Download full-text PDF

Providing Patient Context to Mental Health Professionals Using Mobile Applications

Conference Paper (PDF Available)  · January 2016with48 Reads
DOI: 10.13140/RG.2.1.3793.1923
Conference: Conference: CHI workshop on Computing and Mental Health
Niels van Berkel at University of Melbourne
  • 9.34
  • University of Melbourne
Tony Durkee at Karolinska Institutet
  • 22.27
  • Karolinska Institutet
Vassilis Kostakos at University of Melbourne
  • 30.88
  • University of Melbourne
Providing Patient Context to Mental Health
Professionals Using Mobile Applications
Niels van Berkel1, Simo Hosio1, Tony Durkee2,
Vladimir Carli2, Danuta Wasserman2, Vassilis Kostakos1
1 Center for Ubiquitous Computing, University of Oulu, Finland
2 National Centre for Suicide Research and Prevention of Mental Ill-Health (NASP),
Karolinska Institutet, Stockholm, Sweden
{niels.van.berkel, simo.hosio, vassilis}
{tony.durkee, vladimir.carli, danuta.wasserman}
The quantified-self movement entails self-tracking of
physical activity, often using wearable devices and mobile
applications. In parallel, mobile applications focusing on
mental health are increasingly popular, and they often rely
on active user input to track the user progress and to deliver
feedback and motivation. In this paper we discuss the
potential benefits of bridging these two distinct yet highly
relevant application domains. We argue for the benefits of
combining explicit (user-provided) and implicit (device-
collected) data sources in the context of mental health care.
We argue that this combination allows for improved
methods to observe patients’ lives, and thus provide a more
in-depth overview of their progress. This may enable
mental health professionals to establish more personalised
and adaptive care plans.
Author Keywords
Quantified self; self-care; care trajectory; smartphone;
wearable devices; ESM; EMA; depression; well-being;
mental health
ACM Classification Keywords
H.5.m. Information interfaces and presentation (e.g., HCI):
Mobile applications aiming to improve users’ mental health
are increasingly popular. These applications claim how easy
improving your mental health is by using slogans such as
In just 12 weeks, you can lead the life you deserve[7], or
86% of frequent users get happier in 2 months[15].
These applications typically follow the trend of self-
cultivation, i.e. educating oneself on personal
characteristics and habits to realise personal development.
By collecting data on day-to-day activities, it is possible to
achieve these insights into one’s life.
Self-cultivation applications targeting mental health contain
large similarities in both their approach and objective
compared to the quantified-self (QS) movement. While the
most popular QS applications seemingly focus on physical
activity tracking, there are applications also tailored for
collecting data and providing insights on other areas of life,
such as finances or calorie intake. With an estimated 43.6
million adults facing mental problems in 2014 in the U.S.
alone [4], it is important to explore the possibilities of these
applications also in mental health care.
In contrast to the well-defined tracking objectives in QS
(e.g., distance travelled by bike, personal expenses), mental
health progress can be more abstract and challenging to
measure reliably [8]. Most popular mobile applications for
depression and mental health focus on self-reporting (e.g.,
[7]), but do not fully consider rich sensor data to
complement the self-reports. Doing so could enable a range
of possibilities for treatment and intervention, but also
introduces the challenge of determining which sensor data
to collect and which questions to explicitly ask patients.
In this paper we describe our design of a mobile application
that uses the combination of rich sensor data together with
self-reports from mental health patients to provide a
comprehensive situational awareness for mental health
professionals, e.g. psychotherapist, psychologist,
counsellor. In many countries (including our own), such
patients are typically under continuous treatment of a
qualified mental health professional, with whom they
consult on a regular basis. We discuss how our prototype
can help mental health professionals to augment the
patient's recovery process. We have not begun actual trials
with patients yet in this context, and hope to now raise
discussion especially on the potential limitations, long term
pitfalls, and possibilities of our approach.
A rich literature from various disciplines investigates the
correlation between a person’s context and mental health.
This work has revealed a variety of factors, both on an
Paste the appro priate c opyrigh t/license statement here. ACM now suppo rts
three different publication options:
ACM copyright: ACM holds the copyright on the work. This is the
historical approach.
License: The author(s) retain copyright, but ACM receives an
exclusive publication license.
Open Access: The author(s) wish to pay for the work to be open
access. The additional fee must be paid to ACM.
This text field is large enough to hold the appropriate release statement
assuming it is single-spaced in Times New Roman 8-point font. Please do
not change or modify the size of this text box.
