← Back to Research Papers

Explainable ensemble machine learning for predicting diabetes mellitus and identifying key risk factors: a population-based study in northern Bangladesh.

Authors: Resma MNJ, Kayum MA, Bhowmik P, Khondaker MEA, Kibria MK
Journal: Scientific reports
mental health psychology open access

Abstract

Despite growing evidence of the benefits of physical activity for children’s physical, mental, and social health, global participation rates remain low. Recent harmonized surveillance data confirm that these low levels have remained stable and undesirable over the past decade across most regions. In response, the World Health Organization and UNESCO have called for scalable, context-specific strategies to improve physical activity levels in youth worldwide, emphasizing inclusive, developmentally appropriate, and engaging programs delivered through schools. Schools offer a uniquely powerful setting for such interventions, as they provide structured and equitable access to all children regardless of background, gender, or fitness level. However, physical education classes often vary widely in content and intensity and are insufficient to meet recommended daily activity levels. Pupils with lower fitness tend to be less active and less engaged in typical PE, highlighting the importance of structured programs designed to support participation and perceived competence across fitness levels. Recent meta-analyses have identified several features of school-based physical activity interventions that are most likely to improve components of health-related fitness, including cardiorespiratory fitness, as well as broader health outcomes in youth. Structured and standardized delivery, a frequency and duration typically of two to three sessions per week over at least eight to twelve weeks, moderate-to-vigorous intensity verified through objective measures, and developmentally appropriate design are consistently reported as effective. Programs are also more effective when implemented school-wide, supported by trained staff, and designed to target specific outcomes such as cardiovascular fitness, strength, or motor competence rather than relying on general activity exposure. High-intensity interval formats appear particularly beneficial for cardiorespiratory fitness and cardiometabolic health, while program effectiveness also depends on fidelity and delivery quality. Comprehensive reviews confirm effectiveness when interventions are delivered with structure and inclusivity, although effects may be smaller or absent in adolescents and may differ by gender. Importantly, well-designed interventions should engage not only the already active but also those who are least fit or least confident, who often benefit the most when included but are frequently underserved.