartphone (37/191, 19.4%) and collection of sensor data (35/191, 18.3%). Step data are available for 227 (82.8%,
227/274) participants, and smartphone sensor data are available for 247 (90.1%, 247/274) participants.
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.1http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
Conclusions: This study describes the evidence-based development of a JITAI for increasing physical activity. If components
prove to be efficacious, they will be included in a revised version of the app that offers scalable promotion of physical activity
at low cost.
Trial Registration: ClinicalTrials.gov NCT03384550; https://clinicaltrials.gov/ct2/show/NCT03384550 (Archived by WebCite
at http://www.webcitation.org/74IgCiK3d)
International Registered Report Identifier (IRRID): DERR1-10.2196/11540
(JMIR Res Protoc 2019;8(1):e11540) doi:10.2196/11540
KEYWORDS
physical activity; mHealth; walking; smartphone; incentives; self-regulation
Introduction
Background
Mobile health (mHealth) and sensing technologies recently
sparked excitement because of their capability to deliver
large-scale personalized behavior change interventions at low
cost [1], which can potentially reduce the disease burden
associated with health behaviors, such as diet behavior, smoking,
or physical inactivity [2]. Beyond passive monitoring of health
behavior, smartphones and wearables collect a wealth of
contextual information (such as time, location, communication
logs, or physical activities) that can be used to tailor the delivery
of interventions to participant states that increase the
intervention’s likelihood of success. For example, an
intervention could only be delivered at points in time when the
participant is likely to change her or his behavior (state of
opportunity) or is likely to engage with the intervention content
(state of receptivity) [3]. mHealth apps that utilize this kind of
dynamic tailoring are referred to as just-in-time adaptive
interventions (JITAIs) [3].
During the development process of a JITAI, it is crucial to
decide what key intervention components are needed to affect
the desired intervention outcome and what information should
be used to tailor the delivery of each component to participants
over time [4]. The first question involves an empirical evaluation
of single candidate intervention components. The second
question involves identifying effective time-varying moderators
that indicate in which situations the intervention component is
or is not effective. Unfortunately, these decisions can hardly be
informed by past research because traditional study designs (eg,
randomized controlled trials) rarely evaluate single intervention
components or time-varying moderators of intervention effects.
To facilitate the development of JITAIs, Klasnja et al, therefore,
proposed the microrandomized trial (MRT) [5].
The goals of an MRT are to quantify proximal (short-term) main
effects of single intervention components, to investigate how
these effects change over time, and to identify which contextual
variables moderate the effects of single intervention components.
MRTs use repeated randomization of participants to different
versions, or presence and absence, of individual intervention
components over time, which enables estimation of
time-averaged main effects of single intervention components
on proximal outcomes as well as time-varying effects and their
contextual moderators. Results of an MRT can consequently
inform the researcher which components to include in an
optimized version of the intervention and how to adapt the
delivery of each intervention component to maximize
effectiveness.
Although MRTs are designed to accommodate contextual
moderation, context is likely to be multidimensional —for
example, not just time or location but rather the nexus of time
and location (or other higher order interactions) define opportune
moments for intervention. This limits the approach of
investigating single variables as potential tailoring variables
within MRTs. A potential way of overcoming this limitation is
to train machine learning models that classify the participants’
latent states of intervention receptivity or vulnerability given a
vector of high-resolution smartphone sensor data. Research on
interruptibility, for example, demonstrated that models trained
on smartphone sensor data successfully predict the quality and
quantity of participants’ reactions to notifications on their
smartphone [6-8]. Thus, this approach could allow to
continuously model each participant’s state of receptivity (ie,
the likelihood of engaging with an intervention) from a vast
number of variables. Predictions of these models can in turn be
used to inform intervention delivery of a JITAI.
In this paper, we describe the rationale and design of a 6-week
MRT that evaluates main effects and moderators of 3 different
intervention components (self-monitoring prompts, planning,
and incentives) of the Assistant to Lift your Level of activitY
(Ally), a smartphone app to promote physical activity. Ally
delivers interventions via an interactive text-based chatbot
interface and simultaneously collects contextual data using the
smartphone’s built-in sensors. We also report descriptive
statistics from our remote recruiting process and baseline
characteristics of participants.
Objectives
To inform the evidence-based development of JITAI for physical
activity, the described study has the following objectives:
•To quantify main effects and interactions of main effects
of 3 intervention components of Ally, an mHealth
intervention for physical activity.
•To explore how the effects of intervention components are
moderated by contextual factors and change over time.
•To collect a wide range of high-resolution smartphone
sensor data to predict the participants’ states of receptivity.
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.2http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
Methods
Study Setting
This study is part of a research collaboration between the Center
for Digital Health Interventions, a joint initiative of the
Department of Management, Technology, and Economics at
ETH Zurich and the Institute of Technology Management at
the University of St. Gallen and the CSS insurance, a large
health insurer in Switzerland. Data for this study were collected
from October to December 2017 in the German-speaking part
of Switzerland. The study is registered on ClinicalTrials.gov
(NCT03384550) and was approved by the ethical committee of
ETH Zurich.
The Assistant to Lift Your Level of Activity App
The Ally app focuses on measuring and increasing walking, a
popular and safe activity [9,10] that is known to have positive
health effects independent of other types of physical activity
[11]. Steps per day as an objective measure of walking can be
obtained from the majority of commercially available
smartphones with acceptable accuracy [12]. The Ally
smartphone app tracks participants’ daily step counts and
provides interventions to help participants reach daily step goals.
It contains a dashboard that displays basic information such as
the participants’ current step count and the step goal of the
current day as well as an activity overview of the past 7 days
(Figure 1). Ally runs on the common operating systems Android
and iPhone operating system (iOS). On Android smartphones,
Ally obtains all physical activity–related information from
GoogleFit, a health-tracking platform developed by Google. On
iOS smartphones, the same information is obtained from the
HealthKit, an application programming interface for health apps
provided by Apple. To obtain smartphone sensor data, we used
EmotionSense, a framework to support smartphone-based data
collection originally developed for experimental social
psychology research [13].
Step goals are personalized and calculated daily for each
participant based on the participant’s activity over the past 9
days employing the moving-window percentile-rank algorithm
described by Adams et al [14]. This adaptive goal-setting
algorithm sets the daily step goal to the sixtieth percentile of
the participant’s step count distribution of the past 9 days,
meaning that the participant reaches her or his step goal 40%
of the times when maintaining her or his recent activity level.
Previous studies demonstrated that this adaptive goal setting
outperforms static step goals [14,15]. To facilitate maintenance
of behavior change, adaptive step goals are capped at 10,000
steps per day, which approximates the World Health
Organization recommendations for physical activity [16,17].
To administer the intervention components evaluated in this
study, the Ally app includes a chatbot (Ally) that provides
interactive coaching dialogues similar to other messaging apps
such as Apple’s iMessage, Facebook’s Messenger, or WhatsApp.
The open source behavioral intervention platform MobileCoach
[18] was used to build the chatbot and deliver the interactive
coaching dialogues. In previous studies, MobileCoach-based
interventions have successfully reduced problem drinking in
adolescents [19] and engaged the majority of participants of a
3-month smoking cessation program [20]. Participants interact
with Ally by selecting predefined answer options (Figure 1)
that trigger a response by the chatbot according to the
conversational rules specified in the MobileCoach system.
Beyond specific interventions, the chatbot also communicates
the daily step goal in the morning and feedback regarding the
goal together with informative facts about physical activity at
8 pm in the evening to all participants.
