Smartphone-based Ecological Momentary Intervention for Suicide Prevention: a Randomised Clinical Trial

Introduction: Over 800,000 die every year by suicide, while suicide attempts are estimated to be 20 times more frequent. Suicidal ideation (SI) is estimated to affect 2% of the population. Suicidal thoughts and behaviours (STB) increase the risk of death by suicide, impair people's quality of life and result in high healthcare costs.

New technologies can help us in the management and prevention of STB. Among other advantages, these devices open the possibility of implementing smartphone-based real-time monitoring. There are two main types of smartphone-based monitoring: active and passive. The active mode is also known as Ecological Momentary Assessment (EMA). EMA involves asking daily questions, resulting in real-time responses in the patient's usual environment. EMA provides a more accurate picture of the emotional and cognitive context in which SI appears, offering us the opportunity to identify its immediate predictors.

Passive monitoring -also called passive EMA- involves using the mobile phone's native sensors to collect information such as physical activity, mobility or sleep, which can be useful as proxy factors for mental state. The combination of active and passive monitoring produces valuable information to characterize people's behaviour, giving rise to what is known as an individual's digital phenotype (Barrigón et al., 2019). As it does not require patient effort, passive EMA can increase the acceptability of monitoring and counteract the so-called "EMA fatigue", which occurs when patients stop answering EMA questions.

A considerable number of studies have applied EMA technology to suicide research. In contrast, passive EMA has been used scarcely in suicidology.

A further step is to exploit EMA's potential to design digital therapeutic tools, providing continuous support in the patient's usual environment. This is known as Ecological Momentary Intervention (EMI). EMIs can be a useful add-on to traditional treatment, thanks to their 24-hour availability, low cost, and the possibility of continuing follow-up in a non-presential manner. Some studies have explored the effectiveness of EMI on suicidal behaviour, showing good acceptability and a reduction in SI after the intervention.

This is an up-and-coming area that may contribute to suicide prevention, which has not yet received sufficient attention in the scientific literature.

The SmartCrisis 2.0 randomized clinical trial aims to evaluate the effectiveness of an Ecological Momentary Intervention (EMI) combined with smartphone-based monitoring for the prevention of STB in patients at high risk of suicide.

Our hypotheses are as follows:

1. Before/after comparison: In the intervention group, there will be a significant reduction in STB and an improvement in functionality and quality of life after the follow-up period.
2. Comparison with control group: The intervention group will have a greater reduction in STB and a greater improvement in functioning and quality of life than the control group.
3. Feasibility and acceptability: The use of the eB2 and MEmind mobile applications will be feasible and well accepted by psychiatric patients.

Methods and analysis: Our sample will consist of patients with a history of recent STB treated at either one of these four locations, pertaining to the same hospital Area in the Community of Madrid: University Hospital Fundación Jiménez Díaz (Madrid, Spain), Rey Juan Carlos Hospital (Móstoles, Madrid, Spain), General Hospital of Villalba (Villalba, Madrid, Spain), and University Hospital Infanta Elena (Valdemoro, Madrid, Spain).

Sample size calculation was performed using G*Power software, version 3.1. Based on previous clinical trials exploring the ability of crisis management interventions to reduce SI in the mid-term (Ducasse et al., 2018), the investigators estimated a target sample size enough to have 80% power to detect between-group differences of 6-7 points on the Columbia Suicide Severity Rating Scale (25% decrease in the intensity of SI in the intervention group). The alpha error was set at 5% and the power at 80%. A dropout rate of 20% and a randomization imbalance of 10% were assummed. With these settings, the investigators estimated that a total of 220 participants, 110 in each arm, were needed.

Before starting the clinical trial, a pilot will be conducted, in approximately 40 patients will be recruited. All of them will receive the SmartSafe EMI and be monitored using active and passive EMA. These patients will not be part of the final analysis. After three months of follow-up, when all participants have had time to try out the application, a focus group will be held between the research team, the attending clinicians and the patients in order to get their opinion on the intervention and to identify possible technical and human failures that could be improved before starting the clinical trial.

Once the clinical trial recruitment starts, attending psychiatrists will check their patients' eligibility during their regular appointment, explain the project in detail and invite them to participate. If patients agree to participate, they will be asked to sign the informed consent, after which they will be randomly assigned to the intervention group (EMI + EMA + TAU) or the control group (EMA + TAU).

