BackgroundNeuromodulation techniques (NMTs) have shown potential therapeutic value in Alzheimer’s disease (AD), but the comparative efficacy of different modalities remains unclear. This study aimed to systematically compare and rank available NMTs to identify their relative benefits across cognitive, functional, and neuropsychiatric outcomes.ObjectiveTo evaluate and rank the efficacy of NMTs in improving cognitive function, activities of daily living (ADLs), and neuropsychiatric symptoms in AD patients, thereby providing evidence-based guidance for clinical intervention strategies.MethodsEnglish-language literature was systematically searched in PubMed, Web of Science, Cochrane Library, and EMBASE from database inception to February 2026. Inclusion criteria were randomized controlled trials (RCTs) of AD patients aged ≥18 years. Interventions included repetitive transcranial magnetic stimulation (rTMS), transcranial alternating current stimulation (tACS), transcranial direct current stimulation (tDCS), photobiomodulation (PBM), transcutaneous electrical stimulation (TES), transcranial ultrasound stimulation (TUS), transcranial pulse stimulation (TPS), deep brain stimulation (DBS), and sham stimulation (Sham). Outcome measures included Mini-Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA), Alzheimer’ s Disease Assessment Scale-Cognitive Subscale (ADAS-Cog), Alzheimer’ s Disease Cooperative Study-Activities of Daily Living (ADCS-ADL), Boston Naming Test (BNT), Neuropsychiatric Inventory (NPI), and Geriatric Depression Scale (GDS). Data extraction and risk-of-bias assessment followed Cochrane guidelines. Network meta-analysis (NMA) was conducted in Stata 17.0, with effect sizes ranked by the surface under the cumulative ranking curve (SUCRA).ResultsA total of 36 RCTs involving 1,445 patients were included in the analysis. Compared with Sham, TPS significantly improved MMSE (MD = 3.92, 95% CI: 1.57 to 6.35; SUCRA = 97.50%), MoCA (MD = 4.99, 95% CI: 2.26 to 7.72; SUCRA = 69.25%), and BNT scores (MD = 4.02, 95% CI: 1.81 to 6.23; SUCRA = 81.8%). rTMS also showed favorable effects on ADAS-Cog scores (MD = −11.10, 95% CI: −14.51 to −7.71; SUCRA = 91.00%). Regarding activities of daily living, TES ranked highest on the ADCS-ADL scale (SUCRA = 71.00%). In terms of neuropsychiatric symptoms, TPS achieved the highest ranking for improving NPI scores (SUCRA = 92.3%), whereas tACS showed the best performance on GDS scores (SUCRA = 69.5%).ConclusionNeuromodulation interventions demonstrated domain-specific benefits in AD, with no single approach showing universal superiority. TPS demonstrated the greatest potential for improving global cognitive outcomes, while rTMS showed advantages in ADAS-Cog performance. Nevertheless, limited sample sizes, heterogeneous stimulation protocols, and insufficient long-term evidence restrict the generalizability of current findings. Future multicenter RCTs with standardized protocols and prolonged follow-up are needed to confirm these results and guide clinical implementation.Systematic review registrationPROSPERO, CRD420261364193.
