https://mental.jmir.org/issue/feed 2025-01-03T10:15:04-05:00 JMIR Publications editor@jmir.org Open Journal Systems Unless stated otherwise, all articles are open-access distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work ("first published in the Journal of Medical Internet Research...") is properly cited with original URL and bibliographic citation information. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included. Internet interventions, technologies, and digital innovations for mental health and behavior change. JMIR Mental Health is the official journal of the Society of Digital Psychiatry .  https://mental.jmir.org/2026/1/e79501 2026-06-02T15:15:14-04:00 Laurence Astill Wright Mat Rawsthorne Neil Nixon Boliang Guo Richard Morriss

Background: Ambulatory assessment and active and passive monitoring all offer a real-time, flexible approach to assessing mood and behavior in mood disorders. Despite their potential, concerns remain regarding the performance, usability, adherence, and potential safety of these tools. Objective: This study synthesizes the findings from 7 systematic reviews, integrating quantitative and qualitative data from randomized trials, observational studies, and user experience research to evaluate the performance, feasibility, acceptability, and clinical impact of ambulatory assessment and mood monitoring in people with depression and bipolar disorder. We assessed studies over the medium or long term (3 months or more). Methods: A summary of a series of systematic reviews was carried out by the authors—including meta-analyses (for quantitative data) and meta-syntheses (for qualitative data). Eight electronic databases were searched, and mixed methods studies were included. Studies were assessed for risk of bias. The results were checked for coherence, and recommendations were made by individuals with lived experience, methodologists, and psychiatrists. GRADE (Grading of Recommendations Assessment, Development, and Evaluation) was used to assess the quality and strength of the evidence. Results: The 111 included studies included 19,945 participants and used 69 different ambulatory assessment protocols or mood-monitoring interventions. Key barriers to implementation were identified, including performance inconsistency, adverse effects, and user disengagement. Evidence-based recommendations are provided to guide future clinical and research applications. Conclusions: Ambulatory assessment and mood monitoring hold promise in research and clinical practice, yet their implementation requires more rigorous evaluation, greater personalization, and responsible, user-centered design. Crucially, these measures can add granularity and confirmation, but additional context is often required, and none of these measures are robust enough yet to replace current outcomes.

2026-06-02T15:15:14-04:00 https://mental.jmir.org/2026/1/e87586 2026-06-02T09:45:14-04:00 Andrey Kormilitzin Dan W Joyce Apostolos Tsiachristas Rohan Borschmann Navneet Kapur Galit Geulayov

Background: Self-harm is the strongest risk factor for suicide and an important outcome for mental health care. Although prevalent in clinical populations, it is often imprecisely captured in routinely collected clinical data, where it is often recorded and stored as unstructured free text. Contemporary language models, such as GPT (OpenAI) and Gemini (Google), can analyze free-text clinical notes, but such models may violate data governance of processing sensitive patient data. Objective: This study aimed to evaluate whether a privacy-preserving language model running entirely within an institution’s secure computing infrastructure (here, the UK National Health Service [NHS]) could accurately identify the presence and timing of self-harm using electronic health records from secondary mental health care. Methods: Clinical notes were drawn from Oxford Health NHS Foundation Trust using a multistage workflow: (1) a random sample of 1000 patients with a psychiatric diagnosis, defined according to the (; codes F00–F99); (2) candidate-note identification using a Gemma3-4b language model to flag notes containing self-harm content; and (3) from those candidates, 1352 randomly sampled notes were selected for expert annotation, resulting in gold-standard corpus enriched for self-harm content. Clinical notes were annotated for the presence of self-harm and its timing (≤90 days, >90 days, or unknown). A privacy-preserving locally served 27-billion-parameter Gemma 3 language model (“Gemma3-27b”) was used as the core model. Prompts were systematically developed and refined using a labeled development set to identify self-harm and generate a structured output per clinical record. Gemma3-27b performance was compared against a strong baseline multilabel text classification model based on robustly optimized BERT pretraining approach (RoBERTa), a transformer-based language model architecture. Model performance was evaluated using precision, recall, and the -score (harmonic mean of precision and recall), with 95% CIs estimated from 1000 bootstrap samples with replacement. Results: Gemma3-27b outperformed the RoBERTa classifier across all categories, achieving Precision=0.92, Recall=0.92 (sensitivity), and -score=0.92 for notes containing self-harm, and Precision=0.97, Recall=0.97 (specificity), and -score=0.97 for notes without self-harm. For the 51 notes labeled as recent self-harm in the held-out test set, Gemma3-27b achieved Precision=0.84, Recall=0.75, and -score=0.79. The global weighted -score of Gemma3-27b across all categories was 0.88, compared to 0.85 for RoBERTa. Conclusions: With systematic prompt development on a labeled development set, but no gradient-based fine-tuning, the current Gemma3-27b language model matched or exceeded a fine-tuned RoBERTa classifier for ascertaining self-harm events and their timing. Aggregate gains were modest, while improvements were largest in the most challenging, lower-frequency timing categories. On a simplified binary recent-versus-other task, RoBERTa performed marginally better, indicating that supervised classifiers remain highly effective when the task is simplified and sufficient labeled data exist. This work demonstrates the technical feasibility of privacy-preserving self-harm detection within a secure NHS research environment.

