搜索结果：rollout

We present Clin-JEPA, a multi-phase co-training framework for joint-embedding predictive (JEPA) pretraining on electronic health record (EHR) patient trajectories. JEPA architectures have enabled latent-space planning in robotics and high-quality representation learning in vision, but extending the paradigm to EHR data-to obtain a single backbone that simultaneously forecasts patient trajectories and serves diverse downstream risk-prediction tasks without per-task fine-tuning-remains an open challenge. Existing JEPA frameworks either discard the predictor after pretraining (I-JEPA, V-JEPA) or train it on a frozen pretrained encoder (V-JEPA 2-AC), leaving the encoder unaware of the rollout signal that the retained predictor must use at inference; co-training the encoder and predictor under a shared JEPA prediction objective would supply this grounding, but naïve co-training is unstable, with representation collapse and online/target drift causing autoregressive rollout to diverge. Clin-JEPA's five-phase pretraining curriculum-predictor warmup, joint refinement, EMA target alignment, hard sync, and predictor finalization-addresses each failure mode by phase, stably co-training a Qwen3-8B-based encoder and a 92M-parameter latent trajectory predictor. On MIMIC-IV ICU data, three independent evaluations support the framework: (1) latent l 1 rollout drift uniquely converges (-15.7%) over 48-hour horizons while baselines and ablations diverge (+3% to +4951%); (2) the encoder learns a clinically discriminative latent geometry (deteriorating-patient cohorts displace 4.83× further than stable patients in latent space, vs ≤2.62× for baseline encoders); (3) a single backbone outperforms strong tabular and sequence baselines on multi-task downstream evaluation. Clin-JEPA achieves mean AUROC 0.851 on ICareFM EEP and 0.883 on 8 binary risk tasks (+0.038 and +0.041 vs baseline average).

Microplastics are pervasive pollutants in marine environments that interact with microalgae to form microplastic-microalgae (MP-MA) aggregates. These aggregates can increase effective density, thereby altering the vertical distribution and environmental fate of both microplastics and microalgae. However, the sinking mechanisms of these aggregates are not fully understood, partly due to their irregular morphology and species-specific traits. In this study, a multimodal artificial intelligence (AI) model was developed to estimate the sinking velocity of MP-MA aggregates using a dataset comprising 23 microalgal species. The model incorporated three input modalities: image data (microscopy images), text data (species names), and biological data (species-specific traits and experimental variables). Pretrained Bootstrapping Language-Image Pre-training (BLIP) encoders were used to extract image and text features, which were then concatenated with biological data for regression. Explainable AI and visualization techniques were applied to interpret the model results, including attention rollout visualization and SHapley Additive exPlanations (SHAP). The multimodal AI model outperformed both the Stokes-based formulations and the random forest model, achieving R2 values of 0.857 ± 0.051 for training and 0.567 ± 0.062 for validation across 20 different within-species training-validation data splits, whereas the best-performing Stokes-based formulation achieved an R2 of 0.11. Attention rollout visualization provided qualitative maps highlighting salient regions of the aggregates in the microscopy images. SHAP analysis identified cell wall type, plastic particle count, and swimming mode as influential variables. Overall, this study provides an interpretable basis for understanding how microalgal traits and aggregate structure relate to the sinking behavior of MP-MA aggregates in marine environments.

We analyse the impact of a change in the administration of social security payments, occurring in utero and early infancy, on health in early childhood. We identify this impact through the gradual rollout of the so-called 'income management' policy in Aboriginal communities in Australia's Northern Territory in 2007. This policy changed the delivery method of social security payments but not their value - however, implementation challenges meant that many families did not receive their payments on time. Using linked administrative data, we find that children who were exposed to the policy rollout in utero or in their first three months of life (the 'fourth trimester') were at higher risk of severe infection requiring hospitalisation. These children spent, on average, 4.7 more days in hospital between birth and their 8th birthday. Most of this impact is concentrated in hospitalisations for infection, which increased by 23 percent. These admissions are driven by a range of infection types: bacterial, viral and respiratory. We link our findings to the 'immune programming hypothesis', i.e. maternal stress and poor nutrition during key stages in immune system development can permanently weaken the child's immune system. Our findings highlight the importance of attention to key phases in child development when designing policies that affect households' financial resources, even temporarily.