Large language models for late-life depression: a blinded benchmark of clinical safety, geriatric appropriateness, and triage
BackgroundGeneral-purpose large language models are increasingly used by patients and caregivers to obtain mental health information and guidance about when professional care is required. In late-life depression, broadly accurate information may nevertheless be unsafe when cognitive change, multimorbidity, frailty, polypharmacy, self-neglect, caregiver dependence, or suicide risk is not adequately recognized. This study compared the clinical accuracy, safety, geriatric-specific appropriateness, triage performance, and response consistency of three large language models when answering patient- and caregiver-centered questions about late-life depression.MethodsWe conducted a blinded, paired benchmarking study using 90 questions covering six geriatric psychiatry domains and equally distributed across low-, moderate-, and high-risk strata. Each question was submitted independently to GPT-5.5 Instant via ChatGPT, Gemini 3.5 Flash via Gemini, and Seed2.0 Pro via Doubao, generating 270 primary-round responses. A stratified subset of 30 questions was resubmitted in separate conversations to assess test-retest consistency, yielding 360 responses overall. Two psychiatrists independently evaluated anonymized outputs against prespecified item-specific reference standards, with clinically important disagreements adjudicated by a third senior psychiatrist. The primary outcome was the proportion of clinically acceptable responses, defined using accuracy, clinical safety, geriatric appropriateness, warning-sign recognition, and triage criteria.ResultsClinically acceptable responses were generated for 78.9% of questions by ChatGPT, 72.2% by Gemini, and 60.0% by Doubao (overall P<0.001). The difference between ChatGPT and Gemini was not statistically significant, whereas both outperformed Doubao. This model ranking remained unchanged under alternative core-safety and more stringent optimal-response definitions. Complete geriatric-specific appropriateness was achieved in 64.4%, 55.6%, and 43.3% of responses, respectively. Major safety errors occurred in 5.6% of ChatGPT responses, 10.0% of Gemini responses, and 16.7% of Doubao responses (raw P = 0.015; FDR-adjusted q=0.023). Performance declined substantially with increasing clinical risk. In exploratory post hoc caregiver-centered analyses, clinical acceptability was 76.7% for ChatGPT, 70.0% for Gemini, and 53.3% for Doubao, while explicit caregiver-directed action was present in 80.0%, 70.0%, and 53.3% of responses, respectively. Among high-risk questions, clinically acceptable response rates were 63.3%, 50.0%, and 36.7%, while under-triage occurred in 16.7%, 26.7%, and 36.7% of responses, respectively. Test-retest clinical consistency was highest for ChatGPT (90.0%), followed by Gemini (83.3%) and Doubao (73.3%).ConclusionsThe three models answered many late-life depression questions accurately and safely, but none demonstrated consistently reliable performance across complex or high-risk scenarios. Clinically important weaknesses involved geriatric-specific interpretation, recognition of indirect risk signals, crisis-response completeness, under-triage, and response stability. General-purpose large language models may support selected low-risk educational tasks but should not independently guide emergency triage, medication changes, suicide-risk management, or other safety-critical decisions in older adults.
Recovery-phase symptoms after COVID-19 infection and anxiety levels among college students
ObjectiveThis study examines the relationship between physical symptoms during the recovery phase after COVID-19 infection and anxiety levels among college students, with the aim of providing empirical evidence for post-pandemic mental health screening and health management in universities.MethodsThe study used online survey data collected from college students in March 2023. After excluding straight-line responses, 394 valid questionnaires were retained. Among 372 students with a clear infection status, we examined the association between confirmed COVID-19 infection and anxiety. Among 298 students with confirmed infection, we further analyzed the associations of the number of physical symptoms during recovery, recovery time, and specific physical symptoms with anxiety.ResultsConfirmed COVID-19 infection itself was not significantly associated with a higher anxiety level. Among infected students, however, the number of physical symptoms during recovery was positively associated with higher anxiety levels, whereas recovery time was not significantly associated with anxiety level. Analyses of specific symptoms showed that memory decline and chest pain were significantly associated with higher anxiety levels.ConclusionPost-infection mental health screening in universities should not focus only on whether students have been infected. Greater attention should be paid to the burden of physical symptoms during recovery. Symptom count, cognitive changes, and chest discomfort may serve as useful supplementary indicators for post-infection follow-up and psychological risk identification.
