Biometrics

Automating Hierarchical Condition Category (HCC) assignment directly from unstructured electronic health record (EHR) notes remains an important but understudied problem in clinical informatics. We present HCC-Coder, an end to end NLP system that maps narrative documentation to 115 Centers for Medicare & Medicaid Services(CMS) HCC codes in a multi-label setting. On the test dataset, HCC-Coder achieves a macro-F1 of 0.779 and a micro-F1 of 0.756, with a macro-sensitivity of 0.819 and macro-specificity of 0.998. By contrast, Generative Pre-trained Transformer (GPT)-4o achieves highest score of a macro-F1 of 0.735 and a micro-F1 of 0.708 under five-shot prompting. The fine-tuned model demonstrates consistent absolute improvements of 4%-5% in F1-scores over GPT-4o. To address severe label imbalance, we incorporate inverse-frequency weighting and per-label threshold calibration. These findings suggest that domain-adapted transformers provide more balanced and reliable performance than prompt-based large language models for hierarchical clinical coding and risk adjustment.

Network-based epidemic models have been instrumental in understanding how contact structure shapes infectious disease dynamics, yet widely used frameworks such as Erd[o]s-Renyi, configuration-model, and stochastic block networks do not explicitly capture the combination of fully accessible (saturated) within-group interactions and constrained between-group connectivity characteristic of many real-world settings. Here, we introduce the Multi-Clique (MC) network model, a generative framework in which individuals are organised into fully connected cliques representing stable contact groups (e.g., households, classrooms, or workplaces), with a limited number of external connections governing inter-group transmission. Using stochastic susceptible-infectious-recovered (SIR) simulations on degree-matched networks, we compare epidemic dynamics on MC networks with those on classical random graph models. Despite having an identical mean degree, MC networks exhibit systematically distinct behaviour, including slower epidemic growth, reduced peak prevalence, increased fade-out probability, and delayed time to peak. These effects arise from rapid within but constrained between clique transmission, creating structural bottlenecks that standard models do not capture. The MC framework provides an interpretable, data-driven representation of recurrent contact structure, with parameters that map directly to observable quantities such as household and classroom sizes. By isolating the role of intergroup connectivity, the model offers a basis for evaluating targeted intervention strategies that reduce between-group mixing while preserving within-group interactions. Our results highlight the importance of explicitly representing the real-life clique-based network structure in epidemic models and suggest that classical degree-matched networks may systematically overestimate epidemic speed and intensity in structured populations.

When randomized controlled trials are impractical, interrupted time series designs offer a rigorous quasi-experimental approach to assess population level policies. Indeed, in the context of quasi-experimental designs (QEDs), the Interrupted Time Series (ITS) method is commonly thought of as the most robust. But interrupted time series designs are susceptible to serial correlation and confounding by time-varying factors associated with both the intervention and the outcome, which may result in biased inference. Thus, we provide a simulation-based contrast of controlled interrupted time series (CITS) and multivariable regression (multivariable negative binomial regression) for estimation of policy effects in count time series data. These approaches are widely used in policy evaluations, yet their comparative performance in typical population health settings has rarely been examined directly. We tested both approaches within a variety of data generating situations, differing in the series length, intervention effect size, and magnitude of lag-1 autocorrelation. Bias, standard error calibration, confidence interval coverage, mean squared error, and statistical power were assessed for performance. Both methods gave unbiased estimates for moderate and large intervention effects, although bias was more pronounced for small effects, particularly in short series. Although the point estimate performance was similar, inferential properties varied significantly. CITS always had smaller mean squared error, better consistency between model based and empirical standard errors, and confidence interval coverage near the 95% nominal levels over weak to moderate autocorrelation. By contrast, multivariable regression was more sensitive to serial dependence, leading to underestimated standard errors and undercoverage, especially at moderate to high autocorrelation, regardless of Newey-West adjustments. These findings show the benefits of using a concurrent control series and the importance of structurally accounting for serial correlation when studying population level policies with time series data.

