The Use of Technology in Suicide Prevention*
Mark E. Larsen, Nicholas Cummins, Tjeerd W. Boonstra, Bridianne O’Dea, Joe Tighe, Jennifer
Nicholas, Fiona Shand, Julien Epps, and Helen Christensen

Abstract— Suicide is one of the leading causes of death
globally, and is notably a significant cause of death amongst
young people. A suicide outcome is a complex combination of
personal, social, and health factors, and therefore suicide
prevention is a challenge, requiring a systems approach
incorporating public health strategies, screening at-risk
individuals, targeted interventions, and follow-up for suicide
survivors and those bereaved by suicide. Engineering practice
has been implicated in the hindrance of the adoption of suicide
prevention strategies, such as installing safety barriers at the
Golden Gate Bridge, however technological developments offer
new opportunities in suicide prevention, and the potential to
reduce the number of deaths by suicide.
We present an overview of current technological
developments which are facilitating research in the field of
suicide prevention, including multiple modes of screening such
as network analysis of mobile-phone collected connectivity
data, automatic detection of suicidality from social media
content, and crisis detection from acoustic variability in speech
patterns. The current field of mhealth apps for suicide
prevention is assessed, and an innovative app for an Indigenous
population is presented. From this overview, future challenges
– technical and ethical – are discussed.
I. INTRODUCTION
Suicide is a leading cause of death globally [1], and in
Australia accounts for 2,500 deaths per year – approximately
the same number as breast cancer or skin cancer [2]. Notably,
it is the leading cause of death among Australians aged 15–
44. There is no single causative pathway to a suicide
outcome, rather it is the result of a complex combination of
personal, social and health factors. Suicide prevention is
therefore a challenge, requiring a systems approach
incorporating public health strategies, screening at-risk
individuals, targeted interventions, and follow-up for suicide
survivors and those bereaved by suicide. There have been
instances where it has been suggested that engineering
practice has hindered the adoption of suicide prevention
* Research supported by the NHMRC Centre of Research Excellence in
Suicide Prevention APP1042580, NHMRC John Cade Fellowship
APP1056964, NHMRC Project Grant APP1060477, AFFIRM, SMHR,
NSW Mental Health Commission, Department of Health and Ageing,
Department of Health WA, Young and Well Cooperative Research Centre,
Department of Broadband, Communication and the Digital Economy, ARC
ICT Centre of Excellence program, and ARC Discovery Projects
DP110105240 and DP120100641.
M. E. Larsen, T. W. Boonstra, B. O’Dea, J. Tighe, J. Nicholas, F. Shand
and H. Christensen are with the Black Dog Institute, University of New
South Wales, Sydney, NSW 2031, Australia. T. W. Boonstra is also with
MOVE Research Institute, VU University, Amsterdam, The Netherlands.
Email: {mark.larsen, b.odea, j.tighe, j.nicholas,
h.christensen}@blackdog.org.au, {t.boonstra, fionas}@unsw.edu.au.
N. Cummins and J. Epps are with the School of Electrical Engineering
and Telecommunications, University of New South Wales, Sydney, NSW
2031, Australia and the ATP Research Laboratory, National ICT Australia
(NICTA). Email: n.p.cummins@unsw.edu.au, j.epps@unsw.edu.au.
978-1-4244-9270-1/15/$31.00 ©2015 IEEE
strategies, such as installing safety barriers on the Golden
Gate Bridge [3], however technological advancements offer
new opportunities in suicide prevention, with the potential to
reduce the number of deaths by suicide.
A recent search of the IEEE Xplore database for suicide-
related journal and magazine articles identified 26 potentially
relevant articles. Seven of these reported or mentioned cases
where an individual had died by suicide, four mentioned
suicide bombing, one described an image analysis technique
to detect completed suicides, and six articles discussed
suicide in relation to other mental health topics (linking
surveillance to suicide, assisted suicide, cyber-bullying, the
history of electroconvulsive therapy, forecasting depression
in bipolar disorder, and detecting symptoms of depression
from medical records). Only two articles were closely related
to the prevention of suicide, both detecting depression or
crisis from speech [4, 5]. Although there is a relative lack of
engineering research on suicide prevention, the IEEE EMBS
code of ethics calls for us to “engage in research aimed at
advancing the contribution of science and technology to
improving healthcare provision” [6]. We therefore present an
overview of some of the current technological developments
which are occurring in the field of suicide prevention. Studies
involving human participants were approved by the
institutional ethics committee.
