Sultan Mahmud

Background Bangladesh has experienced a rapid increase in cesarean section (CS) utilization over the past two decades. While previous studies have documented socioeconomic disparities in CS use, evidence on how wealth-related inequalities differ between public and private healthcare facilities remains limited. This study assessed the magnitude and drivers of socioeconomic inequality in CS utilization among facility-based births in Bangladesh. Methods We analyzed data from 3,008 facility-based births reported in the 2022 Bangladesh Demographic and Health Survey (BDHS). Survey-weighted multivariable logistic regression was used to identify factors associated with CS utilization. Wealth-related inequality was assessed using concentration curves and the Erreygers-corrected concentration index (ECCI). Regression-based decomposition of the standard concentration index was performed to quantify the contribution of socioeconomic, demographic, and healthcare-related factors to observed inequalities overall and separately for public and private facilities. Results Overall, 71.2% of facility-based births were delivered by CS, with substantially higher prevalence in private facilities (84.2%) than in public facilities (35.9%). Women delivering in private facilities had markedly higher odds of CS than those delivering in public facilities (adjusted odds ratio [AOR]: 9.07; 95% confidence interval [CI]: 7.17–11.47). Significant pro-rich inequality was observed overall (ECCI: 0.154; 95% CI: 0.117–0.191), with inequality substantially greater in public facilities (ECCI: 0.189; 95% CI: 0.114–0.264) than in private facilities (ECCI: 0.049; 95% CI: 0.014–0.084). Decomposition analysis showed that household wealth was the dominant contributor to inequality, particularly the richest wealth quintile, accounting for 81.5% of overall inequality, 63.8% in public facilities, and 109.7% in private facilities. Conclusions Wealth-related inequalities in CS utilization remain substantial in Bangladesh despite widespread use of the procedure. Although pro-rich inequality exists across both sectors, inequality is considerably greater in public facilities and is driven by different mechanisms across facility types. Policies should simultaneously improve equitable access to medically necessary CS and reduce unnecessary procedures, particularly within the private sector.

Background: Depressive symptoms among reproductive-aged women represent a major public health concern in low- and middle-income countries, yet systematic screening remains limited. In most population survey datasets, the low prevalence of depression results in severe class imbalance, which challenges conventional machine learning models. Therefore, we develop and evaluate a bagging-based ensemble machine learning framework to predict depressive symptoms among reproductive-aged women using highly imbalanced Bangladesh demographic and health survey (BDHS) 2022 data. Methods: The sample comprised women aged 15–49 years drawn from BDHS 2022 data. Depressive symptoms were defined using the Patient Health Questionnaire (PHQ-9 ≥10). Candidate predictors were drawn from sociodemographic, reproductive, nutritional, psychosocial, healthcare access, and environmental domains. Feature selection was performed using Elastic Net (EN), Random Forest (RF), and XGBoost model. Five classifiers (EN, RF, Support Vector Machine (SVM), K-nearest neighbors (KNN), and Gradient Boosting Machine (GBM)) were trained using both oversampling-based approaches and the proposed ensemble framework. Model performance was evaluated on an independent test set using accuracy, sensitivity, specificity, F1-score, and the normalized Matthews correlation coefficient (normMCC). Results: Approximately 4.8% of women were identified with depressive symptoms. The proposed bagging ensemble framework consistently achieved more balanced predictive performance than oversampling-based models. Average normMCC improved from 0.540 (oversampling) to 0.557 (ensemble). RF and GBM ensembles demonstrated notable improvements in identifying depressive cases, while the EN ensemble achieved the highest overall performance and sensitivity. Threshold optimization yielded stable normMCC across models, indicating robust trade-offs between sensitivity and specificity. Conclusions: Bagging-based ensemble learning provides a more robust and balanced approach than synthetic oversampling for predicting depressive symptoms in highly imbalanced population survey data. This approach has important implications for improving early identification and population-level mental health surveillance in resource-constrained settings.