Study Design
From October to December 2017, insurees of a large Swiss
health insurance used the Ally app over a 12-day baseline and
a 6-week intervention period. During the baseline period,
participants only had access to the dashboard of the app, and
no interventions were administered. Over the course of the
6-week intervention period, we randomized participants to
different versions of 3 intervention components: daily
self-monitoring prompts (2 levels; within-subjects), a weekly
planning intervention (3 levels; within-subjects), and daily
incentives (3 levels; between-subjects). The rationale for these
intervention components is described below. To meet study
objective three, we aimed to explore if and how participants’
reaction to intervention components were dependent on their
context. To do so, we ideally need to observe reactions to
intervention notifications in a wide variety of contexts. We,
therefore, sent out intervention and step goal–related
notifications at random points in time but within prespecified
time windows that guaranteed delivery at appropriate times.
For example, daily step goal notifications were delivered at a
random point in time between 8 am and 10 am as users likely
expect to be notified about their goal early in the day.
Participants completed a Web-based questionnaire at baseline
and at postintervention follow-up and received CHF 10 (US
$10 as of 2017) for the successful completion of both
questionnaires. If participants provided consent, they were
invited to participate in exit interviews after the end of the study
that investigate perceptions of participants and mechanisms of
behavior change.
The following subsections first describe details and rationale
for each intervention component as well as for potential
moderators. Subsequently, we outline how each component was
randomized during the intervention period and how we define
the proximal outcome to evaluate each component.
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.3http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
Figure 1. The Ally app: Dashboard with daily (left) and weekly overview (middle) and chat interactions with the Ally chatbot (right).
Intervention Components
Self-Monitoring Prompts
Self-regulatory processes have been identified as a key factor
for health behavior change [21,22]. To support participants’
self-regulation, we designed short dialogue-based self-
monitoring prompts. Self-monitoring prompts remind the
participants of their daily step goal, compare the participants’
current step count to their daily goal, and provide an estimate
of walking minutes necessary to reach the goal together with
an actionable tip on how to increase physical activity. These
dialogues were designed to support the 3 subprocesses of the
self-regulatory construct action control, namely self-monitoring,
awareness of goals or standards, and self-regulatory effort
[23,24]. If a participant had already reached their daily step goal
when starting the dialogue, she/he would receive positive and
encouraging feedback from the Ally chatbot instead.
Participants were randomized to receive a self-monitoring
prompt or no prompt every day during the intervention period
except Sunday, as this day was reserved for the planning
intervention (see below). Self-monitoring prompts were
delivered at a random point in time between 10 am and 6 pm.
Participants’ general tendency to self-monitor their physical
activity may affect the effect of self-monitoring prompts because
the information provided by the prompt is likely to be redundant
to participants who are already aware of their activity level. In
addition, timing of the self-monitoring prompt may be critical.
Research from cognitive psychology demonstrates that people
assign more value to performance increases when their current
performance is close to their goal [25]. Consequently,
self-monitoring prompts may be more effective if they are sent
at times when participants are closer to reaching their step goal.
Planning
Even if motivation to change exists, previous studies show that
on average, 47% of people fail to act upon their good intentions
[26]. Forming specific plans about when and how to act
increases the likelihood of performing the intended behavior
[27,28] and helps to bridge the so-called intention behavior gap.
Planning can be further divided into action planning (AP;
specifying when, where, and how to act) and coping planning
(CP; specifying coping responses for barriers and difficult
situations) [29]. Plans that are articulated in an if-then format
(eg, “if I am tired at work, I will go for a brief walk to get new
energy”) are typically referred to as implementation intentions
[30].
Every Sunday during the intervention period, participants
received either an AP, a CP, or no planning intervention (control;
CC). In the AP condition, Ally asks the participant to plan at
least one and up to 3 walks for the upcoming week. To plan a
single walk, the participants need to specify the day of the week,
the time, and the route that they intend to walk. To create
flexible plans and thus increase the likelihood of adherence,
Ally advises the participant to choose event-related times (eg,
after work) instead of actual times. In the CP condition, Ally
asks the participant to identify barriers for physical activity by
reflecting on the 2 least active days from the previous week.
The participant is then prompted to develop counterstrategies
for each barrier using the if-then format of implementation
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.4http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
intentions [30]. Ally guides this process using examples for
common barriers for physical activity that have been identified
in previous studies [31-33], for example: “If I want to go for a
walk but I lack motivation, I will think of the benefits of walking
for health to motivate myself.” Finally, the participant has the
option to anticipate days of the upcoming week where the barrier
may arise again. Both AP and CP include reminders for the
participant on days when either a walk or a coping reaction was
scheduled. To address the third objective of this study, planning
interventions were sent out on Sundays at a random point in
time between 10 am and 8 pm.
Participants’ activity level and contexts may moderate the effects
of AP and CP. Participants with low activity levels may be more
likely to benefit from AP, which promotes the initiation of
action, whereas participants with high activity levels may benefit
more from CP, which prevents routines from distraction [29,34].
Furthermore, completing the planning intervention can take
several minutes and requires a considerable amount of the
participants’ attention and cognitive capacity. Ideally, the
planning intervention should, therefore, not be delivered in
situations where the participant is involved in an
attention-consuming activity, such as social activities or work.
Incentives
Meta-analyses [35,36] and recent randomized trials [37-39]
have demonstrated the ability of financial incentives to increase
physical activity. However, financial incentives may reduce
intrinsic motivation [40,41]; thus, charity incentives have been
proposed as an alternative incentive strategy. Charity incentives,
that is, donations to a charity organization, could foster
experiences of autonomy and relatedness, which are known to
facilitate rather than impede the buildup of intrinsic motivation
[42]. Moreover, 2 recent studies have so far compared financial
and charity incentives with mixed results [37,43].
In this study, participants were randomly allocated to either a
financial incentive, a charity incentive, or a control condition
using an allocation ratio of 1:1:1. In the financial incentive
condition, participants received CHF 1 (US $1 as of 2017) for
each day they met their personalized step goal. In the charity
incentive condition, instead of keeping the reward to themselves,
participants made a donation of CHF 1 to a charity of their
choice. Participants allocated to the control condition received
no incentives. Earned rewards (maximum of CHF 42) were paid
to participants or donated to charity after completion of the
study.
We hypothesize that the presence of incentives moderates the
effect of the other intervention components. Both planning and
self-monitoring prompts target the participants’ self-regulatory
processes and thus require the participant to be motivated to
reach the provided step goals to produce an effect [44]. As we
expect the incentives to increase the motivation of participants,
we hypothesize that effects of self-monitoring prompts and
planning are more pronounced for participants receiving
financial or charity incentives.
Randomization, Allocation Concealment, and Blinding
The MobileCoach version used in this study requires the time
point of dissemination for all dialogues to be known a priori.
Therefore, randomization had to be performed upon enrollment
of participants for all intervention components. Each participant
was randomized to 1 out of 3 incentive conditions using simple
randomization and an allocation ratio of 1:1:1. In addition,
participants were randomized to 1 out of 9 planning intervention
sequences (S1-S9) that determine the order in which the
participant received the AP intervention, the CP intervention,
or CC intervention throughout the intervention period. We used
blocked randomization with a block size of 9 to achieve balance
between the sequences. The resulting intervention schedule
(Table 1) is uniform and strongly balanced, which controls for
time and carry-over effects [45]. To avoid interference of
self-monitoring prompts and planning, self-monitoring prompts
were not delivered on Sundays. Thus, subtracting 6 from 42 left
36 available days for delivering self-monitoring prompts. To
prevent repetition of content, we created 18 different versions
of self-monitoring prompts that we randomly allocated to the
36 days for each participant. Consequently, at each of the 36
days, half of participants received a self-monitoring prompt (on
average), whereas the other half received no prompt. All
randomizations were performed using random number sequences
generated with the shuffle-array package in JavaScript.
