The COVID-19 pandemic highlighted the importance of timely infectious disease surveillance. In Japan, conventional sentinel and claims-based systems incur reporting lags and capture limited clinical detail, whereas free-text clinical notes in electronic health records (EHRs) hold richer, timelier symptom and vaccination information. Natural language processing (NLP) with large language models (LLMs) offers a way to structure such free text at scale.

We aimed to develop and internally validate an NLP algorithm to extract structured infectious disease-related symptoms and vaccination history from free-text clinical notes in Japanese primary care, as a feasibility step toward low-latency, EHR-based surveillance.

A total of 773 clinical notes, originating from 526 unique patients, were provided by M3 Inc through the Japan Medical Data Survey and used for analysis. Three physicians annotated information related to infectious disease symptoms and vaccination history. The data were divided into 622 (80%) training cases and 151 (20%) evaluation cases with no patient overlap. We compared a physician-designed, rule-based algorithm, few-shot learning (FSL) using commercial and open-source LLMs, and supervised fine-tuning (SFT) of open-source LLMs, using the macroaveraged F₁-score (unweighted mean across 9 clinical categories). Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were also computed, with 95% CIs from a patient-level cluster bootstrap (2000 replicates).

Rule-based extraction achieved a macroaveraged F₁-score of 0.685 (95% CI 0.630-0.736). FSL markedly improved the extraction of high-variability items such as vaccination history and onset date. Anthropic Claude 3.5 Sonnet achieved a macroaveraged F₁-score of 0.875 (95% CI 0.800-0.913; sensitivity 0.929, specificity 0.918). SFT of Google's open-source Gemma 2 27B model with quantized low-rank adaptation (QLoRA) achieved the highest point estimate (macroaveraged F₁-score of 0.906, 95% CI 0.833-0.945; sensitivity 0.921, specificity 0.969, PPV 0.906); the difference from Claude 3.5 Sonnet was small and not statistically distinguishable (ΔF₁-score=0.030, 95% CI -0.035 to 0.140). A small, fine-tuned Gemma 2 2B model reached 0.822 (95% CI 0.752-0.874), significantly lower than that of the 27B model (ΔF₁-score=0.084, 95% CI 0.040-0.163).

A fine-tuned open-source LLM can accurately extract and structure infectious disease-related information from Japanese free-text clinical notes, achieving performance comparable to that of a commercial model while enabling processing within a closed environment. These findings support the feasibility of EHR-based digital surveillance, whose downstream utility remains to be demonstrated.

To estimate the Centers for Medicare and Medicaid Services (CMS) Hierarchical Condition Category (HCC) risk model using Medicare Advantage (MA) encounter data, as an initial step toward recalibrating risk-adjusted MA payments.

A 20% sample of Traditional Medicare (TM) claims and MA encounter data for 2016-2022. Standardized fee schedules provide a measure of resource use in TM claims and MA encounters.

Ordinary least squares regression replication of the CMS HCC version 28 model for relative resource use among community-dwelling, non-dual aged and disabled individuals. Robustness testing includes replication with version 22 model structure, sensitivity to MA chart review records, MA contracts with complete encounter data, and patterns of care during the COVID pandemic.

Using TM data from 2016 to 2022 results in modest changes in estimated coefficients and 1.7% lower average HCC scores, relative to 2018-2019 TM data used for CMS's HCC model v28. Using MA data result in 8.9% lower average scores than TM-based scores. The differences between MA- and TM-based HCC scores vary across the distribution of scores. When re-estimating HCC v28 coefficients, increasing trends in TM and MA diagnosis prevalence are associated with smaller (diluted) coefficients in TM and MA-based HCC risk models. Trends in medical technology can increase (e.g., high-cost targeted cancer therapies) or decrease (e.g., lower-cost biosimilars) HCC model coefficients, with evidence of larger technology-related decreases in an MA-based model. The decrease in MA-based scores does not create new disincentives to enroll beneficiaries who are racial/ethnic minorities or rural residents.

We make an important contribution to the policy debate about MA risk adjustment. Any changes in risk-adjusted MA payment need to be reviewed in the full context of MA payment policy and MA plan enrollment incentives.

Venous thromboembolism (VTE) encompassing deep vein thrombosis (DVT) and pulmonary embolism (PE) poses a significant burden despite advances in diagnostic and therapeutic options. The risk factors and clinical outcomes of VTE are dynamic, highlighting the need for more comprehensive studies, especially in sub-saharan Africa. This study aimed to identify risk factors and clinical outcomes in three major tertiary centers in Kenya.

The HEART Registry was developed in 2021 and recruited patients with VTE prospectively over three years from three facilities in Kenya. Patients' demographics, risk factors for VTE, hospital course, and clinical outcomes over six months were obtained. Descriptive statistics were used to summarize the data, presenting the findings as frequencies, percentages, medians, and interquartile ranges. No pre-specified inferential or comparative analyses were planned, consistent with the descriptive registry design.

