Scholarlink

<p>Pediatric malnutrition remains a major global health challenge. Conventional WHO-based binary classification (Severe Acute Malnutrition (SAM) / Moderate Acute Malnutrition (MAM) relies solely on anthropometric thresholds. It fails to capture the levels of severity and the interaction-driven nature of malnutrition severity. The study aims to develop and validate a six-category malnutrition stratification model, SAM-Trace, that integrates anthropometric indicators with statistically significant interaction effects among maternal, socioeconomic, demographic, and other contextual factors, using fuzzy inference rules. A cross-sectional, real-time dataset of 588 children (1–23 months) was considered for the study. Initially, the model was developed and trained with 90% of the collected dataset and tested with the remaining 10%. The data is analyzed using the proposed SAM-Trace model in a multi-stage process with a Bayesian Network (BN) classifier for six-level anthropometric stratification, significant interaction exploration with False Discovery Rate (FDR), and integration of a fuzzy augmented layer to explore the interaction-driven reclassification of cases within the subcategories. The six-category system achieved 94.55% mean training accuracy (Precision: 94.61%, Recall: 94.13%, F1-score: 94.17). The interaction exploration layer, with specified criteria, revealed five interactions that had a p-value &lt; 0.05, a moderate to strong effect size, and an FDR &lt; 0.10; these interactions served as rules in the subsequent layer of the model. The fuzzy augmented layer in the model built using valid and significant five interactions achieved an accuracy of 83.46%, which represents a decline in accuracy and reclassification of the cases. Further, the SAM-Trace model was tested with the remaining 10% data which is completely remained unseen during all stages of model development, achieved 91.67% accuracy on BN and 81.00% accuracy on fuzzy augmented BN. Various validation techniques were performed, and the results were recorded. These results achieved consistent performance across the training and testing phases and also suggest that contextual factors such as Maternal nutritional indicators, socioeconomic factors, and feeding practices would have an influence on child nutritional status and can push a child from borderline to severe stages. The SAM-Trace model provides a systematic way of integrating fuzzy rules in the model, which acts as a risk-sensitization layer and explores the risk escalation in the reclassification. The proposed model provides evidence-based results that help in policy making, early identification of risk-oriented cases, and optimized resource allocation. The model contributed to the attainment of United Nations Sustainable Development Goals (SDG’s), Zero Hunger and Good health &amp; Well-being, which is SDG2 &amp; 3, respectively.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

SAM-trace: a fuzzy logic approach in identifying interaction effects in the exacerbation of childhood malnutrition severity

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月22日

查看全文
  • S Shruthi,
  • Priya Govindarajan,
  • S. R. Shalini,
  • Pavan John Antony

摘要

Pediatric malnutrition remains a major global health challenge. Conventional WHO-based binary classification (Severe Acute Malnutrition (SAM) / Moderate Acute Malnutrition (MAM) relies solely on anthropometric thresholds. It fails to capture the levels of severity and the interaction-driven nature of malnutrition severity. The study aims to develop and validate a six-category malnutrition stratification model, SAM-Trace, that integrates anthropometric indicators with statistically significant interaction effects among maternal, socioeconomic, demographic, and other contextual factors, using fuzzy inference rules. A cross-sectional, real-time dataset of 588 children (1–23 months) was considered for the study. Initially, the model was developed and trained with 90% of the collected dataset and tested with the remaining 10%. The data is analyzed using the proposed SAM-Trace model in a multi-stage process with a Bayesian Network (BN) classifier for six-level anthropometric stratification, significant interaction exploration with False Discovery Rate (FDR), and integration of a fuzzy augmented layer to explore the interaction-driven reclassification of cases within the subcategories. The six-category system achieved 94.55% mean training accuracy (Precision: 94.61%, Recall: 94.13%, F1-score: 94.17). The interaction exploration layer, with specified criteria, revealed five interactions that had a p-value < 0.05, a moderate to strong effect size, and an FDR < 0.10; these interactions served as rules in the subsequent layer of the model. The fuzzy augmented layer in the model built using valid and significant five interactions achieved an accuracy of 83.46%, which represents a decline in accuracy and reclassification of the cases. Further, the SAM-Trace model was tested with the remaining 10% data which is completely remained unseen during all stages of model development, achieved 91.67% accuracy on BN and 81.00% accuracy on fuzzy augmented BN. Various validation techniques were performed, and the results were recorded. These results achieved consistent performance across the training and testing phases and also suggest that contextual factors such as Maternal nutritional indicators, socioeconomic factors, and feeding practices would have an influence on child nutritional status and can push a child from borderline to severe stages. The SAM-Trace model provides a systematic way of integrating fuzzy rules in the model, which acts as a risk-sensitization layer and explores the risk escalation in the reclassification. The proposed model provides evidence-based results that help in policy making, early identification of risk-oriented cases, and optimized resource allocation. The model contributed to the attainment of United Nations Sustainable Development Goals (SDG’s), Zero Hunger and Good health & Well-being, which is SDG2 & 3, respectively.