At the baseline interview, a psychiatrist will assess the characteristics of the last suicidal event using the Brief Suicide Protocol, which has been previously published, and they will set up the SmartSafe intervention according to the patient's preferences. Next, a trained research assistant, blind to the patient's assigned arm, will administer the standardized questionnaires detailed below.

Patients will be followed for one year, with face-to-face visits at six months and one year. The attending psychiatrist will also check for STB using the Columbia Suicide Severity Rating Scale (CSSRS) at the 4th week, 11th week and 9th-month appointments. If participants cannot attend the follow-up visits, the assessments will be performed by phone. All questionnaires will be completed via the MEmind website.

The piloting is expected to take place between June and December 2021. Recruitment is expected to commence in January 2022 and to be completed by June 2023. Data collection will take place from 2022 to 2024. Data analysis will take place from 2022 to 2024. The investigators expect to publish partial results by the end of 2022 and final results by early 2025.

Digital monitoring will be performed using EMA. Two EMA modalities will be used: active using the MEmind app, and passive using the eB2 app.

The MEmind app asks short daily questions that appear on the mobile screen. Each day, 2-4 random questions will be asked at random times (respecting sleep hours) from the pool of 34 questions that make up the questionnaire based on the Salzburg Suicide Questionnaire.

The EMA questions belong to 6 categories:

• Suicidality (5 questions)
• Non-suicidal self injury (2 questions)
• Affect (9 questions)
• Interpersonal experiences (11 questions)
• Sleep (4 questions)
• Eating (3 questions) The eB2 app runs in the background and does not require user intervention. It collects real-time information from the patient's mobile phone (location, mobility, smartphone usage and sleep), stores it and downloads it to a server. All the information collected from the phone is transformed with a hash function. This hash function makes it impossible to know the original data but allows the analysis of behavioral patterns.

In addition to the smartphone-based monitoring, another digital assessment will be carried out: a neuropsychological assessment using a previously validated Virtual Reality environment, the "Spheres \& Shield Maze Task". This tool recreates a maze with different tasks to assess decision-making, impulsivity, attention and cognitive inhibition.

Non-digital assessments will be carried out by attending psychiatrists and research assistants, all of them previously trained. All questionnaires will be completed on the MEmind website, which allows for easy completion and immediate uploading to databases.

Some of the main tools used will be the following:

• The Brief Suicide Questionnaire (BSQ) will be used to characterize the index event that motivated inclusion in the study (suicide attempt or emergency referral for SI). It includes questions on lethality, method and surrounding circumstances of the attempt, and lifetime history of suicidal behaviour.
• The CSSRS scale will be used to measure STB, both at baseline and follow-up visits.
• The investigators will verify suicide attempts and deaths by suicide through our organization's digital medical records, called the Casiopea software, which integrates information on ER visits, hospitalizations, and visits to specialist consultations. If a patient cannot be located, and there is no information in their record, the investigators will contact the patient's family telephone number. The investigators will also request access to the Spanish register of deaths and causes of death of the "Instituto Nacional de Estadística" [National Statistics Institute].

At the six-month follow-up visit, patients will complete a qualitative satisfaction survey with the following questions referring to each of the applications used (MEmind and eB2) and the face-to-face assessments.

Traditional statistical analyses will be performed using SPSS 24.0 statistical software. To compare SI at baseline with SI at the end of follow-up (CSSRS score), paired samples t-tests will be used. To compare the improvement in SI after follow-up between the intervention group and the control group, an independent samples t-test will be used.

Survival curves (time to a new suicidal event) will be calculated using the Kaplan-Meier test. The intervention group's survival curves and the control group will be compared using the log rank test. To assess the project's feasibility, participation rate, retention and percentage of EMA questions answered will be calculated.

Binary logistic regression will be used to explore the correlation between the study variables, obtaining crude and age- and sex-adjusted ORs. A multivariate regression model will also be built to assess the unique strength of the association of the variables. Cox regression analysis will be performed to explore variables associated with survival time (time until next event). All tests will be two-tailed, with a significance level of p<0.05 and 95% confidence intervals.

In addition to the traditional analyses, with the Signal Theory Department of the Carlos III University's support, unsupervised analyses will be carried out using machine learning techniques to create behavioral models and patient activity profiles.