The fully automated randomization process guarantees allocation
concealment for everyone involved in the study. Variables in
the dataset indicating intervention allocation are encrypted to
blind members of the research team involved in data analysis.
A researcher at the Swiss Federal Institute of Technology in
Zurich who is not involved in data analyses holds the decryption
key and is instructed to safely store the key until the analysis
script has been finalized. Due to the setting of the study, it is
not possible to blind participants to intervention assignments.
To reduce the impact of performance and attrition bias,
participants were not informed about the details of the
intervention components before the study.
Measurements
Primary and Secondary Outcomes
As the intervention components (see Table 2) are randomized
on different timescales, we need to define primary and secondary
proximal outcomes that correspond to these timescales to
correctly evaluate the intervention components. The proportion
of overall participant days that step goals are achieved during
the intervention period is the primary outcome to evaluate the
different incentive conditions. Weekly and daily proportions of
participant days that step goals are achieved during the
intervention period are the primary outcomes of the planning
and self-monitoring prompts, respectively. On the same
timescales, differences in steps per day measured with the
smartphone are investigated as a secondary outcome.
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.5http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
Table 1. Intervention schedule of the planning intervention.
Week 6Week 5Week 4Week 3Week 2Week 1Sequence
CPCCCCc
CPb
APAPa
S1
CCAPAPCCCPCPS2
APCPCPAPCCCCS3
CCCCAPCPCPAPS4
APAPCPCCCCCPS5
CPCPCCAPAPCCS6
APCCCPCPCCAPS7
CPAPCCCCAPCPS8
CCCPAPAPCPCCS9
aAP: action planning.
bCP: coping planning.
cCC: control condition (no planning).
Table 2. Overview of intervention components of the Assistant to Lift your Level of activitY(Ally) app.
Proximal outcomeBehavior change
techniques [48]a
Time of deliveryMode of deliveryRandomizationComponent and interven-
tion options
Self-monitoring prompts
Daily proportion of participant days
that step goals were achieved
1.6; 2.2; 4.1Daily except Sun-
day; randomly be-
tween 10 am and 6
pm
ChatUpon enrollment; allo-
cation ratio 1:1
Prompt
Daily proportion of participant days
that step goals were achieved
N/AN/A
N/Ab
Upon enrollment; allo-
cation ratio 1:1
Control (no prompt)
Planning
Weekly proportion of participant
days that step goals were achieved
1.4Sundays; randomly
between 10 am and
6 pm
ChatUpon enrollment; allo-
cation ratio 1:1:1
Action planning
Weekly proportion of participant
days that step goals were achieved
1.2Sundays; randomly
between 10 am and
6 pm
ChatUpon enrollment; allo-
cation ratio 1:1:1
Coping planning
Weekly proportion of participant
days that step goals were achieved
N/AN/AN/AUpon enrollment; allo-
cation ratio 1:1:1
Control (no plan-
ning)
Incentives
Overall proportion of participant
days that step goals were achieved
10.2DailyDashboard/chatUpon enrollment; allo-
cation ratio 1:1:1
Cash incentives
Overall proportion of participant
days that step goals were achieved
10.3DailyDashboard/chatUpon enrollment; allo-
cation ratio 1:1:1
Charity incentives
Overall proportion of participant
days that step goals were achieved
N/AN/AN/AUpon enrollment; allo-
cation ratio 1:1:
Control (no incen-
tives)
a1.2=problem solving, 1.4=action planning, 1.6=discrepancy between current behavior and goal, 2.2=feedback on behavior, 4.1=instruction on how to
perform a behavior, 10.2=material reward (behavior), and 10.3=nonspecific reward.
bN/A: not applicable
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.6http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
Table 3. Summary of collected sensor data.
Frequencya
Data typeVariableSensor
Every 10 min3D FloatLocationGPSb
ContinuousCategoricalPhysical activityAccelerometer
ContinuousIntegerTimeTime
ContinuousBinary (near and far)Proximity of the phoneProximity
Every 10 minCategorical/stringWi-Fi connectionWi-Fi
Every 10 minCategorical/stringBluetooth connectionBluetooth
ContinuousFloatAmbient lightAmbient light
ContinuousFloat (charged in percentage)Battery statusBattery status
ContinuousBinary (on/off)Screen on/offScreen events
aEstimated frequencies only. Actual frequencies may vary depending on device and operating system.
bGPS: global positioning system.
For financial and charity incentives, postintervention differences
in intrinsic and extrinsic motivation, and differences in app
engagement and nonusage attrition during the intervention
period are evaluated as additional secondary outcomes.
Dimensions of intrinsic and extrinsic motivation are measured
using the Behavioral Regulation for Exercise Questionnaire-2
(BREQ-2) [46]. As the external regulation subscale in the
BREQ-2 exclusively relates to external regulation by other
people, it is substituted by the more generally worded external
regulation subscale of the Situational Motivation Scale (SIMS)
[47]. Subscales of both instruments have shown good reliability
(Cronbach alpha=.73-.86, BREQ-2 [46] and Cronbach
alpha=.86, SIMS external regulation subscale [47]). Validity
has been confirmed by factor analysis (BREQ-2) [46] and
correlational analysis (SIMS) [47]. We measure engagement
with the Ally app using the number and length of app launch
sessions per day. An app launch session is defined as any
interaction of the participant with the Ally app, separated by 5
min between events. If a participant left the app open and did
not take action for 5 min or more, then the next interaction with
the app counts as a new session. We coded a participant as
“non-usage attrition observed” when she/he stopped using the
Ally app at least 7 days before the end of the study.
Other Outcomes
As a preliminary pre-post evaluation of the Ally app,
self-reported health outcomes and targeted mediators of behavior
change were assessed at baseline and at postintervention
follow-up. In addition, we assessed participant’s perceptions of
the Ally app, of intervention components, and of the chatbot in
addition to predictors of technology acceptance at
postintervention follow-up. An overview of all measured
variables is available in Multimedia Appendix 1 ([49-57]).
Sensor Data
Drawing on previous literature on context-aware mobile
notification management systems [58], we identified smartphone
sensors that may aid with predicting the participants’ state of
receptivity. Sensor data were obtained from participants during
the intervention period. Table 3 provides a summary of these
sensors, their collected data, and their sensing frequency. In line
with previous studies, we operationalize state of receptivity by
using the response rate (ie, whether a participant responds to a
notification or not) and the response time (ie, time between
notification and response) to notifications of the Ally app.
Sample Size
We used a simulation-based approach to estimate the power of
our study design and determine the necessary sample size. As
interaction effects require a greater number of participants to
be detected with adequate power [59], we focused the power
analysis on the two-way interaction effect of the between-subject
factor incentives and the within-subject factor planning. We
systematically varied the probability of reaching the step goal
p (SG) when no intervention is provided (0.30, 0.40, and 0.50).