Adolescent and adult migrants in low- and middle-income countries (LMICs) are often under-immunised because of disrupted vaccination in childhood and barriers to care, increasing the risk of vaccine-preventable diseases. WHO's Immunization Agenda 2030 strengthens calls for life-course immunisation and inclusion of marginalised groups in catch-up initiatives (including Td/Tdap, IPV, MR/MMR, HepB, HPV). Despite evidence from Europe that migrants are frequently missed, it is unclear how far LMIC policies include adolescents and adults beyond child-focused programmes. We analysed catch-up and life-course vaccination policies, operational guidance, laws and frameworks across 136 LMICs. Using framework analysis, we conducted a comparative policy analysis of national and subnational policies, operational guidance, laws and frameworks on catch-up and life-course vaccination for adolescents (10-19 years) and/or adults (>19 years) migrants (including refugees, undocumented migrants, asylum seekers, Internally Displaced Person (IDPs) and migrants workers) across LMICs. We did systematic database and grey-literature searches including and Ministry of Health websites, key international organisations (Jan 2000-Feb 2025), and contacted health ministries and national experts to locate unpublished policies and guidelines. 33 policy and guidance sources met the inclusion criteria (15 peer-reviewed; 18 grey literature), reporting vaccination policies for migrants across 57 LMICs. Evidence clustered into two policy domains, routine vaccinations and COVID-19. We found that migrant-inclusive catch-up or life-course vaccination policies for routine vaccines were rare, identified in only 9 (6.6%) of 136 LMICs. 3 (33.3%) of these countries focused on outbreak-responsive or ad hoc provisions only, 4 (44.4%) applied general national immunisation schedules with opportunistic catch-up to migrants (but without migrant-specific operational guidance), and 2 (22.2%) recognised migrant entitlement through formal policy frameworks or structured partnerships. Policies targeted both adolescents and adults in 9 (6.6%) of 136 LMICs, and refugees in 9 (6.6%) LMICs, while other migrant groups were less consistently specified. Vaccines most commonly appearing in guidelines for adolescent and adult migrants in LMICs were HPV in 6 (4.4% of 136 LMICs) and MR/MMR in 5 (3.7%), followed by OCV and HepB in 3 (2.2%) countries each. Polio, tetanus and typhoid were also included in rare cases. Where vaccination history was uncertain, only 3 (33.3%) countries explicitly recommended presumptive catch-up vaccination. For COVID-19, vaccination policies demostrated broad migrant inclusion: 57 (41.9%) of 136 LMICs documented migrant-inclusive policies or rollout targeting adolescents and/or adult migrants through national deployment plans or emergency strategies. Most LMICs lack policies ensuring adolescent and adult migrants access catch-up and life-course vaccination, despite ample evidence that these are under-immunised populations. Governments should clarify entitlements, adopt migrant-inclusive guidance, and strengthen delivery and monitoring systems to reach people who miss vaccines, doses, and boosters in origin countries or during transit, and who face barriers to care.

South Korea became a super-aged society in 2024, and this demographic shift is unfolding alongside the depopulation of rural municipalities across the country. How spatial inequality and community social capital jointly relate to elderly health-and whether those relationships look different for younger versus older elderly-remains an open question. We investigated associations between two dimensions of community social capital (sense of belonging and neighbor communication), subjective perception of capital-provincial inequality, and self-rated health among Korean elderly, with separate analyses for the Young-Old (aged 60-69) and Old-Old (aged 70+). We used the 2024 Social Integration Survey from the Korea Institute of Public Administration (full sample N = 2588; elderly subsample N = 1020). Random intercept hierarchical linear models accounted for the nesting of individuals within 17 metropolitan cities and provinces. Stepwise models examined social capital antecedents, a healthcare satisfaction indirect association pathway, and the direct association of spatial inequality perception with health. The elderly subsample was stratified into Young-Old (N = 289) and Old-Old (N = 731). A mixed-effects ordered logistic regression with Liang-Zeger cluster-robust standard errors was estimated as a robustness check. Sense of belonging was positively associated with subjective health among the elderly (B = 0.065, p < 0.05) as a net of rurality and socioeconomic controls. Perceived spatial inequality showed a negative association (B = -0.070, p < 0.05). The indirect association pathway through healthcare satisfaction was not supported (Sobel Z = -1.458, p = 0.144). Age-stratified models revealed a striking split: belonging was the dominant predictor for the Young-Old (B = 0.149, p < 0.01), while neighbor communication (B = 0.078, p < 0.05) and spatial inequality perception (B = -0.092, p < 0.01) were significant only among the Old-Old. The ordered logistic robustness check confirmed the negative association of perceived spatial inequality across all specifications. What predicts health in the younger elderly is not what predicts health in the older elderly. Korea's Integrated Community Care Act, set for nationwide rollout in 2026, should account for this divergence-prioritizing psychological community attachment for the Young-Old and face-to-face social contact combined with regional equity for the Old-Old.

