Background: Unmet need for postpartum contraceptive (PPC) remains a major public health challenge in Sub-Saharan Africa (SSA). Traditional regression models often fail to capture complex, non-linear interactions among predictors. This study developed an explainable machine learning framework combined with association rule mining to predict postpartum unmet need and extract policy-relevant multi-factor risk profiles. Methods: We analyzed pooled Demographic and Health Survey (DHS) data from 66,454 postpartum women across 27 SSA countries (2016–2024). Logistic Regression, Random Forest, and XGBoost models were optimized using stratified five-fold cross-validation. SMOTE was applied strictly to training data to resolve class imbalance. Model performance was evaluated on an independent 20% test set using accuracy, precision, recall, F1-score, and AUC-ROC. SHapley Additive exPlanations (SHAP) prioritized key features for Apriori-based association rule mining (thresholds: Support ≥15%, Confidence ≥45%, Lift >1.0). Results: In SSA, the overall prevalence of unmet need for PPC was 24.0%. Random Forest achieved the highest predictive performance (AUC-ROC: 0.84, 95% CI: 0.83–0.85; Accuracy: 86.0%; Recall: 85.0%; F1-score: 83.0%). SHAP analysis identified marital status (married), higher parity, maternal age (15-19), higher education attainment, and healthcare-access barriers as key predictors of unmet need. Association rule mining revealed multi-variable risk patterns; notably, being married combined with financial barriers to healthcare yielded a support of 18.02%, confidence of 48.34%, and a lift of 1.37 (a 37% increased risk above baseline). Conclusions: Postpartum unmet need is associated with intersecting structural, economic, and systemic access barriers. Combining machine learning, SHAP, and association rule mining provides robust predictive power alongside interpretable, actionable insights to guide targeted reproductive health policy in low-resource settings.
Background/Objectives Indoor air quality (IAQ) in healthcare settings significantly influences the health, comfort, and productivity of healthcare workers (HCWs). Malaysia’s tropical climate poses significant occupational health risks in hospital environments where inadequate ventilation, high humidity and pollutant accumulation intensify exposure hazards for healthcare workers. This study aimed to assess IAQ-related conditions across different hospital work areas (open and enclosed) and to describe environmental discomfort and self-reported health-related symptoms among healthcare workers in a major specialist hospital in northern Malaysia. Methods A mixed-methods design was employed, comprising a systematic walkthrough inspection guided by the Industry Code of Practice on Indoor Air Quality (ICOP-IAQ, 2010) and a cross-sectional survey using a validated Questionnaire for Building Occupants. Data from 265 HCWs were analysed using descriptive and inferential statistics (Chi-square test) to explore differences in environmental discomfort and reported symptoms across workstation types. Results Walkthrough inspection revealed visible fungal growth, dust accumulation, and poor ventilation across several clinical, support and administrative areas. The most commonly reported environmental discomforts were temperature variability (78.1%), dust (71.7%), and low temperature (68.3%), while the most frequently reported symptoms were fatigue (78.1%), headache (73.6%), and cough (69.1%). Open-concept workstations were significantly associated with high temperature, stuffy air, dry air, and unpleasant odours. Conclusions: The findings indicate notable IAQ-related concerns across hospital work areas, accompanied by widespread environmental discomfort and a high prevalence of self-reported SBS-like symptoms among healthcare workers.
