Shaheed Ziaur Rahman Medical College

Abstract Selecting an appropriate supportive response for a mental health question post is a key step toward scalable mental health question answering, yet it remains difficult due to semantic mismatch, noisy informal language, and the need to align supportive intent beyond surface lexical overlap. We propose MF-GAT, a novel multi view graph attention matching framework that constructs a Concept Interaction Graph (CIG) to explicitly encode post response concept alignments and their interaction structure. MF-GAT learns three complementary evidence streams, including local interaction features, multi view fusion features, and global context features, and applies view specific graph attention to propagate and reweight informative relational signals over the CIG. A gated fusion module then adaptively integrates the view representations into a unified matching vector for prediction. We evaluate MF-GAT both as a pair classification model and as a retrieval ranking model for selecting the best support from a candidate pool, reporting Accuracy and F1 together with standard ranking metrics including MRR and NDCG. On the MHQA benchmark, MF-GAT achieves 0.95 Accuracy and 0.85 F1, outperforming BERT (0.89, 0.67), CIG-GCN (0.92, 0.79), ARC-II (0.85, 0.60), and MatchPyramid (0.82, 0.59). These results show that novel multi view interaction graph modeling with attention based propagation improves both supportive response classification and practical retrieval quality for mental health support selection.

Sleep problems are frequently found among the children, and it is more among the intellectually disabled children than those who are not intellectually disabled. This study was aimed to assess the sleep problems of intellectually disabled children. It was a cross-sectional descriptive study. The study period was from February 2019 to December 2019. The study was conducted in two organizations and two schools of Dhaka city. Data collection instrument was, ‘Children’s Sleep Habit Questionnaire’ developed by Judith A. Owens, Anthony Spirito and Melissa Mc Guinn. It was translated into Bangla for convenience. It was distributed among the parents of the intellectually disabled and non-intellectually disabled children and later collected from them having the answers. The sample size was in total 148, of which 74 were intellectually disabled and 74 were non-intellectually disabled children. The sampling technique was a purposive one and data was analyzed manually. Among the non-intellectually disabled children sleep problems were found 25%, whereas among the intellectually disabled children it was found more than 95%. It was statistically highly significant (p value was p<0.00001). This study concludes that intellectually disabled children suffer more from sleep problems than the non-intellectually disabled children. So due medical and psycho-social care and proper sleep hygiene should be arranged for them. Bang J Psychiatry 2026;40(1): 15-18

Depression is a major health concern among the elderly, affecting their quality of life. The Geriatric Depression Scale - Short Form (GDS-15) is commonly used to screen for depression in older adults. This study aimed to translate, culturally adapt, and validate the GDS-15 into Bengali for use among elderly people in Bangladesh. The GDS-15 was translated into Bengali, followed by cultural adaptation. A cross-sectional study was conducted with 100 elderly participants aged 60 years, and above in Chittagong Medical College Hospital. Psychometric validation was performed using reliability testing, chi-square analysis, ANOVA, and principal component analysis (PCA). The Bengali version of the GDS-15 demonstrated strong construct validity, with significant correlations between the original and translated versions of the items (r = 0.9788, p < 0.001). However, reliability analysis showed a low Cronbach's alpha (0.4753), indicating suboptimal internal consistency. The PCA revealed multiple components, suggesting the multidimensionality of the scale in this population. The Bengali GDS-15 showed low internal consistency which warrants the need for further cultural refinement and psychometric validation before routine use among the population. However, the Bengali GDS-15 has the potential to serve as a useful screening tool for geriatric depression in Bangladesh with further improvements.

Background and Objectives: An updated 2026-version of the 2026 International Diabetes Federation and Diabetes & Ramadan (IDF-DAR) Alliance risk calculator has replaced the 2021-version. This study aimed to compare pre-Ramadan risk categorisation and complication occurrences during Ramadan using both versions of the calculator among Muslims with diabetes mellitus (DM) Methods: This prospective observational study included 638 adults with DM (mean age: 52 years; 67% female) who intended to fast during Ramadan 2026 and attended public and private endocrine centres in Bangladesh. Pre-Ramadan baseline data were evaluated to categorize participants into low (0–3), moderate (3.5–6), or high (>6) risk groups using both tools. The investigators conducted follow-up evaluations within two weeks post-Ramadan. Results: The 2026 calculator reclassified significantly more participants upward into higher-risk categories than the 2021 version (15% vs. 8% downward; P<0.001). A total of 59 people were lost to follow-up, and three died. Of 576 participants completing follow-up, 51 (8.9%) developed hypoglycemia and 95 (16.5%) developed hyperglycemia. Both calculators demonstrated modest (AUC: 0.6–0.7) and comparable predictive accuracy for hypoglycemia (z-statistic=-0.21, P=0.834) and hyperglycemia (z-statistic=-0.5, P=0.594). High-risk individuals had significantly higher odds of hypoglycemia (OR: 2.7 for 2021, P=0.001 vs 2.4 for 2026, P<0.001) and hyperglycemia (OR: 2.2 for 2021, P=0.014 vs 2.3 for 2026, P<0.001) than non-high-risk individuals (low and moderate risk), using both calculators. Conclusions: The updated 2026 IDF-DAR calculator stratifies more patients into higher-risk categories while maintaining comparable predictive ability for glycemic complications compared with the 2021 version. [J Assoc Clin Endocrinol Diabetol Bangladesh, 2026;5(Suppl 1): S41]

The increasing trend of the adoption of digital technology in the modern workplace has given rise to opportunities as well as challenges for mental health, necessitating the applicability of machine learning for mental health screening. Mental health data obtained from digital technology is high dimensional and heterogeneous in nature and is not effective for predictive models. This problem is tackled by introducing a hybrid model of Autoencoder and Light Gradient Boosting Machine (LightGBN) for mental health screening in a technologydominated workplace. The autoencoder is used for efficient extraction of latent vectors from survey forms, and nonlinear classification can be obtained by LightGBM. This method is tested for accuracy on the Mental Health in Tech Survey dataset by implementing Logistic Regression, Support Vector Machines (SVM), Random Forest (RF), and XGBoost Models as baseline models. Experimental results show that the proposed framework outperforms all baseline methods, achieving 96 percent accuracy, an F1-score of 0.960, and a ROC-AUC of 0.98. The results demonstrate that combining deep representation learning with gradientboosted ensembles enables accurate, reliable, and interpretable mental health screening in digitally mediated work settings.