Symptom propagation occurs when the symptoms of secondary cases are related to those of the primary case as a result of epidemiological mechanisms. Determining whether - and to what extent - symptom propagation occurs requires data-driven methods. Here we quantify the strength of symptom propagation as the increase in risk of a secondary case developing severe symptoms if the primary case has severe symptoms. We first used synthetic results to determine the data requirements to robustly estimate the strength of symptom propagation and to investigate the effect of severity-dependent reporting bias. Categorising symptom severity into two group (mild or severe; asymptomatic or symptomatic), our estimation requires only four summary statistics - the number of primary-secondary case pairs of each combination of symptom presentations. Our analysis showed that a relatively small number (100) of synthetic primary-secondary case pairs was sufficient to obtain a reasonable estimate of the strength of symptom propagation and 1,000 pairs meant errors were consistently small across replicates. Our estimates were robust to severity-dependent reporting bias. We also explored how symptom propagation can be separated from other individual-level factors affecting severity, using age dependence as an example. Although synthetic data generated from an age-structured model led to overestimations of the strength of symptom propagation, allowing disease severity to be age-dependent restored the accuracy of parameter estimation. Finally, we applied our methodology to estimate the strength of symptom propagation from three publicly available data collected during the COVID-19 pandemic with data on presence or absence of symptoms: England households, Israel households, and Norway contact tracing. Our age-free methodology indicated a 12-17% increase in the risk of being symptomatic if infected by someone symptomatic. Our positive estimates for the strength of symptom propagation persisted when applying our age-dependent methodology to the two household data sets with age-structured information (England and Israel). These findings demonstrate evidence for symptom propagation of SARS-CoV-2 and provide consistent estimates for its strength. Our synthetic data analysis supports the conclusion that these correlations are not a result of reporting bias or age-dependent effects. This work provides a practical tool for estimating the strength of symptom propagation that has minimal data requirements, enabling application across a wide range of pathogens and epidemiological settings.

Machine learning models trained on observational data from one environment frequently fail when deployed in another, because standard learning algorithms exploit spurious correlations alongside causal ones. Invariant learning methods address this problem by seeking representations that support stable prediction across training environments, but their behavior on tabular data remains poorly characterized. We present CausTab, a gradient variance regularization framework for causal invariant representation learning on mixed tabular data. CausTab penalizes the variance of parameter gradients across training environments, providing a richer invariance signal than the scalar penalty used by Invariant Risk Minimization (IRM). We provide formal results showing that the gradient variance penalty is zero at causally invariant solutions and positive at solutions that rely on spurious features. Through experiments on synthetic data across three spurious-correlation regimes, four cycles of the National Health and Nutrition Examination Survey (NHANES), and four hospital systems in the UCI Heart Disease dataset, we demonstrate that: (1) IRM consistently degrades relative to standard empirical risk minimization (ERM) on tabular data, losing up to 13.8 AUC points in spurious-dominant settings, a failure we trace mechanistically to penalty collapse during training; (2) CausTab matches or exceeds ERM in every experimental condition; (3) CausTab achieves consistently better probability calibration than both ERM and IRM; and (4) invariant learning methods fail when environments differ in outcome prevalence rather than in spurious feature correlations, a boundary condition we characterize both empirically and theoretically. We introduce the Spurious Dominance Index (SDI), a practical scalar diagnostic for determining whether a dataset requires invariant learning, and validate it across all experimental settings

BackgroundThe rapid proliferation of digital platforms has transformed health information access but has also led to increased exposure to misinformation. Existing research lacks standardized tools to quantify individual-level exposure to health misinformation in a comprehensive manner. ObjectiveTo develop a novel composite index--the Misinformation Exposure Index (MEI)--to measure multidimensional exposure to health misinformation among social media users. MethodsA questionnaire-based pilot study was conducted among a young adult population to assess patterns of health information exposure, source utilization, trust, and behavioural responses. The MEI was developed using a multi-domain framework comprising Exposure Frequency, Source Diversity and Risk, Trust in Information, and Behavioural Response. Responses were scored using Likert scales and weighted domain contributions to generate a composite score ranging from 0 to 100. ResultsParticipants demonstrated moderate to high engagement with digital platforms for health information, with reliance on both formal and informal sources. Variability in trust and verification behaviours was observed, with a proportion of participants reporting adoption of health-related practices without professional consultation. Composite MEI scores indicated that most individuals fell within the moderate exposure category, with a subset exhibiting high exposure characterized by frequent engagement with high-risk sources and behavioural influence. ConclusionThe MEI provides a novel and comprehensive framework for quantifying health misinformation exposure by integrating exposure patterns, source characteristics, trust, and behavioural outcomes. The index has potential applications in public health surveillance and intervention design. Further validation through large-scale studies is warranted to establish its reliability and generalizability.

