Research Appraisals
Evidence-based critical appraisals of the latest medical research, systematically evaluated using Oxford CEBM methodology.
Showing 4 appraisals
The Spanish journal of psychology
A Systematic Review of the Risk and Protective Factors for Suicide in Autism Spectrum Disorder
The shorter lifespan and increased prevalence of physical and psychiatric comorbidities in autistic people is a challenge for mental health. There is a high risk of suicide-related behaviors in autism, which are related to mental health issues, social isolation, and other factors. Here we aimed to systematically review the evidence on risk and protective factors of suicidal behaviors in autistic people. Electronic datasets (PubMed, Scopus, Web of Science, and PsycInfo) were searched for empirical articles on the risk and protective factors for suicide in autism. The PRISMA guidelines were followed, and a series of inclusion and exclusion criteria were applied to the retrieved data. This review included 19 studies reporting multiple risk factors associated with suicide-related behaviors. Noteworthy contributors among these factors include psychiatric comorbidities, bullying, cyberbullying, teacher harassment, and cognitive functions. The evidence for protective factors is limited; however, social relationships are highlighted. Identifying risk and protective factors is paramount for detecting early warning signs and enabling the prevention and intervention of suicidal behavior. Our review identified key sociodemographic, clinical, and cognitive risk factors to take into account as alarm-signs when identifying suicidal behaviors in autism.
31 July 2026
Read appraisal →Journal of affective disorders
Emotion regulation-targeted interventions initiated during hospitalization for suicidal crisis: a systematic review and exploratory meta-analysis
INTRODUCTION: Suicide is a leading cause of premature death, with risk peaking around psychiatric hospitalization. Emotion dysregulation (ED) is recognized as a key factor contributing to suicidal ideation and behavior across psychiatric diagnoses. The efficacy of interventions that target ED for suicidality reduction remains unclear, so we aimed to assess these interventions in psychiatric inpatients. METHODS: We conducted a PRISMA-compliant, PROSPERO-registered (CRD420251140949) systematic review and meta-analysis of randomized and non-randomized trials evaluating interventions targeting emotion regulation (ER) for psychiatric inpatients with suicidal ideation and/or behavior. Searches covered PubMed, Embase, and PsycINFO to September 18, 2025. Standardized mean differences (Hedges' g) were pooled under random-effects models and heterogeneity was assessed (Q, I2, τ2). Exploratory meta-regressions examined age, sex, and study quality; small-study effects were explored with funnel plots. RESULTS: Twelve studies met inclusion (n = 1708), spanning dialectical behavior therapy (DBT), mindfulness-based interventions, acceptance and commitment therapy (ACT), and other ER-related protocols. Across modalities, feasibility and acceptability were high and interventions consistently improved suicide-related outcomes in transdiagnostic adolescents and adult populations. Suicide attempts and ideation met outcome criteria for exploratory meta-analyses, and, for both, ER-targeted interventions outperformed usual care, despite heterogeneity and small samples. Exploratory meta-regressions were non-significant; funnel plots did not suggest marked publication bias, though power was limited. CONCLUSIONS: ER-targeted interventions delivered during psychiatric hospitalization are feasible and show promising benefits on suicide-related outcomes. Given heterogeneity and limited sample sizes, larger, well-powered randomized trials using standardized suicidality endpoints are needed to establish efficacy, refine inpatient protocols, and inform acute care pathways.