These values seem reasonable given the fact the probability of
step goal achievement according to the goal setting algorithm
is 0.40. We further varied the increase in probability because
of incentive and planning main effects (0.05, 0.10, and 0.15)
and the interaction effect (0.05, 0.10, and 0.15) for sample sizes
ranging from n=20 to n=400. These effect sizes were based on
previous studies on the use of incentives to promote physical
activity [38,39]. A total of 100 simulations were generated for
each scenario. Pvalues of interaction effects were obtained by
fitting generalized estimating equations (GEEs) models to the
simulated data, and power was calculated as the proportion of
Pvalues below the significance level of alpha=.05. Figure 2
displays simplified results of this simulation with constant main
effects of .15 and different values for p (SG) and the interaction
effect. The black horizontal line indicates the recommended
level of power of 1-beta=.80.
Simulations indicate that a sample size of roughly 220 is
sufficient to detect an interaction effect of .05 with a power of
1-beta=.80 and alpha=.05 for p (SG)=.50. Sample sizes to detect
an interaction effect .05 considerably increase for smaller values
of p (SG) and smaller main effects (not shown). We, therefore,
considered a sample size of 220 to be most feasible, and
accounting for dropout, we set the target sample size for our
study to 300.
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.7http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
Figure 2. Results of the simulation-based power analysis. p (SG): probability of reaching the step goal.
Recruitment and Eligibility
We invited insurees via email to participate in our study. On
the basis of a previous study in the same population and with a
similar recruiting process [60], we expected a participation rate
of approximately 3%. We initially sent the invitation email to
10,000 insurees. However, because initial participation was
lower than expected, an additional 20,000 insurees were invited
to meet the required sample size.
The invitation email contained brief information about the study,
eligibility criteria, and emphasized the benefits of participation.
No details about the different intervention conditions were
disclosed to the insurees. By following a link in the invitation
mail, interested insurees were forwarded to an online survey
platform, where they were screened for eligibility. Eligibility
criteria were as follows: (1) German-speaking, (2) aged 18 years
or older, (3) enrolled in a complementary insurance program,
(4) being free of any medical condition that prohibits increased
levels of physical activity, (5) not actively using an activity
tracker or a comparable smartphone app, and (6) not working
night shifts.
As meeting the first 3 eligibility criteria could be determined
from the insurance company’s database, only insurees meeting
these criteria were invited to participate. Due to legal regulations
in Switzerland, the Ally app could be offered to insurees with
complementary health insurance plans only. Note, however,
that in Switzerland, 75% of people are enrolled in
complementary insurance plans [61]. We excluded insurees
working night shifts because interventions were sent out on
prespecified times during the day only. Eligible insurees could
subsequently obtain detailed information about the goals and
study procedures, provide consent to participate, and enroll in
the study. After enrollment, participants completed the first
online questionnaire and subsequently received a 6-digit code,
together with instructions on how to download and install the
Ally app. Participants had to enter the code once upon first
opening the Ally app to connect survey data and app data and
to ensure that only study participants were using the app.
Statistical Analyses
All analyses were prespecified before enrolling participants into
the study. After completion of the study but before starting data
analyses, the statistical methods for analyzing the effects of
intervention components were changed from hierarchical linear
modeling to a GEE-based approach to avoid biased effect
estimates [62].
Primary Analyses
To evaluate main effects and interactions of intervention
components, we will use the centered and weighted GEE
approach described in the study by Boruvka et al [62]. This
approach guarantees unbiased effect estimates when treatment
and moderator variables are time-varying. Statistical models
will evaluate each main effect and interaction of intervention
components of interest on the components appropriate proximal
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.8http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
outcome. For all main effects and interactions that include
comparisons of multiple conditions, the main comparisons of
interest are between the respective intervention and control
conditions.
Missing data on covariates and on the dependent variable will
be imputed using multiple imputation, provided the missing at
random assumption is justified. We will perform sensitivity
analyses to assess the robustness of the results of the primary
analyses. These analyses include a per-protocol analysis and an
adjusted analysis in which effect estimates are adjusted for a
linear trend of time, baseline step count, and covariates of
physical activity. For all tests, we use 2-sided Pvalues with
alpha<.05 level of significance.
Secondary Analyses
Secondary analyses focus on the analysis of intervention
components on participants’ step counts and on the effects of
incentives on app engagement, nonusage attrition, and
motivation. Steps per day are analyzed using the same modeling
approach as described above. Again, if missing data can be
assumed to be missing at random, we plan to impute missing
step counts using multiple imputations. As evidence suggests
that participant days with less than 1000 steps are unlikely to
represent accurate activity data [63,64], those days will be set
to missing before imputation.
Generalized linear models will be used to analyze the effect of
incentives on engagement and nonusage attrition. One-way
analysis of variance (ANOVA) is performed for each subscale
of the BREQ-2 to analyze the effect of incentives on the
different forms of intrinsic and extrinsic motivation. Pvalues
will be adjusted according to the Holm-Bonferroni method [65].
If the omnibus test of the ANOVA is significant, we will
investigate contrasts between the 3 incentive groups. Again, the
main comparison of interest is between the intervention groups
and the control group. An overview of all planned statistical
analysis is available in Multimedia Appendix 2.
Moderators
Due to the lack of existing research in this field, the moderation
analyses of main effects are exploratory and may investigate
various moderators of intervention components, different forms
of operationalizing these moderators, or varying types of
relationships (eg, linear or quadratic). Moderations of main
effects are investigated by adding a term for the interaction
between the main effect and the respective moderator to the
statistical model.
State of Receptivity
We will compare several different methodological approaches
to predict the participants’ state of receptivity. First, we plan to
evaluate the performance of supervised learning algorithms in
predicting response rate and response time. These algorithms
have produced predictions of acceptable accuracy in previous
studies on interruptibility [58]. Second, we plan to frame the
problem at hand as a classification problem. A classifier will
be trained to learn to differentiate between contexts in which
the notification is sent (and are assumed to represent
nonreceptive contexts) and contexts in which the participant
interacts with the app (and in turn are assumed to represent
receptive contexts). To this end, we aim to use generalized linear
models as a starting point before exploring online learning
algorithms that can learn and adapt to each participant’s
preferences, and any change thereof. This analysis strategy,
however, is preliminary at the time of writing, as the final
analysis will consider additional factors such as the quality and
distribution of collected data.
Results
Recruitment
Of all 30,000 invited insurees, 749 (2.50%) clicked the link in
the invitation mail and were subsequently screened for
eligibility. Of those, 694 (92.7%) were eligible and 382 (51.0%)
provided informed consent to participate. Of all insurees who
provided informed consent, 274 (71.7%) successfully completed
the baseline survey and installed the Ally app on their
smartphone (Figure 3). Invited insurees were given the
opportunity to select reasons why they declined participation
from a list of predefined answer options using a separate survey
(n=191). A link to this survey was included in the invitation
mail and placed on the informed consent screen. Possession of
an incompatible smartphone (37/191, 19.4%) and unwillingness
to share smartphone sensor data (35/191, 18.3%) were the most
frequently stated reasons to decline participation.
Of 274 participants, 32 (11.7%) did not receive any interventions
because they stopped using the app before the start of the
intervention period. Due to technical errors, 6 participants did
not receive the interventions they were randomized to (eg, a
self-monitoring prompt was sent out on a day where the
participant was randomized to not receiving a prompt). For the
6 participants, these errors affected between 1 and 25 out of 42
participant days. Steps per day measured with the smartphone
are available for 227 (82.8%, 227/274) participants, and
smartphone sensor data are available for 247 (90.1%, 247/274)
participants. After completing the 6-week intervention period,
181 (66.1%, 181/274) participants filled out the Web-based
follow-up survey. Data collection finished in January 2018.