Genome-wide association studies (GWAS) have identified numerous genetic variants associated with complex traits. However, linkage disequilibrium (LD) confounds these associations, leading to false positives where non-causal variants appear associated because they are correlated with nearby causal variants. This is particularly the case in highly polygenic traits where the genome can be saturated in causal variants. To address this issue, we propose LDeconv a method based on truncated singular value decomposition (SVD) that adjust GWAS summary statistics without requiring individual-level genotype data. This approach accounts for LD structure, isolates causal variants in high-LD regions, and improve the reliability of effect size estimates. We assess its performance through simulations across various LD scenarios, conduct extensive sensitivity analyses, and apply them to real GWAS data from the UK Biobank. Our results demonstrate that LDeconv effectively reduces false discoveries while preserving true associations, offering a robust framework for post-GWAS analysis.

Gene networks (GNs) encode diverse molecular relationships and are central to interpreting cellular function and disease. The heterogeneity of interaction types has led to computational methods specialized for particular network contexts. Large language models (LLMs) offer a unified, language-based formulation of GN inference by leveraging biological knowledge from large-scale text corpora, yet their effectiveness remains sensitive to prompt design. Here, we introduce Gene-Relation Adaptive Soft Prompt (GRASP), a parameter-efficient and trainable framework that conditions inference on each gene pair through only three virtual tokens. Using factorized gene-specific and relation-aware components, GRASP learns to map each pair's biological context into compact soft prompts that combine pair-specific signals with shared interaction patterns. Across diverse GN inference tasks, GRASP consistently outperforms alternative prompting strategies. It also shows a stronger ability to recover unannotated interactions from synthetic negative sets, suggesting its capacity to identify biologically meaningful relationships beyond existing databases. Together, these results establish GRASP as a scalable and generalizable prompting framework for LLM-based GN inference.

Background: The urgent care departments in Europe face a structural paradox: accelerating digitalisation is accompanied by a patient population that is disproportionately unable to engage with standard digital tools. An internal analysis at the Emergency Department (Akutafdelingen) of Nordsjaellands Hospital in Hilleroed, Denmark found that 43% of emergency patients struggle with digital solutions - a figure that reflects the predictable composition of acute care populations rather than any individual failing. Objective: This paper presents the design, iterative development, and secondary validation of the ED Adaptive Interface (v5): a prototype adaptive patient terminal developed in response to this challenge. The system operationalises what the author terms impairment-first design - a methodology that treats the most constrained patient experience as the primary design problem and derives the standard experience as a subset. The interface configures itself in under ten seconds via nurse-led setup, adapting across four axes of impairment: visual, motor, speech, and cognitive. System: Version 4 supports five accessibility modes, a heatmap pain assessment grid, a Privacy and Dignity panel, a live workflow tracker with care notifications, structured dual-category help requests, and plain-language medical term definitions across four languages. Version 5, reported here for the first time, introduces a Condition Worsening Escalation button, a Referral Pathway Display, a "Why Am I Waiting?" triage explainer, a Symptom Progression Log, MinSP/Yellow Card Scan simulation, expanded language support (seven languages: English, Danish, Arabic with full RTL layout, Turkish, Romanian, Polish, and Somali), and an expanded ten-item Communication Board. The entire system runs as a single 79-kilobyte HTML file with zero infrastructure requirements. Methods: To base the design on patient-generated evidence, two independent social media threads were subjected to an inductive thematic analysis (Braun and Clarke, 2006): a primary corpus of 83 entries in the Facebook group Foreigners in Denmark (collected March 2026) and a corroborating corpus in an international community group in the Aarhus region (collected April 2026). All identifiers in both datasets were fully anonymised under GDPR Article 89 research provisions prior to analysis. No participants were contacted. Generative AI tools were used to assist with drafting, writing, and prototype code development; all scientific content, data collection, analysis, and conclusions are the sole responsibility of the authors. Results: The first discourse corpus produced five major themes corresponding to the five problem areas the prototype was designed to address: system navigation and triage literacy gaps (31 entries); language and cultural barriers (6 entries); communication failures during care (5 entries); staff overload and capacity constraints (8 entries); and pain and severity assessment failures (14 entries). The corroborating dataset supported all five themes and introduced two additional themes: differential treatment of international patients and medical gaslighting as a long-term pattern of patient advocacy failure. One structural finding - the five most-liked comments incorrectly criticised the original poster for self-referring when she had received explicit 1813 telephone triage approval - directly inspired the Referral Pathway Display and "Why Am I Waiting?" features in v5. Conclusions: The convergence of design rationale and independent social evidence across all five problem categories suggests that impairment-first design is not a niche accessibility concern but a structural approach to healthcare interface quality. The prototype is ready for a structured clinical pilot using the System Usability Scale (SUS) and semi-structured staff interviews. The long-term roadmap includes full MinSP integration, hospital PMS connectivity, and clinical validation.