Background This study aimed to evaluate the prognostic accuracy of the cerebroplacental ratio (CPR) for predicting adverse neonatal outcomes in pregnancies complicated by fetal growth restriction (FGR). Methods A prospective cohort study was conducted in …., from March 2021 to March 2023, including 114 women with estimated fetal weight <10th percentile between 28 and 38 weeks of gestation. The primary objective was to assess the prognostic accuracy of CPR for adverse neonatal outcomes. The secondary objective was to identify clinical and neonatal predictors of adverse outcomes. CPR was calculated as the ratio of the middle cerebral artery pulsatility index to the umbilical artery pulsatility index, with abnormal values defined as ≤1.1. Participants were categorized into normal and abnormal CPR groups. Adverse neonatal outcomes included umbilical artery pH <7.1, 5-minute Apgar score <7, neonatal hypoglycemia (<45 mg/dL), and NICU admission >10 days. ROC curve analysis assessed prognostic accuracy, while logistic regression identified significant predictors. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of CPR were calculated. Results A total of 98 patients were included in the analysis (49 per group). Neonates in the normal CPR group had significantly higher mean birth weight compared to the abnormal CPR group (2372.5 ± 649 g vs. 1802.5 ± 579 g, p < 0.001). ROC analysis indicated an area under the curve (AUC) of 0.71 (95% CI: 0.56–0.75, p = 0.011) for CPR in predicting adverse neonatal outcomes. Using a cutoff value of 1.087, CPR demonstrated 78% sensitivity, 73% specificity, 77% positive predictive value, and 68% negative predictive value. Logistic regression showed that higher birth weight was protective against adverse outcomes, with each 100-g decrease associated with a 2% higher odds of an adverse neonatal event (adjusted OR 1.02; 95% CI 1.01–1.03; p < 0.001). Conclusion CPR is a valuable, non-invasive tool for predicting adverse neonatal outcomes in FGR pregnancies. Lower CPR values were associated with poorer neonatal condition and increased NICU admissions. These findings support the use of CPR for risk assessment and clinical decision-making in high-risk pregnancies.
Background Beta thalassemia trait (BTT) screening is challenged by a diverse mutation spectrum and by individuals with borderline HbA2 values who may carry pathogenic variants despite near-normal screening results. This study aimed to characterise the genotypic spectrum of BTT carriers in a heterogeneous Indian population, compare hematological parameters across β0, β+, and rare HBB variant groups, and assess the diagnostic performance of red cell parameters and discrimination indices for carrier identification, with particular focus on the borderline HbA2 range. Methods Within a population-based cohort of 4,886 young adults (Delhi NCR), 209 individuals (115 with HbA2 4–9%; 94 with borderline HbA2 3.2–3.9%) underwent molecular analysis by ARMS-PCR and Sanger sequencing. Results The β0 mutations (31.7%) showed the most pronounced thalassemic profile, and MCV, MCH, and HbA2 differed significantly across β0, β+, and rare HBB variant groups (p ≤ 0.002). In the full cohort, MCV was the strongest single discriminator (AUC 0.88), followed by HbA2 (0.86) and MCH (0.84); the Shine and Lal index performed best among the indices (AUC 0.88). Within the borderline subgroup, MCH and MCV retained the greatest discriminatory value (AUC 0.84 and 0.83), with high specificity (92.2% and 87.8%) at cut-offs of 20.8 pg and 73.9 fL. Conclusion MCV, MCH, and HbA2 are the most informative first-tier measures for flagging BTT carriers, though molecular confirmation remains essential. In the borderline HbA2 range, MCH and MCV best prioritise individuals for molecular confirmation. Future studies should evaluate better screening criteria for this under-studied group to reliably identify carriers.
Problem considered Liver cirrhosis (LC) is linked to ischemic heart disease (IHD). This study investigates mortality trends due to concurrent LC and IHD using national-level data in the U.S. Methods This retrospective, population-based descriptive study utilized the CDC WONDER (Centers for Disease Control and Prevention Wide-ranging Online Data for Epidemiologic Research) database to retrieve mortality data among adults aged ≥25 years from 1999 to 2023. Crude rates (CRs), age-adjusted mortality rates (AAMRs) per 100,000 population, and annual percentage changes (APCs) with 95% confidence intervals (CIs) were calculated, and data were stratified by year, sex, race, urbanization, and state. Results Among 113,089 deaths, AAMRs decreased from 1999 to 2009 (APC: -1.16; 95% CI: -1.97 to -0.60), increased from 2009 to 2018 (APC: 2.71; 95% CI: 1.80 to 3.51), and rose sharply from 2018 to 2021 (APC: 9.72; 95% CI: 7.56 to 11.15). Rates between 2021 and 2023 were stable. (APC: 0.35; 95% CI: -1.95 to 2.93) Males had triple the AAMRs compared to females (3.08 vs 1.01). Non-Hispanic (NH) American Indian/Alaska Natives had the highest AAMRs (4.05), while NH Asian/Pacific Islanders (0.99) had the lowest. AAMRs in the West (2.14) were the highest, followed by the South (2.12), Midwest (1.77), and Northeast (1.66). Nonmetropolitan areas had higher AAMRs than metropolitan ones (1.96 vs 1.84). States in the top 90th percentile of crude mortality rates included West Virginia (CR: 3.89), Rhode Island (CR: 3.63), Oklahoma (CR: 3.29), Texas (CR: 3.01), and Vermont (CR: 3.01). Conclusion Mortality due to concurrent LC and IHD has risen in recent years, remaining higher in men, NH American Indian/Alaska Natives, nonmetropolitan areas, and the West, underscoring the need for strategic interventions to mitigate these disparities.