Health systems in low- and middle-income countries (LMICs) like Bangladesh face persistent challenges in delivering timely and equitable care, often exacerbated by poor planning and inefficient resource allocation. Forecasting service utilization using routine health data can support more responsive and data-driven health system planning, yet such approaches remain under utilized in Bangladesh. By analyzing service utilization trends and projecting future service volume at national and regional levels, we aim to improve region-specific health planning. This can promote more efficient and equitable service provision. We analyzed monthly routine health service data reported into the District Health Information Software 2 (DHIS2) platform between January 2021 and March 2025 in Bangladesh. We examined key indicators across maternal, newborn, child and hospital-based services. Bayesian log-linear Poisson regression models, adjusted for seasonality and autocorrelation, were applied to forecast service utilization for the final nine months of 2025 and all of 2026. Relative changes in 2025 and 2026 were calculated using 2024 as the reference year. The analysis revealed rising trends across most service areas relative to 2024 levels. Kangaroo Mother Care (KMC) has the highest projected expansion, with coverage forecast to rise by over 75% by 2026. Over the same time period, outpatient visits and pneumonia treatment are also expected to increase by about 30%. More moderate increases are seen in low birth weight (LBW) deliveries, cesarean sections, and normal deliveries. Notable regional disparities persist, with Dhaka and Chittagong showing the highest service utilization, while Barishal and Sylhet consistently report the lowest levels. Bangladesh’s health system must prepare for increasing service utilization across all service categories. Forecasting using DHIS2 data supports for proactive planning and equitable resource allocation. Strategic investments in infrastructure, workforce, and data-driven planning are essential for building a resilient health system.

Background: Child mental health is a critical public health concern in the United States. This study examines the multifaceted determinants of child mental health, including demographic, socioeconomic, family, healthcare access, and chronic health condition information. Methods: This study utilized data from the National Survey of Children’s Health (NSCH). Univariate, bivariate, and multivariable logistic regression models were fitted to identify the determinants of each outcome variable. Results: We pooled data from two years (2020 & 2021) of NSCH and included a nationally representative sample of 6- to 17-year-old U.S. children (N = 60,809; male, 51.1%; female, 48.9%; 12-17-year-old, 50.9%; 6-11-year-old, 49.1%). There was an increase in the prevalence of depression, anxiety, and ADHD/ADD among 6-17-year-old U.S. children from 2020 to 2021, while the prevalence of behavioral problems decreased during the same period (the prevalences were 4.8%, 10.7%, 8.8%, and 11% respectively for the year of 2020 and 5.3%, 11%, 7.9% and 11.6% in 2021). Our findings reveal significant gender disparities, with girls experiencing higher rates of depression (AOR: 1.64; 95% CI: 1.34-2.01, P <0.001) and anxiety (AOR: 1.62; 95% CI: 1.43-1.84, P <0.001). Teens (12-17 years) were more susceptible to depression and anxiety than younger children (6-11 years). Ethnic and racial disparities were observed, with lower rates of these conditions among non-Hispanic Asian and Hispanic/Latino children. Neighborhood support was a protective factor against anxiety and behavioral problems. Physical activity for at least 60 minutes every day was significantly associated with lower odds of depression (AOR: 0.51; 95% CI: 0.37-0.71; P <0.001) and anxiety (AOR: 0.67; 95% CI: 0.54 - 0.83; P <0.001) as opposed to no physical activities. Adverse childhood experiences (ACEs) were consistently linked to a higher likelihood of mental health conditions. Conclusion: This study underscores the complex interplay of factors influencing child mental health and offers actionable insights for policy and practice. Future research should explore causal relationships and longitudinal trajectories to guide effective interventions and policies.

