Preventing Suicide With Digital Phenotyping and Pharmacogenetics-Based Interventions

Overview

The goal of this protocol for an observational retrospective multi-site cohort study is to develop a predictive algorithm for suicidal behavior integrating genetic risk markers, digital phenotypes, and exposomic data in people with a lifetime history of suicide attempt. Participants will be aged from 12 onwards (requiring parental consent if under 18) and only be excluded if they are unable to provide a genetic sample or do not consent to the study. The main question\[s\] it aims to answer is: \- Can genotyping/omics analysis of individuals with a history of suicidal behaviuor reveal potential genetic factors associated with suicide risk? Secondary questions include: 1. Can behavioral factors associated with suicide risk be explored through data obtained from Google Takeout? 2. Is there an association between medication changes and suicidal behavior, as identified through the Unified Prescription Module and the Digital Health Record? 3. Do different suicidal phenotypes differ based on the collected variables? 4. Can the investigators construct an exposome using geolocated and time-stamped data from Google Takeout, combined anonymously with data from the National Statistics and Meteorology Institutes? The investigators hypothesize that: 1. Individuals with a history of suicidal behavior will show significant genotypic differences compared to the general population (based on Spanish Genome Project data). 2. Suicidal phenotypes-especially between single and multiple attempters-will differ across collected variables, including genotype, omics, and exposome data. participants will complete genetic assessments, digital phenotypes, clinical questionnaires and exposomic data

Participant recruitment will occur in person at outpatient mental health clinics, emergency departments, and short-stay hospital units. Mental health professionals (psychiatrists, psychologists, and nurses) will assess eligibility during clinical encounters, provide study information, and obtain informed consent. Data will be entered into the MeMind environment for data monitoring, and patient validation, ensuring eligibility criteria and noting any missing information. Prior to recruitment, training sessions and standardized guidelines are provided to healthcare professionals to ensure adherence to the study protocol. Data will be verified against INE statistics and hospital records to assess the representativeness and completeness of the data. The primary outcome is suicidal behavior risk, assessed using the Columbia-Suicide Severity Rating Scale (C-SSRS), the Brugha scale, documented suicidal events, healthcare utilization, behavioral patterns, digital phenotyping, and pharmacological treatment modifications. Secondary outcomes include evaluating the efficacy of pharmacological treatments and the cost-effectiveness of genetic markers in guiding antidepressant selection. Baseline data collection includes: 1. Genetic sampling for genomic and metabolic profiling. 2. A sociodemographic and clinical assessment. 3. Consent-based extraction of digital behavioral data via Google Takeout. To enrich participant health profiles, retrospective data from electronic health records (EHRs), prescription registries, and sociodemographic indicators from the Spanish National Institute of Statistics (INE) will be integrated. EHR data will include the Basic Minimum Set of Data (BMSD), a standardized clinical-administrative dataset. Blood samples will allow for DNA/RNA extraction and metabolomic analyses. Genetic findings will be compared with GWAS summary statistics from the Psychiatric Genomics Consortium (PGC) on suicide attempts and validated against control samples from the Banco Nacional de ADN and the Madrid Manic Group cohort. Google Takeout data will be used to construct digital phenotypes. Patients will choose which data to share, supported by a research assistant. The selected data will be pseudonymized via a secure script before being stored on servers at Universidad Carlos III. Prescription histories from various autonomous regions will be analyzed to monitor treatment trajectories over time. This study is part of a national consortium targeting a sample of at least 5,000 participants. Power calculations, informed by a recent meta-analysis (Li, 2023), indicate that this sample size would yield 93.5% power at a genome-wide significance level (p \< 5×10-⁸), assuming a minor allele frequency of 0.2 and an odds ratio of 1.3, accounting for gender, diagnosis, and treatment response variability. Statistical analyses will be conducted using SPSS software version 29.0. Logistic regression models will be used to examine factors associated with the primary outcomes, and a multivariate regression model will be developed to assess independent associations. All tests will be two-tailed, with statistical significance defined as p \< 0.05 and 95% confidence intervals reported. To identify digital behavioral biomarkers, advanced statistical and machine learning techniques will be applied. Individualized suicide-related profiles will be generated and compared using integrated data from genetic, metabolic, medical, and digital phenotyping sources. Additionally, a personalized and anonymized exposome will be constructed by combining geo-temporal data from Google Takeout with contextual information from public sociodemographic datasets (e.g., INE).