(1) Background/Objectives: Over one billion individuals globally live with mental health conditions, yet the treatment gap exceeds 75% in low- and middle-income countries. Large language model (LLM)-based conversational agents have emerged as a potentially scalable solution, though the evidence base remains nascent and largely pre-clinical. This review synthesises barriers and facilitators to their implementation in mental healthcare using the Consolidated Framework for Implementation Research (CFIR). (2) Methods: Eight databases were searched from January 2022 to January 2026. Study selection was managed using Covidence. Two reviewers independently screened, extracted, and appraised studies using the Mixed Methods Appraisal Tool. Directed content analysis guided by CFIR was used for synthesis. (3) Results: Twenty-seven studies (three RCTs, nine mixed methods, eight qualitative, four cross-sectional, three observational) comprising >22,000 participants across 12 countries met inclusion criteria. Five barrier domains (27 sub-themes) and four facilitator domains (22 sub-themes) were identified. Inadequate crisis detection (reported in 21/27 studies) and 24/7 availability (reported in 26/27 studies) are the most frequently reported barriers and facilitators, respectively. These figures represent study-level reporting frequencies, not population-level prevalence estimates. CFIR mapping revealed universal coverage for Knowledge and Beliefs (100%) and Patient Needs and Resources (96%) but critical gaps in the Process domain (Evaluating: 7%; Champions: 11%). (4) Conclusions: LLM-based conversational agents demonstrate substantial promise but present critical safety deficiencies. A tiered implementation framework, independent safety certification, and equity-sensitive design are recommended.

Background: The NHS Improving Access to Psychological Therapies (IAPT) programme, now rebranded as NHS Talking Therapies, faces persistent capacity constraints with average wait times exceeding 90 days for cognitive behavioral therapy (CBT) in many Clinical Commissioning Group areas. AI-powered CBT platforms have been introduced as a digital adjunct within stepped care, yet longitudinal evidence on anxiety symptom trajectories and their predictors in routine NHS settings remains limited. Objective: To model individual anxiety symptom trajectories among patients referred to an AI-powered CBT platform within NHS primary care, identify distinct trajectory classes, and examine patient-level and practice-level predictors of differential treatment response using multilevel growth curve modeling. Methods: A prospective cohort study was conducted using linked clinical and administrative data from 6,284 patients (aged 18-65) referred to the CalmLogic AI-CBT platform across 187 general practices in four NHS England Integrated Care Systems (ICSs) between April 2023 and September 2025. Patients completed GAD-7 assessments at baseline, 4 weeks, 8 weeks, 12 weeks, and 24 weeks. Three-level growth curve models (assessments nested within patients nested within practices) with random intercepts and random slopes were fitted. Growth mixture modeling (GMM) was subsequently applied to identify latent trajectory classes. Predictors were examined at Level 2 (patient demographics, baseline severity, comorbidities, digital literacy, engagement intensity) and Level 3 (practice deprivation index, list size, urban/rural classification, and IAPT wait time). Results: The unconditional growth model revealed a significant average linear decline in GAD-7 scores of -0.94 points per month (p < .001), with substantial between-patient variation in both intercepts (variance = 14.82, p < .001) and slopes (variance = 0.38, p < .001). Significant between-practice variation accounted for 8.7% of intercept variance (ICC = 0.087). Growth mixture modeling identified four distinct trajectory classes: Rapid Responders (28.4%, steep early decline stabilising by week 8); Gradual Improvers (34.1%, steady linear decline through 24 weeks); Partial Responders (22.8%, modest early improvement followed by a plateau at clinically significant levels); and Non-Responders (14.7%, minimal change or slight deterioration). Higher baseline severity, female gender, and greater module completion predicted membership in the Rapid Responder class. Practice-level IAPT wait times exceeding 90 days independently predicted faster improvement trajectories (coefficient = -0.31, p = .003), suggesting that AI-CBT has its greatest incremental value in capacity-constrained areas. Patients in the most deprived quintile showed slower trajectories (coefficient = 0.22, p = .011) despite equivalent engagement levels, indicating a deprivation-related treatment response gap. Conclusions: AI-powered CBT platforms integrated within NHS primary care produce significant anxiety symptom reduction on average, but treatment response is heterogeneous, with four distinct trajectory classes identified. The finding that longer IAPT wait times predict better AI-CBT outcomes supports the platform’s positioning as a scalable bridge intervention for capacity-constrained services. The deprivation-related response gap warrants targeted support strategies for patients in the most disadvantaged communities.