As health systems begin deploying autonomous AI agents that make independent clinical decisions and take direct actions within care workflows, ensuring patient safety and care quality requires governance standards that go beyond existing medical device frameworks designed for human-in-the-loop prediction tools. This paper introduces the Healthcare AI Agents Regulatory Framework (HAARF), a comprehensive verification standard for autonomous AI systems in clinical environments, developed collaboratively with 40+ international experts spanning regulatory authorities, clinical organizations, and AI security specialists. HAARF synthesizes requirements from nine major regulatory frameworks (FDA, EU AI Act, Health Canada, UK MHRA, NIST AI RMF, WHO GI-AI4H, ISO/IEC 42001, OWASP AISVS, IMDRF GMLP) into eight core verification categories comprising 279 specific requirements across three risk-based implementation levels. The framework addresses critical gaps in health system readiness for autonomous AI including: (1) progressive autonomy governance with clinical accountability, (2) tool-use security for agents that independently access EHRs, medical devices, and clinical systems, (3) continuous equity monitoring and bias mitigation across diverse patient populations, and (4) clinical decision traceability preserving human oversight authority. We validate HAARFs enforcement capabilities through a scenario-based red-team evaluation comprising six adversarial scenarios executed under baseline (no middleware) and HAARF- guardrailed conditions (N = 50 trials each, Gemini 2.5 Flash primary with Claude Sonnet 4.6 cross-model validation). In baseline conditions, the agent model executes unauthorized tools in 56-60% of adversarial trials. Under the HAARF condition, deterministic middleware enforcement reduces the unauthorized-tool success rate to 0%, with 0% contraindication misses and 0% policy-injection success (95% Wilson CI [0.00, 0.07]). Cross-model validation confirms identical security metrics, supporting HAARFs model-agnostic design. Mapping analysis demonstrates 48-88% coverage of major regulatory frameworks, with per-category FDA alignment ranging from 73% (C5, Agent Registration) to 91% (C3, Cybersecurity; C7, Bias & Equity). Initial validation with healthcare organizations shows a 40-60% reduction in multi-jurisdictional compliance burden and improved clinical safety governance outcomes. HAARF provides health systems with a practical, risk-stratified pathway for safe AI agent deployment--shifting from reactive compliance to proactive quality governance while maintaining rigorous patient safety standards and human-centered care principles.

Objectives: Longer lifespans lead to longer time on retirement, despite the efforts to raise the retirement age. Therefore, it is important to study how the retirement years can be spent without diseases. This study examined socioeconomic and sociodemographic differences in healthy years spent on retirement. Methods: We followed a cohort of retired Finnish municipal employees (N=4231, average follow-up 15.4 years) on national administrative registers for major chronic diseases: cancer, coronary heart disease, cerebrovascular disease, diabetes, asthma or chronic obstructive pulmonary disease, dementia, mental disorders, and alcohol-related disorders. Median healthy years on retirement and age at first occurrence of illness (ICD-10 and ATC-based) in each combination of sex, occupational class, and age of retirement were predicted using Royston-Parmar models. Prevalence rates for each diagnostic group were calculated. Results: Most healthy years on retirement were spent by women having worked in semi-professional jobs who retired at age 60-62 (median predicted healthy years 11.6, 95% CI 10.4-12.7). The least healthy years on retirement were spent by men having worked in routine non-manual jobs who retired after age 62 (median predicted healthy years 6.5, 95% CI 4.4-9.5). Diabetes was slightly more common among lower occupational class women, and dementia among manual working women having retired at age 60-62. Discussion: Healthy years on retirement are not enjoyed equally by women and men and those who retire early or later. Policies aiming to increase the retirement age should consider the effects of these gaps on retirees and the equitability of those effects.

The global ambition to end the human immunodeficiency virus (HIV) epidemic requires understanding which system-level policy levers, enacted under the framework of Universal Health Coverage (UHC), are most effective in achieving both transmission reduction and diagnostic coverage. This study addresses an important evidence gap by quantifying the within-country association between measurable UHC policy indicators and the estimated rate of new HIV infections across nine Southeast Asian countries between 2013 and 2022. Employing a Fixed-Effects panel data methodology, the analysis controls for time-invariant national heterogeneity, ensuring reliable estimates of policy impact. We found that marginal changes in total current health expenditure (CHE) as a percentage of gross domestic product (GDP) were not statistically significantly associated with changes in HIV incidence. However, increases in the UHC Infectious Disease Service Coverage Index were statistically significantly associated with concurrent reductions in HIV incidence (p < 0.001), suggesting the efficacy of targeted service implementation as the principal driver of curbing new HIV infections. In addition, the UHC Reproductive, Maternal, Newborn, and Child Health Service Coverage Index exhibited a statistically significant positive association with changes in HIV incidence (p < 0.01), which is interpreted as a vital surveillance artefact resulting from expanded detection and reporting of previously undiagnosed HIV cases. Furthermore, out-of-pocket (OOP) health expenditure as a percentage of CHE showed a counter-intuitive negative association with changes in HIV incidence (p < 0.01), suggesting this metric primarily shows ongoing indirect cost burdens on the established patient cohort, or, alternatively, presents a diagnostic access barrier that results in lower case finding. These findings suggest that policymakers should prioritise investment in targeted infectious disease service efficacy over aggregate fiscal commitment and utilise integrated sexual health platforms for strengthened HIV surveillance and case identification.

