Jmir Mental Health最新文献

Background: Patient safety incidents are a leading cause of harm in psychiatric settings, yet early warning systems (EWS) tailored to mental health remain underdeveloped. Traditional risk tools such as the Dynamic Appraisal of Situational Aggression-Inpatient Version (DASA-IV) offer limited predictive accuracy and are reactive rather than proactive.

Objective: We introduce the Predictive Risk Identification for Mental Health Events (PRIME) tool, a deep learning-based EWS trained on longitudinal psychiatric electronic medical record (EMR) data to anticipate adverse events in 24-hour windows.

Methods: A retrospective cohort study using routinely collected EMR data to train and validate machine learning (ML) models for short-term risk prediction was conducted. This study took place at Waypoint Centre for Mental Health Care, a large inpatient psychiatric hospital in Ontario, Canada, serving both high-security forensic and nonforensic patient populations. A total of 4651 patients and 403,098 encounters from January 2020 to August 2024 were included. For model evaluation, the 2024 test set included 900 patients and 48,313 encounters. PRIME was trained using recurrent neural networks with attention mechanisms on multivariate time-series data. The model used an autoregressive design to forecast risk based on 7 days of prior patient data and was benchmarked against the DASA-IV clinical tool and other ML baselines. The primary outcome was the occurrence of an adverse mental health event recorded in the EMR within the following 24 hours. Model performance was assessed using area under the receiver operating characteristic curve (AUC) and recall, alongside subgroup analyses and interpretability assessments using integrated gradients.

Results: The long short-term memory with attention mechanism achieved the highest predictive performance (AUC=0.83), outperforming existing tools such as DASA-IV by 0.20 AUC (0.81 vs 0.61) and demonstrating the potential of ML-based models to support proactive risk management in mental health settings.

Conclusions: The PRIME tool is one of the first developed and evaluated deep learning-based EWS for psychiatric inpatient care. By outperforming existing clinical tools and providing interpretable, rolling predictions, PRIME offers a pathway toward safer, more proactive mental health interventions. Future work should assess its equity implications and integration into routine psychiatric workflows.

Background: Digital mental health interventions (DMHIs) offer scalable and cost-effective support for mental health but are predominantly developed in WEIRD (western, educated, industrialized, rich, democratic) contexts, raising questions about their global applicability. Dropout, attrition, and adherence rates critically influence DMHI effectiveness yet remain poorly characterized in culturally adapted formats.

Objective: This systematic review aimed to (1) synthesize evidence on dropout, attrition, and adherence in culturally adapted DMHIs delivered to non-WEIRD adult populations and (2) assess the methodological quality of the included studies.

Methods: PsycINFO, PubMed, and ScienceDirect were systematically searched for randomized controlled trials published in English between January 2014 and April 2024. Screening and data extraction followed PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, and methodological quality was evaluated using the Appraisal Tool for Cross-Sectional Studies tool. Extracted variables included dropout, attrition, adherence, adaptation techniques, and clinical outcomes.

Results: Twenty-three randomized controlled trials (n=4656) from diverse regions met inclusion criteria. Attrition ranged from 5.3% to 87% (median 18.4%), dropout from 0% to 66% (median 18.7%), and adherence from 26.3% to 100% (median 71%). Deep, participatory adaptations-such as language translation combined with culturally resonant content, stakeholder engagement, and iterative refinement-were consistently associated with lower dropout (<11%) and higher adherence (>75%). In contrast, surface-level adaptations (eg, translation only) showed higher dropout (up to 56%). Studies that incorporated both cultural tailoring and human support reported the most favorable engagement and clinical outcomes (eg, reductions in insomnia, depression, and anxiety). Most studies (91%) were rated as "Good" quality, although some lacked representative sampling or objective engagement metrics.

Conclusions: Comprehensive and participatory cultural adaptation is associated with engagement and effectiveness of DMHIs among non-WEIRD populations. Future research should integrate hybrid human-digital delivery models, objective engagement metrics, and larger multicenter trials to improve generalizability and scalability.

Background: Depression and anxiety are prevalent but commonly missed and misdiagnosed, an important concern because many patients do not experience spontaneous recovery, and the duration of untreated illness is associated with worse outcomes.

Objective: This study aims to explore the potential of using smartphone-tracked behavioral markers to support diagnostics and improve recognition of these disorders.

Methods: We used the dedicated Behapp digital phenotyping platform to passively track location and app usage in 217 individuals, comprising symptomatic (n=109; depression/anxiety diagnosis or symptoms) and asymptomatic individuals (n=108; no diagnosis/symptoms). After quantifying 46 behavioral markers (eg, % time at home), we applied a machine learning approach to (1) determine which markers are relevant for depression/anxiety recognition and (2) develop and evaluate diagnostic prediction models for doing so.