Problem considered Environmental enteric dysfunction (EED) is a subclinical inflammatory disorder of the small intestine that results in nutrient malabsorption, stunting, impaired cognitive development, and adverse pregnancy outcomes. It disproportionately affects populations in low- and middle-income countries, where chronic exposure to enteric pathogens and poor sanitation is common. Despite growing recognition as a global health concern, EED remains poorly defined, with no standardized diagnostic tools. This bibliometric analysis examines research trends, leading contributors, and emerging themes to inform future research directions. Methods The study analyzed publications related to EED retrieved from the Scopus database, covering the period 1972 to 2024. Data cleaning was performed using OpenRefine. Keyword co-occurrence network analysis was conducted using VOSviewer, while trend and collaboration analyses were carried out in R Studio, and geographic visualization in Tableau. Results From 1972 to 2024, 484 publications on EED were identified, with an annual growth rate of 5.06%. Output increased markedly after 2010, driven by growing recognition of EED in child growth and malnutrition. The United States and the United Kingdom led publication output, while Bangladesh, Pakistan, and India were the most active LMICs, though their contributions remain low relative to disease burden. Research focus has shifted from basic pathology to multidisciplinary areas, including microbiota, biomarkers, and WASH-related determinants. Conclusion EED research has expanded over five decades, evolving from a focus on malnutrition to biomarkers and mechanistic studies. However, geographic imbalances persist, underscoring the need to strengthen research capacity, promote equitable collaborations, and translate mechanistic insights into scalable, context-appropriate interventions.
Background Unmet need for postpartum family planning (PPFP) remains a major public health challenge in Sub-Saharan Africa (SSA). Traditional regression models often fail to capture complex, non-linear interactions among predictors. This study developed an explainable machine learning framework combined with association rule mining to predict postpartum unmet need and extract policy-relevant multi-factor risk profiles. Methods We analyzed pooled Demographic and Health Survey (DHS) data from 66,454 postpartum women across 27 SSA countries (2016–2024). Logistic Regression, Random Forest, and XGBoost models were optimized using stratified five-fold cross-validation. SMOTE was applied strictly to training data to resolve class imbalance. Model performance was evaluated on an independent 20% test set using accuracy, precision, recall, F1-score, and AUC-ROC. SHapley Additive exPlanations (SHAP) prioritized key features for Apriori-based association rule mining (thresholds: Support , Confidence , Lift ). Results Overall unmet need for PPFP was 24.0%. Random Forest achieved the highest predictive performance (AUC-ROC: 0.84, 95% CI: 0.83–0.85; Accuracy: 86.0%; Recall: 85.0%; F1-score: 83.0%). SHAP analysis identified marital status, parity, maternal age, education, and healthcare-access barriers as key predictors. Association rule mining revealed critical multi-variable risk patterns; notably, being married combined with severe financial barriers to healthcare yielded a support of 18.02%, confidence of 48.34%, and a lift of 1.37 (a 37% increased risk above baseline). Conclusions Postpartum unmet need is driven by intersecting structural, economic, and systemic access barriers. Combining machine learning, SHAP, and association rule mining provides robust predictive power alongside interpretable, actionable insights to guide targeted reproductive health policy in low-resource settings.