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.9http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
Figure 3. Participant flow.
Baseline Characteristics
Baseline and demographic characteristics of participants are
presented in Table 4. Participants (mean age 41.73 years; 57.7%
[158/274] female) were mostly Swiss (246/274, 89.8%) and
walked on average 6336 (SD 2701) steps per day during the
baseline period. The distribution of age and gender is
comparable with those of other studies evaluating physical
activity apps [66,67]. Self-reported physical activity and
comparisons of self-reported health with the German 12-item
Short Form norm sample indicate that on average, participants
in our study may be healthier and more active than the general
population.
Expected Results and Dissemination
We will start data analyses after publication of this study
protocol. We anticipate submitting results to a peer-reviewed
journal in 2019. Preliminary results of the study may be
presented at conferences, workshops, symposia etc. Results of
the analysis of sensor data to predict the participants’ state of
receptivity will be published separately in a peer-reviewed
journal or conference proceedings.
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.10http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
Table 4. Baseline and demographic characteristics of participants (N=274).
StatisticsCharacteristics
41.73 (13.54)Age in years, mean (SD)
Sex, n (%)
158 (57.7)Female
111 (40.5)Male
5 (1.8)
N/Aa
Education, n (%)
3 (1.1)Compulsory education
97 (35.4)High school
164 (59.9)University
10 (3.7)N/A
Nationality, n (%)
246 (89.8)Swiss
13 (4.7)German
12 (4.4)Other
3 (1.1)N/A
Employment, n (%)
152 (55.5)Full-time
76 (27.7)Part-time
22 (8.0)Retired
2 (0.7)Unable to work
14 (5.1)Unemployed
8 (2.9)N/A
Income, n (%)
30 (11.0)<CHF 2500
53 (19.3)CHF 2501-5000
86 (31.4)CHF 5001-7500
37 (13.5)CHF 7501-10,000
24 (8.8)>CHF 10,000
Smartphone, n (%)
186 (67.9)iPhone operating system
88 (32.1)Android
Step count, n (%)
74 (27.0)<5000
68 (24.8)5000-7499
35 (12.8)7500-9999
21 (7.7)>10,000
76 (27.7)N/A
IPAQb, , n (%)
31 (11.3)Low
115 (42.0)Moderate
122 (44.5)High
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.11http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
StatisticsCharacteristics
6 (2.2)N/A
24.44 (4.15)
BMIc, mean (SD)
53.32 (4.58)
SF-12dphysical component summary, mean (SD)
51.17 (8.11)SF-12 mental component summary, mean (SD)
aN/A: not applicable.
bIPAQ: International Physical Activity Questionnaire (short form) [68].
cBMI: body mass index.
dSF-12: 12-item Short Form.
Discussion
Summary
This study protocol describes the design of an MRT that
investigates the effectiveness of 3 intervention components as
well as associated moderators to guide the design of a
smartphone app to promote physical activity. This study is
among the first to generate data for the evidence-based
development of a JITAI for physical activity. In addition, a data
collection strategy is described that enables the parallel
collection of sensor data needed to build predictive models that,
when implemented into a JITAI, allow real-time prediction of
the state of receptivity. These predictions allow to better inform
adaptive intervention delivery by highlighting situations where
users are likely to respond to intervention notifications. Insights
from this study are of value for anyone involved in the
development of mHealth interventions and to support important
decisions, such as which components to include in an mHealth
intervention or how to tailor intervention delivery to participants
over time.
Strengths and Limitations
Our study illustrates potential and challenges associated with
mHealth studies. The study’s remote recruitment and data
collection process allowed recruiting more than 270 participants
in less than a week and collecting a unique and powerful
high-resolution dataset that contains real-world behavioral and
contextual sensor data. In line with other mHealth studies [69],
we observed a larger drop in app usage at the beginning of the
study, potentially complicating interpretation of our findings.
Likewise, step and sensor data were missing for some
participants. Explanations for missing data include never
reacting to a message of the Ally chatbot, which was required
to request step counts from GoogleFit or the HealthKit, or
denying app permissions to collect sensor data. Even though
the Ally app instructed participants to carry their smartphone
whenever possible, other studies observed an underestimation
of smartphone-based step counts because smartphones are often
not carried consistently in free-living conditions [70]. This may
lead to conservative effect estimates, if increases in step counts
are not recorded by the Ally app. Sending invitations via email
and to insurees of one insurer only, the restricted range of
compatible smartphones, and the requirement to share sensitive
data (eg, global positioning system sensor data) are likely
contributing to a self-selection of participants in our study. This
limits the generalizability of our findings and conclusions.
Although all participants indicated upon enrollment that they
were using no comparable app or device for tracking physical
activity, we cannot exclude that such apps or devices were used
or that participants primarily used the Apple Health or GoogleFit
apps that were required for the Ally app to count steps correctly.
Use of such additional apps or devices could potentially affect
the use of the Ally app and the effectiveness of intervention
components.
If intervention components prove to be effective, we plan to
include them in a revised version of the Ally app that provides
just-in-time adaptive support depending on identified moderators
and predicted states of receptivity. We plan to evaluate this
revised version in a randomized controlled trial.
Acknowledgments
This study is funded by CSS insurance, Switzerland. The CSS insurance supported the recruitment of participants but had no role
in app development, study design, data collection, data analysis and interpretation, writing the manuscript, or in reviewing and
approving the manuscript for publication.
Authors' Contributions
JNK, TK, and US developed the concept for intervention components and for the Ally app. FK, VM, and TK were responsible
for app design and implementation. JNK, FK, TK, and SS developed the study design described in this protocol. JNK and SS
developed the methodological approach for the analyses of the different intervention components, and FK, VM, DK, and TK
developed the methodological approach for the analyses of smartphone sensor data. BP developed the concept and methodology
for the qualitative exit interviews. JNK wrote the manuscript incorporating critical reviews from all authors. All authors reviewed
and approved the manuscript before submission.
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.12http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
Conflicts of Interest
JNK, FK, and TK are affiliated with the Center for Digital Health Interventions, a joint initiative of the Department of Management,
Technology, and Economics at ETH Zurich and the Institute of Technology Management at the University of St. Gallen, which
is funded in part by the Swiss health insurer CSS. TK is also cofounder of Pathmate Technologies, a university spin-off company
that creates and delivers digital clinical pathways and has used the open source MobileCoach platform for that purpose, too.
However, Pathmate Technologies is not involved in the intervention described in this paper. No other conflicts of interests are
declared.
Multimedia Appendix 1
Study timeline including intervention components and assessment of outcomes.
[PDF File (Adobe PDF File), 69KB - resprot_v8i1e11540_app1.pdf ]
Multimedia Appendix 2
Overview of variables, measures and methods of analysis.