Results: Our analysis identifies the total number of GPS-based trajectories as a potential marker of depression/anxiety, where individuals with fewer trajectories are more likely to be symptomatic. Models using this feature in combination with demographics or in isolation outperformed demographics-only models (area under the receiver operating characteristic curveMdn=0.60 vs 0.60 vs 0.51).

Conclusions: Collectively, these findings indicate that smartphone-tracked behavioral markers have limited discriminant ability in our study but potential to support future depression/anxiety diagnostics.

Background: Posttraumatic stress disorder (PTSD) is highly prevalent in conflict-affected regions like eastern Democratic Republic of Congo; yet, cultural stigma and lack of psychoeducation limit public understanding and help-seeking behaviors.

Objective: This study evaluates the effect of a short, culturally adapted animated video on mental health perception, knowledge, and attitudes toward trauma.

Methods: A community-based quasi-experimental pre-post design was implemented among 239 participants from South Kivu. The intervention involved viewing a 3-minute animated psychoeducational video portraying locally relevant PTSD symptoms and resilience strategies. Perception, knowledge, and attitude scores were measured before and after the intervention, alongside PTSD prevalence and video appreciation.

Results: Out of 239, 40% (n=96) of the participants screened positively for PTSD. Post intervention, significant improvements were observed in perception (P=.01), knowledge (P<.001), and attitudes (P=.001) toward trauma. Appreciation was high; 82% (n= 195) expressed empathy for the characters, and 74% (n= 176) were likely to share the video. Linear regression showed that having PTSD symptoms (β coefficient=3.29, SE=1.09; P=.003), years of education (β coefficient=0.54, SE=0.08; P<.001), empathy toward the portrayed situations (β coefficient=5.07, SE=0.56; P<.001), perceived acquisition of new knowledge (β coefficient=2.58, SE=0.59; P<.001) and willingness to share the video (β coefficient=1.75, SE=0.50; P=.001) predicted stronger positive effect. A multiple linear regression including all predictors revealed that PTSD symptoms (β coefficient=1.93, SE=0.90; P=.03), years of education (β coefficient=0.47, SE=0.07; P<.001), empathy toward the portrayed situations (β coefficient=3.50, SE=0.55; P<.001), and willingness to share the video (β coefficient=1.75, SE=0.50; P=.001) remained significant predictors of video impact. Age and perceived acquisition of new knowledge were not significant in the multivariate model. This model accounted for 44.6% of the variance in video impact scores (R²=0.446, F_6,231=30.99, P<.001).

Conclusions: This study highlights the effectiveness of culturally grounded, low-cost digital media for improving mental health literacy in postconflict settings. Video-based tools may serve as scalable components of trauma-informed care and public health communication in low-resource, high-need areas.

Background: Depressive disorder affects over 300 million people globally, with only 30% to 40% of patients achieving remission with initial antidepressant monotherapy. This low response rate highlights the critical need for digital mental health tools that can identify treatment response early in the clinical pathway.

Objective: This study aimed to evaluate whether reasoning-based large language models (LLMs) could accurately predict 12-week remission in patients with depressive disorder undergoing antidepressant monotherapy and to assess the clinical validity and interpretability of model-generated rationales for integration into digital mental health workflows.

Methods: We analyzed data from 390 patients in the MAKE Biomarker discovery study who were undergoing first-step antidepressant monotherapy with 12 different medications, including escitalopram, paroxetine, sertraline, duloxetine, venlafaxine, desvenlafaxine, milnacipran, mirtazapine, bupropion, vortioxetine, tianeptine, and trazodone, after excluding those with uncommon medications (n=9) or missing biomarker data (n=32). Three LLMs (ChatGPT o1, o3-mini, and Claude 3.7 Sonnet) were tested using advanced prompting strategies, including zero-shot chain-of-thought, atom-of-thoughts, and our novel referencing of deep research prompt. Model performance was evaluated using balanced accuracy, sensitivity, specificity, positive predictive value, and negative predictive value. Three psychiatrists independently assessed model outputs for clinical validity using 5-point Likert scales across multiple dimensions.

Results: Claude 3.7 Sonnet with 32,000 reasoning tokens using the referencing of deep research prompt achieved the highest performance (balanced accuracy=0.6697, sensitivity=0.7183, and specificity=0.6210). Medication-specific analysis revealed negative predictive values of 0.75 or higher across major antidepressants, indicating particular utility in identifying likely nonresponders. Clinical evaluation by psychiatrists showed favorable mean ratings for correctness (4.3, SD 0.7), consistency (4.2, SD 0.8), specificity (4.2, SD 0.7), helpfulness (4.2, SD 1.0), and human likeness (3.6, SD 1.7) on 5-point scales.

Conclusions: These findings demonstrate that reasoning-based LLMs, particularly when enhanced with research-informed prompting, show promise for predicting antidepressant response and could serve as interpretable adjunctive tools in depressive disorder treatment planning, although prospective validation in real-world clinical settings remains essential.