II. SCREENING FOR SUICIDALITY
Detecting the presence of suicide ideation or behaviours
is a challenge, even for physicians [7]. The potential to aid in
the detection of a suicidal crisis or the onset of a pre-crisis
state is an important outcome, as it has been estimated that up
to 66% of people who die by suicide have contact with a
primary health service in the month prior to their death, but
their suicidality is often undetected [8]. The increasing
ubiquity of sensor-enabled smartphones and the use of online
social media platforms offer the potential for passive,
background data collection, and automatic detection of
suicidality and its associated risk factors.
A. Detecting Social Withdrawal with Smartphones
Social withdrawal is a key warning sign for episodes of
depression and suicidal behaviour [9], and one which may be
detectable by other members of an individual’s social
network. However, the emergence of this behaviour is not
always obvious. Passive monitoring using smartphones
provides a tool that can map social networks and detect such
withdrawal.
Several studies have used proximity or location-based
features as a proxy for social ties [10]. Whereas these
previous studies have used specific study-provided handsets,
which limits the scalability of such a system, researchers at
the Black Dog Institute have developed a custom app for
7316
Android and iPhones that participants installed on their own
handset. The app performed a periodic Bluetooth discovery
scan to detect other nearby users, and was tested in a pilot
study with staff and students within the organisation. Based
on the scanning data, patterns in the connectivity between
individual participants in close proximity can be observed
(Fig. 1). A social network can then be defined by combining
the scanning statistics of the participants. The weights of the
network edges were determined from the number of times
each pair of devices observed one another (Fig. 2a), which in
turn define the network topology (Fig. 2b).
Figure 1 – Connectivity between one participant and other nearby
devices, over one week. The colour map represents the number of times
the Bluetooth devices were in proximity during a one-hour window.
Figure 2 – (a) Connectivity matrix showing the strength of each
network edge, and (b) the resulting topography of the social network.
Previous studies have established the connection between
social connectivity and mental health, indicating that people
with depressive symptoms tend to group together [11] and
that adolescent girls with depressive symptoms have fewer
social connections [12]. Therefore, tracking how a social
network changes over time and detecting changes in social
connectivity may provide important information about
emerging episodes of depression and suicidal behaviour. The
ability to map changes in social networks using participants’
own smartphones offers a non-intrusive, platform-
independent, and scalable method to investigate the effect of
changes in social interactions on mental health.
B. Detecting Suicidality on Twitter
In addition to the correlations with real-world social
networks, mental health is also associated with the patterns of
usage and behaviour within online social networks [13, 14].
Online social networking platforms, particularly social media
sites, have been used to detect the presence of a suicide crisis,
as evidenced by the recent attempts to deploy detection tools
for use in Twitter [15] and Facebook [16]. However, due to
the large volume of content shared in these sites, the ability to
7317
accurately and automatically detect social media posts that
are indicative of genuine risk is of great interest.
Machine learning algorithms for the automatic detection
of strongly concerning suicidal tweets have been developed
[17]. In this study, 14,701 suicide-related tweets were
acquired, 2,000 of which were selected for manual labelling
as either “strongly concerning”, “possibly concerning”, or
“safe to ignore”. A total of 14% were considered strongly
concerning. These labelled tweets were then applied to a
Support Vector Machine, which demonstrated an accuracy of
80% when applied to a sample of unseen “strongly
concerning” tweets – an accuracy equivalent to that of the
inter-rater agreement among the human coders in the manual
labelling process. This demonstrates the feasibility of
establishing automated detection systems for concerning
suicidal content posted on Twitter and has the potential to be
adapted for other social media sites.
C. Crisis Detection from Speech Patterns
Speech can be a key marker for suicide crisis detection.
The linguistic content of speech is used by clinicians to aid
crisis detection, however gaining reliable linguistic
information from a patient who is in crisis can be difficult.
The frankness and honesty of an individual’s response
depends on their desire and ability to communicate at a time
when their outlook and motivation may be impaired [18, 19].
Paralinguistics are the vocal phenomena that occur
alongside the linguistic information, and these non-linguistic
(and hence language-independent) cues can be used as an
objective marker in crisis detection [20]. Paralinguistic
features include biological traits (such as ethnicity, sex, age
and height) and social signals (such as conversational sounds
and intimacy). Of particular interest for suicide crisis
detection is that paralinguistic analysis has been shown to be
useful in objectively monitoring emotional [21] and mental
states [20, 22].