Human papillomavirus (HPV) testing is the primary method for cervical cancer screening, and a negative HPV test is associated with a very low subsequent risk of invasive cancer. Nevertheless, a small number of cervical cancers are diagnosed following an HPV-negative testing result, posing challenges within HPV-based screening pathways. Using nationwide Swedish registry data of HPV testing, we identified women diagnosed with invasive cervical cancer between 2019 and 2024 and reconstructed HPV testing histories from the National Cervical Screening Registry (NKCx). The most recent HPV test prior to diagnosis was defined as the index test, and longitudinal HPV testing trajectories were classified among women with an HPV-negative index test. Of 3,000 women diagnosed with invasive cancer, 243 (8.1%) had an HPV-negative index test. These women were older at diagnosis and more frequently diagnosed at advanced stages compared with women with an HPV-positive index test. Most HPV-negative index tests (66.3%) were performed in the peri-diagnostic period (+/- 30 days). Among women with an HPV-negative index test, 52.7% (128/243) had no prior HPV testing recorded, while the remainder had consistently HPV-negative histories (33.3%, 83/243) or evidence of prior HPV positivity before the index negative test (14%, 32/243). Possible recurrent HPV positivity following an intervening negative test was rare (0.4%, 1/243). HPV-negative screening results preceding invasive cancer reflect heterogeneous screening histories and cannot be explained solely by test failure. Findings highlighting the importance of reaching women earlier in screening programs and show that fluctuating HPV detectability is rare.

PurposeAntibiotic use (ABU) is a major driver of antimicrobial resistance (AMR), but ABU patterns are poorly understood in low-income countries where the burden of AMR is great and ABU is insufficiently regulated. Here, we report ABU from ten sites ranging from rural villages to small cities in Madagascar, a country with high AMR levels, and present results from modeling to identify factors that may be associated with ABU in this setting. MethodsWe conducted surveys of 290 individuals from ten sites in the SAVA Region of northeast Madagascar to gather data on sociodemographic characteristics, agricultural and animal husbandry practices, recent antibiotic use, the antibiotics that participants recalled using in their lifetimes, and the sources of their antibiotics. Using these data, we conducted statistical analyses with a mixed-effects logistic model to determine which characteristics were associated with recent antibiotic use. ResultsNearly all respondents (N=283, 97.6%) reported ABU in their lifetimes, with amoxicillin being the most widely reported antibiotic (N=255, 90.1% of those reporting ABU). All recalled antibiotics were classified as frontline drugs except for ciprofloxacin. Most respondents who reported antibiotic use also reported obtaining antibiotics without prescriptions from local stores (N=273, 96.5%), while only 52.3% (N=148) reported obtaining antibiotics through a prescriptive route, such as from a health clinic or private doctor. Of the 127 individuals (44.9%) who reported recent ABU, men were found to be significantly less likely to have recently taken antibiotics than women. ConclusionsOur findings provide new insights into ABU in agricultural settings in low-income countries, which have historically been understudied in AMR and pharmacoepidemiologic research. Knowledge of ABU patterns supports understanding of AMR dynamics and AMR control efforts in these contexts, such as interventions on inappropriate antibiotic dispensing. Key pointsO_LIAntibiotic use (ABU) in Madagascar is largely unstudied despite its role in antimicrobial resistance (AMR), which Madagascar faces a high burden of. C_LIO_LIABU was widespread among livestock owners in northeast Madagascar, with the majority of study participants reporting ABU in their lifetimes and most people reporting ABU also having taken antibiotics in the previous three months. C_LIO_LIMost respondents reported obtaining their antibiotics from non-pharmaceutical stores, indicating high levels of unregulated ABU, though more than half also reported sourcing their antibiotics through prescriptive means (like doctors and health clinics). C_LIO_LIMen were less likely than women to have taken antibiotics in the previous three months. C_LIO_LIThese findings support the development of interventions to mitigate the burden of AMR in Madagascar and similar contexts while underscoring the need for more comprehensive research on the drivers and patterns of ABU. C_LI Plain language summaryIn this study, we provide basic information on antibiotic use (ABU) patterns in Madagascar, a country that experiences high levels of resistance but has been particularly understudied in AMR and pharmacological research. We surveyed 290 farmers with livestock from ten sites across northeast Madagascar about their ABU and found that nearly all study participants (N=283, 97.6%) have used antibiotics in their lifetimes, while a little under half of those who reported ABU also reported using antibiotics in the previous three months (N=127, 44.9%). The most used antibiotic was amoxicillin (N=255, 90.1%). Most people obtained their antibiotics from sources that do not require prescriptions, like general stores, indicating that most ABU is unregulated. Through modeling, we also found that men were less likely than women to have taken antibiotics in the previous three months (OR=0.50, CI 0.30-0.82). These findings help us better understand the dynamics of ABU in low-income countries, which have historically been understudied in AMR and pharmacological research. They also support efforts to mitigate the burden of AMR by revealing ABU dynamics that may contribute to the emergence and spread of AMR, as well as identifying targets for intervention to curb inappropriate ABU.