[PDF File (Adobe PDF File), 45KB - resprot_v8i1e11540_app2.pdf ]
References
1. Steinhubl SR, Muse ED, Topol EJ. The emerging field of mobile health. Sci Transl Med 2015 Apr 15;7(283):283rv3 [FREE
Full text] [doi: 10.1126/scitranslmed.aaa3487] [Medline: 25877894]
2. World Health Organization. World health statistics 2017: monitoring health for the SDGs. 2017. URL: https://www.who.int/
gho/publications/world_health_statistics/2017/en/ [accessed 2018-11-30] [WebCite Cache ID 74JjNiLk7]
3. Nahum-Shani I, Smith SN, Spring BJ, Collins LM, Witkiewitz K, Tewari A, et al. Just-in-Time Adaptive Interventions
(JITAIs) in mobile health: key components and design principles for ongoing health behavior support. Ann Behav Med
2018 May 18;52(6):446-462. [doi: 10.1007/s12160-016-9830-8] [Medline: 27663578]
4. Murray E, Hekler EB, Andersson G, Collins LM, Doherty A, Hollis C, et al. Evaluating digital health interventions: key
questions and approaches. Am J Prev Med 2016 Dec;51(5):843-851 [FREE Full text] [doi: 10.1016/j.amepre.2016.06.008]
[Medline: 27745684]
5. Klasnja P, Hekler EB, Shiffman S, Boruvka A, Almirall D, Tewari A, et al. Microrandomized trials: an experimental design
for developing just-in-time adaptive interventions. Health Psychol 2015 Dec;34S:1220-1228 [FREE Full text] [doi:
10.1037/hea0000305] [Medline: 26651463]
6. Pielot M, Dingler T, Pedro JS, Oliver N. When attention is not scarce-detecting boredom from mobile phone usage. In:
Proceedings of the 2015 ACM International Joint Conference on Pervasive and Ubiquitous Computing. New York: ACM;
2015 Presented at: UbiComp '15; September 07-11, 2015; Osaka, Japan. [doi: 10.1145/2750858.2804252]
7. Fischer JE, Greenhalgh C, Benford S. Investigating episodes of mobile phone activity as indicators of opportune moments
to deliver notifications. In: Proceedings of the 13th International Conference on Human Computer Interaction with Mobile
Devices and Services. New York: ACM; 2011 Presented at: MobileHCI '11; August 30-September 2, 2011; Stockholm,
Sweden. [doi: 10.1145/2037373.2037402]
8. Pejovic V, Musolesi M. InterruptMe: designing intelligent prompting mechanisms for pervasive applications. In: Proceedings
of the 2014 ACM International Joint Conference on Pervasive and Ubiquitous Computing. 2014 Presented at: UbiComp
'14; September 13-17, 2014; Seattle, USA. [doi: 10.1145/2632048.2632062]
9. Lamprecht M, Fischer A, Stamm H. Federal Office of Sport. 2014. Sport Schweiz 2014: Sport activity and sport interest
of the Swiss population [in German] URL: https://www.baspo.admin.ch/de/dokumentation/publikationen/sport-schweiz-2014.
html [accessed 2018-11-30] [WebCite Cache ID 74Jkch0JZ]
10. Hootman JM, Macera CA, Ainsworth BE, Martin M, Addy CL, Blair SN. Association among physical activity level,
cardiorespiratory fitness, and risk of musculoskeletal injury. Am J Epidemiol 2001 Aug 01;154(3):251-258. [doi:
10.1093/aje/154.3.251] [Medline: 11479190]
11. Kelly P, Kahlmeier S, Götschi T, Orsini N, Richards J, Roberts N, et al. Systematic review and meta-analysis of reduction
in all-cause mortality from walking and cycling and shape of dose response relationship. Int J Behav Nutr Phys Act 2014
Oct 24;11:132. [doi: 10.1186/s12966-014-0132-x] [Medline: 25344355]
12. Case MA, Burwick HA, Volpp KG, Patel MS. Accuracy of smartphone applications and wearable devices for tracking
physical activity data. J Am Med Assoc 2015 Feb 10;313(6):625-626. [doi: 10.1001/jama.2014.17841] [Medline: 25668268]
13. Rachuri K, Musolesi M, Mascolo C, Rentfrow PJ, Longworth C, Aucinas A. EmotionSense: a mobile phones based adaptive
platform for experimental social psychology research. In: Proceedings of the 12th ACM international conference on
Ubiquitous computing. 2010 Presented at: UbiComp '10; September 26-29, 2010; Copenhagen, Denmark. [doi:
10.1145/1864349.1864393]
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.13http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
14. Adams MA, Hurley JC, Todd M, Bhuiyan N, Jarrett CL, Tucker WJ, et al. Adaptive goal setting and financial incentives:
a 2 × 2 factorial randomized controlled trial to increase adults' physical activity. BMC Public Health 2017 Dec 29;17(1):286
[FREE Full text] [doi: 10.1186/s12889-017-4197-8] [Medline: 28356097]
15. Adams MA, Sallis JF, Norman GJ, Hovell MF, Hekler EB, Perata E. An adaptive physical activity intervention for overweight
adults: a randomized controlled trial. PLoS One 2013;8(12):e82901 [FREE Full text] [doi: 10.1371/journal.pone.0082901]
[Medline: 24349392]
16. Tudor-Locke C, Bassett DR. How many steps/day are enough? Preliminary pedometer indices for public health. Sports
Med 2004;34(1):1-8. [doi: 10.2165/00007256-200434010-00001] [Medline: 14715035]
17. Tudor-Locke C, Hatano Y, Pangrazi RP, Kang M. Revisiting "how many steps are enough?". Med Sci Sports Exerc 2008
Jul;40(7 Suppl):S537-S543. [doi: 10.1249/MSS.0b013e31817c7133] [Medline: 18562971]
18. Filler A, Kowatsch T, Haug S, Wahle F, Staake T, Fleisch E. MobileCoach: A Novel Open Source Platform for the Design
of Evidence-based, Scalable and Low-Cost Behavioral Health Interventions - Overview and Preliminary Evaluation in the
Public Health Context. : IEEE; 2015 Presented at: 2015 Wireless Telecommunications Symposium (WTS); April 15-17,
2015; New York, USA. [doi: 10.1109/WTS.2015.7117255]
19. Haug S, Paz Castro R, Kowatsch T, Filler A, Dey M, Schaub MP. Efficacy of a web- and text messaging-based intervention
to reduce problem drinking in adolescents: results of a cluster-randomized controlled trial. J Consult Clin Psychol 2017
Dec;85(2):147-159. [doi: 10.1037/ccp0000138] [Medline: 27606700]
20. Paz Castro R, Haug S, Filler A, Kowatsch T, Schaub MP. Engagement within a mobile phone-based smoking cessation
intervention for adolescents and its association with participant characteristics and outcomes. J Med Intern Res 2017 Nov
01;19(11):e356 [FREE Full text] [doi: 10.2196/jmir.7928] [Medline: 29092811]
21. Michie S, Abraham C, Whittington C, McAteer J, Gupta S. Effective techniques in healthy eating and physical activity
interventions: a meta-regression. Health Psychol 2009 Nov;28(6):690-701. [doi: 10.1037/a0016136] [Medline: 19916637]
22. Mann T, de Ridder D, Fujita K. Self-regulation of health behavior: social psychological approaches to goal setting and goal
striving. Health Psychol 2013 May;32(5):487-498. [doi: 10.1037/a0028533] [Medline: 23646832]
23. Carver CS, Scheier MF. Control theory: a useful conceptual framework for personality-social, clinical, and health psychology.