The rapid growth of digital health initiatives has heightened reliance on frontline health workers (FLHWs) to deliver, document, and manage services through digital tools, particularly in low- and middleincome countries (LMICs). In India, the widespread rollout of platforms under the Ayushman Bharat Digital Mission (ABDM) is not yet matched by a standardized digital health competency framework (DHCF) for FLHWs, hindering systematic skill development, assessment, and integration. This study designed, developed, and evaluated a theory-driven, evidence-based, and scalable DHCF for India's health workforce. Framed as a feasibility and proof-ofconcept study, it was piloted among FLHWs in Uttar Pradesh using a three-stage approach comprising design, implementation, and evaluation. The framework development drew on a systematic literature review and the Government of India's Framework for Roles, Activities, and Competencies (FRAC). A cadre-agnostic competency dictionary was created, spanning functional, behavioral, domainspecific, and intervention-specific skills across graded proficiency levels. Competencies were mapped to FLHW roles, and aligned training materials and assessments were developed. The framework was piloted through in-person, instructor-led sessions for Auxiliary Nurse Midwives (ANMs) in two districts (n&#x2009;=&#x2009;70), alongside baseline assessments for Accredited Social Health Activists (ASHAs; n&#x2009;=&#x2009;32). The resulting DHCF comprises a three-component package: (i) a cadre-agnostic competency dictionary with progressive proficiency levels; (ii) systematic role-tocompetency mapping using the FRAC methodology; and (iii) integrated training content and assessment scaffolding designed for institutional embedding. The framework defined 10 core competencies, enabling role-specific mapping across cadres. Feasibility testing demonstrated significant gains in ANMs' knowledge and digital skills: Wilcoxon signed-rank tests showed significant improvements in two of four competency levels (C1L1 and C2L1; both P&#x2009;<&#x2009;.001), with the largest effect for data collection basics (r&#x2009;=&#x2009;0.84). ASHA baseline assessments revealed substantial foundational literacy gaps (mean total score 11.97/30 [39.9%]; data collection was the weakest competency at 32.5%, with no ASHA scoring above 60% on C2L1). Stakeholders affirmed the framework's relevance, feasibility, and adaptability, while identifying the need for hybrid training models and stronger institutional embedding.The DHCF offers a structured, scalable approach to standardizing digital health training for FLHWs and strengthening workforce preparedness in resource-limited settings during India's digital health transition. This feasibility study establishes the framework's relevance and applicability; future work is needed to evaluate effectiveness at scale, long-term competency retention, and linkage to service delivery outcomes. Parallel attention to digital tool design and usability will be essential to complement competency-building efforts.

Using the staggered rollout of China's one-child policy (OCP) across provinces and birth cohorts as a quasi-natural experiment, we demonstrate that differential fertility between richer and poorer households exacerbates intergenerational income inequality. Rural/poorer families, who are less constrained by the OCP than their urban/richer counterparts, tend to have more children but invest less in each child's human capital. This reduction in mobility is primarily driven by the rising economic status of children born to urban/wealthier families. Our estimates suggest that the OCP accounts for approximately 25% of the observed decline in intergenerational income mobility in China and thus highlight a demographic channel through which economic inequality persists across generations.