Each submission will be assigned a DOI string to be included here.
individual and community level. Individuals growing up in
‘risky’ families, characterised as aggressive and/or conflict-
prone and with low levels of support, develop a different
bio-behavioural profile (resulting, for example, in
problematic emotional processing and increased likelihood
of substance abuse) [13]. Also, depression severity has been
shown to correlate strongly with that of friends and
neighbours [14], extending up to three degrees of separation
(friend of a friend’s friend). Depression has also been
shown to effect interpersonal communication [17], resulting
in lower levels of personal engagement, and experienced
levels of discomfort during human-to-human interaction.
Diagnosis and analysis of depression through social
network usage characteristics has recently attracted more
attention. De Choudhury et al. [5] show that various aspects
of an individuals’ Twitter activity (e.g., raise in negative
affect, decrease in social activity) can characterise the onset
of depression. Other work has investigated the use of social
media to track multiple ailments and symptoms, including
allergies, depression, and obesity [11]. As such, social
media does not only allow for analysis of what is happening
at this instance in time, but can also be used in the form of a
digital diary to provide insight on past events.
Using mobile sensor data to support mental healthcare has
been explored in various applications. Burns et al. [1]
explore the use of mobile sensors to develop a context-
aware system named Mobilyze!. This mobile application is
able to detect when users need assistance and can directly
offer support. Combining concurrent sensor values and
machine learning techniques, the application achieves a
60% to 91% accuracy rate in predicting various categorical
contextual states (e.g., user activity, location). However,
states rated on scales (e.g., mood, emotions) did not attain
high accuracy. Nevertheless, following usage of the
application, the mental health of participants significantly
improved. Ma et al. [10] developed MoodMiner, a mobile
application aimed to collect a real-time, daily mood
assessment using various sensors of the mobile phone
including hardware sensors (e.g., GPS, accelerometer) and
communication channels (e.g., SMS, phone calls). Over a
30-day experiment period (N=15), the researchers are able
to achieve a 50% accuracy over the three recorded
dimensions (displeasure, tiredness, and tensity) compared to
user self-reports. Similar to [1], Ma et al. state that “people
show significant difference in daily behavior style and use
pattern of mobile phone, making it difficult to build a model
working well for anyone[10].
In collaboration with mental healthcare professionals, we
have designed and developed the first version of our
prototype as a plugin for the AWARE framework [6].
AWARE is a mobile sensing platform that makes collecting
raw sensor data from Android devices easy, and the data
can also be synced to a server in near real-time. Our goal is
to allow for the simultaneous collection of self-reported
mental states and sensor-collected data on Android devices.
We propose using machine learning to uncover relations in
the data. As an example, consider the combination of
location data and self-reported mental states, that together
indicate a change in the patient’s mental state at certain
locations (similar approach as [3]). Interpretation of these
classification results makes sense in the treatment process,
and we plan to make this type of insight available to the
mental health professional. It is then the professional’s task
to highlight and discuss these observations with the patient.
The technology required to achieve this is already available
and, based on previous work [1], we are exploring to
combine self-reports with the following sensors for further
higher-level insights: location, ambient light,
accelerometer, mobile application usage, and physical
activity. Naturally, creating classifiers and refining data per
each sensor type requires rigorous analysis and justification.
Information presentation
Healthcare professionals typically do not have the
experience required to interpret mobile sensor data.
Classifying and visualising data in a meaningful manner is
challenging, despite the recent advances in making machine
learning easier for the masses using cloud-based
approaches. Furthermore, professionals are required to link
the emerging insights to the patients’ overall situation. This
requires a thorough understanding of both the collected
sensor data and the patient’s mental states. To facilitate
professionals in this challenging task, we have implemented
a dashboard that allows the professional to configure and
visualise the various parameters of the patient’s (mobile)
context. The dashboard is a component of AWARE [6].
Personalised data collection
The treatment plan of the patient should be continuously re-
evaluated and updated to maximise its efficiency [1]. This
is usually achieved following a consultation session
between patient and clinician, or between clinicians alone.
This process of continuous interplay is presented in Figure
1. The model follows a repeating process that fits well
within the existing mental healthcare process, where
patients regularly visit their counsellor. Information
obtained from both the collected sensor data and the data
submitted by patients will allow the mental health
professional to either bring up or more accurately discuss
relevant events, thoughts, and experiences of the past. In
turn, the mental health professional can recommend
adjustments to the data collection configuration, to obtain
more relevant data for future consultation needs. These
consultations help to determine the next steps in the
patient’s treatment, and can therefore possibly result in a
change of the currently relevant metrics in the patients’
treatment process.
Figure 1. Model of mobile context in mental healthcare
process with mental health professional
Issuing the Right Question at the Right Time
The answers provided by the application users are key
sources of information regarding their mental state. It is
therefore vital that the questions issued by the mobile
application are adequate and carefully considered. In our
prototype, we use the PHQ-9 Patient Health Questionnaire
[8], a validated scale consisting of nine question items
therefore allowing for relatively fast user input. PHQ-9 is
used to track developments in the patients’ mental state
over longer periods of time. Since we rely on explicit input
for the collection of this data, both the quality and quantity
(response rate) of the answers depend on the users of the
application. We discuss two elements of user-generated
data collection that are of high importance, namely timing
and content of the questions.