Psychol Bull 1982 Jul;92(1):111-135. [doi: 10.1037/0033-2909.92.1.111] [Medline: 7134324]
24. Sniehotta FF, Nagy G, Scholz U, Schwarzer R. The role of action control in implementing intentions during the first weeks
of behaviour change. Br J Soc Psychol 2006 Mar;45(Pt 1):87-106. [doi: 10.1348/014466605X62460] [Medline: 16573874]
25. Heath C, Larrick RP, Wu G. Goals as reference points. Cogn Psychol 1999 Feb;38(1):79-109. [doi: 10.1006/cogp.1998.0708]
[Medline: 10090799]
26. Sheeran P. Intention—behavior relations: a conceptual and empirical review. Eur Rev Soc Psychol 2002;12(1):1-36. [doi:
10.1080/14792772143000003]
27. Gollwitzer P, Sheeran P. Implementation intentions and goal achievement: a meta-analysis of effects and processes. Adv
Exp Soc Psychol 2006;38(1):69-119. [doi: 10.1016/S0065-2601(06)38002-1]
28. Bélanger-Gravel A, Godin G, Amireault S. A meta-analytic review of the effect of implementation intentions on physical
activity. Health Psychol Rev 2013;7(1):23-54. [doi: 10.1080/17437199.2011.560095]
29. Sniehotta FF, Schwarzer R, Scholz U, Schüz B. Action planning and coping planning for long-term lifestyle change: theory
and assessment. Eur J Soc Psychol 2005;35(4):565-576. [doi: 10.1002/ejsp.258] [Medline: 25855820]
30. Gollwitzer PM. Implementation intentions: strong effects of simple plans. Am Psychol 1999;54(7):493-503. [doi:
10.1037/0003-066X.54.7.493]
31. Reichert FF, Barros AJ, Domingues MR, Hallal PC. The role of perceived personal barriers to engagement in leisure-time
physical activity. Am J Public Health 2007 Mar;97(3):515-519. [doi: 10.2105/AJPH.2005.070144] [Medline: 17267731]
32. Cerin E, Leslie E, Sugiyama T, Owen N. Perceived barriers to leisure-time physical activity in adults: an ecological
perspective. J Phys Act Health 2010 Jul;7(4):451-459. [doi: 10.1123/jpah.7.4.451] [Medline: 20683086]
33. Zunft HJ, Friebe D, Seppelt B, Widhalm K, Remaut de Winter AM, Vaz de Almeida MD, et al. Perceived benefits and
barriers to physical activity in a nationally representative sample in the European Union. Public Health Nutr 1999
Mar;2(1A):153-160. [Medline: 10933635]
34. Ziegelmann JP, Lippke S, Schwarzer R. Adoption and maintenance of physical activity: planning interventions in young,
middle-aged, and older adults. Psychol Health 2006;21(2):145-163. [doi: 10.1080/1476832050018891] [Medline: 21985115]
35. Mitchell MS, Goodman JM, Alter DA, John LK, Oh PI, Pakosh MT, et al. Financial incentives for exercise adherence in
adults: systematic review and meta-analysis. Am J Prev Med 2013 Nov;45(5):658-667. [doi: 10.1016/j.amepre.2013.06.017]
[Medline: 24139781]
36. Barte JC, Wendel-Vos GW. A systematic review of financial incentives for physical activity: the effects on physical activity
and related outcomes. Behav Med 2017;43(2):79-90. [doi: 10.1080/08964289.2015.1074880] [Medline: 26431076]
37. Finkelstein EA, Haaland BA, Bilger M, Sahasranaman A, Sloan RA, Nang EE, et al. Effectiveness of activity trackers with
and without incentives to increase physical activity (TRIPPA): a randomised controlled trial. Lancet Diabetes Endocrinol
2016 Dec;4(12):983-995. [doi: 10.1016/S2213-8587(16)30284-4] [Medline: 27717766]
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.14http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
38. Patel MS, Asch DA, Rosin R, Small DS, Bellamy SL, Heuer J, et al. Framing financial incentives to increase physical
activity among overweight and obese adults: a randomized, controlled trial. Ann Intern Med 2016 Mar 15;164(6):385-394
[FREE Full text] [doi: 10.7326/M15-1635] [Medline: 26881417]
39. Patel MS, Asch DA, Rosin R, Small DS, Bellamy SL, Eberbach K, et al. Individual versus team-based financial incentives
to increase physical activity: a randomized, controlled trial. J Gen Intern Med 2016 Dec;31(7):746-754 [FREE Full text]
[doi: 10.1007/s11606-016-3627-0] [Medline: 26976287]
40. Deci EL, Koestner R, Ryan RM. A meta-analytic review of experiments examining the effects of extrinsic rewards on
intrinsic motivation. Psychol Bull 1999 Nov;125(6):627-668. [doi: 10.1037/0033-2909.125.6.627] [Medline: 10589297]
41. Cerasoli CP, Nicklin JM, Ford MT. Intrinsic motivation and extrinsic incentives jointly predict performance: a 40-year
meta-analysis. Psychol Bull 2014 Jul;140(4):980-1008. [doi: 10.1037/a0035661] [Medline: 24491020]
42. Deci E, Ryan R. Intrinsic Motivation and Self-Determination in Human Behavior. New York, USA: Plenum Press; 1985.
43. Harkins KA, Kullgren JT, Bellamy SL, Karlawish J, Glanz K. A trial of financial and social incentives to increase older
adults' walking. Am J Prev Med 2017 May;52(5):e123-e130. [doi: 10.1016/j.amepre.2016.11.011] [Medline: 28062271]
44. Schwarzer R, Luszczynska A. How to overcome health-compromising behaviors: the health action process approach. Eur
Psychol 2008 Jan;13(2):141-151. [doi: 10.1027/1016-9040.13.2.141]
45. Hills M, Armitage P. The two-period cross-over clinical trial. Br J Clin Pharmacol 1979;8(1):7-20. [doi:
10.1111/j.1365-2125.1979.tb05903.x]
46. Markland D, Tobin V. A modification to the behavioural regulation in exercise questionnaire to include an assessment of
amotivation. J Sport Exerc Psychol 2004 Jun;26(2):191-196. [doi: 10.1123/jsep.26.2.191]
47. Guay F, Vallerand RJ, Blanchard C. On the assessment of situational intrinsic and extrinsic motivation: the Situational
Motivation Scale (SIMS). Motiv Emot 2000;24(3):175-213. [doi: 10.1023/A:1005614228250]
48. Michie S, Richardson M, Johnston M, Abraham C, Francis J, Hardeman W, et al. The behavior change technique taxonomy
(v1) of 93 hierarchically clustered techniques: building an international consensus for the reporting of behavior change
interventions. Ann Behav Med 2013 Aug;46(1):81-95. [doi: 10.1007/s12160-013-9486-6] [Medline: 23512568]
49. Ware J, Kosinski M, Keller SD. A 12-Item Short-Form Health Survey: construction of scales and preliminary tests of
reliability and validity. Med Care 1996 Mar;34(3):220-233. [doi: 10.2307/3766749] [Medline: 8628042]
50. Lippke S, Schwarzer R, Ziegelmann JP, Scholz U, Schüz B. Testing stage-specific effects of a stage-matched intervention:
a randomized controlled trial targeting physical exercise and its predictors. Health Educ Behav 2010 Aug;37(4):533-546.