BackgroundCurable sexually transmitted infections (STIs), including Chlamydia trachomatis, Neisseria gonorrhoeae, and Trichomonas vaginalis, remain highly prevalent among pregnant women in South Africa. Despite poor diagnostic performance in pregnancy, syndromic management remains standard care. Point-of-care (POC) screening enables aetiological diagnosis and same-visit treatment but is not yet included in national guidelines. We conducted a mixed-methods process evaluation to examine determinants of antenatal POC STI screening implementation in public facilities. MethodsThis evaluation was embedded within the three-arm Philani Ndiphile randomized trial (March 2021-February 2025) across four public clinics in the Eastern Cape. Screening used a near-POC, electricity-dependent nucleic acid amplification test with a 90-minute turnaround time. Reach, Adoption, Implementation, and Maintenance were assessed using the RE-AIM framework. Quantitative indicators included uptake of screening, treatment, and follow-up attendance. Qualitative data included in-depth interviews with 20 pregnant women and five focus group discussions with 21 research staff and government healthcare workers. The Consolidated Framework for Implementation Research guided qualitative analysis. Findings were integrated using narrative weaving. ResultsScreening uptake was high (99.0%), with treatment coverage of 95.2% at baseline and 93.5% at repeat screening. Same-day treatment was lower (50.7% and 69.8%) and varied substantially by facility, reflecting operational constraints including turnaround time, patient volume, infrastructure, and electricity. Attendance was higher when screening was integrated into routine ANC. Women valued screening for infant health, while providers recognised advantages over syndromic management but highlighted workforce, resource, and maintenance constraints. Socioeconomic factors, including transport costs, hunger, and work commitments, influenced retention and waiting. ConclusionsAntenatal POC STI screening was acceptable and achieved high treatment coverage in a research setting. However, same-day treatment was constrained by operational requirements of the testing platform. Scale-up will require workflow integration, strengthened health system capacity, and faster diagnostics suited to routine antenatal care. Key MessagesO_ST_ABSWhat is already known on this topicC_ST_ABSSyndromic management remains standard antenatal care in many low-resource settings despite failing to capture up to 89% of infections that remain asymptomatic. Point-of-care aetiological screening has demonstrated feasibility, acceptability, and potential clinical benefit in research settings, yet has not been widely adopted into national policy. Limited evidence exists on the health system requirements and contextual determinants influencing scale-up within routine public facilities. What this study addsThis mixed-methods process evaluation demonstrates high uptake and treatment coverage of antenatal POC STI screening in a trial setting, while identifying facility-level, structural, and socioeconomic factors shaping same-day treatment and retention. We show that implementation success varies substantially across clinics and depends on assay characteristics, workflow integration, human resources, infrastructure reliability, and follow-up capacity. How this study might affect research, practice or policyThese findings provide implementation-relevant evidence to inform national policy deliberations on integrating POC STI screening into antenatal care. Sustainable scale-up will require context-adapted delivery models, strengthened workforce and supply systems, faster diagnostics, and alignment with existing ANC workflows to ensure equitable and durable impact.