2026-06-02T09:45:14-04:00 https://mental.jmir.org/2026/1/e70207 2026-05-28T18:00:05-04:00 Leah H Rubin Pauline M Maki Joan Severson Adam Lieberman Eran F Shorer Sabina A Haberlen Deborah R Gustafson Michelle Floris-Moore Valentina Stosor Matthew J Mimiaga Jamie Peven Deborah L Jones Anandi N Sheth Kathleen M Weber Amanda B Spence Anjali Sharma David E Vance Raha M Dastgheyb

Background: Digital cognitive assessments are increasingly used in large-scale studies to assess brain health, offering scalable, standardized, and self-directed testing solutions. Cognitive function remains a concern for people with HIV despite antiretroviral therapy. The BRACE (BrainBaseline Assessment of Cognition and Everyday Functioning) is a validated tablet-based screener for cognition in people with HIV. Preliminary pilot norms were established in a small sample (n=144), but full regression-based normative data have not yet been developed. Consequently, HIV serostatus differences based on standardized BRACE scores and cognitive correlates have not been systematically examined. Objective: This study aims to develop regression-based normative data for BRACE performance in people without HIV who were demographically and behaviorally comparable to people with HIV within biological sex; to examine differences in cognitive performance by HIV status and biological sex; and to evaluate sociodemographic, behavioral, and clinical correlates of BRACE performance. Methods: A total of 2937 participants (1063 people without HIV [499 women] and 1874 people with HIV [1053 women]) in the Multicenter AIDS Cohort Study/Women’s Interagency HIV Study Combined Cohort Study completed BRACE once between November 2020 and March 2025. BRACE includes the Trail Making Test (A and B), Stroop-Color, and visual spatial learning. Regression-based norms were derived from people without HIV using multiple demographic models (eg, age-only, age + education, and age + education + sex). The age + education model was selected for primary analyses because it provided the best balance of interpretability, parsimony, and generalizability while avoiding race-based corrections. HIV serostatus and sex differences were examined using ANOVA and tests, with effect sizes calculated using Cohen’s . Results: Cognitive performance was largely comparable between people with HIV and people without HIV across all BRACE outcome measures. Statistically significant differences were very small in magnitude (all effect sizes<0.11) and primarily observed among men on Stroop-Color. Across groups, older age and fewer years of education were associated with poorer raw BRACE performance, although these associations attenuated after demographic adjustment using T-scores. Most clinical and behavioral factors (eg, hypertension, smoking, and noncannabis substance use) were related to poorer raw scores but not standardized performance. However, diabetes and cannabis use remained independently associated with T-scores across multiple measures—diabetes with poorer scores and cannabis use with higher scores, an association that should be interpreted cautiously. HIV-specific clinical factors, such as nadir CD4 count and antiretroviral therapy duration, were linked primarily to raw scores. Conclusions: This study establishes the first regression-based normative data for BRACE, derived from a large, demographically diverse people without HIV, and demonstrates its applicability for evaluating cognitive function in people with HIV. Findings indicate minimal cognitive differences between people with HIV and people without HIV and highlight the influence of common sociodemographic and metabolic factors. These results support BRACE as a scalable, reliable, and self-administered digital tool for assessing cognitive health in diverse populations and underscore its potential for longitudinal monitoring and precision phenotyping in both research and clinical contexts.