Paralinguistic analysis may be particularly well suited to
implementation in mobile devices or call centres, in view of
the fact that 80% of suicide attempts involve some form of
prior warning, and people who call crisis centres are around
30 times more likely than average to attempt suicide [23].
Speech production is the output of a complex cognitive
and muscular system, and slight physiological and cognitive
changes, which can be expected due to the intense changes in
affective state during a crisis, can produce noticeable acoustic
changes. These changes alter speech production systems in
such a way that is measurable using prosodic, articulatory,
and acoustic speech features. For example, reduced acoustic
variability has been found to be significantly linked to the
presence of depression (Fig. 3) [24].
A range of prosodic alterations have been linked with
individuals in a suicidal crisis, including unusual and
unexpected vocal patterns and audible changes in acoustic
properties [25]. Suicidality has also been linked with shifts in
spectral energy, either from lower to higher frequencies [4,
5], or the reverse [26]. Recently statistically significant
differences have been reported between voice quality features
obtained from the speech of either suicidal adolescent voices
or matched controls [27].
Figure 3 – A comparison of spectrograms showing four repetitions of
the phrase ‘PATAKA’ for an individual in a normal mood state (top)
and very severe depressed state (bottom). Low energy regions are
darker, high energy regions are lighter. Notice the lack of sustained
energy in the 1kHz and 3kHz regions in the very severe example.
Figure reproduced from [24].
Research into links between changes in suicide risk and
changes in paralinguistic cues is still in its infancy. Many
results have been found in small datasets recorded under non-
ideal conditions, raising issues relating to their validity and
generalisability. However they show future speech research
can potentially yield a simple, low cost, automated, and
objective crisis detection aid.
III. INTERVENTIONS
Following the identification of a potentially suicidal
individual, effective interventions are required in order to
manage the crisis situation. The screening methods identified
above are still in development and have not yet been
incorporated into an intervention, however, they are all well-
suited to incorporation into mobile applications (apps). Apps
provide a flexible platform from which a variety of
intervention strategies can be employed – including public
health and mental health techniques.
A. Review of Mobile Apps
Mobile apps are a popular way of accessing health
information and resources, despite a lack of evidence for
publicly available apps across a spectrum of mental health
disorders [28]. As there is no clear link between research and
public implementation, we have conducted a review to assess
whether existing apps for suicide prevention are consistent
with established evidence-based practice.
The review identified 123 apps from the iTunes and
Google Play app stores which were related to suicide. Of
these 123 apps, 12 (9.8%) contained potentially harmful
content which may encourage self-harming or suicidal
behaviours. Exhaustive testing of the apps was not part of the
review process, however 31 (25.2%) of the apps exhibited
obvious bugs or reliability issues, excluding those related to
general usability. Despite the potential for apps to provide an
interactive experience for users and to harness data from
various sensors, only 54 (43.9%) of the apps offered any
interactive features beyond the passive consumption of, or
navigation through, content. Mapping individual components
of these apps to the evidence base is currently underway.
7318
B. ibobbly
Although there is a general lack of evidence-based,
interactive apps currently publicly available, this situation is
changing. One example of an innovative suicide prevention
app is ibobbly, which was designed to reduce the suicide rate
amongst young Aboriginal Australians [29]. This is a
particularly high-risk group, in which the suicide rate is four
times higher than the non-Indigenous population, and one
which has a high rate of adoption of smartphones and social
media use [30].
The ibobbly app was designed with extensive
consultation with the local community, and in a culturally-
relevant manner. The app is based on proven Acceptance and
Commitment Therapy, but presented in a highly graphical
form (see Fig. 4) with artwork by local Aboriginal artists. It
also incorporates audio playback of the text components, to
aid accessibility in a population with various levels of
English literacy. Self-reported data were collected through
the app using custom measures and standard
psychometrically validated measures of depression, distress,
suicidality, and impulsiveness. The ibobbly app has been
piloted in a randomised controlled trial in eight communities
in the Kimberley region of Western Australia and the results
will be available shortly.
Figure 4 – Screenshot from one of the ibobbly modules.
IV. DISCUSSION
The field of suicide prevention is a relatively new
application for the engineering community. Nevertheless,
advancements are being made in multiple modes of
screening individuals who may be at risk, and the detection
of possible suicidal crisis – we have described detection
through passive sensor data collection, analysis of social
media content, and paralinguistic processing of speech.
Mobile apps are being used as a convenient platform for the
delivery of suicide prevention programs, and will be
enhanced by the incorporation of these novel screening
techniques, for example in the use of adaptive interventions
in response to these measures.
