Research paperDisturbed sleep as a clinical marker of wish to die: A smartphone monitoring study over three months of observation
Introduction
Suicidal behaviour is a major public health problem. Over 800,000 people take their own lives every year worldwide, and approximately 20 times more people attempt suicide (WHO, 2020). Suicide risk assessment and monitoring are crucial for the prevention of suicidal behaviour (Wortzel et al., 2017). However, a direct assessment of suicide risk can be difficult in contexts where a specialized clinical support cannot be guaranteed, such as in primary care settings, or when using self-report questionnaires (Saini et al., 2014; Barnes et al., 2017).
Clinical proxies of suicidal behaviour can facilitate assessment and be less overwhelming for patients. Wish to die —or passive suicidal ideation— has been shown to increase the risk for suicide attempts and death by suicide, even in the absence of active suicidal ideation (Baca-García et al., 2011; Suokas et al., 2001; Palacio et al., 2007; Liu et al., 2020). A recent systematic review and meta-analysis deeply characterized passive suicide ideation, concluding that it was highly similar to active suicide ideation in terms of psychological correlates, and that it was strongly associated with suicide attempts (Liu et al., 2020).
Smartphones are especially apt for monitoring symptoms, given their increasing ubiquity, their versatility, and the users’ widespread habit of carrying them at all times (Konok et al., 2016; Davidson et al., 2016). Smartphone monitoring is being increasingly used in biomedical research. There are two modalities of smartphone monitoring: passive —by using the smartphone's native sensors—, and active —by asking questions to participants via the device, resulting in a real-time and takes place in the participant's usual environment. Smartphone monitoring reduces recall bias, respect ecological validity, and facilitate the simultaneous collection of data (Kleiman & Nock, 2018; Melbye et al., 2020).
There are some previous smartphone monitoring studies in suicide research, which have looked into different aspects such as fluctuation of suicide ideation over time or feasibility and acceptability of monitoring systems (Ben Zeev et al., 2017; Kleiman et al., 2017; Hadzic et al., 2020b; Vine et al., 2020; Hallensleben et al., 2019; Glenn et al., 2020; Gratch et al., 2021; Oquendo et al., 2020; Porras-Segovia et al., 2020). However, most of these previous studies have follow-up periods of less than a month and sample sizes of less than a hundred participants (Kleiman et al., 2017; Hadzic et al., 2019; Hallensleben et al., 2019; Glenn et al., 2020; Gratch et al., 2020; Vine et al., 2020). Another limitation is the use of economic incentives to increase engagement, a practice that limits the applicability of results (Singer and Couper, 2008; Groth, 2010). However, smartphone monitoring can be feasible in real-world conditions, as shown by a feasibility study by our research group: we tested the MEmind application in 457 participants and obtained a retention rate of 66.6% and an overall 68.0% compliance with questions over two months of observation, without using economic incentives (Porras-Segovia et al., 2020).
Sleep is recently emerging as a promising clinical marker. Several forms of disturbed sleep —including insomnia, nightmares, poor sleep quality and reduced sleep quantity— have been associated with several forms of suicidal thoughts and behaviours (STB)—including wish to die, suicidal ideation, suicide attempts and death by suicide ( Bernert et al., 2014; Li et al., 2016; Bernert et al., 2017; Mirsu-Paun et al., 2017; Littlewood et al., 2018). This association has been confirmed in a number of systematic reviews and meta-analyses, which have found a significant and independent association between sleep disturbances and STB (Pigeon et al., 2012; Malik et al., 2014; Porras-Segovia et al., 2019; Chaib et al., 2020).
The association between sleep problems and STB seems relevant in both the long term and the short term. For example, a 13-year retrospective study of 479,967 patients showed that insomnia tripled the risk of suicide attempt (Lin et al., 2018). Looking at the short-term relationship, a prospective study with a 21-day follow-up showed that actigraphy-measured sleep variability was a significant predictor of suicidal ideation (Bernert et al., 2017). Another study, where participants completed a sleep diary over the course of 1 week showed that short sleep duration and poor sleep quality increased the severity of next day suicide ideation (Littlewood et al., 2018). Thus, sleep may increase the risk of suicide both over several years and over the course of just a few hours. The long-term association may be mediated by increased risk for other mental disorders, such as depressive disorders, among other factors, while the short-term association may result from emotional dysregulation and impulsivity (Porras-Segovia et al., 2019).
Negative feelings and appetite alterations have also been associated with STB in previous studies. Negative feelings have been explored in some previous smartphone monitoring studies (Kleiman et al., 2017; Forkman & Teismann, 2017; Glenn et al., 2020; Vine et al., 2020; Czyz et al., 2019; Hallensleben et al., 2019; Peters et al., 2020). For instance, in the study by Hallensleben et al., (2019), hopelessness and perceived burdensomeness were prospectively associated with suicide ideation. However, this is not the case for appetite, which has been explored with more traditional methodologies —for instance, Kitagawa et al. (2017) found that adolescents with appetite loss had a five-fold risk for suicidal ideation compared to those with normal appetite— but has not yet been measured through smartphone monitoring.
Smartphone monitoring may be particularly suitable for measuring the relationship between sleep and suicide, given the huge variability of suicidal ideation over short periods of time and the probable role of sleep as a short-term marker of STB.
As shown in the meta-analysis by Franklin et al., (2017), there has been little progress in the search for valid risk factors for suicidal behavior in the last decades. The authors of this meta-analysis highlight the potentials of machine learning to advance in the field of suicidology (Franklin et al., 2017).
