Predictive modelling on handgrip strength in adolescents

Introduction: Handgrip strength represents a key measure of upper limb function and general physical fitness in adolescents, reflecting muscular health and potential risk for future health conditions. Objetive: This study aimed at developing a predictive model of adolescent HGS using anthropometric variables to identify those at risk of future health issues. Methods: Sociodemographic data, Body Max Index (BMI), and body fat percentage were collected during Physical Education classes. HGS was measured twice using the dominant arm, with the best result recorded. Additional data from the Course Navette test, the Physical Activity and Leisure Motivation Scale (PALMS), the International Physical Activity Questionnaire (IPAQ), and the Questionnaire of Health and Well-Being (QHWB) were also collected. A nested 10-fold cross-validation pipeline was used, incorporating Boruta feature selection, and ElasticNet regression. Model evaluation included Mean Average Error, Root Mean Square Error, R², and bootstrapped confidence intervals. The sample included 867 secondary school students (mean age = 14.03 ± 1.19 years; 53.9% boys). Results: The model showed good predictive performance: MAE = 3.76 (0.29), RMSE = 4.73 (0.36), R² = 0.48 (0.08), and Average Normalized MAE = 9.90%. Selected predictors included Age (b = 1.86), Sex (b = -1.03), BMI (b = 4.16), Body Fat (b = -3.94), and Navette Stages (b = 1.06). Conclusions: This study provides preliminary evidence for a predictive model that estimates HGS in adolescents using indirectly measurable predictors, employing a rigorous machine learning approach that retains only robust predictors for population screening.