Developing tools for suicide prevention remains a
challenge. It is a difficult population to access, and one
which requires a high degree of sensitivity. Best practice for
responding to an automatic indicator of suicidality is not yet
clear [31], and the expectations of privacy and the
implications of false positives must be addressed. Ensuring
appropriate safeguards and community consultation are
crucial to facilitating this important research topic.
Expanding the scope and availability of suicide prevention
programs and translation into wider practice requires
community and clinical engagement, as well as
technological innovation and advancements, and engineering
skills will be a key component in this interdisciplinary field.
ACKNOWLEDGMENT
M.E.L., T.W.B. and H.C. thank the staff and students at
the Black Dog Institute who participated in the pilot of the
social connectivity project. N.C. and J.E. thank National ICT
Australia who support their work. B.O. and H.C.
acknowledge the efforts of Dr. Cecile Paris, Dr. Stephen
Wan, Dr. David Milne (CSIRO) as well as Dr. Philip J
Batterham and Dr. Alison L Calear (National Institute for
Mental Health Research) who significantly contributed to
the ‘Detecting Suicidality on Twitter’ project. J.T., F.S. and
H.C. thank Alive and Kicking Goals Suicide Prevention
Program (WA), the Young and Well Cooperative Research
Centre, HITnet Innovations, Thought Works, Muru Marri
Indigenous Health Unit UNSW, Kimberley Aboriginal
community members, Men’s Outreach Service (WA),
Aboriginal Artists – Martha Lee, Danica Manado and Esah
Coffin, Nyamba Buru Yawuru Aboriginal Corporation,
Rubibi (Broome, WA), Goolarri Media (WA), Dr Kathy
McKay, BackTrack (NSW) and Samsung for their help and
support in the ibobbly study.
REFERENCES
[1] World Health Organization, "Disease and injury regional
mortality estimates, 2000–2012," 2012.
[2] Australian Bureau of Statistics, "3303.0 - Causes of Death,
Australia, 2013," 2015.
[3] Suicide Prevention Australia, "Overcoming the stigma of
suicide," 2010.
[4] D. J. France, R. G. Shiavi, S. Silverman, M. Silverman, and D.
M. Wilkes, "Acoustical properties of speech as indicators of
depression and suicidal risk," IEEE Trans Biomed Eng, vol. 47,
pp. 829-37, Jul 2000.
[5] A. Ozdas, R. G. Shiavi, S. E. Silverman, M. K. Silverman, and
D. M. Wilkes, "Investigation of vocal jitter and glottal flow
spectrum as possible cues for depression and near-term suicidal
risk," IEEE Trans Biomed Eng, vol. 51, pp. 1530-40, Sep 2004.
[6] IEEE Engineering in Medicine & Biology Society, "IEEE
EMBS Code of Ethics."
[7] P. A. Nutting, L. M. Dickinson, L. V. Rubenstein, R. D. Keeley,
J. L. Smith, and C. E. Elliott, "Improving detection of suicidal
ideation among depressed patients in primary care," Ann Fam
Med, vol. 3, pp. 529-36, Nov-Dec 2005.
[8] J. J. Mann, A. Apter, J. Bertolote, A. Beautrais, D. Currier, A.
Haas, et al., "Suicide prevention strategies: a systematic review,"
JAMA, vol. 294, pp. 2064-74, Oct 26 2005.
[9] K. A. Van Orden, T. K. Witte, K. C. Cukrowicz, S. R.
Braithwaite, E. A. Selby, and T. E. Joiner, Jr., "The interpersonal
theory of suicide," Psychol Rev, vol. 117, pp. 575-600, Apr
2010.
[10] N. Eagle, A. S. Pentland, and D. Lazer, "Inferring friendship
network structure by using mobile phone data," Proc Natl Acad
Sci U S A, vol. 106, pp. 15274-8, Sep 8 2009.
[11] J. N. Rosenquist, J. H. Fowler, and N. A. Christakis, "Social
network determinants of depression," Mol Psychiatry, vol. 16,
pp. 273-81, Mar 2011.
[12] M. C. Pachucki, E. J. Ozer, A. Barrat, and C. Cattuto, "Mental
health and social networks in early adolescence: a dynamic study
of objectively-measured social interaction behaviors," Soc Sci
Med, vol. 125, pp. 40-50, Jan 2015.
7319
[13] M. Burke, C. Marlow, and T. Lento, "Social Network Activity
and Social Well-Being," Chi2010: Proceedings of the 28th
Annual Chi Conference on Human Factors in Computing
Systems, Vols 1-4, pp. 1909-1912, 2010.