2026-05-28T18:00:05-04:00 https://mental.jmir.org/2026/1/e96894 2026-05-26T12:00:24-04:00 Hina Tahseen

Research on artificial intelligence (AI) and mental health has focused largely on harms at deployment, including chatbot safety, sycophancy, and AI-associated delusions. Less attention has been paid to a prior question: whether the human-generated text and preference judgments that shape large language models are themselves clinically reliable, particularly when self-report may be distorted. This Viewpoint aims to develop the clinical psychiatric construct of collusion—the uncritical acceptance of an unreliable account—as an analytic lens for AI training and deployment, and to argue that the clinical reliability of training and preference data should be treated as an explicit trustworthy-AI criterion in mental-health–relevant systems. A conceptual synthesis of psychiatry, clinical psychology, and AI safety literature was undertaken. The analysis distinguishes three pipeline layers: pretraining corpora, preference data and posttraining methods, and deployment-time interaction. It maps the clinical construct of collusion against adjacent technical concepts, including sycophancy, reward overoptimization, grounding, refusal training, red-teaming, and live monitoring. The synthesis suggests that collusion-like dynamics are least applicable at the pretraining layer and most applicable at the preference-data and deployment layers, where unassessed user or labeler input can be reinforced without corroboration. Existing mitigations, including data curation, Constitutional AI, reward-model evaluation, grounded generation, refusal training, red-teaming, and postdeployment monitoring, address parts of this problem. However, these approaches are not yet organized around a clinically informed account of when self-report is unreliable. The central novelty is therefore not a generic claim about bias, but the proposal that clinical self-report reliability should be assessed as a distinct data-quality and governance dimension. Trustworthy-AI frameworks for mental-health–relevant applications should incorporate clinical expertise in self-report reliability into preference-data design, red-teaming, and postmarket surveillance. Adding the clinical reliability of training and preference data as an explicit criterion could complement existing technical safeguards while leaving empirical evaluation of clinician involvement as an open research agenda.

2026-05-26T12:00:24-04:00 https://mental.jmir.org/2026/1/e89928 2026-05-20T16:30:15-04:00 Angela Nickerson Gulsah Kurt Philippa Specker Anna Camilleri Dessy Susanty Rizka Argadianti David Keegan Randy Nandyatama Atika Yuanita Angga Putra Reynady Hermawan Richard A Bryant Meaghan O'Donnell Mahdi Rafei Rahmatullah Bayani Belinda J Liddell