Health systems in low- and middle-income countries (LMICs) like Bangladesh face persistent challenges in delivering timely and equitable care, often exacerbated by poor planning and inefficient resource allocation. Forecasting service utilisation using routine health data can support more responsive and data-driven health system planning, yet such approaches remain underutilised in Bangladesh. By analysing service utilization trends and projecting future service volume at national and regional level, we aim to improve region-specific health planning and promote more efficient and equitable health service provision. We analyzed monthly routine health service data reported into the District Health Information Software 2 (DHIS2) platform between January 2021 and March 2025 in Bangladesh. We examined key indicators across maternal, newborn, child, and hospital-based services. Bayesian log-linear Poisson regression models, adjusted for seasonality and autocorrelation, were applied to forecast service utilisation for the final nine months of 2025 and all of 2026. Relative changes in 2025 and 2026 were calculated using 2024 as the reference year.The analysis revealed rising trends across most service areas relative to 2024 levels. Compared to levels in 2024, normal deliveries are projected to increase by 6% in 2025 and 10% in 2026, while caesarean sections are expected to rise by 4% and 12%, respectively. Low birth weight (LBW) deliveries are forecasted to increase by 10% in 2025 and 25% in 2026, with the largest relative growth in Rajshahi and Dhaka. Coverage of Kangaroo Mother Care (KMC) is projected to rise substantially—by 20% in 2025 and 74% in 2026. Pneumonia treatment is expected to increase by 31% by 2026, particularly in Dhaka. Outpatient visits are forecasted to grow by 32% by 2026. Notable regional disparities persist, with Dhaka and Chattogram showing the highest service utilisation, while Barisal and Sylhet consistently report the lowest levels. Bangladesh’s health system must prepare for increasing service utilisation across all service categories. Forecasting using DHIS2 data supports for proactive planning and equitable resource allocation. Strategic investments in infrastructure, workforce, and data-driven planning are essential for building a resilient health system.

Background: While maternal education is linked to antenatal care (ANC) use, its causal effect remains uncertain. This study applies a machine learning approach, Causal Forests, to estimate the causal impact of maternal education on adequate ANC utilization in Bangladesh. Methods: We analyzed data from 6,815 ever-married women aged 15–49 years who had a live birth within five years preceding the 2022 Bangladesh Demographic and Health Survey (BDHS). The outcome was adequate ANC utilization, defined as receiving four or more ANC visits from skilled providers. Maternal education was dichotomized as p < 0.001), the estimated Average Treatment Effect (ATE) after matching was modest and statistically non-significant (ATE = 0.006, 95% CI: –0.002 to 0.016, p = 0.17). However, significant heterogeneity in Individual Treatment Effects (ITEs) was detected across subgroups, with higher effects among women aged 20–40 at first birth, urban residents, those with media exposure, and those with wealthier or more educated spouses. The sensitivity analysis indicated moderate robustness to hidden bias. Conclusions: Maternal education alone may not significantly impact ANC use once socioeconomic and contextual factors are accounted for. However, its benefits are amplified among specific subgroups, suggesting the need for integrated and context-sensitive maternal health interventions. Advanced machine learning methods can enhance causal inference and inform equity-oriented policy strategies.

Despite advances in antiretroviral therapy (ART), the early and accurate diagnosis of Human Immunodeficiency Virus (HIV) infection remains a significant public health challenge. Traditional biomarkers, such as CD4+ T cell counts and viral load, are limited in capturing the complex biological mechanisms underlying HIV pathogenesis. This study proposes a machine learning (ML) and interpretable artificial intelligence (AI) framework to identify transcriptomic biomarkers for HIV diagnosis using immune cell–specific gene expression data. We utilized the GSE6740 dataset, which includes microarray profiles from CD4+ and CD8+ T cells of HIV-positive (treatment-naïve) and HIV-negative individuals. Feature selection was performed through differential gene expression analysis, weighted gene co-expression network analysis (WGCNA), and protein–protein interaction (PPI) network construction. Hub genes identified through these methods were used to train six supervised ML classifiers. The best-performing model was selected based on cross-validated metrics, including accuracy, Kappa, ROC, sensitivity, and specificity, and further interpreted using SHapley Additive exPlanations (SHAP) values. Seven co-expression modules were identified, with the red and green modules showing strong positive and negative correlations with HIV status, respectively. From the intersection of WGCNA modules, differentially expressed genes (DEGs), and PPI networks, ten hub genes were prioritized. Among the trained models, the regularized regression model (GLMNET) demonstrated the highest diagnostic performance (ROC = 0.97, accuracy = 91%). SHAP analysis highlighted GBP1, ISG15, OAS2, OAS1, and DDX60 as the most influential genes contributing to model predictions, thereby enhancing interpretability and biological relevance. By integrating transcriptomic profiling with interpretable ML, this study identifies novel gene-based biomarkers for HIV diagnosis and underscores the potential of explainable AI in advancing precision medicine approaches for infectious diseases.

