Hsien-Ho Lin

Pairwise analysis of genomic and spatial data offers opportunities to identify and estimate the associations between covariates and the transmission of pathogens between individuals. However, such pairwise analyses are computationally intensive, and may not be feasible to conduct given the high dyad count in even moderately sized datasets. Here we compare two approaches to increase the efficiency of pairwise analysis for large datasets. We quantify and compare the performance of divide-and-conquer Bayesian model fitting and pairwise case-control approaches for estimating associations between individual- and pair-level covariates and shared membership in a transmission cluster. We utilize a large dataset (n = 4,154) of spatially-referenced, genomically-sequenced Mycobacterium tuberculosis isolates collected from a single city for this analysis, and conduct simulation studies under three representative tuberculosis genomic clustering settings. Across all simulation studies, the case-control approach produced negligible bias and expected 95% credible interval coverage, and performed comparably to the divide-and-conquer approach, with a somewhat narrower distribution of bias estimates and fewer outliers when effect sizes were large. Thus, we recommend using the case-control approach with five controls per case to downscale datasets for pairwise analysis when analysis of the entire dataset is not possible. This approach mitigates the computational challenges of pairwise Bayesian modeling on datasets that require significant computational resources while maintaining desired inferential properties.

Background: Economic burdens of tuberculosis (TB) in high-income settings remain overlooked. None of the 40 countries reporting TB cost surveys to the World Health Organization (WHO) were high-income. We conducted the first such survey in a high-income country, Taiwan, to assess WHO End TB Strategy’s zero catastrophic total cost target.Methods: Nationwide cross-sectional survey data were collected from households sampled from 2018 National TB Registry. Following WHO’s protocol, we extrapolated costs throughout the TB episode. Catastrophic total cost was defined as total costs, consisting of direct medical and direct non-medical and indirect costs, exceeding 20% of annual household income. Predictors of catastrophic total cost were assessed using multivariable logistic regression.Findings: Of 481 people with drug-susceptible TB (DS‑TB) and 64 with multidrug-resistant TB (MDR‑TB), 72% were male; mean age was 65. Mean total costs (95% Confidence Interval, CI) were US$2,316 (US$2,083-2,550) for DS‑TB and US$6,665 (US$4,460-8,869) for MDR‑TB. For DS‑TB, indirect, non‑medical, and medical costs comprised 57%, 29%, and 14%; for MDR‑TB, 67%, 29%, and 4%. Catastrophic total cost affected 23% (95% CI: 19-27%) of DS‑TB households and 52% (39-64%) of MDR‑TB households. Household income below the minimum cost of living, diagnosis of MDR-TB, and lack of private health insurance were the predictors of catastrophic total costs.Interpretation: Despite the well-functioning national TB program and national health insurance, a subset of TB-affected households in Taiwan faced substantial financial burdens. Integrated strategies tailored to different cost components are essential to achieve the End TB goal.

Diabetes prevalence is increasing in Taiwan, raising concern over vision impairment among affected individuals. This study aimed to assess the burden of cause-specific vision impairment among individuals with diabetes, overall and stratified by age, sex, geographic region, and time, with a particular focus on diabetic retinopathy (DR). A cross-sectional, annual analysis was conducted from 2010 to 2020. Population prevalence of major eye diseases among adults aged ≥20 years with diabetes was obtained from the National Health Insurance Research Database. Ophthalmological data from a multicentre electronic medical record system across various hospital levels were used to characterise cause-specific vision impairment. Hierarchical Bayesian models integrated these datasets to estimate the prevalence of vision impairment and corresponding years lived with disability. In 2019, DR accounted for 1283.9 years lived with disability (95% credible interval: 992.9-1, 644.6), comprising 50.8% of the total vision impairment burden among individuals with diabetes, followed by glaucoma (11.7%) and cataract (9.9%). The age-standardised prevalence per 1000 individuals for DR-related vision impairment was 0.58 (0.43-0.76), 0.50 (0.36-0.67), 0.12 (0.06-0.20), and 0.09 (0.05-0.17) for mild, moderate, severe, and blindness, respectively. Higher burdens were observed in females, individuals aged 20-39, and residents of eastern Taiwan. Between 2010 and 2020, the prevalence of DR-related vision impairment decreased in the 55-69 age group and increased in the 20-39 age group. This study highlights high-risk subgroups for vision impairment among individuals with diabetes, providing evidence to inform targeted public health and diabetes eye-care strategies.