Background: Common mental health disorders are highly prevalent among refugees. There is an urgent need to address the mental health burden in this population. Objective: This study tested the efficacy of an individually supported self-help stress-management intervention developed by the World Health Organization—Doing What Matters in Times of Stress (DWM)—in reducing psychological distress and improving functioning among refugees in Indonesia, a major transit country in the Asia-Pacific region. Methods: A single-blind randomized controlled trial with 303 Farsi-speaking refugees was conducted between June 2024 and June 2025. Participants with moderate to high psychological distress (Kessler Psychological Distress Scale [K10] score≥20) were randomly allocated to the facilitator-guided individual DWM condition (n=202) or a repeated assessment control condition (n=101). The primary outcome was psychological distress (K10 score) at the posttreatment assessment. Secondary outcomes were posttraumatic stress disorder symptoms, functional impairment, social functioning, and personally identified problems. Results: Intent-to-treat analysis indicated that participants in the DWM condition showed greater reductions in K10 scores than those in the repeated assessment control condition (posttreatment: =−.563, SE=0.124; <.001; Cohen =0.56; 1-month follow-up: =−.447, SE=0.140; =.002; Cohen =0.45). Similarly, those participants in the DWM condition reported greater improvements in posttraumatic stress disorder symptoms, well-being, social functioning, functional impairment, and personally identified psychological problems. No serious adverse events were reported. Conclusions: The findings provide the first evidence for the effectiveness of DWM in reducing psychological distress and improving overall functioning among urban refugees living in a transit setting. Individually supported self-help interventions such as DWM mayoffer an effective, feasible, and scalable approach to improving mental health for refugees. Trial Registration: ANZCTR ACTRN12624000609550; https://www.anzctr.org.au/Trial/Registration/TrialReview.aspx?id=387637&isReview=true

2026-05-20T16:30:15-04:00 https://mental.jmir.org/2026/1/e91367 2026-05-19T17:01:46-04:00 Daniel Szoke Ilana Hutzler Jerry Liu Samantha Addante Zuhaib Akhtar Dale L Smith Kirsten Dickins Charles Small Sarah Pridgen Philip Held

<strong>Background:</strong> Artificial intelligence (AI)–based conversational tools are rapidly expanding within mental health care as a means of increasing access and scalability. At the same time, these systems introduce distinct safety risks arising from both user disclosures (eg, self-harm ideation) and inappropriate or inadequate AI responses. <strong>Objective:</strong> This study aimed to develop and evaluate the Automated Safety Testing and Reporting Application (ASTRA), an external system intended to identify clinically relevant risk behaviors across entire AI-mediated mental health conversations. <strong>Methods:</strong> ASTRA was tested on a dataset of 100 synthetic therapeutic conversations written by licensed clinicians to reflect risk behaviors and harmful responses between users and AI tools. Conversations varied in length and included both subtle and overt risk behavior examples across 8 predefined categories. Human coder consensus ratings served as the reference standard. ASTRA’s classifications were evaluated across 2 prompt iterations using standard diagnostic performance metrics and agreement statistics. <strong>Results:</strong> ASTRA demonstrated consistently high concordance with expert human ratings across all categories. Accuracy exceeded 0.90 for all risk behavior categories examined, with specificity uniformly high and sensitivity varying by category (range 0.55-1.00). Agreement beyond chance was substantial to almost perfect between ASTRA and human raters (κ=0.65-1.00). Detection of user self-harm indicators was particularly accurate, even in conversations where risk was expressed subtly. <strong>Conclusions:</strong> In this initial validation study, ASTRA reliably identified multiple forms of mental health–related risk behaviors at the conversation level. These findings support the feasibility of independent safety monitoring systems as a complement to AI tools used in mental health contexts and underscore the need for further evaluation using larger and real-world datasets.

2026-05-19T17:01:46-04:00 https://mental.jmir.org/2026/1/e82370 2026-05-19T16:30:03-04:00 Brian D Kiluk Lara A Ray Omeed Tartak Lovisa Werner Thomas A Trikalinos Molly Magill