Background: AI powered cognitive behavioral therapy CBT chatbots represent a scalable approach to addressing the global mental health treatment gap However causal evidence on their population level effectiveness in low and middle income countries LMICs remains limited and patient perspectives on acceptability and engagement are critical determinants of sustained use Brazils Estrategia de Saude da Familia ESF deployed an AI powered CBT chatbot Saude Mental Digital SMD to registered patients aged 18 and older at participating primary care units with eligibility determined by a composite vulnerability score exceeding a predetermined threshold Objective: To estimate the causal effect of AI powered CBT chatbot access on anxiety and depressive symptoms among primary care patients in Minas Gerais Brazil leveraging the eligibility score threshold as an exogenous source of variation Methods: We conducted a fuzzy regression discontinuity design fuzzy RDD study using linked administrative and clinical data from 312 ESF primary care units across Minas Gerais N 43287 patients January 2022 December 2024 The running variable was the composite vulnerability score with a threshold of 60 points determining chatbot eligibility The primary outcome was the 12 week change in the Patient Health Questionnaire Anxiety and Depression Scale PHQ ADS composite score Two stage least squares 2SLS estimation was used with local polynomial regression and triangular kernel weighting Bandwidth selection followed the Calonico Cattaneo Titiunik CCT optimal procedure Results: The fuzzy RDD estimated a local average treatment effect LATE of 473 points 95 CI 691 to 255 p 0001 on the PHQ ADS composite score at the eligibility threshold indicating clinically meaningful symptom reduction among compliers First stage estimates confirmed a strong 312 percentage point jump in chatbot uptake at the threshold F statistic 1274 Subgroup analyses revealed larger treatment effects among patients in rural municipalities 618 95 CI 902 to 334 those with lower educational attainment 582 95 CI 844 to 320 and women 537 95 CI 761 to 313 McCrary density tests confirmed no evidence of running variable manipulation p 067 Results were robust across alternative bandwidths polynomial orders and kernel specifications Conclusions: AI powered CBT chatbot access causally reduces anxiety and depressive symptoms among primary care patients near the eligibility threshold in Brazil with particularly pronounced benefits for rural less educated and female populations These findings provide quasi experimental evidence supporting the scalable deployment of AI powered CBT tools within public primary care systems in LMICs while underscoring the importance of incorporating patient perspectives on acceptability to maximize engagement and sustained therapeutic benefit

Background: AI-powered cognitive behavioural therapy (AI-CBT) tools hold significant promise for addressing the global mental health treatment gap, yet sustained user engagement remains critically low. While patient attitudes and experiential factors have been qualitatively documented, the psychological mechanisms through which AI literacy translates into long-term engagement remain poorly understood. Existing systematic evidence highlights trust, perceived therapeutic alliance, and stigma as salient themes, but no large-scale quantitative study has modelled these as a mediated pathway. Objective: This study aimed to (1) examine whether trust in AI systems and perceived therapeutic alliance mediate the relationship between AI literacy and sustained AI-CBT engagement, and (2) determine whether mental health stigma moderates these mediated pathways. Methods: A cross-sectional national online survey was conducted in the United Kingdom (N = 1,247). Eligible adults (18+) with a history of anxiety or depression who had used an AI-CBT tool in the preceding 12 months were recruited via stratified random sampling. Structural equation modelling (SEM) with moderated mediation was conducted in R (lavaan 0.6-17). Moderated mediation was evaluated using the PROCESS macro framework adapted for SEM, with 5,000 bootstrap replications for bias-corrected confidence intervals. Model fit was assessed using CFI, TLI, RMSEA, and SRMR indices. Results: The final SEM demonstrated excellent fit (CFI = 0.967, TLI = 0.959, RMSEA = 0.043 [90% CI: 0.036–0.051], SRMR = 0.052). AI literacy exerted a significant indirect effect on sustained engagement through trust in AI (β = 0.213, SE = 0.031, p < .001) and perceived therapeutic alliance (β = 0.187, SE = 0.028, p < .001). Mental health stigma significantly moderated the trust→engagement pathway (ΔR ² = 0.042, p = .003), with the indirect effect being stronger among individuals with lower stigma scores. The total indirect effect accounted for 58.4% of the total effect of AI literacy on engagement. Conclusions: AI literacy promotes sustained AI-CBT engagement primarily through its effects on trust and perceived therapeutic alliance, pathways that are attenuated by mental health stigma. These findings underscore the need for stigma-reduction interventions and AI literacy programmes as implementation strategies. Findings have direct implications for the design and deployment of AI-CBT tools across UK NHS digital mental health services.