16 June 2026
Read appraisal →Journal of medical Internet research
Combining Machine Learning Models and Screening to Enhance Suicide Risk Identification for American Indian Patients: Retrospective Cohort Study
BACKGROUND: American Indian and Alaska Native communities experience disproportionately high suicide rates. While machine learning (ML) models leveraging electronic health records have emerged as promising tools for suicide risk identification, the optimal integration of these models with existing screening practices remains unclear. OBJECTIVE: The objective of this study was to compare parallel and serial testing strategies that combine an ML suicide risk model and the Ask Suicide-Screening Questions (ASQ) against using the ASQ alone. To achieve this, we conducted a retrospective secondary analysis of electronic health record data. The cohort consisted of adult emergency department visits at an Indian Health Service facility between October 1, 2019, and October 2, 2021. METHODS: Sensitivity, specificity, predictive values, and 95% CIs were averaged across 10 cross-validated patient-level folds. The final sample included 7897 American Indian patients with 26,896 visits, 824 (3.1%) of which had a positive ASQ result and 102 (0.4%) of which had the outcome of suicide attempt or death within 90 days of the visit. The logistic regression ML model previously developed using Indian Health Service-specific data was operationalized at the 95th and 75th percentiles to evaluate high-risk and medium-risk thresholds, respectively. A sensitivity analysis was performed to evaluate identification approaches across all emergency department visits during this period. RESULTS: The ML medium-risk threshold alone identified the most true positives (sensitivity: 0.782, 95% CI 0.648-0.915; specificity: 0.751, 95% CI 0.725-0.777; positive predictive value [PPV]: 0.012, 95% CI 0.009-0.014; negative predictive value [NPV]: 0.999, 95% CI 0.998-0.999) in comparison to the ML high-risk threshold alone (sensitivity: 0.429, 95% CI 0.287-0.572; specificity: 0.955, 95% CI 0.948-0.961; PPV: 0.035, 95% CI 0.022-0.048; NPV: 0.998, 95% CI 0.997-0.999) or the ASQ alone (sensitivity: 0.178, 95% CI 0.073-0.282; specificity: 0.970, 95% CI 0.968-0.971; PPV: 0.022, 95% CI 0.010-0.034; NPV: 0.997, 95% CI 0.996-0.998). Combining the ML high-risk threshold with the ASQ in series yielded the greatest positive predictive ability (PPV: 0.050, 95% CI 0.014-0.086) at the cost of reduced sensitivity (0.129, 95% CI 0.036-0.221). Finally, the parallel testing approach using the ML medium-risk threshold yielded the greatest sensitivity (0.795, 95% CI 0.671-0.920; specificity: 0.742, 95% CI 0.716-0.767; PPV: 0.012, 95% CI 0.009-0.014; NPV: 0.999, 95% CI 0.998-0.999) without missing any cases identified through screening. CONCLUSIONS: Unlike existing studies that evaluate ML and screening tools in isolation, this study innovates by assessing combined parallel and serial testing strategies in a real-world setting. We demonstrated that, while serial testing maximizes predictive accuracy, it is often infeasible. Instead, parallel testing brings value as a clinical "safety net" to catch at-risk patients missed by standard practices. Ultimately, integrating ML in suicide prevention requires balancing statistical accuracy with setting-specific, real-world workflows.
13 May 2026
Read appraisal →JMIR mental health
Predicting Momentary Suicidal Ideation From Smartphone Screenshots Using Vision-Language Models: Prospective Machine Learning Study
BACKGROUND: Passive smartphone sensing shows promise for suicide prevention, but behavioral metadata (GPS, screen time, and accelerometry) often lacks the contextual information needed to detect acute psychological distress. Analyzing what people actually see, read, and type on their phones-rather than just usage patterns-may provide more proximal signals of risk. OBJECTIVE: This study aimed to test whether vision-language models (VLMs) applied to passively captured smartphone screenshots can predict momentary suicidal ideation (SI). METHODS: Seventy-nine adults with past month suicidal thoughts or behaviors completed ecological momentary assessments (EMA) over 28 days while screenshots were captured every 5 seconds during active phone use. We fine-tuned open-source VLMs (Qwen2.5-VL [Alibaba Cloud], LFM2-VL [Liquid AI]), and text-only models (Qwen3 [Alibaba Cloud]) to predict SI from screenshots captured in the 2 hours preceding each EMA. We evaluated performance with temporal and subject holdouts. RESULTS: The analytic sample comprised 2.5 million screenshots from 70 participants. Temporal holdout models achieved strong discrimination at the EMA level (AUC=0.83; AUPRC=0.77), with image-based models outperforming text-only models (AUC=0.83 vs 0.79; 95% CI 0.003-0.07). Subject holdout generalization was near chance (AUC≈0.50), though a simple lexical screening method retained modest discrimination (AUC=0.62). Smaller models performed comparably to larger models, supporting feasible on-device deployment. CONCLUSIONS: Screen content predicts short-term SI with clinically meaningful accuracy when models are personalized but does not generalize across individuals. These findings support a 2-stage clinical architecture, coarse lexical screening for new patients, with personalized VLM-based monitoring after a calibration period. On-device inference may enable privacy-preserving deployment.
10 Apr 2026
Read appraisal →