Introduction: Patients with multiple myeloma (MM) face heightened infection susceptibility, particularly severe risks from COVID-19. This study, the first systematic review in its domain, seeks to assess the impacts of COVID-19 on MM patients. Method: Adhering to PRISMA guidelines and PROSPERO registration (ID: CRD42023407784), this study conducted an exhaustive literature search from January 1, 2020, to April 12, 2024, using specified search terms in major databases (PubMed, EMBASE, and Web of Science). Quality assessment utilized the JBI Critical checklist, while publication bias was assessed using Egger’s test and funnel plot. The leave-one-out sensitivity analyses were performed to assess the robustness of the results by excluding one study at a time to identify studies with a high risk of bias or those that significantly influenced the overall effect size. Data synthesis involved fitting a random-effects model and estimating meta-regression coefficients. Results: A total of 14 studies, encompassing a sample size of 3214 yielded pooled estimates indicating a hospitalization rate of 53% (95% CI: 40.81, 65.93) with considerable heterogeneity across studies (I2 = 99%). The ICU admission rate was 17% (95% CI: 11.74, 21.37), also with significant heterogeneity (I2 = 94%). The pooled mortality rate was 22% (95% CI: 15.33, 28.93), showing high heterogeneity (I2 = 97%). The pooled survival rate stood at 78% (95% CI: 71.07, 84.67), again exhibiting substantial heterogeneity (I2 = 97%). Subgroup analysis and meta-regression highlighted that study types, demographic factors, and patient comorbidities significantly contributed to the observed outcome heterogeneity, revealing distinct patterns. Mortality rates increased by 15% for participants with a median age above 67 years. ICU admission rates were positively correlated with obesity, with a 20% increase for groups with at least 19% obesity. Mortality rates rose by 33% for the group of patients with at least 19% obesity, while survival rates decreased by 33% in the same group. Conclusion: Our meta-analysis sheds light on diverse COVID-19 outcomes in multiple myeloma. Heterogeneity underscores complexities, and study types, demographics, and co-morbidities significantly influence results, emphasizing the nuanced interplay of factors.

Introduction: The people who inject drugs (PWID) are attributed to high-risk groups for transmission of the Hepatitis C virus (HCV). This study assessed the prevalence and associated factors of current HCV infection (CHI) among U.S. general population and PWID of ages between 20 and 59 years old. Methods: This study utilized cross-sectional data from the 2009–2018 National Health and Nutrition Examination Survey, conducting separate analyses for the U.S. general population, including PWID and non-PWID, as well as specific analyses focusing solely on PWID. The analytical methods included the estimation of CHI prevalence, Rao-Scott chi-square test to compare CHI-positive and CHI-negative groups, and univariate and multivariable logistic regressions models to evaluate the associated risk factors of CHI. Results: The prevalence of CHI among general population and PWID were 1% and 19%, respectively. Compared to non-PWID, the odds of CHI were significantly higher among PWID (OR = 32.6, 95% CI = 17.7–60.3) in general population. Among PWID, male vs. female (OR = 2.6, 95% CI = 1.1–5.9), adults aged 40–59 vs. 20–39 years old (OR = 2.9, 95% CI = 1.2–7.3), Non-Hispanic Black vs. White (OR = 4.6, 95% CI = 1.5–13.6), with high school diploma or less educational attainment vs. above college degree (OR = 3.5, 95% CI = 1.4–9.2) showed higher odds of having CHI. Conclusion: The prevalence of CHI was found to be higher among PWID especially those who were male, aged 40–59 years old, Non-Hispanic Black, and had lower educational attainment. Targeted intervention such as screening and awareness program among PWID population is recommended to reduce the burden of new HCV infections in the U.S.