Introduction Breast cancer is a major public health problem among Indonesian women. Chronological age is widely used in risk assessment but may not capture biological variation from reproductive, familial, lifestyle, anthropometric, and dietary factors. This study compared deficit-index-derived biological age with chronological age for discriminating breast cancer status, and assessed its relationship to a conventional risk score. Materials and methods A hospital-based case-control study was conducted in West Sumatra, Indonesia (June–August 2025), including 682 histopathologically confirmed cases and 682 age-matched controls. A deficit index was built from candidate risk factors, the regression of log-transformed deficit index on chronological age was estimated using controls only, to avoid outcome contamination. Discrimination was assessed by AUC (DeLong test for comparisons), adjusted associations used multivariable logistic regression, with bootstrap internal validation. Results Based on 12 deficits, mean biological age was 64.14 ± 14.34 years versus chronological age 49.57 ± 10.12 years (gap 14.57). Biological age discriminated cases from controls better than chronological age (AUC 0.776 vs 0.509). Combining both gave only a marginal gain (AUC 0.783; P = 0.011). A conventional risk score from the same variables discriminated significantly better than biological age (AUC 0.813; P < 0.001), consistent with the index aggregating established risk factors. Conclusion Deficit-index-derived biological age discriminated breast cancer status better than chronological age but did not exceed a conventional risk score from the same variables. It is best interpreted as an integrative summary of established risk, with value for risk communication pending prospective and external validation.
Background: Tuberculosis (TB) remains a global public health challenge affecting all age groups. Children and adolescents are vulnerable group that contribute to the overall TB burden. This study examined the epidemiological characteristics and factors associated with mortality among pediatric patients received DS-TB treatment. Methods: This retrospective study analyzed data from 2300 pediatric TB patients registered for treatment in the Nikshay portal at the Department of TB and Respiratory Diseases, Sir Sunder Lal Hospital, Banaras Hindu University, Varanasi, India, between January 2017 and December 2023. Survival patterns were evaluated using the Kaplan–Meier estimates, Cox regression, and accelerated failure time (AFT) models with Weibull, log–logistic, and log-normal distributions. Model performance was assessed using Cox–Snell residual plots, log-likelihood, AIC, and BIC values. Results: Out of the 2300 pediatric TB patients included, 92 (4.0%) died during treatment. The log-normal AFT model provided the best fit and showed that Age (β=0.228, p<0.001) and Site of disease (pulmonary TB) (β=1.375, p<0.001) were associated with longer survival time. Whereas, microbiologically confirmed TB patients (β=−2.255, p<0.001) were associated with accelerated time to death. Kaplan–Meier analysis also showed higher mortality in children aged 1–6 years, patients with microbiologically confirmed TB, those with extrapulmonary disease, and patients whose follow-up was not done. Conclusions: Paediatric patients with this predictors represent high-risk groups requiring closer monitoring and targeted clinical management. Early identification and timely treatment of severe TB cases may improve survival outcomes and support progress toward TB elimination goals.
Background The rising prevalence of hypertension and diabetes presents significant challenges for management, especially in rural areas where healthcare resources are limited. This systematic review examined the potential of digital health technologies in managing these conditions in rural populations. Methods Search was conducted across PubMed, Scopus, EMBASE, and Google Scholar for studies published from the earliest records indexed in each database up to September 2025 that assessed the effectiveness of digital health interventions in managing diabetes and/or hypertension in rural areas. Quality assessment was performed using Revised Cochrane Risk-of-Bias Tool (RoB 2.0) and Risk of Bias in Non-Randomized Studies of Interventions (ROBINS-I) tool. Results A total of 11 studies were included, with 3 focusing on hypertension, 7 on diabetes, and 1 on both conditions. Digital interventions for hypertension were associated with reductions in blood pressure and improvements in quality of life, treatment adherence, and self-efficacy. Similarly, interventions for diabetes were associated with reductions in blood glucose levels or HbA1c and improvements in LDL levels and self‐management skills. However, no significant changes were observed in HDL levels, triglycerides, albumin-to-creatinine ratio, body mass index, or waist circumference. Conclusion Digital health interventions show potential to support the management of hypertension and diabetes in rural populations by improving clinical and behavioral outcomes. However, factors such as low digital literacy, language barriers, poor internet connectivity, and limited access to digital devices may affect implementation. Future research should evaluate interventions that are appropriate for rural settings to improve the management of these chronic conditions.