Hepatocellular carcinoma (HCC) is a fatal disease of the young in sub-Saharan Africa (SSA) and chronic hepatitis B virus (HBV) infection remains the predominant aetiology. There is paucity of data regarding HCC in the adolescent population globally. Adolescents, defined as individuals aged 10 to 19 years according to the World Health Organization (WHO), with HCC treated at Groote Schuur Hospital (GSH) in South Africa from 1 January 2012 and 31 December 2024 were studied. Five (0.5%) of the 726 HCC patients managed at GSH during the study period were adolescents. The median age was 18 (13-19) years and three were female. All five had chronic HBV infection and most presented with pain (60%) and/ or an abdominal mass (40%). All had advanced disease, with four (80%) having Barcelona Clinic Liver Cancer (BCLC) stage C and one (20%) with BCLC stage D. Two (40%) had extrahepatic metastases and three (60%) had portal vein tumour thrombosis. Treatment included liver resection (1), sorafenib (1), lenvatinib (1), and best supportive care (2). At the time of the study, only one patient was alive. The median survival was 137 (25-425) days. Despite national HBV vaccination programmes in South Africa, in our experience adolescent HCC was HBV-related in all five patients. Extrahepatic metastases and macrovascular invasion were frequently reported and restricted patient access to curative-intended therapies (ablation, liver resection, transplantation). These findings highlight the urgent need for improved early detection and prevention strategies against perinatal HBV transmission in South Africa, including the rollout of the WHO-recommended universal hepatitis B birth-dose vaccination rather than the current targeted prevention approach.

Social prescribing is a growing community health intervention and has been associated with improved patient outcomes. However, the evidence around inequalities in referrals shows varying patterns. We aim to examine referrals to social prescribing link workers funded by the Additional Roles Reimbursement Scheme in England. We conducted a retrospective observational population-based study, using primary care data from Clinical Practice Research Datalink (CPRD) Aurum from 1st July 2019-31st March 2024. Participants included over 12 million patients aged 16 years or older. We examined the likelihood of offers and subsequent referrals to social prescribing by patient- and area-level characteristics using logistic regression and report odds ratios (ORs). Since July 2019, approximately 4% of the CPRD population have been offered a referral to social prescribing. 77&#xb7;7% of those were referred. Patients who are: female, older, living in less deprived areas and have multiple long-term conditions have higher odds of being offered social prescribing (Female OR = 1&#xb7;35, 95% CI [1&#xb7;32 to 1&#xb7;38] p&#x2009;<&#x2009;0.001). Factors such as region, rurality, and ethnicity do not result in inequalities in offers compared to the general population. Of those offered, we find that those who are female, those from non-white ethnicities (Black, Asian and Mixed), and have multiple long-term conditions had higher odds of accepting offers of referrals (being referred). Referrals to social prescribing have increased following the national rollout of link workers. However, inequalities in offers and referrals to social prescribing have been identified by patient and area-level factors. Our findings indicate that policies should improve awareness of social prescribing in deprived areas and direct certain patient groups, such as ethnic minorities, males and those older to the benefits of being referred to social prescribing.

In terms of rollout, comprehensiveness, and strategy, Singapore's regulatory landscape governing the ethical use of Artificial Intelligence (AI) in healthcare has generally kept pace with other global leaders in AI advancement. However, establishing a robust and holistic regulatory framework that evolves along with emerging technologies is not easy-especially in healthcare, where the stakes are high and resources may be limited. We conducted a structured scoping analysis of key AI regulatory and professional documents in Singapore, selected using predefined inclusion criteria. Documents were systematically mapped against Savulescu et al.'s nine categories of ethical risk, followed by cross-document comparison to identify integration gaps and inconsistencies, and benchmarking against international AI governance frameworks. These recommendations are generalizable beyond Singapore for developers, implementers, healthcare professionals and patients and include dealing with bias in AI, enhancing human productivity without deskilling, facilitating more informed decision-making, and cultivating greater knowledge exchange between clinicians and patients, to name a few.

We study the long-run consequences of fertility policy for aging and survival, exploiting the staggered provincial rollout of China's 1970s Later, Longer, Fewer campaign. Linking variation in exposure to the Chinese Longitudinal Healthy Longevity Survey (1998-2021), we find that cohorts subject to fertility restrictions experienced significantly higher late-life mortality, over 10% on average, and worse cognitive and psychological outcomes. These effects arise because smaller sibships, delayed childbearing, and wider spacing reduced both the supply and timing of intergenerational care. While the subsequent introduction of LTCI mitigated some adverse impacts, the effects were driven by in-kind benefits, whereas cash-based LTCI provided little offset, underscoring the limits of formal substitution for kin-based support. Our findings reveal a fundamental intergenerational trade-off in fertility control and highlight the enduring demographic costs of policies that reshape family structure.