Background: This study aims to detect self-harm or suicide (SH-S) ideation language used by youth (aged 13-21 y) in their private Instagram (Meta) conversations. While automated mental health tools have shown promise, there remains a gap in understanding how nuanced youth language around SH-S can be effectively identified.

Objective: Our work aimed to develop interpretable models that go beyond binary classification to recognize the spectrum of SH-S expressions.

Methods: We analyzed a dataset of Instagram private conversations donated by youth. A range of traditional machine learning models (support vector machine, random forest, Naive Bayes, and extreme gradient boosting) and transformer-based architectures (Bidirectional Encoder Representations from Transformers and Distilled Bidirectional Encoder Representations from Transformers) were trained and evaluated. In addition to raw text, we incorporated contextual, psycholinguistic (linguistic injury word count), sentiment (Valence Aware Dictionary and Sentiment Reasoner), and lexical (term frequency-inverse document frequency) features to improve detection accuracy. We further explored how increasing conversational context-from message-level to subconversation level-affected model performance.

Results: Distilled Bidirectional Encoder Representations from Transformers demonstrated a good performance in identifying the presence of SH-S behaviors within individual messages, achieving an accuracy of 99%. However, when tasked with a more fine-grained classification-differentiating among "self" (personal accounts of SH-S), "other" (references to SH-S experiences involving others), and "hyperbole" (sarcastic, humorous, or exaggerated mentions not indicative of genuine risk)-the model's accuracy declined to 89%. Notably, by expanding the input window to include a broader conversational context, the model's performance on these granular categories improved to 91%, highlighting the importance of contextual understanding when distinguishing between subtle variations in SH-S discourse.

Conclusions: Our findings underscore the importance of designing SH-S automatic detection systems sensitive to the dynamic language of youth and social media. Contextual and sentiment-aware models improve detection and provide a nuanced understanding of SH-S risk expression. This research lays the foundation for developing inclusive and ethically grounded interventions, while also calling for future work to validate these models across platforms and populations.

Background: University students face high levels of stress with limited support for coping and well-being. Campus mental health services are increasingly using digital resources to support students' stress management and coping capacity. However, the effectiveness of providing this support through web-based, self-directed means remains unclear.

Objective: Using a randomized controlled design, this study examined the acceptability and effectiveness of a self-directed, web-based resource containing evidence-based strategies for stress management and healthy coping for university students. The study additionally explored the potential benefits of screening and directing students to personalized resources aligned with their needs.

Methods: Participants consisted of 242 university students (193/242, 79.9% women; mean age 21.15 years) assigned to one of 3 groups (ie, automatically directed to personalized resources, nondirected, and waitlist comparison). They completed pre, post (4 wk), and follow-up (8 wk) measures for stress, coping, and well-being. The resource groups also completed acceptability measures at 2, 4, and 8 weeks after the web-based resource access.

Results: Results indicate high acceptability, reflecting students' satisfaction with the resource. Furthermore, significant decreases in stress and unhealthy coping, as well as significant increases in coping self-efficacy and healthy coping in the resource groups relative to the comparison group, were found. Interestingly, the directed approach showed no added benefit over nondirected resource access.

Conclusions: In summary, this study demonstrates the acceptability and effectiveness of a self-directed digital resource platform as a viable support option for university student stress and coping.

Background: The global demand for mental health services has significantly increased over the past decade, exacerbated by the COVID-19 pandemic. Digital resources, particularly smartphone apps, offer a flexible and scalable means of addressing the research-to-practice gap in mental health care. Clinicians play a crucial role in integrating these apps into mental health care, although practitioner-guided digital interventions have traditionally been considered more effective than stand-alone apps.

Objective: This scoping review explored mental health practitioners' views on potential use or integration of smartphone apps into clinical practice. We asked, "What is known about how mental health practitioners view the integration of smartphone apps into their practice?" Further, this scoping review explored the factors that might influence integration of smartphone apps into practice, such as practitioner and client characteristics, app design and functionality, and practitioner views.

Methods: We conducted a systematic search of 3 databases that yielded 38 studies published between 2018 and 2025, involving 1894 participants across various mental health disciplines, most predominantly psychologists and psychiatrists. Data were collected on practitioner and client characteristics, app functionality, and factors deemed important or influencing practitioners' opinions about app integration.

Results: The included studies were most likely to explore use of apps outside the clinical session and focused on self-management apps for mental health monitoring and tracking, and for collecting data from the patient. Fewer studies explored use of apps within-session, or practitioner-guided apps. Practitioners prioritized app features aligned with the American Psychological Association's evaluation criteria, with practitioners prioritizing engagement and interoperability, but also noted the importance of training and resourcing to support integration.

Conclusions: While practitioners recognize the potential of apps in mental health care, integration into clinical practice remains limited. This study highlights the need for further research on practical implementation, clinical effectiveness, and practitioner training to facilitate the transition from potential to actual use of apps in mental health care settings. Recommendations include evaluating effectiveness of app integration through experimental studies and developing training modules to develop practitioners' digital competencies and confidence in app use.