An Evidence-Based AI Virtual Assistant for Young People With Attention Deficit Hyperactivity Disorder: Co-Design and Prototype Development
<strong>Background:</strong> Though attention deficit hyperactivity disorder (ADHD) is thought to be the most prevalent neurodevelopmental disorder in young people worldwide, there are inequalities in access to psychoeducation and health care support. One way to improve access, potentially increase engagement, reduce health care inequalities, and enhance care is by co-developing digital responsive interventions. These have the potential to support long-term condition management and to act as an adjunct to usual care. Virtual assistants that use large language models can provide information in response to questions and learn to tailor communication to suit an individual user’s needs. This can be especially valuable for people with ADHD who often struggle to regulate attention and can experience communication challenges. Involving people with lived experience in the co-design process is crucial for the development of effective digital interventions. Therefore, this article explores the views and preferences of young people with ADHD and their supporters from the United Kingdom who collaborated with researchers to co-design a prototype chatbot. <strong>Objective:</strong> This study aimed to co-develop an evidence-based chatbot prototype, intended to help young people with ADHD thrive through improved access to health care information, psychoeducation, and self-management strategies. <strong>Methods:</strong> An interdisciplinary team was established, including researchers, software developers, clinicians, and lived experience collaborators. Research advisory and working groups were set up in ways that facilitated flexible involvement. Following the person-based approach, guiding principles were established, and workshops were held with young people with ADHD and supporters of young people with ADHD to co-develop an early prototype. Feedback was sought via think-aloud interviews with lived experience collaborators. <strong>Results:</strong> In total, 9 experts by lived experience and 3 health care professionals chose to engage in workshops, and this feedback informed the development of a SmartADHD chatbot prototype. An off-the-shelf chatbot (GPT-4o hosted on Convai) was trained using resources from the National Health Service (NHS). Overall, 6 experts by lived experience engaged with think-aloud interviews, providing feedback on the prototype conversational flow and feel, the avatar, the text-to-speech, the chatbox feature, and the content of the messages. Seven recommendations are made for future development, which will inform the SmartADHD program of work. <strong>Conclusions:</strong> These findings provide rich data on the preferences of people with ADHD. Specific recommendations for a chatbot for young adults with ADHD have not been investigated before with young people, making this study a novel contribution to the field. These findings provide an excellent foundation for chatbot development for this group and may be relevant for those developing digital tools for people with ADHD across the lifespan and other neurodevelopmental conditions. Further work is required to elucidate the views of health care professionals and identify the limits of the technology before subsequent evaluation. <strong>Trial Registration:</strong>
Tool or Companion? Reframing Conversational AI to Prevent Psychological Harm
People increasingly turn to conversational AI for companionship, emotional support, and well-being, using both purpose-built companion apps, such as Replika and Character.AI, and general-purpose assistants, such as ChatGPT and Claude. While some evidence suggests potential benefits, including short-term reductions in loneliness and mood improvement, several adverse outcomes have been reported in both clinical and nonclinical populations, including emotional dependence, exacerbation of symptoms, and self-harm. The fluent and apparently empathic responses from these models lead users to engage with them not only as tools but also as if they were social entities. This framing is conceptually misleading and may pose risks across different user profiles, particularly for vulnerable individuals. Drawing on research in AI, psychiatry, psychology, and network science, we highlight mechanisms through which emotional reliance develops and the boundary between tool and companion erodes. Design choices that evoke personality and warmth encourage users to anthropomorphize these systems. Simulated empathy, generated through probabilistic language patterns rather than genuine emotional experience, creates a structurally asymmetric interaction, in which the user discloses and the system responds, but without reciprocity, vulnerability, or accountability. Overvalidation and sycophancy can reinforce maladaptive cognitions, delusional ideation, and distorted perceptions of reality, as they tend to reinforce people’s beliefs, even at the expense of the accuracy of models’ responses. These mechanisms are not incidental: they emerge from alignment procedures that reward responses perceived as warm and empathic. The result is a self-reinforcing feedback loop between the model and the user that may amplify maladaptive beliefs, delusional ideation, and emotional distress, even in those who engage for largely functional purposes. Understanding these dynamics requires an examination of both what these agents can do—considering their technical limitations and implementations—and what humans believe they can do, including social and psychological impacts. We argue that conversational AI should be treated primarily as a tool supporting human systems rather than as a substitute for human relationships. Perhaps more importantly, reviewing the current hype surrounding AI interactions can help reformulate a paradigm that contributes to human well-being and societal value, while minimizing misconceptions, maladaptive interactions, or social disintegration.