Background: Optimal rehabilitation dosing after ischaemic stroke remains contested. Linear assumptions underlying conventional regression models may mask clinically important threshold effects, whereby functional gains accelerate or plateau beyond specific intensity thresholds. This study applied panel threshold regression to Australian hospital administrative data to identify endogenous breakpoints in the dose-response relationship between rehabilitation intensity and functional recovery. Methods: We used a retrospective longitudinal cohort derived from the Australian Stroke Clinical Registry (AuSCR) and the National Hospital Cost Data Collection (NHCDC) for fiscal years 2018-2019 to 2022-2023. The analytical sample comprised 18,742 hospitalised ischaemic stroke patients across 48 public hospitals in five Australian states. The primary exposure was daily rehabilitation intensity (minutes of physiotherapy, occupational therapy, and speech pathology per inpatient day). The primary outcome was change in the modified Rankin Scale (mRS) score from admission to discharge. We employed Hansen’s (1999) panel threshold regression framework to test for single, double, and triple threshold effects, using bootstrap p-values (n=500) to establish statistical significance. Fixed-effects estimation controlled for unobserved hospital heterogeneity. Secondary outcomes included acute length of stay and discharge destination. Cost-related parameters were benchmarked against published Australian cost-effectiveness data. Results: The panel threshold model identified two statistically significant breakpoints in the intensity-recovery relationship (p<0.001 for both). Below the first threshold (27.4 minutes/day; 95% CI: 24.8-29.6), each additional minute of daily rehabilitation was associated with a 0.008-point reduction in mRS score (beta = -0.008, 95% CI: -0.011 to -0.005, p<0.001). Between the two thresholds (27.4 to 54.7 minutes/day; 95% CI: 51.2-58.9), the marginal benefit approximately doubled (beta = -0.018, 95% CI: -0.022 to -0.013, p54.7 minutes/day), the marginal effect diminished substantially (beta = -0.004, 95% CI: -0.009 to 0.002, p=0.186), suggesting a ceiling effect. These dose-response patterns were consistent across age subgroups, stroke severity strata, and hospital volume tertiles. Conclusions: Rehabilitation intensity thresholds exist in stroke inpatient recovery and are non-linear. Patients receiving between 27 and 55 minutes of daily multidisciplinary therapy derive disproportionate functional benefit per unit of resource investment. Scheduling rehabilitation below the lower threshold represents a clinically and economically suboptimal allocation of inpatient resources. These findings have direct implications for workforce planning, clinical pathway design, and value-based commissioning in Australian public hospitals.

Background: Robotic exoskeleton-assisted gait training (RAGT) has emerged as a promising modality for post-stroke rehabilitation. However, the longitudinal trajectory of functional recovery and its association with clinically meaningful milestones such as independent community ambulation remain poorly characterised. Standard analytical approaches that treat longitudinal and survival outcomes separately may yield biased estimates due to informative dropout and the endogenous nature of time-varying functional measures. Objective: To jointly model the longitudinal trajectory of lower-extremity motor function and time to independent community ambulation following RAGT for stroke survivors, while accounting for the mutual dependence between the two processes using a Bayesian joint modelling framework. Methods: A multi-centre retrospective cohort study was conducted across four Canadian rehabilitation hospitals (2019 to 2024). A total of 327 adults with first-ever ischaemic or haemorrhagic stroke who received RAGT using the EksoNR or Indego exoskeleton platforms were included. The primary longitudinal outcome was the Fugl-Meyer Assessment Lower Extremity (FMA-LE) score measured at baseline, weeks 4, 8, 12, 24, and 52. The survival outcome was time to achieving independent community ambulation (Functional Ambulation Category score of 4 or higher sustained for at least two consecutive assessments). A Bayesian shared-parameter joint model was specified, linking a nonlinear mixed-effects longitudinal submodel to a Weibull proportional hazards survival submodel through the current value and slope of the subject-specific FMA-LE trajectory. Estimation was performed using Hamiltonian Monte Carlo sampling with four chains of 5,000 iterations each (2,500 warmup). Results: The median age was 62.4 years (IQR 54.1 to 71.8), 58.1% were male, and 63.0% had ischaemic stroke aetiology. The longitudinal submodel revealed a nonlinear recovery pattern best described by a three-knot restricted cubic spline, with rapid improvement during the first 12 weeks (mean gain 8.7 FMA-LE points, 95% CrI 7.2 to 10.3) followed by a plateau phase. The association parameter linking the current FMA-LE value to the hazard of achieving community ambulation was 0.084 (95% CrI 0.061 to 0.109), indicating that each one-point increase in the subject-specific FMA-LE trajectory was associated with an 8.8% increase in the instantaneous hazard (HR = 1.088, 95% CrI 1.063 to 1.115). The trajectory slope parameter was also significant (0.043, 95% CrI 0.012 to 0.078), suggesting that patients with steeper recovery gradients had additional survival advantages beyond their current functional level. At 52 weeks, 54.7% of participants achieved independent community ambulation. Haemorrhagic stroke (HR = 0.68, 95% CrI 0.49 to 0.93), older age (HR per decade = 0.81, 95% CrI 0.70 to 0.94), and higher baseline NIHSS score (HR per point = 0.94, 95% CrI 0.91 to 0.97) were associated with lower hazards of achieving the ambulation milestone. Conclusions: The Bayesian joint model revealed that both the current functional level and the rate of functional change are independently predictive of achieving community ambulation following RAGT. These findings support individualised rehabilitation planning where treatment intensity may be dynamically adjusted based on the evolving recovery trajectory, and provide further evidence for the clinical value of robotic exoskeleton interventions in stroke rehabilitation.