PROBLEM CONSIDERED Cancer-related cognitive impairment or CRCI is an adverse effect of chemotherapy that reduces executive function, processing speed and memory of cancer patients. Various factors that are associated with CRCI in turn leads to a significant reduction in the quality of life of cancer patients. This study aims to identify the proportion of cancer patients having CRCI and reduced QoL, and assess the respective factors related to the decline of the same. METHODS A cross-sectional study was conducted in a period of 6 months among 138 cancer patients undergoing chemotherapy. Necessary data pertaining to sociodemographic and clinical aspects were collected, and questionnaires Functional Assessment of Cancer Therapy – Cognitive Function and European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30 were administered to assess the cognitive function and QoL, respectively. The data was entered and analysed using SPSS Software version 29. Pearson correlation and Linear regression were used to evaluate the data. RESULTS A weak positive correlation (Pearson’s r = 0.236, p = 0.005) was observed between cognitive function anf quality of life. On univariate analysis, age, comorbidities, alcohol and tobacco use, pre-medications and chemotherapy agents were associated with cognitive function and QoL. After adjustment for covariates, age, comorbidities, and tobacco/areca nut use remained significantly associated with cognitive impairment, whereas none of the assessed factors remained significantly associated with QoL. CONCLUSION Age, comorbidities, and tobacco/areca nut use were independently associated with cognitive impairment among patients receiving chemotherapy, and cognitive function was positively correlated with quality of life; no factors showed an independent association with QoL after adjustment. These findings support individualized screening and management strategies for chemotherapy-related cognitive impairment (CRCI).
Problem considered Tobacco pouch keratosis (TPK), an oral potentially malignant disorder (OPMD), remains poorly studied in India, with fragmented and geographically heterogeneous evidence. We estimated the community-based pooled prevalence (CBPP) of TPK and examined variation by region and population characteristics. Methods This review was registered in PROSPERO and conducted in accordance with the PRISMA guidelines. PubMed, EMBASE, and Scopus were searched for community-based studies published since 2000, supplemented by citation tracking. A random-effects meta-analysis (logit transformation) was performed to estimate the CBPP of TPK with 95% confidence intervals (CIs). Prediction intervals, subgroup analyses, publication bias, and sensitivity analysis were assessed. Results Thirteen studies comprising 16,169 participants and 668 TPK cases were included. The CBPP of TPK cross the included studies was 3.0% (95% CI: 1.0-8.0), with substantial heterogeneity (I2 = 99.2%). Prevalence was higher in the Eastern region (23.0%; 95% CI: 17.0%–29.0%) followed by the Northern region (9.0%; 95% CI: 6.0%–13.0%). Tribal population had a higher prevalence (9.0%; 95% CI: 2.0%–28.0%) than occupational groups (5.0%; 95% CI: 1.0%–28.0%). Leave-one-out analysis supported the stability of the pooled estimate, while Egger’s test suggested small-study effects (p = 0.0009). Conclusion The pooled estimate reflects heterogeneous community-based evidence rather than a precise national prevalence, with variation across geographical regions and population groups. Targeted tobacco-control and prevention strategies, alongside longitudinal community-based studies, are needed to better characterize TPK prevalence and its malignant transformation potential.
Problem considered:This study aims at assessment of comorbidities, which may be associated with asthma severity. Methods This cross-sectional, prospective observational, hospital-based study was conducted from November 2022–July 2024 at a medical college hospital from south India. Study included consecutive inpatient children aged 5–18 years diagnosed with asthma. Data on demographics, asthma severity, and details of comorbidities were collected. Univariate analysis was conducted to evaluate the association between individual risk factors and the severity of asthma. Results One hundred thirty-nine consecutive children with asthma were enrolled. Among them, 90 (64.7%) had moderate persistent asthma, 31 (22.3%) mild persistent, 10 (7.2%) severe persistent, and 8 (5.8%) intermittent asthma. The most common comorbidities were allergic rhinitis (88.5%), adenoid hypertrophy (66.9%), and sinusitis (66.2%). The presence of three or more comorbidities was associated with increased asthma severity (p = 0.043). Additionally, concomitant sinusitis was associated with moderate-to-severe persistent asthma in the univariate analysis (p = 0.020). Conclusion In hospitalized children with asthma, conditions like allergic rhinitis, adenoid hypertrophy, and sinusitis are very common. The presence of sinusitis was linked to more severe forms of asthma. Although this study did not examine how these comorbidities affect asthma control, having multiple comorbidities was linked to a higher likelihood of severe persistent asthma. Our results emphasise the need for routine screening of comorbidities in hospitalized children with asthma and support future prospective research to determine if targeted treatments can enhance asthma severity and control.