Clinical trials evaluating population-based screening tests or other interventions likely to affect care delivery in real-world settings often do not consider spillover effects, such as whether the intervention rollout affects access to limited health care services. To examine whether regional participation in a population-based screening trial (NHS-Galleri) of a cell-free DNA-based multicancer early detection (MCED) test was associated with changes in cancer diagnostic delay rates. Cross-sectional study of all 21 cancer alliance regions in England, 8 of which participated in the population-based MCED screening trial. An event study using difference-in-differences design evaluated changes from 6 months before (April 2021) to 3 years after trial start (September 2024). Regional participation in the population-based MCED screening trial. The primary outcome was diagnostic delay rates (percentage of patients referred for suspected cancer evaluation taking longer than 28 days to reach diagnostic resolution), a surrogate measure for system-level spillover effects; the secondary outcome was patient referral rates. Analysis focused on a primary group of 3 cancer types (head and neck, lung, and upper gastrointestinal) that were identified in the trial protocol and were not subject to routine screening. Overall, 1&#x202f;875&#x202f;236 patient referrals for suspected head and neck, lung, or upper gastrointestinal cancers were recorded across all 21 regions. In the first 6 months of the population-based screening trial, diagnostic delay rates increased in participating regions (28.6% before trial start and 29.6% after) and decreased in nonparticipating regions (28.9% to 26.3%), an adjusted difference-in-differences estimate of 3.4 percentage points (95% CI, 1.9-5.0; P&#x2009;<&#x2009;.001). This increase persisted during the second 6-month period (adjusted difference-in-differences estimate of 4.8 percentage points [95% CI, 1.9-7.7; P&#x2009;=&#x2009;.003]) and was no longer statistically significant thereafter. Patient referral rates for suspected head and neck, lung, and upper gastrointestinal cancers were also higher in participating regions in the first 6 months (adjusted difference-in-differences estimate of 23.8 per 100&#x202f;000 population [95% CI, 0.9-46.8; P&#x2009;=&#x2009;.04]). Regional participation in a population-based MCED screening trial was associated with a modest increase in diagnostic delay rates for patients referred for suspected head and neck, lung, and upper gastrointestinal cancers. This increase is unlikely to have materially affected interpretation of the MCED screening trial primary findings. Future trials of population-based screening interventions likely to affect demand for limited health care resources should consider monitoring for system-level spillover effects.

Menstrual hygiene management (MHM) can be challenging for women and girls living in conflict-affected low- and middle-income countries, where access to menstrual products, soap, water, and private sanitation facilities is often limited. In response to these challenges, the Canadian Red Cross, in partnership with the South Sudan Red Crescent and the South Sudanese Government, implemented the Healthy Bodies Healthy Minds (HBHM) program from 2020-2023 to support MHM in South Sudan. Despite growing attention to MHM needs, few studies have examined the on-the-ground realities of implementing MHM programming in fragile settings. To address this gap, this study aimed to (i) describe the implementation of HBHM, (ii) identify barriers faced during implementation, (iii) identify potential solutions to overcome barriers, and (iv) understand the impact of the COVID-19 pandemic on program delivery. Data was collected through key informant interviews with individuals involved in program design and implementation and were analyzed via thematic analysis. A total of 14 respondents were interviewed. Respondents identified several interrelated and compounding barriers, including resource constraints, logistical challenges, competing stakeholder interests/priorities, ongoing conflict, and climate change. To mitigate these barriers and improve the delivery of MHM programs, respondents recommended solutions and strategies such as extending implementation timelines, planning program activities around anticipated climate events, and strengthening inter-organizational collaboration. The COVID-19 pandemic introduced additional barriers, particularly by restricting the movement of people and goods. However, it also yielded unexpected benefits, notably expanding the project's reach beyond the target group of girls to the broader community. Implementation of the HBHM program was impacted by several interrelated and compounding factors including resource constraints and logistical challenges, protracted conflict, climate change, and COVID-19, however, collaboration between implementing partners and agility in the approach to program rollout supported enhanced program reach. International partnerships enabled community-level implementation, which proved especially effective for continued program implementation. Additionally, the shift to a more community-based model resulted in benefits to the broader community, offering another key insight that should be considered in the design of future MHM programs in South Sudan and similar contexts.

