BACKGROUND: The use of large language model (LLM)-based artificial intelligence (AI) to assess the quality of outpatient mental health documentation represents an understudied application. Critical questions remain about the effects of exposing human auditors to AI-generated recommendations on auditor efficiency, agreement, and patterns of decision-making.

OBJECTIVES: The objective of this study is to examine how exposure to LLM-generated quality assessments influences auditor performance during reviews of documents of outpatient mental health intake sessions.

METHODS: We conducted a pre-post observational evaluation of 10 trained human auditors who reviewed outpatient mental health intake notes and assigned quality ratings using a standardized 13-item chart audit rubric before and after implementation of LLM support. The AI system used Anthropic Claude Sonnet 4 with a fixed chain-prompting strategy. We assessed item-level pass frequency for each auditor across the pre-AI and post-AI periods, AI-human agreement (the alignment between auditor pass/fail determinations and AI-generated recommendations), and auditing time. The pre-AI period included 9,711 notes (126,243 item-level observations) and the post-AI period included 2,677 notes (34,801 item-level observations).

RESULTS: Following implementation of AI-supported auditing, the mean item-level pass rate increased by approximately 2%. The distribution of auditor-level changes was broad (range: -35 to +35%, standard deviation [SD]: 10%) indicating substantial heterogeneity in individual auditor responses to AI support. High auditor agreement with AI recommendations was associated with larger pass-rate changes for selected rubric items, with some auditors increasing and others decreasing pass rates relative to the pre-AI period. The average time for an auditor to review notes decreased from 8.77 min (SD: 9.64) per review during the pre-AI period to 8.06 min (SD: 10.28) during the post-AI period (p = 0.002), an 8% improvement equivalent to 71 min saved per 100 notes reviewed.

CONCLUSION: LLM-based AI tools may improve the efficiency of clinical quality assurance workflows, but exposure to AI recommendations may meaningfully change human auditor behavior.

The Suicide Safety Planning Intervention (SPI) is an evidence-based intervention associated with reduced suicidal behavior and improved treatment engagement, yet provider fidelity in delivering SPI remains inconsistent and opportunities for structured skills practice are limited. This study evaluated whether Virtual Standardized Patient (VSP) training improved SPI delivery skills among Veterans Health Administration (VA) behavioral health providers compared with training as usual (UT).

Thirty VA behavioral health professionals without prior formal SPI training were randomized to VSP training or UT. Primary outcomes were changes in SPI fidelity measured by blinded expert ratings using the Safety Planning Intervention Rating Scale (SPIRS) during standardized patient encounters. Secondary outcomes included self-reported knowledge, confidence, satisfaction, and usability.

Participants receiving VSP training demonstrated greater improvements in SPI fidelity than UT participants for both general safety planning skills and safety plan construction skills. More VSP participants achieved competency thresholds in general safety planning skills at follow-up. VSP participants reported significantly greater training satisfaction and usability than UT participants.

VSP training improved provider fidelity in delivering SPI and may represent a scalable strategy for disseminating evidence-based suicide prevention practices across healthcare systems.

Backgroundthis study analyses psychopathological symptomatology and post-traumatic stress disorder (PTSD) in victims of intimate partner violence against women (IPVAW) and examines the presence of school, behavioural and emotional problems in their children.ObjectivesThe aim of the study was to (1) analyse psychopathological symptomatology and PTSD in women survivors of IPVAW, in comparison to women unexposed; (2) analyse the connection between the experience of this type of violence and the presence of psychopathological symptomatology and PTSD; (3) examine the presence of school, behavioural and emotional problems in the children of women survivors of IPVAW compared to the children of unexposed women; and to (4) examine the relationship between psychopathological symptomatology in women survivors of IPVAW and the presence of school, behavioural and emotional problems in their children.DesignA cross-sectional comparative study. Participants were recruited through Women's Care Institutions (experimental group) and schools (control group).MethodThe sample included 29 women survivors of IPVAW and 30 non-exposed women, along with their respective children (n = 60). Maternal psychopathology was assessed with the SCL-90-R and ITQ; violence exposure with the CAS-R-SF; and children's behavioural, emotional and school outcomes with the BASC-3.ResultsWomen survivors of IPVAW and their children present significantly more psychopathological symptoms compared to women and children unexposed to this violence. Furthermore, a relationship was found between the symptomatology in mothers and school, behavioural and emotional problems of their children.ConclusionIPVAW has a broad psychopathological impact on both survivors and their children. This study shows the importance of specifically working on psychological intervention with both.