BackgroundThis study aimed to assess caregiver knowledge of Infant and Young Child Feeding (IYCF), child health, severe acute malnutrition (SAM) screening, and Community-Based Management of Acute Malnutrition (CMAM), its determinants, and associations with IYCF/ WaSH (water, sanitation, and hygiene) practices among caregivers of children 6-59 months with SAM in Ethiopian agrarian and pastoralist settings. MethodData were from the baseline survey of the R-SWITCH Ethiopia cluster-randomized controlled trial (cRCT), which screened [~]28,000 children aged 6-59 months and identified 686 SAM cases. Caregiver knowledge was evaluated using a validated 32-item questionnaire (Cronbachs for internal reliability) and analyzed via linear mixed-effects and Poisson regression models in Stata 17. ResultsCaregiver knowledge was positively associated with improved IYCF/WaSH practices among children aged 6-23 months with SAM, including higher minimum dietary diversity (MDD: IRR=1.50), minimum acceptable diet (MAD: IRR=1.63), and reduced zero vegetable/fruit intake (IRR=0.77), as well as MDD in children aged 24-59 months, improved water access (IRR=1.19), water treatment (IRR=2.02), and handwashing stations (IRR=1.41). Literate ({beta} = 4.1; 95% CI:1.5-6.6, p= 0.016), pregnant({beta} = 4.4; 95% CI:0.9-7.8, 0.018), having child weighing at a health post/ health center ({beta} = 8.9;95% CI:3.5-14.2,p [&le;] 0.001), and higher household wealth index ({beta} = 11.8;95% CI:3.6-20.1,p= 0.005) were associated with higher knowledge, while possible depression ({beta} = -0.3;95% CI: -0.5 to 0.0, p= 0.015) was associated with lower knowledge. ConclusionCaregiver knowledge determines better IYCF/WaSH practices among children aged 6-59 months with SAM. Literacy, pregnancy, having child weighing at a health post or health center, and greater household wealth were associated with caregivers knowledge, whereas possible depression was associated with lower knowledge. Integrating context-specific caregiver education and mental health support into CMAM, GMP(Growth monitoring and promotion), and primary care services could enhance feeding/WaSH practices in Ethiopia.

BackgroundInsecticides remain the cornerstone of mosquito vector control for malaria, dengue, and other mosquito-borne diseases, yet global patterns of deployment and their socioeconomic and environmental drivers are poorly characterized. Understanding where and why insecticides are used is essential for better targeting control efforts and ensuring they are effective, equitable, and efficient. MethodsWe analyzed annual country-level insecticide-use data from 122 countries (1990-2019), reported as standard spray coverage for insecticide-treated nets (ITNs), residual spraying (RS), spatial spraying (SS), and larviciding (LA). Generalized linear mixed models and hurdle models quantified associations between deployment and disease incidence, human development index (HDI), human population density, temperature, and precipitation. Models were evaluated using repeated cross-validation and applied to generate downscaled predictions of insecticide use at subnational administrative region level 2 (ADM2) globally. FindingsInsecticide deployment increased with malaria and dengue incidence, but this response was substantially stronger in higher-HDI countries, indicating that deployment depends on socioeconomic capacity as well as disease burden that leads to weaker scaling in lower-resource settings. Intervention types exhibited distinct patterns; ITN use tracked malaria burden, whereas infrastructure-intensive approaches (e.g., RS and SS) were concentrated in higher-HDI settings and increased with Aedes-borne disease incidence. Downscaled ADM2-level maps uncovered substantial within-country heterogeneity that is obscured at the national scale, highlighting regions where predicted deployment remains low relative to disease risk across sub-Saharan Africa, South Asia, and parts of Latin America. InterpretationGlobal insecticide deployment reflects not only epidemiological need but also economic and logistical capacity, creating mismatches between risk and control. High-resolution mapping can support more equitable allocation of interventions, guide insecticide resistance stewardship, and improve strategic planning as climate and urbanization reshape mosquito-borne disease risk.