Background: Nutrition is a critical determinant of tuberculosis (TB), providing a protective effect at high body mass index (BMI) and incurring an increased risk of TB disease at low BMI. Global nutritional transition and interventions to end hunger could directly affect the TB epidemic in high TB burden countries. Methods: We constructed dynamic TB transmission models for 12 high TB burden countries with low HIV prevalence. We explicitly accounted for the effects of BMI on TB disease progression and treatment outcomes using a meta-analysis of longitudinal cohort studies, incorporating the effect of BMI mediated through diabetes. The models were calibrated to historical trends in TB epidemiology and mean BMI. We estimated potential changes in TB incidence and mortality between 2015 and 2030 under different scenarios of population nutrition. Findings: Compared with a scenario where mean BMI remained at 2015 levels, if past trends in mean BMI continued then by 2030 TB incidence and mortality would decline by a cumulative 14.7% (95% credible interval: 12.7%–16.7%) and 15.6% (12.5%–19.2%), respectively. In comparison, achieving zero hunger by 2030 would reduce incidence and mortality by 32.0% (20.0%–43.8%) and 37.3% (26.1%–49.6%), respectively. If past trends continued and zero hunger was also achieved, incidence and mortality would be reduced by 38.2% (27.0%–49.1%) and 42.4% (32.1%–53.5%), respectively, equivalent to preventing 20.6 million people developing TB disease and averting 5.4 million TB deaths over 15 years in the 12 high-burden countries. Conclusions: Nutrition transitions and interventions to end hunger could have a major impact on the future epidemiology of TB in high-burden countries. Investment is urgently required to implement and scale up nutritional interventions.

Background: This study assessed whether higher levels of fine particulate matter (PM _2.5 ) reduce the protective effects of leisure-time physical activity (LTPA) on all-cause, cardiovascular, and cancer mortality, and explored the PM _2.5 threshold beyond which attenuation occurs. Methods: We conducted two complementary investigations. First, a systematic review and meta-analysis (per PRISMA guidelines) identified eligible cohort studies from PubMed, Web of Science, Embase, and SPORTDiscus (from inception to 6 January 2025) that examined the independent or joint associations of LTPA and PM₂.₅ with mortality among adults (≥ 18 years). Second, an individual-level pooled analysis using harmonized data from three cohorts was performed using Cox regression modeling to assess the associations observed in the meta-analysis. Results: In Study One, a total of seven cohort studies ( n = 1,515,094; deaths = 115,196) were included in the meta-analysis, revealing that the reduction in all-cause mortality risk diminished with higher PM _2.5 exposure. Meeting the recommended LTPA level (7.5–15 MET-h/week) reduced all-cause mortality risk by approximately 30% at PM _2.5 < 25 μg/m ³ but only 12–15% at 25 + μg/m ³ . Study Two (three cohorts; n = 869,038; deaths = 45,080) confirmed this pattern. Individuals meeting the recommended LTPA level (7.5–15 MET-h/week) had a lower risk of all-cause mortality compared to those in the highest-risk group (reference: < 1 MET-h/week and PM _2.5 : 35–50 μg/m ³ ). Hazard ratios (HRs) varied by PM _2.5 exposure, with lower HRs indicating a greater protective effect: 35–50 μg/m ³ (HR = 0.75, 95% CI: 0.61–0.93), 25–35 μg/m ³ (HR = 0.67, 95% CI: 0.57–0.79), 15–25 μg/m ³ (HR = 0.34, 95% CI: 0.29–0.39), 10–15 μg/m ³ (HR = 0.34, 95% CI: 0.28–0.41), and < 10 μg/m ³ (HR = 0.30, 95% CI: 0.25–0.37). Higher levels of LTPA were generally associated with lower all-cause and cause-specific mortality across most PM₂.₅ exposure categories, but the protective effects were attenuated at PM₂.₅ levels 25 + μg/m ³ for all outcomes and became non-significant for cancer mortality at 35–50 μg/m ³ . Conclusions: LTPA is beneficial for all-cause, cardiovascular, and cancer mortality even at relatively high PM _2.5 levels, with greater benefits observed under cleaner air conditions. However, its protective effects are attenuated at 25 + μg/m ³ for all outcomes and become less evident at 35–50 μg/m ³ , particularly for cancer mortality. PROSPERO Registration Number: CRD42023395364.

Objectives: The COVID-19 pandemic was associated with significant changes in preventive behaviors worldwide. We conducted a cohort follow-up survey to understand these behavioral changes in Taiwan during 2020 and 2021. Methods: A population-based survey with three follow-up rounds was implemented among individuals aged 20 and older. The survey included questions about perceptions of COVID-19 risk and adherence to preventive behaviors promoted by the Taiwan CDC in its New Life Strategies. Results: Between July and September 2020, a period with no local COVID-19 cases, adherence to indoor and outdoor social distancing was 51.5% and 63.5%, respectively, while 81.4% of people indicated they would wear facial masks if social distancing could not be practiced. In stratified analysis, females and the elderly exhibited significantly higher adherence to most of the preventive measures. We observed an increase in adherence to preventive behaviors from 2020 to the middle of 2021, with no signs of fatigue. Conclusions: The survey provided critical empirical evidence on the adherence to preventive behaviors in the general population in Taiwan during the first two years of the pandemic. As non-pharmaceutical interventions (NPI) are essential measures during the early stage of any pandemic, we believe that regular monitoring of preventive behaviors could serve as a foundation for pandemic preparedness. More efforts are needed to identify the most feasible and cost-effective methods for surveillance of preventive behaviors.