Health and education are interdependent pillars of human capital development, yet they are often treated in isolation within policy and economic models. This study presents a conceptual and empirical framework that integrates health and education interventions to assess their joint impact on educational attainment, population health, productivity, and economic development. Using a human capital approach, the study models the effects of four school-based health interventions-nutrition, deworming, water, sanitation, and hygiene (WASH), and mental health programmes-on education outcomes in 75 low- and middle-income countries. It also quantifies the long-term health benefits from reduced adolescent pregnancies, due to increased years of schooling, in the form of lower under-five and adult mortality. The immediate health benefits for adolescents of these school-based programmes are not modelled. Using an updated version of an education model combined with an employment benefits model, the study estimates the costs and returns of investing in these interventions from 2025 to 2050. Results show the interventions alone yield over 422 million additional school completions, with economic benefits exceeding US$15.3 trillion and a benefit-cost ratio (BCR) of 7.0. When health benefits of education in the form of reduced mortality are included, benefits rise to over US$25.2 trillion, with a BCR of 11.5. These findings demonstrate the powerful bidirectional relationship between health and education and support the case for integrated investments. The study also outlines methodological advancements in modelling the combined impact of education and health, while acknowledging limitations such as the focus on market-based outcomes and assumptions of uniform intervention effectiveness. Future research should include immediate economic benefits of health interventions, broader dimensions of human development, including equity, institutional quality, and social capital. The findings offer compelling evidence for policymakers to prioritize coordinated, cross-sectoral strategies that yield high returns in both education and health.

Adolescents aged 10-19 years represent the largest cohort in human history and a critical global resource-yet realizing their potential requires deliberate, sustained investment in their health and well-being. This editorial introduces a six-article supplement in Health Policy and Planning addressing key priorities for adolescent health and well-being policy and programming, with particular attention to young people's own perspectives and evidence from low- and middle-income countries. The supplement opens with findings from the 'What Young People Want Survey', drawing on responses from over 1.5 million young people across 89 countries. Results reveal that adolescents' priorities span multiple domains-especially learning and employability, safety and supportive environments, and health and nutrition-underscoring the necessity of multi-sectoral approaches. Three articles address financing. Together, they demonstrate that integrated health-education interventions can yield returns of at least 11 dollars per dollar invested and that scaling up treatment for adolescent depression and anxiety produces benefit-cost ratios of 15.4 and 13.9 in Colombia and South Africa, respectively-while cautioning that investments must be designed with an explicit equity focus to avoid deepening existing inequalities. Another article presents new World Health Organization and Partnership for Maternal Newborn and Child Health guidance for monitoring adolescent health and well-being holistically across the five domains of the UN conceptual framework, supported by an Excel-based data tool that maximizes use of existing data. The supplement concludes with an early assessment of commitments made at the 2023 Global Forum for Adolescents across four countries, finding that progress is hindered by fragmented systems, funding gaps, and insufficient disaggregated data. The evidence presented in this supplement makes clear that improving adolescent health and well-being is neither aspirational nor unaffordable-it is an investment with compelling returns. Realizing those returns demands that adolescents' voices shape policy and that progress is rigorously monitored.

Despite being established risk factors for poor mental or brain health outcomes in aging, rest-activity rhythm (RAR) disturbances are not routinely monitored or treated. This is, in part, due to a lack of clinician-friendly RAR monitoring systems.

We tested the feasibility and acceptability of personalizing a 6-week behavioral intervention using RAR monitoring from a consumer wearable device (Apple Watch). We selected a target population study of people with dementia and their family caregivers, given that rest-activity pattern disturbances are common in these groups.

This single-arm trial enhanced a behavioral activation rhythm treatment with the Apple Watch-based app myRhythmWatch, providing users and their therapists with objective RAR monitoring for customizing therapy. Therapists used information from the myRhythmWatch app to visualize the participants' behavioral patterns, identify treatment targets, and track progress. Participants included 21 older adults (15 dementia caregivers: mean age 61.8, SD 8.4 years; 6 people with dementia: mean age 81.45, SD 8.5 years). Feasibility outcomes were as follows: (1) proportion adherent enough to assess RARs (defined as ≥3 consecutive valid days with ≥20 hours per day) and (2) the total number of valid days. Acceptability was measured via the Likert scale to gauge participants' satisfaction with the intervention. We secondarily examined preintervention and postintervention changes in depression (9-item Patient Health Questionnaire scores) and insomnia (Insomnia Severity Index) scores among a smaller group of 11 caregivers who completed these measures.

All 21 participants obtained the minimum data requirement for characterizing an RAR snapshot. Caregivers averaged 35 valid RAR monitoring days, and all 15 caregivers were still using the app at week 6 of the trial. In contrast, people with dementia averaged 30 valid days, and of the 6 people with dementia, only 6 were active users at the end of the trial. On average, caregivers completed 5.7 (SD 0.46) of the 6 therapy sessions offered, and participant satisfaction with program components was "high." Depression symptoms improved with medium preintervention and postintervention effect sizes (t₁₀=2.417; P=.02; Hedges g=0.67, 95% CI 0.04-1.28), and there were large effect size improvements in insomnia symptoms (t₁₀=3.377; P=.004; Hedges g=0.94, 95% CI 0.25-1.61).

These findings show that older adults without dementia were highly engaged with RAR monitoring. This supports the feasibility of personalizing interventions for older adults with objective RAR monitoring. Randomized controlled trials are warranted to determine whether adding RAR monitoring improves intervention efficiency, efficacy, or durability. While feasible in a subset of people with dementia, we observed lower use rates indicating that there are more barriers to implementing long-term consumer wearable-based RAR monitoring in people with dementia.