[14] M. Larsen, T. Boonstra, P. Batterham, B. O'Dea, C. Paris, and H.
Christensen, "We Feel: Mapping emotion on Twitter," IEEE J
Biomed Health Inform, In press 2015.
[15] Samaritans. (2014). Samaritans Launches Twitter App To Help
Identify Vulnerable People. Available:
http://www.samaritans.org/news/samaritans-launches-twitter-
app-help-identify-vulnerable-people
[16] Facebook. (2015). Facebook Safety. Available:
https://www.facebook.com/fbsafety/posts/817724748265365
[17] B. O'Dea, S. Wan, P. J. Batterham, A. L. Calear, C. Paris, and H.
Christensen, "Detecting suicidality on Twitter," Internet
Interventions, vol. 2, pp. 183-188, 2015.
[18] H. Hendin, J. T. Maltsberger, A. Lipschitz, A. P. Haas, and J.
Kyle, "Recognizing and responding to a suicide crisis," Ann N Y
Acad Sci, vol. 932, pp. 169-86; discussion 186-7, Apr 2001.
[19] R. W. Maris, "Suicide," Lancet, vol. 360, pp. 319-26, Jul 27
2002.
[20] N. Cummins, S. Scherer, J. Krajewski, S. Schnieder, J. Epps,
and T. Quatieri, "A Review of Depression and Suicide Risk
Assessment using Speech Analysis," Speech Communication (In
Press), 2015.
[21] V. Sethu, J. Epps, and E. Ambikairajah, "Speech based emotion
recognition," in Speech and audio processing for coding
enhancement and recognition, T. Ogunfunmi, R. Togneri, and
M. Narasimhai, Eds., ed New York: Springer, 2014, pp. 197-
228.
[22] Z. N. Karam, E. M. Provost, S. Singh, J. Montgomery, C.
Archer, G. Harrington, et al., "Ecologically Valid Long-Term
Mood Monitoring of Individuals with Bipolar Disorder Using
Speech," 2014 IEEE International Conference on Acoustics,
Speech and Signal Processing (ICASSP), 2014.
[23] Mental Health America. Suicide. Available:
http://www.mentalhealthamerica.net/suicide
[24] N. Cummins, J. Epps, V. Sethu, M. Breakspear, and R. Goecke,
"Modeling spectral variability for the classification of depressed
speech," in 14th Annual Conference of the International Speech
Communication Association, INTERSPEECH 2013, August 25,
2013 - August 29, 2013, Lyon, France, 2013, pp. 857-861.
[25] N. W. Hashim, M. Wilkes, R. Salomon, and J. Meggs, "Analysis
of Timing Pattern of Speech as Possible Indicator for Near-Term
Suicidal Risk and Depression in Male Patients," in 2012
International Conference on Conference on Signal Processing
Systems (ICSPS 2012), 2012, pp. 6-13.
[26] T. Yingthawornsuk, H. Kaymaz Keskinpala, D. M. Wilkes, R.
G. Shiavi, and R. M. Salomon, "Direct acoustic feature using
iterative EM algorithm and spectral energy for classifying
suicidal speech," in 8th Annual Conference of the International
Speech Communication Association, Interspeech 2007, August
27, 2007 - August 31, 2007, Antwerp, Belgium, 2007, pp. 749-
752.
[27] S. Scherer, J. Pestian, and L.-P. Morency, "Investigating the
speech characteristics of suicidal adolescents," in 2013 38th
IEEE International Conference on Acoustics, Speech, and Signal
Processing, ICASSP 2013, May 26, 2013 - May 31, 2013,
Vancouver, BC, Canada, 2013, pp. 709-713.
[28] T. Donker, K. Petrie, J. Proudfoot, J. Clarke, M. R. Birch, and H.
Christensen, "Smartphones for smarter delivery of mental health
programs: a systematic review," J Med Internet Res, vol. 15, p.
e247, 2013.
[29] Black Dog Institute. (2015). Overview of the iBobbly trial.
Available: http://digitaldog.org.au/programs/ibobbly-black-dog-
institute/
[30] T. Callinan. (2014). Remote Indigenous Australians rely on
Facebook to stay in touch. Available:
http://www.sbs.com.au/news/article/2014/08/26/remote-
indigenous-australians-rely-facebook-stay-touch
[31] N. Lee, "Trouble on the radar," Lancet, vol. 384, p. 1917, 2014.