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/31acb2c2de9a6960a54052d3939a044a" />
Renaclic for Chronic Kidney Disease Self-Management: Formative User-Centered Co-Design Study
Background: Chronic kidney disease (CKD) affects over 850 million individuals worldwide and requires sustained patient education, shared decision-making, and self-management support. Digital health platforms may facilitate these needs; however, many lack comprehensive, user-centered design and fail to address practical and psychosocial patient concerns. Objective: This study aimed to describe the formative co-design of Renaclic, a digital platform developed with patients and health care professionals (HCPs) to support CKD education and self-management in the French health care context. Methods: A three-stage user-centered design (UCD) study was conducted between April and August 2021. Stage 1 included benchmarking of 10 French-language CKD platforms and semistructured interviews with 8 stakeholders (3 HCPs: 1 nephrologist, 1 general practitioner, 1 nurse; 5 patients: 2 novice and 3 expert users). Stage 2 consisted of two virtual co-design focus groups incorporating card-sorting and interface design activities. Stage 3 translated identified needs into structured content categories and a low-fidelity prototype. Qualitative data were analyzed thematically. Results: Benchmarking revealed fragmented educational content, limited practical guidance on home dialysis logistics, insufficient coverage of lifestyle and psychosocial concerns, and minimal interactive features. Interviews identified five main themes: need for reliable centralized information, anxiety during care transitions, demand for practical lifestyle guidance, importance of reassuring language, and value of moderated peer exchange. More than 30 thematic content items were generated and organized through card-sorting into five consolidated content categories that structured the platform’s information architecture. Co-design sessions informed mobile accessibility, simplified navigation pathways, and integration of educational, symptom-tracking, and community-oriented functionalities. Conclusions: This formative study demonstrates how participatory UCD with diverse CKD stakeholders can generate a structured, context-sensitive digital platform framework. Renaclic provides a replicable model for early-stage co-design of digital health interventions in chronic disease management. Future research will evaluate usability, engagement, and implementation in real-world settings.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/ba7e69d2448a2fd5498156332504853e" />
Autistic traits and obsessive–compulsive symptom severity: a retrospective six-month follow-up study
Autistic traits have been increasingly associated with obsessive–compulsive disorder (OCD), yet their influence on longitudinal symptom severity and treatment-related change remains poorly understood. This retrospective longitudinal study included 38 adults with DSM-5 OCD who completed six months of guideline-based pharmacological treatment. Obsessive–compulsive symptom severity was assessed using the Yale–Brown Obsessive–Compulsive Scale (Y-BOCS) at baseline, one month, and six months. Autistic traits were evaluated at the six-month follow-up using the Autism Spectrum Quotient (AQ). Linear mixed-effects models served as the primary analyses, while repeated-measures analyses of covariance were performed as sensitivity analyses. Exploratory analyses examined the contribution of individual AQ subscales. Higher autistic trait levels were independently associated with persistently greater OCD symptom severity across the six-month follow-up, whereas the rate of treatment-related improvement did not differ according to autistic trait levels. This association was primarily driven by compulsive rather than obsessive symptoms. Exploratory analyses further identified social skills and attention switching as the AQ dimensions most strongly associated with overall OCD severity and compulsive symptom burden, whereas attention to detail was not significantly associated with symptom severity. These findings suggest that autistic traits represent stable markers of greater clinical burden in OCD without necessarily conferring poorer pharmacological treatment responsiveness. The results further support a dimensional conceptualization of autistic traits and highlight social functioning as a potentially important contributor to OCD symptom burden.
