The proposed study will use ecological momentary assessment (EMA) via smartphone applications
(apps) and wearable trackers to examine the relationship between suicidality (wish to die,
suicidal ideation and suicide attempt) and changes in sleep quality and disturbed appetite.
These behavioral markers, if the hypothesis is proven true, could help predict increased
suicidal risk in real-time within a vulnerable population across different cultures.
The study aims to: (1) Establish the extent to which quality of sleep is related to suicide
ideation and suicide attempts; (2) Establish the extent to which change in appetite is
related to suicide ideation and suicide attempts; (3) Determine the emotional impact of the
app when suicidality is assessed; (4) Clarify the timeline of the relationship between sleep
disturbances and suicidal behavior; (5) Develop personalized algorithms based on EMA protocol
and motor activity markers or "signatures" to assess the risk of suicide attempts.
The hypothesis is that variations in sleep quality will correlate with increased wish to die,
suicide ideation and suicide attempts. It is expected that a decrease in sleep quality will
be a suicide risk marker especially among young individuals.
This prospective cross-national study will use the infrastructure of an existing network
(WORECA). Woreca has defined a common protocol of suicide assessment, data sharing and
analysis strategy.
1044 suicide attempters will be included and followed for 6 months. Each participant will be
assessed using an EMA protocol via two smartphone apps: (1) One app will ask everyday
questions following a dynamic protocol to assess quality of sleep, appetite, suicidal
ideation and psychopathology; (2) the other app will record activity using smartphone
sensors. Additionally, 300 participants (150 in France and 150 in Spain) will have their
sleep phases and other physiological changes during sleep monitored with a wearable armband.
Study outcomes include wish to live, wish to die, suicidal ideation, and suicide attempt
during the follow-up period.
A multi-level logit regression analysis will be used to account for multiple observations per
individual, to identify individual-level (sleep, appetite, socio-demographic, clinical data,
treatment data) and site-level characteristics associated with death desire, suicidal
ideation or suicide attempt (aim 1 and 2). Hazards models will also be used to relate
covariate characteristics (sleep, appetite, sociodemographic, clinical data, treatment data)
with time to suicide reattempt during the follow-up period (aim 1 and 2). Data mining
(machine learning) techniques will be used to examine risk factors, patterns of illness
evolution (aim 3 and 4) and patient stratification by level of suicidal risk (aim 5).
Identifying surrogate markers of suicidality related with physiological functions, which
carry less or no stigma for the patients and are easier to report, or have a lower reporting
threshold is an essential task. These markers would allow to predict in real-time an increase
in suicidal risk within a vulnerable population and ultimately help to prevent and even
personalize treatment.Suicidal behaviours, including suicidal ideation, are preventable but
to be efficient, prevention needs to rely on the identication of specific risk factors.