A total of 422 patients with VTE were prospectively enrolled. The median age was 46.5 years notably younger than typically reported in Western populations. DVT was the most common presentation (54%) followed by PE (38.4%). A substantial proportion of patients (49.5%) did not have any identifiable risk factors (unprovoked VTE). The most common acute-phase treatment was low molecular weight heparin (57.5%), followed by warfarin (28.6%). Direct oral anticoagulants recommended as first line therapy by current international guidelines were available but underutilized reflecting access constraints. Only 1.7% of participants received thrombolytic therapy. The median length of hospital stay was eight days. In the acute phase, 10.4% (n=31) of admitted participants died. At six-month follow-up, death occurred in an additional 39 patients.

A substantial proportion of VTE patients in Kenya had no identifiable risk factors which is a pattern consistent with global data on unprovoked VTE. This highlights the need for caution in attributing VTE solely to preventable causes. VTE associated acute and six month mortality was high, driven predominantly by underlying comorbidities. Preventing VTE in acute medical conditions remains a major area of concern and future prospective studies are needed to identify novel population specific risk factors and optimize outcomes in this setting.

Pneumocystis jirovecii pneumonia (PJP) is a life-threatening opportunistic infection in immunocompromised patients. BALF cytomegalovirus (CMV) co-detection is frequently observed in PJP patients, but its clinical characteristics and prognostic impact in non-HIV populations remain unclear.

In this single-center retrospective cohort study, we enrolled 62 non-HIV patients with confirmed PJP between 2019 and 2023. BALF CMV co-detection was defined as detectable CMV DNA in bronchoalveolar lavage fluid via metagenomic next-generation sequencing (mNGS), combined with compatible respiratory symptoms and chest computed tomography abnormalities. The Benjamini-Hochberg false discovery rate (FDR) correction was applied for multiple comparisons.

Overall, 31 patients (50.0%) had BALF CMV co-detection. The 28-day all-cause mortality was significantly higher in the CMV co-detection group than in the PJP-only group (54.84% vs. 19.35%, p = 0.008), and dyspnea was more prevalent (p = 0.024). After FDR correction for 39 laboratory parameters, only fibrinogen remained significantly lower in the co-detection group (q = 0.039), while (1,3)-β-D-glucan (BDG) and D-dimer showed independent associations with CMV co-detection in multivariable analyses. mNGS revealed more concurrent viral and fungal pathogens in the co-detection group, and multiple co-pathogens were more frequent in non-survivors within this subgroup. Interleukin-6 (IL-6), procalcitonin (PCT) and D-dimer were identified as independent prognostic factors for 28-day mortality in the CMV co-detection subgroup.

In this single-center retrospective cohort, BALF CMV co-detection is associated with substantially elevated unadjusted 28-day mortality in non-HIV patients with PJP. These findings are hypothesis-generating; IL-6, PCT and D-dimer show potential as exploratory prognostic markers. Comprehensive mNGS-based pathogen screening combined with biomarker monitoring may facilitate risk stratification in this high-risk population, pending validation in larger prospective cohorts.

Between January and March 2022, the UK Public Health Rapid Support Team delivered a master's-level short course, Pandemics: Emergence, Spread and Response, designed to strengthen interdisciplinary knowledge and skills for pandemic preparedness. This study evaluated the course's perceived usefulness, outcomes, and longer-term impacts on professional practice, one year after completion.

A participatory convergent mixed-methods design was used, involving course alumni as co-evaluators. Quantitative data were collected through an online survey (n = 31; 62% response rate), while qualitative data were generated through focus group discussions (FGDs), value creation stories (VCS), and iterative workshops. Wenger, Trayner and de Laat's Value Creation Framework informed the evaluation design and interpretation of findings. Quantitative analyses were exploratory and intended to identify patterns rather than test formal hypothesis.

Participants consistently reported that the course was relevant and useful to their professional roles. Exploratory analyses identified associations between perceived course usefulness and engagement, course content, delivery mode, learning platform, and facilitator diversity; these findings should be interpreted cautiously given the small sample size and exploratory design and should not be interpreted as causal. Qualitative findings provided explanatory depth by demonstrating how interactive pedagogy, interdisciplinary learning, practical case-based teaching, and exposure to diverse perspectives facilitated knowledge transfer and application in professional practice, generating value across Wenger's cycles of immediate, potential, applied, realised and reframing value.

Overall, the findings suggest that interdisciplinary, learner-centred online training may support professional practice change when engagement, inclusivity, and contextual relevance are prioritised. This study demonstrates the value of participatory, mixed-methods evaluation and Wenger's Value Creation Framework for understanding long-term learning outcomes in pandemic preparedness training.