Association of iron deficiency and iron-deficiency anemia with behavioral problems in children with autism: a cross-sectional study
PurposeIron deficiency (ID) and iron deficiency anemia (IDA) are common in children with autism spectrum disorder (ASD) and may contribute to behavioral problems. We examined their association with internalizing and externalizing behavioral problems.MethodsThis cross-sectional study (October 2023 to March 2025) was conducted at a tertiary hospital in India. Of 179 children aged 2 to 18 years with newly diagnosed ASD (DSM-5 criteria), 173 were analyzed after six exclusions for missing data. ID and IDA were defined using age-specific World Health Organization thresholds and modeled as a three-category variable with normal iron status as the reference. The Child Behavior Checklist was used to assess internalizing and externalizing problems (T-score > 63). Multivariable logistic regression was used to estimate adjusted odds ratios (aORs), controlling for age, sex, maternal education and family type; ferritin and hemoglobin were modeled using restricted cubic splines. Bonferroni correction was applied across the six adjusted comparisons.ResultsAmong 173 children (median age 4.0 years [IQR 3.0–5.0]; 79.8% male), 19.7% had ID, 22.5% had IDA and 42.2% had either; internalizing and externalizing problems occurred in 39.9% and 35.8% of the children, respectively. Compared with normal iron status, both ID and IDA were associated with higher odds of internalizing problems (ID aOR 3.03, 95% confidence interval [CI] 1.34–6.86; IDA aOR 2.73, 95% CI 1.23–6.05) and externalizing problems (ID aOR 3.71, 95% CI 1.63–8.46; IDA aOR 3.24, 95% CI 1.45–7.24), corresponding to prevalence differences of 25 to 29 percentage points. An interquartile increase in ferritin (9.2 to 21.4 µg/L) was associated with lower odds of both outcomes (internalizing aOR 0.23, 95% CI 0.11–0.46; externalizing aOR 0.30, 95% CI 0.15–0.60). Five of six associations survived Bonferroni correction; IDA with internalizing problems did not. Hemoglobin was not consistently associated with either outcome.ConclusionsID and IDA are prevalent in Indian children with ASD and are associated with clinically significant behavioral problems, with higher serum ferritin inversely associated with both outcomes. Routine iron screening may be considered during initial ASD evaluation, particularly in high-burden settings, although randomized trials are needed to determine whether correcting deficiency improves behavioral outcomes.
Neurodiversity as an epistemic stress test for psychotherapy and health care: a neuroaffirmative conceptual analysis of relational-proximity priors in clinical, institutional, and AI-supported models
Neurodiversity-affirming mental health care requires not only adapting support to neurodivergent people but also examining the relational assumptions through which alliance, empathy, cooperation, progress, and risk are interpreted. This Conceptual Analysis develops the heuristic concept of relational-proximity priors: implicit assumptions that emotional closeness, rapid reciprocity, eye contact, visible warmth, and ready acceptance of help are generally reliable indicators of therapeutic fit or improvement. Drawing primarily on autism research and, more cautiously, on broader neurodivergence-informed literature, the paper synthesizes work on neurodiversity, double empathy, minority stress, healthcare accessibility, interpersonal distance, epistemic injustice, and psychotherapy process. The argument is that when such priors remain unaudited, clinically meaningful behaviors such as distance, reduced affect display, written communication, slower pacing, or literal style may be misread as resistance, poor alliance, or lack of insight. The paper further proposes that these priors can become sedimented in service routines, documentation practices, and AI-supported systems if observation and interpretation are not adequately separated. The article does not present a validated empirical construct or a systematic review; rather, it offers a theory-anchored conceptual synthesis and a set of heuristic audit tools intended to slow interpretation and widen clinically plausible hypotheses. Implications are discussed for psychotherapy, service design, documentation, participatory evaluation, and AI governance in public mental health.