Introduction Burnout among nurses is a critical occupational health issue associated with reduced quality of care, increased turnover, and adverse patient outcomes. Valid and culturally appropriate instruments are essential for surveillance and intervention planning. The Oldenburg Burnout Inventory assesses exhaustion and disengagement but has limited psychometric validation in Indian nursing populations. Methods A cross-sectional psychometric validation study was conducted among 247 staff nurses. Internal consistency was assessed using Cronbach's alpha and McDonald's omega. Convergent validity was evaluated through correlations with DASS-21 depression, anxiety, and stress scores. Exploratory factor analysis using principal axis factoring with oblimin rotation was performed. Sampling adequacy was evaluated using the Kaiser–Meyer–Olkin test and Bartlett’s test of sphericity. Parallel analysis and scree plot inspection determined factor retention. Confirmatory factor analysis tested the theoretical two-factor model. Results The OLBI demonstrated good internal consistency (α = 0.835; ω = 0.832). Mean disengagement score was 2.31 (SD = 0.36), mean exhaustion score was 2.50 (SD = 0.43), and mean total burnout score was 2.41 (SD = 0.36). KMO was 0.838 and Bartlett’s test was significant (χ2(120) = 1615.97, p < .001). Parallel analysis supported retention of two factors explaining 37.1% of total variance. However, items clustered primarily according to wording direction rather than theoretical dimensions. The predefined two-factor CFA model demonstrated poor fit. Discussion & Conclusion The OLBI shows good reliability and acceptable exploratory structural validity among staff nurses; however, the theoretical two-factor structure was not confirmed. Wording and contextual influences may affect dimensional stability in this population.
Aims To determine the prevalence of invasive pulmonary aspergillosis (IPA) among patients with chronic obstructive pulmonary disease (COPD) using the BULPA diagnostic criteria at a tertiary care hospital in Delhi. Methods A cross-sectional study was conducted among COPD patients attending or admitted to the hospital between April 2024 and August 2025. A total of 100 eligible patients were enrolled after informed consent. Clinical, radiological, and laboratory data were collected. Sputum samples were examined by KOH mount, fungal culture, and Aspergillus genus PCR. Serum samples were tested for galactomannan antigen and Aspergillus fumigatus-specific IgG and IgM antibodies. Patients were classified as proven, probable, possible IPA, or colonization according to the BULPA criteria. Statistical analysis was performed using SPSS version 20.0, with p < 0.05 considered significant. Results Of the 100 COPD patients, 54% fulfilled the BULPA criteria for suspected IPA (probable or possible IPA). IPA was significantly more frequent in patients with advanced COPD (GOLD stages III and IV), showing a strong association between disease severity and IPA (p < 0.001). Although corticosteroid exposure was more common among IPA patients, the association was not statistically significant (p = 0.158). Radiological findings, particularly nodules and cavitary lesions, were predominantly observed in probable IPA cases. Conclusion A substantial burden of IPA was identified among COPD patients, especially those with advanced disease. Early application of the BULPA criteria may facilitate timely diagnosis and prompt antifungal management in high-risk patients.
Background/Objectives Perimenopause is characterized by cardiometabolic and hormonal changes that increase cardiovascular disease (CVD) risk. Mobile health (mHealth) technologies may support cardiovascular health. This scoping review mapped the evidence on mHealth interventions for CVD prevention among perimenopausal women. Methods Interventional and observational studies were included. MEDLINE Complete, ScienceDirect, Scopus, and Web of Science were searched for studies published between January 2014 and January 2026, following JBI methodology. Results Twelve studies were included, mostly from high-income countries. Only one intervention was tailored to menopause. Mobile app–based interventions were most common; outcomes were mainly assessed by individual risk factors rather than composite CVD risk scores. Conclusions Menopause-tailored pragmatic studies in LMICs, including Arab countries, are needed.