BackgroundDespite their potential to serve as a reduced-harm alternative to combustible tobacco, e-cigarette take-up remains low among older (45+) adult smokers, especially in the U.S. While social media is a known driver of vaping attitudes and behaviors in younger populations, its influence on older smokers is poorly understood. This paper provides the first focused analysis of e-cigarette-related social media exposure in this population, documenting its prevalence, characteristics, and attitudinal correlates. MethodsData come from an opt-in survey of U.S. adults (N = 974) recruited via Prolific, comprising three groups: (i) non-vaping smokers aged 45+ (N = 484), (ii) former-smoking vapers aged 45+ (N = 149), and (iii) any-vaping-status smokers aged 18-35 (N = 341). Descriptive statistics, weighted to U.S. population benchmarks, characterize self-reported exposure to e-cigarette-related content on social media. Logistic regressions estimate associations between exposure and intentions for future e-cigarette use, e-cigarette harm perceptions, and related attitudes. ResultsOlder smokers (35.3%) reported exposure to e-cigarette-related content on social media less frequently than both older vapers (44.0%) and younger smokers (72.0%). For older smokers, e-cigarette health risks were the most frequently reported topic of content viewed, followed by youth vaping and e-cigarette addiction. Among this group, exposure was positively associated with stated intentions for future e-cigarette use. Exposure was not significantly associated with perceived e-cigarette harms for any group. ConclusionsFindings provide suggestive evidence that social media exposure may promote e-cigarette adoption among older smokers. However, the cross-sectional design limits causal inference, and the observed associations may reflect selection bias or reverse causality. If a causal relationship exists, the patterns observed suggest that exposure influences e-cigarette adoption through mechanisms other than updating beliefs about e-cigarette risks. While these results tentatively support the potential of social media as a channel for older-smoker harm reduction, any policy applications must carefully weigh privacy concerns and risks to youth. Rigorous experimental studies are needed to confirm these findings and clarify how social media might be leveraged to improve public health outcomes among older smokers.

Background Higher levels of sedentary behaviour, such as leisure screen time (LST), and lower levels of physical activity are associated with diseases across multiple body systems which contribute to a large global health burden. Whether these associations are causal is unclear. The primary aim of this study is to investigate the causal effects of higher LST (given greater power) and, secondarily, lower moderate-to-vigorous intensity physical activity (MVPA), on a wide range of diseases in a hypothesis-free approach. Methods A two-sample Mendelian randomisation phenome-wide association study was conducted for the main analyses. Genetic single nucleotide polymorphisms (SNPs) were first selected as exposure genetic instruments for LST (hours of television watched per day; 117 SNPs) and MVPA (higher vs. lower; 18 SNPs) based on the genome-wide significant threshold (p < 5*10-8) from the largest relevant genome-wide association study (GWAS). For disease outcomes, we used summary results from FinnGen GWAS, including 1,719 diseases defined by hospital discharge International Classification of Diseases (ICD) codes in 453,733 European participants. For the main analyses, we used the inverse-variance weighting method with a Bonferroni corrected p-value of p [&le;] 3.47*10-4. Sensitivity analyses included Steiger filtering, MR-Egger and weighted median analyses, and data from UK Biobank were used to explore replication. Findings Genetically predicted higher LST was associated with increased risk of 87 (5.1% of the 1,719) diseases. Most of these diseases were in musculoskeletal and connective tissue (n=37), genitourinary (n=12) and respiratory (n=8) systems. Genetic liability to lower MVPA was associated with six diseases: three in musculoskeletal and connective tissue and genitourinary systems (with greater risk of these diseases also identified with higher LST), and three in respiratory and genitourinary systems. Sensitivity analyses largely supported the main analyses. Results replicated in UK Biobank, where data available. Conclusions Higher levels of sedentary behaviour, and lower levels of physical activity, causally increase the risk of diseases across multiple body systems, making them promising targets for reducing multimorbidity.

This study evaluates the feasibility of implementing artificial intelligence (AI)-driven disease surveillance systems at Julius Nyerere International Airport (JNIA) in Tanzania, a key hub for regional and international travel. Through a mixed-methods approach combining qualitative interviews and quantitative surveys, the research assesses the infrastructure, human resource capacity, and regulatory frameworks necessary for AI integration. Findings indicate that while Port Health Officers are strongly optimistic about AIs potential to enhance disease detection, the airport faces significant barriers, including outdated infrastructure, insufficient technical resources, and a lack of trained personnel. Ethical and privacy concerns, particularly surrounding data security, also emerged as key challenges, compounded by limited public awareness and the socio-cultural acceptability of AI systems. Furthermore, the study identifies gaps in national policies and inter-agency coordination that hinder the effective implementation of AI technologies. The research concludes that while current conditions render AI adoption infeasible, strategic investments in infrastructure, workforce training, and policy development could pave the way for future integration, enhancing public health surveillance at JNIA and potentially other airports in low- and middle-income countries. This study contributes critical insights into the barriers and opportunities for AI-driven disease surveillance in low-resource settings, specifically focusing on a high-priority transit point, international airports. It emphasizes the importance of region-specific solutions to enhance health security in East Africa and supports the broader global health agenda by advocating for international collaboration and the development of scalable disease surveillance systems. Future research should explore pilot AI implementations at other airports to evaluate real-world challenges and refine AI systems for broader applicability, including cost-effectiveness analyses and integration of public perspectives on AI.