[doi: 10.1177/1090198109359386] [Medline: 20547760]
51. Scholz U, Keller R, Perren S. Predicting behavioral intentions and physical exercise: a test of the health action process
approach at the intrapersonal level. Health Psychol 2009 Nov;28(6):702-708. [doi: 10.1037/a0016088] [Medline: 19916638]
52. Schwarzer R, Lippke S, Luszczynska A. Mechanisms of health behavior change in persons with chronic illness or disability:
the Health Action Process Approach (HAPA). Rehabil Psychol 2011 Aug;56(3):161-170. [doi: 10.1037/a0024509] [Medline:
21767036]
53. Venkatesh V, Morris MG, Davis GB, Davis FD. User acceptance of information technology: toward a unified view. MIS
Q 2003;27(3):425-478. [doi: 10.2307/30036540]
54. Venkatesh V, Thong J, Xu X. Consumer acceptance and use of information technology: extending the unified theory of
acceptance and use of technology. MIS Q 2012;36(1):157-178. [doi: 10.2307/41410412]
55. Bickmore TW, Mitchell SE, Jack BW, Paasche-Orlow MK, Pfeifer LM, Odonnell J. Response to a relational agent by
hospital patients with depressive symptoms. Interact Comput 2010 Jul 01;22(4):289-298 [FREE Full text] [doi:
10.1016/j.intcom.2009.12.001] [Medline: 20628581]
56. Spittaels H, Verloigne M, Gidlow C, Gloanec J, Titze S, Foster C, et al. Measuring physical activity-related environmental
factors: reliability and predictive validity of the European environmental questionnaire ALPHA. Int J Behav Nutr Phys Act
2010 May 26;7(1):48 [FREE Full text] [doi: 10.1186/1479-5868-7-48] [Medline: 20504339]
57. Rammstedt B, John OP. Measuring personality in one minute or less: a 10-item short version of the Big Five Inventory in
English and German. J Res Pers 2007 Feb;41(1):203-212. [doi: 10.1016/j.jrp.2006.02.001]
58. Künzler F, Kramer JN, Kowatsch T. Efficacy of mobile context-aware notification management systems: A systematic
literature review and meta-analysis. 2017 Presented at: 13th International Conference on Wireless and Mobile Computing,
Networking and Communications (WiMob); October 09-11, 2017; Rome, Italy.
59. Montgomery AA, Peters TJ, Little P. Design, analysis and presentation of factorial randomised controlled trials. BMC Med
Res Methodol 2003 Nov 24;3:26 [FREE Full text] [doi: 10.1186/1471-2288-3-26] [Medline: 14633287]
60. Kowatsch T, Kramer JN, Kehr F, Wahle F, Elser N, Fleisch E. Effects of charitable versus monetary incentives on the
acceptance of and adherence to a pedometer-based health intervention: study protocol and baseline characteristics of a
cluster-randomized controlled trial. JMIR Res Protoc 2016 Sep 13;5(3):e181 [FREE Full text] [doi: 10.2196/resprot.6089]
[Medline: 27624645]
61. Eisler R, Lüber A. Comparis. 2006. [How important is Swiss supplemental hospital insurance?] URL: https://www.
comparis.ch/~/media/files/mediencorner/studies/2006/krankenkassen/spitalzusatzversicherungen_studie.pdf [accessed
2018-11-30] [WebCite Cache ID 74Jn7Pq0k]
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.15http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
62. Boruvka A, Almirall D, Witkiewitz K, Murphy SA. Assessing time-varying causal effect moderation in mobile health. J
Am Stat Assoc 2017 Mar 31;113(523):1112-1121. [doi: 10.1080/01621459.2017.1305274]
63. Sequeira MM, Rickenbach M, Wietlisbach V, Tullen B, Schutz Y. Physical activity assessment using a pedometer and its
comparison with a questionnaire in a large population survey. Am J Epidemiol 1995 Nov 01;142(9):989-999. [doi:
10.1093/oxfordjournals.aje.a117748] [Medline: 7572981]
64. Bassett DR, Wyatt HR, Thompson H, Peters JC, Hill JO. Pedometer-measured physical activity and health behaviors in
US adults. Med Sci Sports Exerc 2010 Oct;42(10):1819-1825 [FREE Full text] [doi: 10.1249/MSS.0b013e3181dc2e54]
[Medline: 20305579]
65. Holm S. A simple sequentially rejective multiple test procedure. Scand Stat Theory Appl 1979;6(2):65-70. [doi:
10.2307/4615733]
66. Safran Naimark J, Madar Z, Shahar DR. The impact of a web-based app (eBalance) in promoting healthy lifestyles:
randomized controlled trial. J Med Internet Res 2015 Mar 02;17(3):e56 [FREE Full text] [doi: 10.2196/jmir.3682] [Medline:
25732936]
67. Glynn LG, Hayes PS, Casey M, Glynn F, Alvarez-Iglesias A, Newell J, et al. Effectiveness of a smartphone application to
promote physical activity in primary care: the SMART MOVE randomised controlled trial. Br J Gen Pract 2014
Jul;64(624):e384-e391 [FREE Full text] [doi: 10.3399/bjgp14X680461] [Medline: 24982490]
68. Craig CL, Marshall AL, Sjöström M, Bauman AE, Booth ML, Ainsworth BE, et al. International physical activity
questionnaire: 12-country reliability and validity. Med Sci Sports Exerc 2003 Aug;35(8):1381-1395. [doi:
10.1249/01.MSS.0000078924.61453.FB] [Medline: 12900694]
69. Modena BD, Bellahsen O, Nikzad N, Chieh A, Parikh N, Dufek DM, et al. Advanced and accurate mobile health tracking
devices record new cardiac vital signs. Hypertension 2018 Aug;72(2):503-510. [doi:
10.1161/HYPERTENSIONAHA.118.11177] [Medline: 29967036]
70. Duncan MJ, Wunderlich K, Zhao Y, Faulkner G. Walk this way: validity evidence of iphone health application step count
in laboratory and free-living conditions. J Sports Sci 2018 Aug;36(15):1695-1704. [doi: 10.1080/02640414.2017.1409855]
[Medline: 29179653]
Abbreviations
Ally: Assistant to Lift your Level of activitY
ANOVA: analysis of variance
AP: action planning
BREQ-2: Behavioral Regulation for Exercise Questionnaire-2
CP: coping planning
CC: control condition (no planning)
GEE: generalized estimating equation
JITAI: just-in-time adaptive intervention
mHealth: mobile health
MRT: microrandomized trial
p (SG): probability of reaching the step goal
SIMS: Situational Motivation Scale
Edited by N Kuter, G Eysenbach; submitted 11.07.18; peer-reviewed by S Thurnheer, S Ferguson, TT Luk; comments to author
27.08.18; revised version received 26.10.18; accepted 31.10.18; published 31.01.19
Please cite as:
Kramer JN, Künzler F, Mishra V, Presset B, Kotz D, Smith S, Scholz U, Kowatsch T
Investigating Intervention Components and Exploring States of Receptivity for a Smartphone App to Promote Physical Activity:
Protocol of a Microrandomized Trial
JMIR Res Protoc 2019;8(1):e11540
URL: http://www.researchprotocols.org/2019/1/e11540/
doi:10.2196/11540
PMID:
©Jan-Niklas Kramer, Florian Künzler, Varun Mishra, Bastien Presset, David Kotz, Shawna Smith, Urte Scholz, Tobias Kowatsch.
Originally published in JMIR Research Protocols (http://www.researchprotocols.org), 31.01.2019. This is an open-access article
distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which
permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.16http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
Research Protocols, is properly cited. The complete bibliographic information, a link to the original publication on
http://www.researchprotocols.org, as well as this copyright and license information must be included.
JMIR Res Protoc 2019 | vol. 8 | iss. 1 | e11540 | p.17http://www.researchprotocols.org/2019/1/e11540/
(page number not for citation purposes)
Kramer et alJMIR RESEARCH PROTOCOLS
XSL
•
FO
RenderX