In this study, we use smartphone monitoring and machine learning techniques to explore the associations between wish to die, disturbed sleep, negative feelings and altered appetite, in non-incentivized psychiatric patients over three months of observation. Our hypotheses are: 1. That we will be able to detect relevant latent features across the dataset, 2. That these latent features will show a prominent association between sleep problems and STB (i.e. these variables will be present in a short time window).
Section snippets
Settings & Design
This is a prospective cohort study of psychiatric patients receiving mental health care at the Hospital Universitario Fundación Jiménez Díaz, which has a catchment area of 450,000 people. This study comprises a sub-set of the cross-national multicentre study SmartCrisis. SmartCrisis’ study protocol has been published elsewhere (Berrouiguet et al., 2019).
The study was approved by Ethics Committee of the Jimenez Díaz Foundation University Hospital and was conducted according to the principles set
Characteristics of the sample
Of the 189 patients approached, 165 (87.3%) agreed to participate in the study. The mobile application could not be installed in 26 (13.8%) participants due to technical issues. 139 (73.5%) had the application installed. Participants who answered for less than five days were excluded, resulting in 110 participants included in the final analysis. There were no significant differences regarding age, sex, or history of STB between compliant and non-compliant participants. Total number of responses
Comparison with previous findings
In our study, we found that ‘wish to die' and sleep problems tend to be present at the same time window of 96 hours. Our results add to the evidence about the association between sleep disturbances and different forms of suicidal behavior (Pigeon et al., 2012; Malik et al., 2014; Porras-Segovia et al., 2019). Particularly, our study adds to the evidence that sleep problems are associated with STB in the short term (Bernert et al., 2017; Littlewood et al., 2018). Although in our study we have
Conclusions
Using smartphone monitoring and machine learning techniques we found that disturbed sleep was associated with wish to die among psychiatric patients in the short-term (within 96 hours). Our findings stress the importance of evaluating sleep as part of the screening for suicidal behavior. In the presence of other factors, such as high-risk psychiatric diagnoses or a history of previous attempts, it is relevant to ask about the quality and quantity of sleep, as it can be a precipitating factor.
Contributors
Enrique Baca-García designed the study. Aurora Cobo and Isaac Díaz-Oliván were in charge of sample recruitment and data collection. Alejandro Porras-Segovia, Antonio Artés-Rodríguez, Enrique Baca-García and Maria Luisa Barrigon carried out data analysis. Aurora Cobo, Alejandro Porras-Segovia and Isaac Díaz-Oliván designed the figures. Alejandro Porras-Segovia wrote the manuscript. Sofian Berrouiguet, Jorge Lopez-Castroman, Philippe Courtet and María A. Oquendo made substantial contributions to
Role of the funding source
This study received grant support from Instituto de Salud Carlos III (ISCIII PI13/02200; PI16/01852; CM19/00026), Delegación del Gobierno para el Plan Nacional de Drogas (20151073), the American Foundation for Suicide Prevention (LSRG-1-005-16), the Ministerio de Ciencia, Innovación y Universidades (RTI2018-099655-B-I00; TEC2017-92552-EXP), and by the Comunidad de Madrid (Y2018/TCS-4705, PRACTICO-CM).
Declaration of Competing Interest
Financial Disclosure: This study received grant support from the institutions specified in the Acknowledgements section. MAO receives royalties for the commercial use of the Columbia-Suicide Severity Rating Scale. Her family owns stock in Bristol Myers Squibb.
Non-financial Disclosure: EBG and AAR designed the MEmind application.
Acknowledgements
This study received grant support from Instituto de Salud Carlos III (ISCIII PI13/02200; PI16/01852; CM19/00026), Delegación del Gobierno para el Plan Nacional de Drogas (20151073), the American Foundation for Suicide Prevention (LSRG-1-005-16), the Ministerio de Ciencia, Innovación y Universidades (RTI2018-099655-B-I00; TEC2017-92552-EXP), and by the Comunidad de Madrid (Y2018/TCS-4705, PRACTICO-CM).
References (65)
- et al.
Estimating risk for suicide attempt: Are we asking the right questions?
Journal Of Affective Disorders
(2011) - et al.
Validation of the insomnia severity index as an outcome measure for insomnia research
Sleep Medicine
(2001) - et al.
The Pittsburgh sleep quality index: A new instrument for psychiatric practice and research
Psychiatry Research
(1989) - et al.
Daily patterns in non-suicidal self-injury and coping among recently hospitalized youth at risk for suicide
Psychiatry Research
(2019) - et al.
Entrapment, perceived burdensomeness and thwarted belongingness as predictors of suicide ideation
Psychiatry Research
(2017) - et al.
Insomnia and daytime cognitive performance: a meta-analysis
Sleep medicine reviews
(2012) - et al.
The association of trait impulsivity and suicidal ideation and its fluctuation in the context of the Interpersonal Theory of Suicide
Comprehensive Psychiatry
(2020) - et al.
Appetite loss as a potential predictor of suicidal ideation and self-harm in adolescents: A school-based study
Appetite
(2017) - et al.
Real-time assessment of suicidal thoughts and behaviors
Current Opinion in Psychology
(2018) - et al.
Humans' attachment to their mobile phones and its relationship with interpersonal attachment style
Computers In Human Behavior
(2016)