Problem considered Predicting child mortality in Tanzania is challenged by the class imbalance in the Demographic and Health Survey (DHS) dataset, which compromises the predictive performance and interpretability of machine learning models intended to support policy decision-making. In this study, the imbalanced DHS dataset was balanced using an appropriate machine learning-based resampling technique to improve child mortality prediction. Methods Seven machine learning models, such as Random Forest, Support Vector Machine, Decision Tree, Linear Discriminant Analysis, Logistic Regression, LightGBM, and CatBoost, were evaluated for predictive performance before and after applying SMOTE for data balancing. Results After applying SMOTE, the Random Forest model outperformed other machine learning models for predicting child mortality in Tanzania, achieving an area under the curve (AUC) of 94%, an accuracy of 90%, a recall of 0.86 and an F1-Score of 0.86. Model interpretability was achieved by using the SHAP technique, which confirmed that the top predictors included total children ever born, mother's age group, Zone, wealth index, current marital status, and mother's level of education. The SHAP analysis shows that the most influential feature among the identified predictors was total children ever born. This variable should be used as a meaningful predictor in the machine learning framework, but not as a cause of child mortality, due to its potential endogeneity. Conclusion The SMOTE-balanced Random Forest model with SHAP provides a superior explainable framework for child mortality prediction, offering policymakers and stakeholders in Tanzania evidence-based insights to target key important predictors and strengthen healthcare interventions.
Problem considered This study sought to identify predictors of spontaneous abortion among women admitted with this condition at Butajira General Hospital in Ethiopia, addressing a gap in research on this global health issue, which contributes to maternal mortality and morbidity. Methods An unmatched hospital-based case-control study was conducted among 125 women who experienced spontaneous abortions as cases and 250 pregnant women attending antenatal care with no spontaneous abortion as controls. Cases were selected consecutively, while controls were randomly chosen, maintaining a 1:2 ratio of cases to controls. Data were entered into Epi Info 7 and exported to SPSS 26 for analysis. Multivariable logistic regression was used to identify the determinants of spontaneous abortion, with P < 0.05. Results The study achieved a 97.3% response rate with 365 women participating. Maternal age ≥35 years (AOR = 4.97, 95% CI 1.93-12.79) and ages 30-34 years (AOR 2.94, 95% CI 1.36-6.35), rural residence (AOR = 3.97, 95% CI 2.07-7.60), parity ≤1 (AOR = 1.75, 95% CI 1.01-3.05), previous cesarean delivery (AOR = 2.71, 95% CI 1.14-6.41), short birth interval (AOR = 3.28, 95% CI 1.84-5.83), and husband's alcohol use (AOR = 2.27, 95% CI 1.13-4.56) were independent determinants. Conclusion Promoting birth spacing education, offering specialized antenatal care for older mothers and those with prior caesarean births, and educating male partners on preconception health and alcohol risks may reduce spontaneous abortion.
Problem considered A Patient Information Leaflet (PIL) is an educational tool designed to inform patients about their disease management to improve health outcomes. This study aimed to develop, validate, and conduct user testing of MDR-TB PIL. Methods Study was conducted over four months at the District Tuberculosis Centre, BIMS, Belagavi, Karnataka, with prior ethical approval (Ref. No. KAHER/EC/24-25/D-743). The PIL was developed in English, translated into Kannada, Hindi, and Marathi, and validated by subject experts. It was then evaluated for readability, design, layout, and effectiveness. User testing was carried out using pre and post-test assessments. Results The PILs were first developed in English and translated into regional languages with support from language experts. Readability assessments showed a Flesch Reading Ease score of 55.49 and a Flesch-Kincaid Grade Level of 7.78, indicating suitability for patient use. The Gunning Fog Index (10.16) and SMOG Index (10.74) reflected a high school reading level, while the BALD score (40/48) demonstrated good design and layout. User testing showed a significant improvement in patient knowledge (p = 0.001), with pre-to post-test scores increasing across all versions: English (33.7 ± 2.05 to 42 ± 2.07), Hindi (32.9 ± 2.72 to 41 ± 3.12), Kannada (33.1 ± 1.96 to 41.3 ± 3.12), and Marathi (32.2 ± 2.39 to 41 ± 2.90). Conclusion The study concludes that the developed Patient Information Leaflets (PILs) were effective educational tools for MDR-TB patients. They met acceptable readability standards, demonstrated appropriate design quality, and significantly improved patient knowledge across all tested languages.