Rheumatoid arthritis-associated interstitial lung disease (RA-ILD) is one of the most severe extra-articular complications of rheumatoid arthritis (RA). Although tumor necrosis factor-α (TNF-α) plays a central role in RA pathogenesis, its relationship with RA-ILD remains uncertain. This study aimed to investigate the association between serum TNF-α levels and RA-ILD, with a focus on sex-stratified patterns.

We performed a cross-sectional study that initially included 1,191 consecutive RA inpatients at Xingtai People's Hospital between March 2022 and December 2024. Clinical and laboratory data were extracted from electronic medical records. RA-ILD was diagnosed using high-resolution computed tomography, with subtypes classified by consensus. Multivariable logistic regression and generalized additive models were used to assess the association of TNF-α with RA-ILD. Serum TNF-α was natural log-transformed (ln) to improve linearity and reduce the influence of extreme values.

After exclusions, 790 patients were included in the final analysis, of whom 149 (18.86%) had RA-ILD, with a higher prevalence in males than in females (30.68% vs 15.47%). In sex-stratified analyses, after adjusting for age, disease duration, rheumatoid factor, and anti-citrullinated protein antibody in both sexes, and additionally adjusting for smoking in males, serum TNF-α was associated with RA-ILD in females (highest vs lowest tertile: OR = 1.96, 95% CI: 1.10 - 3.48; P for trend < 0.01). This association followed a nonlinear pattern, with an exploratory inflection point above which elevated TNF-α showed a positive association with increased odds of RA-ILD. In males, no statistically significant association was observed across all models. However, the formal sex ×ln(TNF-α) interaction term was not statistically significant (P = 0.08).

In sex-stratified analyses, serum TNF-α was associated with RA-ILD in females, exhibiting a nonlinear relationship with a potential inflection point; no statistically significant association was observed in males. However, as the interaction testing did not confirm a statistically significant sex difference, these findings are exploratory and warrant further prospective validation in larger cohorts.

Wastewater-based epidemiology (WBE) is a cornerstone of post-pandemic SARS-CoV-2 surveillance, yet the environmental covariates and spatiotemporal heterogeneities of viral signals remain poorly characterized.

We conducted a 35-month longitudinal study across 19 sites in Chongqing, China, quantifying SARS-CoV-2N gene concentrations via RT-qPCR to assess the value of WBE and the correlation between viral concentrations and meteorological fluctuations. We employed Gaussian generalized additive models with REML estimation to quantify seasonal dynamics and lagged effects of diurnal temperature range (DTR).

A total of 2,520 wastewater samples with complete data were obtained, with an overall N gene detection rate of 62.02%, a median flow-population normalized viral load of 8.40 log10 gc/day per 1,000 inhabitants, and median 1-day, 3-day and 7-day lagged DTR of 6.05 °C, 6.05 °C and 5.67 °C, respectively. Our results demonstrate a surveillance lead time for wastewater signals relative to clinical notifications. WBE detected a 2025 resurgence 2 months earlier than official reports. Wastewater SARS-CoV-2 circulation exhibited a distinct seasonal pattern, characterized by a primary peak in June and two secondary peaks in March and August. The spatiotemporal distribution patterns revealed by monthly heatmaps were highly consistent with the city-wide seasonal trends, and significant spatial heterogeneity was observed. The 7-day moving average DTR exhibited a complex nonlinear relationship with viral loads, initially decreasing, then increasing to a peak at 7-8 °C before declining again.

These findings indicate that WBE serves as a useful population-level proxy for infection burden and trends, complementing clinical surveillance. The statistical association between DTR and viral dynamics offers a predictive framework for refining early-warning systems and informs targeted interventions in complex urban environments.

Pediatric idiopathic nephrotic syndrome (INS) involves complex immunological mechanisms, and clinical presentation may be influenced by environmental factors and access to medical services. The impact of the COVID-19 pandemic on its clinical and evolutionary profile remains insufficiently characterized.

To compare pandemic and post-pandemic pediatric INS regarding demographic, clinical and biological characteristics and early disease evolution, and to explore factors associated with infection-related onset and corticosteroid response.

Single-center retrospective study on 59 pediatric patients with INS, diagnosed during the pandemic (n=29) and post-pandemic (n=30) period, analyzing demographic, clinical, biological and therapeutic variables.

