Machine learning analysis of physical factors associated with low back pain in male high school soft tennis players: Emphasis on nondominant hip internal rotation.

OBJECTIVE: Low back pain (LBP) is common among adolescent racket sport athletes; however, simultaneous assessment of multiple sport-specific physical factors remains limited. This study aimed to identify physical factors associated with LBP in male high school soft tennis players with national-level competitive experience using a machine learning approach. DESIGN: Cross-sectional study. SETTING: Preseason intensive training camps conducted at a single standardized training center in Japan, with participants recruited from elite teams across multiple regions. PARTICIPANTS: One-hundred and sixty male high school soft tennis players. MAIN OUTCOMES: LBP history was assessed using a self-administered questionnaire. Candidate variables included shoulder and hip rotational range-of-motion (ROM) indices and training-related factors. Feature selection was performed within a nested cross-validation framework. Six machine learning algorithms were evaluated using stratified five-fold nested cross-validation. Model performance was assessed using accuracy and area under the receiver operating characteristic curve (AUC), and feature importance was quantified using permutation importance. RESULTS: Logistic regression showed the most stable performance (mean AUC = 0.81). Across models, nondominant hip internal rotation ROM was consistently the most important feature. CONCLUSIONS: Restricted nondominant hip internal rotation ROM was consistently identified as the factor most strongly associated with LBP across models. This may represent a key physical characteristic associated with LBP in male high school soft tennis players.