Background　To compare the prevalence of depressive symptoms among older adults with and without rheumatoid arthritis (RA) in 2019 and 2022 using large-scale population-based data.Methods　This repeated cross-sectional study used data from the Japan Gerontological Evaluation Study 2019 and 2022. A questionnaire was mailed to community-dwelling adults aged ≥65 years, identifying patients with RA undergoing treatment using self-reported information. Depressive symptoms were defined as a Geriatric Depression Scale (GDS) score ≥5. Multivariable logistic regression included an interaction term (RA × survey year) to assess whether the association between RA and depressive symptoms differed by year, adjusting for sociodemographic, health, and lifestyle factors.Results　The number of participants with and without RA were 433 and 16,958 in 2019 and 472 and 16,881 in 2022 (total: 34,744). From 2019 to 2022, depressive symptoms increased among those with RA (35.5% to 40.4%), but changed minimally among those without RA (25.6% to 26.2%). The RA × year interaction was positive (OR 1.37, 95% CI 1.02-1.83; P = 0.036), indicating a greater increase in depressive symptoms among older adults with RA than among non-RA peers. Exploratory stratified analyses showed significant RA-by-year interactions among participants aged ≥75 years, those living alone or unmarried, those with lower educational attainment, never-smokers, and urban residents.Conclusions　The difference in depressive symptoms between older adults with and without RA was greater in 2022 than that in 2019. Public health and clinical strategies that incorporate mental health and social support should be considered for vulnerable older adults with RA.

Attention-deficit/hyperactivity disorder (ADHD) is characterized by gray matter alterations and executive dysfunction. Whether the spatially distributed regional abnormalities reported in previous meta-analyses converge within large-scale networks of the developing brain remains unclear. Using coordinate-informed functional connectivity network mapping (FCNM), we mapped the network architecture of meta-analytic gray matter volume (GMV) and executive dysfunction coordinates in pediatric ADHD and evaluated its reproducibility and cross-disorder specificity.

We extracted neuroimaging coordinates from published meta-analyses reporting GMV alterations and task-based executive dysfunction in pediatric ADHD (n_cases = 5,015, n_controls = 5,915). To map these disparate regional coordinates onto common functional circuits, we applied FCNM utilizing a large-scale, high-quality pediatric normative connectome from the Chinese Child Brain Development (CCBD) project (n = 2,120). The spatial overlap between the derived abnormality networks and canonical brain networks was quantified using Dice coefficients. Furthermore, network robustness was validated using independent pediatric clinical cohorts (from the ADHD-200 dataset and CCBD), and disease specificity was assessed via cross-disorder connectomic comparison with autism spectrum disorder (ASD).

Despite the spatial dispersion of the input coordinates, the FCNM-derived connectivity maps showed differential overlap with large-scale brain networks. Specifically, the network associated with gray matter alterations localized predominantly to the default mode network (DMN) (Dice = 0.457; P_NCT = 0.001) and the ventral attention network (VAN) (Dice = 0.237; P_NCT = 0.03), while the network underlying executive dysfunction localized primarily to the VAN (Dice = 0.319; P_NCT = 0.001). Importantly, this dual-network architecture demonstrated robust reproducibility across independent clinical validation cohorts and exhibited clear topological specificity when compared to ASD-related networks.

Using a large pediatric normative connectome, coordinate-informed FCNM mapped ADHD-related abnormalities onto partially overlapping DMN and VAN architectures, providing a reproducible framework for understanding distributed abnormalities in ADHD and identifying biologically defined targets for neuromodulation and early intervention.

Functional outcomes after manic episodes in bipolar disorder (BD) are heterogeneous and likely reflect the interaction between illness burden and biological processes, including systemic inflammation. However, the contribution of immunometabolic factors to functional remission remains unclear.

To identify clinical and immunometabolic predictors of functional remission in patients hospitalized for manic episodes.

We conducted a retrospective cohort study of 680 manic episodes. Functional outcomes at discharge were assessed using the Global Assessment of Functioning (GAF score ≥ 70). Clinical, demographic, and laboratory data were extracted from routine records. Multivariable logistic regression models were used to identify independent predictors of functional remission.

Functional remission was achieved in 72.6% of manic episodes. In the final multivariable logistic regression model, remission was independently associated with shorter illness duration, shorter duration of untreated bipolar disorder (DUBD), higher GAF scores at admission, higher platelet-to-lymphocyte ratio (PLR), lower leukocyte levels, female sex, and absence of valproate treatment. However, in mixed-effects analyses accounting for repeated manic episodes, only illness duration, DUBD, GAF score at admission, and PLR remained statistically significant predictors of functional remission. The combined model showed acceptable discriminative ability for predicting functional remission (AUC = 0.780 95% CI 0.740-0.819).

Functional remission in BD appears to be primarily determined by illness burden, particularly indicators of illness chronicity. Among immunometabolic factors examined, PLR showed the most consistent association with remission outcomes. These findings suggest that immunometabolic markers may contribute to our understanding of the biological mechanisms underlying functional recovery, although their added predictive value beyond clinical variables appears to be limited.