Patient distribution was similar between periods. Hospital stay was longer during the pandemic (median 10, IQR 5 vs. median 8, IQR 4; p = 0.015), whereas the interval from symptom onset to presentation did not differ significantly between periods. Infections associated with onset were less frequent during the pandemic compared to the post-pandemic period (31% vs. 60%; p=0.026), without differences in severity. Corticosteroid-response categories are reported descriptively because follow-up duration differed substantially between the two cohorts, limiting complete ascertainment of corticosteroid dependence in the post-pandemic group. A lower proportion of corticosteroid-sensitive patients was observed during the pandemic, while corticosteroid responsiveness was associated with time to remission (ρ = 0.439; p = 0.003). Exploratory logistic regression suggested that the post-pandemic period, anemia at diagnosis, and weight status were associated with infection-associated onset. In an exploratory logistic regression model, macroscopic hematuria was the only significant clinical variable associated with the observed corticosteroid response (p = 0.039). Biological parameters were similar between periods, except for higher fibrinogen (p = 0.021) and complement C4 (p = 0.034) levels during the pandemic. Exploratory descriptive analyses of the subgroup with available C4 measurements did not identify statistically significant associations between C4 levels and corticosteroid-response categories or other examined clinical characteristics.

Children diagnosed during the pandemic experienced longer hospitalizations, whereas most demographic and biological characteristics were comparable between periods. Infection-associated onset was less frequent during the pandemic than during the post-pandemic period, whereas the number of diagnosed cases was similar between the two study periods. Complement C4 differed between the pandemic and post-pandemic cohorts, but exploratory descriptive analyses did not identify statistically significant associations between C4 levels and corticosteroid response or other examined clinical characteristics. Hematuria was associated with corticosteroid response, whereas anemia and weight status were associated with infection-related onset in exploratory analyses. However, these findings should be interpreted cautiously because of the limited sample size and require confirmation in larger prospective studies.

Severe pneumonia is frequently associated with dysregulated immune-inflammatory responses and immunothrombotic activation, contributing to myocardial injury and adverse clinical outcomes. Early identification of patients at high risk for myocardial injury and mortality across varied infectious etiologies remains challenging. We aimed to characterize inflammatory and coagulation-related signatures associated with myocardial injury and construct interpretable machine learning models for risk stratification in patients with severe pneumonia.

This retrospective cohort study enrolled 287 adult patients with severe pneumonia admitted to the intensive care unit from 2018 to 2024. All patients were stratified into groups with bacterial infection, COVID-19 and bacterial co-infection, and influenza and bacterial co-infection. Clinical variables collected within 48 h after admission were analyzed using multivariable regression, competing-risk models, and interpretable machine learning. SHapley Additive exPlanations (SHAP) were used to identify key predictive features.

Myocardial injury was highly prevalent across all infectious subgroups. Patients with bacterial infection exhibited an exacerbated inflammatory and coagulation burden, characterized by elevated leukocyte counts, D-dimer levels, and prolonged prothrombin time. Multivariate analysis confirmed that D-dimer, prothrombin time, and creatinine were independently associated with myocardial injury. Machine learning analyses identified coagulation and inflammatory markers as major contributors to myocardial injury risk. For mortality prediction, the XGBoost model yielded optimal predictive performance with an AUC of 0.85. SHAP analysis revealed that vasopressor administration, prothrombin time, advanced cardiovascular support, age, and hypoxemia were the top prognostic determinants for mortality. Although infectious etiology was not independently associated with mortality, patients with myocardial injury and bacterial infection exhibited the poorest survival outcomes.

Inflammatory and immunothrombotic signatures are closely associated with myocardial injury and adverse outcomes in severe pneumonia. Interpretable machine learning models exhibited promising discriminative performance for both myocardial injury and all-cause mortality in our cohort. Although the initial training dataset was derived from a single center with a relatively limited sample size, we performed temporal validation in this study, which preliminarily suggested the potential clinical applicability of these models for the early identification of high-risk patients.

Limited evidence exists regarding user satisfaction with 151 call center service established by the Ministry of Health in Kuwait. This study explores user satisfaction with a healthcare call center by analyzing user experiences with Kuwait's Ministry of Health 151 call center.

A national anonymous online survey was conducted through Twitter, WhatsApp, and Facebook. A 5-point Likert scale was used to obtain user ratings of their satisfaction with the various aspects of the 151 call center. Mean ratings and their standard deviations summarized the ratings, as One-way ANOVA tested the existence of statistically significant differences in mean ratings across groups to identify factors influencing the ratings.

Out of 1020 respondents, 28.4% (n = 290) had contacted the 151 service with vaccination-related inquiries, general inquiries, and COVID-19-related inquiries. Overall, female participants expressed statistically significantly higher satisfaction levels with quality (p < 0.021) and communication with doctors (p < 0.030). Response to calls, forwarding of complains, and representatives' knowledge and skills were rated highly by respondents with high school diploma and bachelor's degree compared to individuals with less than high school education, master's degree, and doctorate degree (p = 0.032, p = 0.012, and p = 0.007, respectively).

Nationally-representative Kuwaiti embraced the 151 call center, with gender and education level influencing the participants' satisfaction with 151. The findings revealed the need for the call center to tailor-make services by gender and education diversities to optimize user experiences.