<strong>Background:</strong> Cognitive behaviorally based interventions have broad appeal and potential for impact when treating adult alcohol and other drug use. Digitally delivered cognitive behaviorally based interventions (dCBIs) may offer this impact with the benefit of increased accessibility. Although prior reviews have indicated the benefits of dCBIs on substance use outcomes, the extension to psychosocial functioning outcomes is unknown. <strong>Objective:</strong> This meta-analysis provides an overview of dCBI effects across a range of functional end points. <strong>Methods:</strong> A literature search was conducted through October 2024. All primary and secondary reports of clinical trials of dCBI were obtained, and all available study end points were eligible for meta-analysis. Descriptive data were extracted and categorized into 1 of 13 different outcome types (eg, abstinence, quantity, cognitive, and quality of life) and into 2 broader outcome classes (ie, consumption and psychosocial). Robust variance estimation was used to conduct hypothesis tests on random effects pooled estimates with outcome class and comparison type as the primary subgroup variables of interest. <strong>Results:</strong> The study sample included 65 randomized trials (<i>K</i>=110 publications; 753 effect sizes) of dCBI for adult alcohol and other drug use. With respect to efficacy, dCBI as a stand-alone treatment in contrast to a minimal treatment control showed positive and statistically significant effects for consumption (<i>g</i>=0.27; <i>P&lt;</i>.001; <i>I</i>²=85.1%; <i>k</i>=31; <i>k_es</i>=134) and psychosocial (<i>g</i>=0.16; <i>P</i>=.008; <i>I</i>²=75.2%; <i>k</i>=16; <i>k_es</i>=60) outcomes. As an addition to usual care, efficacy was demonstrated for consumption (<i>g</i>=0.23; <i>P&lt;</i>.001; <i>I</i>²=9.8%; <i>k</i>=20; <i>k_es</i>=65), but not psychosocial functioning. Efficacy compared to another digital or in-person intervention or cognitive behaviorally based intervention delivered by a therapist was not observed. Within the dCBI condition, large effect sizes were observed for both outcome classes (ie, 60%-80% of participants showed improvement relative to baseline), and effect size magnitude and statistical heterogeneity varied by the type of outcome examined. <strong>Conclusions:</strong> These results show a benefit for dCBI as a stand-alone therapy and an addition to usual care. Importantly, stand-alone effects were observed for both consumption and some psychosocial outcomes. This study is the first to offer a comprehensive look at dCBI intervention effects across a range of functional end points.

2026-05-19T16:30:03-04:00 https://mental.jmir.org/2026/1/e91716 2026-05-19T15:30:10-04:00 Ibrahim Arpaci Ismail Kuşci Kasim Karataş Mustafa Baloglu

Background: Positive aging, a concept found in positive psychology, serves as the theoretical foundation for this study. To age positively, one must manage hidden or unrecognized challenges, show flexibility in behavior and thought, adopt a positive outlook on problems involving regression, and make decisions that promote one’s well-being. Objective: This study examined the role of wisdom and life purpose in the mental well-being of middle-aged and older adults. More specifically, we tested 4 hypotheses: wisdom would exhibit a positive correlation with mental well-being, quality of life would exhibit a positive correlation with mental well-being, meaning and purpose would exhibit a positive correlation with mental well-being, and freedom would exhibit a positive correlation with mental well-being. Methods: The research used a multianalytical methodology combining covariance-based structural equation modeling and artificial neural network techniques to analyze data from 377 individuals aged 50 to 102 years. Results: Results from the covariance-based structural equation modeling indicate that meaning and purpose, wisdom, and quality of life were significantly associated with the mental well-being, accounting for 71% of the explained variance. Additionally, the artificial network analysis yielded exact forecasts of mental well-being. The artificial network model achieved an accuracy of 82.1% and 73% on the training and test sets, respectively, for predicting mental well-being. Sensitivity analysis revealed that meaning and purpose were the most critical factors in explaining participants’ mental well-being. Conclusions: These findings have prominent theoretical implications for social psychology researchers and practical consequences for authorities involved in the care of older adults, who can use the results to develop strategic plans and take necessary actions.