Data-driven modeling of nonlinear industrial processes is often complicated by heterogeneous temporal dynamics, measurement noise, and fixed-rate data acquisition. Under such conditions, direct regression on raw time-series data may become sensitive to sampling imbalance and fast transient behavior, leading to degraded predictive performance. This work proposes a structured reduced-order modeling framework for constructing compact and numerically stable predictive surrogates. The approach integrates adaptive resampling to redistribute temporal information, spline-based smoothing for stable derivative estimation, delay embedding to incorporate short-term temporal structure, kernel-based dimensionality reduction to extract dominant patterns, and sparse regression in a latent coordinate space to obtain parsimonious dynamical models. Within this framework, sparsity is used primarily to control model complexity in the reduced representation rather than to recover explicit governing equations. The method is evaluated on two benchmark reactor systems and a real grinding-classification process using chronological train/test splits and multi-step rollout prediction. The results indicate that the proposed approach can improve predictive robustness and numerical stability compared with direct sparse regression and its partial variants, particularly in the presence of multiscale temporal behavior and moderate measurement noise. The framework provides a practical strategy for predictive reduced-order modeling under realistic industrial data constraints. Its design emphasizes stability and compactness of the learned dynamics, while acknowledging that the resulting models are defined in a latent representation and are not intended as exact reconstructions of physical governing equations.

Musculoskeletal (MSK) anatomy education is a critical foundation for developing competency among radiologists, physiatrists, rheumatologists, and orthopedic surgeons. However, current undergraduate medical curricula often exhibit significant deficiencies in instructional hours, integration of diverse teaching modalities, and clinical relevance. This narrative review synthesizes recent evidence (March 2021-March 2026) identified through a targeted search of Medline, Embase, and Scopus, with an emphasis on consensus guidelines, validation studies, and clinically focused publications related to MSK anatomy, imaging modalities (ultrasound, magnetic resonance imaging, computed tomography), and curriculum design for medical students in relevant specialties. Multimodal interventions, including cadaveric dissection, radiological anatomy, case-based rheumatologic and rehabilitation modules, and technology-enhanced platforms such as 3D virtual models and AI-driven adaptive learning, have been associated with improvements in knowledge retention, spatial reasoning, diagnostic accuracy, and procedural confidence compared with didactic instruction alone. Persistent knowledge gaps undermine interpretive proficiency in MSK imaging, including the identification of synovitis and enthesopathy, and are linked to reduced clinical preparedness. Objective assessments reveal suboptimal performance despite completion of conventional preclinical training. Cadaveric dissection fosters practical skills and ethical professionalism, while early integration of imaging connects theoretical morphology with three-dimensional relational understanding and pathological correlations in MSK and rheumatic diseases. Implementation frameworks recommend phased rollouts that incorporate stakeholder needs assessments, faculty development through train-the-trainer models, resource reallocation for point-of-care ultrasound and virtual reality, and strategies to address barriers, such as grants for low-resource settings and modularization to reduce curricular congestion. These evidence-based approaches support scalable reforms that produce MSK-literate clinicians prepared for precision diagnostics and interventional practice.

Hepatitis B surface antigen (HBsAg) loss, a functional cure in chronic hepatitis (CHB), is rare. Temporal changes in HBsAg loss rates were evaluated in relation to COVID-19 vaccination rollout in South Korea. An interrupted time series (ITS) analysis was performed using electronic health records of 59,946 adults with CHB (2002-2025). Monthly HBsAg loss incidence rates were estimated using a competing risk framework. National COVID-19 vaccination coverage was used as a surrogate for population-level exposure. A three-phase segmental regression model assessed changes before vaccination, during a predefined vaccination effect period (18&#x2009;months following &#x2265;&#x2009;80% of two-dose coverage), and after this period. Hepatocellular carcinoma (HCC) incidence was analysed as a negative control. Over 382,509 person-years, 2483 HBsAg loss events occurred. Pre-vaccination rates were stable (0.11-0.80 per 100 person-years). Following the COVID-19 vaccination rollout, the incidence increased to 0.96 in 2021, peaking at 1.27 in 2022, before declining to 0.93 in 2024. In the ITS model, no baseline trend was observed (p&#x2009;=&#x2009;0.606). At the intervention point (November 2021), HBsAg loss rates increased significantly (incidence rate ratio; 1.56; 95% CI 1.26-1.93; p&#x2009;<&#x2009;0.001). No significant slope change was observed during the vaccination period, while a negative slope change was observed post-vaccination (coefficient -0.039 per month; p&#x2009;=&#x2009;0.003). Findings were consistent across sensitivity and subgroup analyses. HCC incidence showed no significant changes. COVID-19 vaccination was temporally associated with transiently higher HBsAg loss rates in CHB, followed by attenuation, suggesting a potential population-level immunomodulatory effect on HBV dynamics.