Originally we considered two main approaches to collecting
user-provided data. The first approach, active probing,
consists of notifying the user throughout the day to provide
information. In this form, data collection is based on the
Experience Sampling Method (ESM) [9] or similar
methodologies. In the second approach, no probing takes
place and users are required to proactively open the
application and input information. We have chosen to
utilise active probing, as it (in our case the ESM), allows us
to obtain […] reports about people’s experience as it
occurs, thereby minimizing the effects of reliance on
memory and reconstruction. [9]. With active probing,
asking questions at the right moment is crucial. Failing to
do so can result in a reduced usefulness of the answer (out
of context), no answer provided by the user, and increased
level of user annoyance.
Context sensing can be used to not only passively infer in
which context an answer was provided, or determine the
timing of a question, but can also actively steer the question
content. This way, the questions are in line with the current
activity and/or location of the patient. For example, the
mental health professional and patient might agree that
increased physical activity would be beneficial to the
patient. Based on the detected activity context, the mobile
application is then able to ask questions predetermined to
follow the physical activity. Furthermore, by storing user
answers the number of questions asked to the patient can be
reduced. For example, after asking the user to assign a label
to the current location (e.g., home, work), the location can
be saved to the device and used to steer further questions.
Health professionals, including those working in mental
health, benefit from a complete picture on the daily life and
progress of their patients. Having such understanding has
multiple advantages to the treatment process, including
tailoring of the treatment plan, identifying potential barriers
to treatments, patient compliance, as well as the ability to
set goals that are appropriate to the patient’s current
(mental) state [1].
Since consultations in our environment usually follow a
regular schedule (e.g., weekly, bi-weekly, monthly), the
picture that professionals can form regarding their patients’
lives can potentially be improved using a more real-time
approach: by leveraging the patients’ personal mobile
devices as the proverbial right hand of the mental health
professional. Employing the context obtained through the
users’ mobile phone, it becomes feasible to obtain a much
more accurate picture of patients’ lives.
The widespread proliferation of mobile phones allows for
new application possibilities that serve both individuals and
the society. The popularity of mental self-care applications
is just one example. Particularly in the mental health
domain, we see potential in using mobile phones to collect
data about users, without the need of constant appointments
with clinicians. Further, phones can simultaneously collect
other data about patients’ lives, to provide new type of
information to the care personnel. Using a reflective
method (Figure 1), mental healthcare professionals are able
to adjust data collection as they see appropriate, and thus
benefit the patient.
Various applications offer the user notes of encouragement
throughout the day (e.g. Joyable [7]). Though, as stated by
Calvo et al., much is yet to be discovered about how and
when to provide mental health support [2]. Both the
already existing self-care applications and the prototyped
mental care trajectory application bear strong parallels to
the QS movement. We should therefore be wary to fall for
the same mistakes that threaten (long-term) usage of QS
applications [16]. Figure 2 shows the different stages of a
typical QS process. The process starts with an initial
question and, ultimately, results in a certain action by the
user. To accomplish this, the application should be able to
construct a data cycle that leads from a user need to an
insight provided to the user. As noted by Van Berkel et al.,
a QS-tool needs to adapt to the users’ dynamics of
Figure 2. Stages of QS [16].
In the case of typical QS applications, these users’
dynamics of questions are derived from a single user. By
involving the mental healthcare professional, as proposed in
this paper, the various stages in the QS process become
shared between stakeholders. For example, the patient and
clinician may come up with a certain question in
consultation but analysis of the collected data is
performed by the clinician. User needs are therefore derived
directly from the consultation between patient and clinician.
Following the popularity of mental self-care applications in
the mobile market, we discuss the opportunities for mobile
applications in the professional mental healthcare process.
Through the use of context sensing, the mental healthcare
professional is able to construct a better-rounded overview
on the patient’s progress. We have developed a mobile
prototype application that can be used to collect this
information, present it to a professional, and allow the
professional to adjust the parameters of data collection. We
stress the integration in the patient’s treatment plan as a key
to success. By integrating application usage with the
ongoing treatment, the mental health professional can
continuously (re-)configure the area of relevance for the
application (Figure 1). This way, a continuous relevance of
the collected data to the treatment process is ensured.