2026-05-19T15:30:10-04:00 https://mental.jmir.org/2026/1/e88731 2026-05-15T17:30:17-04:00 Nan Cheng Mary K Lam Christine Grove Monica Wachowicz

Background: Digital mental health tools (DMHTs) offer scalable support, but engagement varies. Understanding the shapes of initiation and ongoing use is essential for effective design and implementation. Objective: This study aims to synthesize determinants of adults’ initiation and engagement with DMHTs, organized through two lenses: (1) psychological factors aligned with the theory of planned behavior (TPB) and (2) design and access features. Methods: A systematic search of 9 databases (June 2025) identified qualitative and mixed methods primary studies reporting end-users’ experiences with DMHTs. Studies were screened and reported in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. Quality appraisal used quality assessment with diverse studies (QuADS). Data were synthesized using a framework-guided thematic approach, mapping findings to TPB constructs and complementary design and access domains. Results: A total of 22 studies met inclusion criteria. Findings clustered into 2 interdependent domains. TPB constructs explained how beliefs, social expectations, and perceived control shaped decisions to start and persist with DMHTs. Design and access features frequently acted through these same pathways, especially by altering perceived behavioral control (PBC), with cost, connectivity, device constraints, and time flexibility affecting feasibility, with content design and privacy shaping perceived value and trust. Perceived fit (goals, cultural or linguistic relevance, and routine alignment) consistently influenced both initiation and continuation. Several features operated bidirectionally; depending on context, the same feature could facilitate or hinder engagement. Conclusions: Engagement with DMHTs is jointly determined by users’ beliefs and the design and access conditions within which tools are offered. Implementation should pursue a dual strategy, strengthening willingness to seek support (addressing attitudes, norms, and perceived control) while engineering low-effort, trustworthy, and context-appropriate experiences. Priorities include equity-focused policies (data costs, devices, and connectivity), transparent data practices, co-design with diverse communities, and consistent, theory-informed outcome measures.

2026-05-15T17:30:17-04:00 https://mental.jmir.org/2026/1/e88057 2026-05-15T13:30:29-04:00 Matheus Calvin Lokadjaja Jordon Junyang Kho Peter Johannes Schulz Wilson Wen Bin Goh

Background: Large language models (LLMs) are poised to transform mental health care, offering advanced capabilities in diagnosis, prognosis, and decision support. Since their inception, numerous mental health-focused LLMs have emerged in the scientific literature, reflecting the growing interest in leveraging these models across various clinical applications. With a broad range of models available, diverse optimization strategies, and multiple use cases, reviewing the current landscape is critical to understanding where future impact lies. Objective: This study aimed to conduct a scoping review investigating the use of LLMs in mental health across diagnostic, prognostic, and decision support tasks. Methods: We screened 3121 papers from PubMed, Scopus, and Web of Science for studies published between January 2023 and October 2025, using terms related to LLM and mental health. After removing duplicates, 2 reviewers (MCL and WWBG) independently screened the studies, with a third (JJK) to resolve conflicting opinions. We extracted and synthesized information on the models, use cases, datasets, and adaptation methods from selected papers. Results: In total, 41 papers were selected. Many studies included evaluations on OpenAI’s GPT series applications: GPT-4 (24 studies, 58.5%) and GPT-3.5 (16 studies, 39%). Others included Bidirectional Encoder Representations from Transformers-derived models (9 studies, 22%), LLaMA (8 studies, 19.5%), and RoBERTa-derived models (6 studies, 14.6%). While all studies initially applied out-of-the-box LLMs, several adapted them through few-shot learning or fine-tuning to better align with specific research goals. The most common use case was in diagnostics (31 studies, 75.6%), while the most common target condition was depression (11 studies, 26.8%). While many studies reported superior performance of LLMs, only a minority of studies (13 studies, 31.7%) validated LLM performance against clinician assessments using real patient data, with the majority relying on proxy outcomes such as clinical vignettes, examination questions, or social media posts. Conclusions: Despite rapid growth and diversity of LLM applications in mental health, the field remains nascent and exploratory. Future developments must emphasize consistent model adaptation procedures to ensure safety and clinical workflow alignment. Models must also be evaluated on robust evaluation criteria by using standardized protocols and real clinical outcome measures.

2026-05-15T13:30:29-04:00