This work is partially funded by the Academy of Finland
(Grants 276786-AWARE, 285062-iCYCLE, 286386-
CPDSS, 285459-iSCIENCE), and the European
Commission (Grants PCIG11-GA-2012-322138 and
1. Michelle N. Burns, Mark Begale, Jennifer Duffecy,
Darren Gergle, Chris J. Karr, Emily Giangrande and
David C. Mohr. 2011. Harnessing Context Sensing to
Develop a Mobile Intervention for Depression. J.
Medical Internet Research 13, 3.
2. Rafael A. Calvo, Rosalind Picard, Karthik Dinakar and
Pattie Maes. 2016. Computing in Mental Health. In
CHI ’16 Extended Abstracts.
3. Luca Canzian and Mirco Musolesi. 2015. Trajectories
of Depression: Unobtrusive Monitoring of Depressive
States by means of Smartphone Mobility Traces
Analysis. In Proc. UbiComp 2015, 1293-1304.
4. Center for Behavioral Health Statistics and Quality.
2015. Behavioral Health Trends in the United States:
Results from the 2014 National Survey on Drug Use
and Health.
5. Munmun De Choudhury, Scott Counts, Eric Horvitz
and Michael Gamon. 2013. Predicting Depression via
Social Media. In ICWSM '13.
6. Denzil Ferreira, Vassilis Kostakos and Anind K. Dey.
2015. AWARE: mobile context instrumentation
framework. Frontiers in ICT 2, 6: 1-9.
7. Joyable | Outcomes. Retrieved 06/01/2016 from
8. K Kroenke, R. L. Spitzer and J. B. Williams. 2001. The
PHQ-9: validity of a brief depression severity measure.
J. Gen Intern Med 16, 9: 606-13.
9. Reed Larson and Mihaly Csikszentmihalyi. 1983. The
Experience Sampling Method. In Flow and the
Foundations of Positive Psychology (eds.). Wiley
Jossey-Bass, San Francisco, 15, 41-56.
10. Yuanchao Ma, Bin Xu, Yin Bai, Guodong Sun and
Run Zhu. 2012. Daily Mood Assessment Based on
Mobile Phone Sensing. In BSN '12, IEEE, 142-147.
11. Michael J. Paul and Mark Dredze. 2011. You Are
What You Tweet: Analyzing Twitter for Public Health.
In ICWSM '11, 265 - 272.
13. Rena L. Repetti, Shelley E. Taylor and Teresa E.
Seeman. 2002. Risky Families: Family Social
Environments and the Mental and Physical Health of
Offspring. Psychological Bulletin 128, 2: 330-366.
14. J.N. Rosenquist, J.H. Fowler and N.A. Christakis.
2011. Social network determinants of depression.
Molecular psychiatry 16, 3.
15. Science-Based Happiness Games & Activities.
Retrieved 21/12/2015 from
16. Niels van Berkel, Chu Luo, Denzil Ferreira, Jorge
Goncalves and Vassilis Kostakos. 2015. The Curse of
Quantified-Self: An Endless Quest for Answers. In Adj.
Proc. UbiComp '15, 973-978.
17. Mary A. Youngren and Peter M. Lewinsohn. 1980. The
functional relation between depression and problematic
interpersonal behavior. J. Abnormal Psychology 89, 3:
  • ... This requires additional software implementation, and is highly dependent on the nature of the collected data. We give the example of participants' own self-reports being used in discussion with their professional caregiver, providing a direct benefit to data collection (van Berkel et al. 2016a). Another example is found in the visualisation of self-reports-allowing participants to measure progress and answer their personal (quantified-self) questions (van Berkel et al. 2015). ...
Aim is to quantify how users interact with their devices. Currently creating a software to understand notification interactions and manage arriving notifications.
The objective of this project was to develop a standardized multidisciplinary framework for experimentally evaluating real-world ubicomp systems with increased external validity. The realization of…" [more]
Full-text available
September 2017
    The Experience Sampling Method is used to capture high-quality in situ data from study participants. This method has become popular in studies involving smartphones, where it is often adapted to motivate participation through the use of gamification techniques. However, no work to date has evaluated whether gamification actually affects the quality and quantity of data collected through... [Show full abstract]
    Conference Paper
    Full-text available
    September 2016
      Mobile instrumentation provides researchers and professionals the opportunity to collect data on several aspects of human life. In this paper we discuss our initial experiences on collecting data via mobile instrumentation in an elementary school. We augmented a classroom with mobile phones and Bluetooth beacons to capture student experiences as well as their relative distance to each other... [Show full abstract]
      Conference Paper
      Full-text available
      April 2018
        The Internet offers plenty of options for those who want to lose weight. Choosing among the practically unlimited number of weight loss diets, exercises, and pills, advertised as borderline magical, is however challenging. We present The Diet Explorer, a crowd-powered, knowledge base that can be queried in real-time to discover weight loss diets that best match personal needs. Our long-term... [Show full abstract]
        Discover more