Comparison Between Blood Exams and Questionnaire Features for Low Bone Mineral Density Classification in Female Athletes

抄録

Low bone mineral density is a risk factor that increases injuries, bone fractures, and other problems, particularly for female athletes. Thus, it is essential to be able to identify athletes with osteopenia and which factors most impact it. In this study, logistic regression models were trained separately on data derived from blood exams and questionnaires, and their performances were compared. The dataset consisted of 219 Japanese female athletes. The multivariate prediction model based on the questionnaires achieved a ROC-AUC of 0.77±0.06. Contrastingly, the multivariate model based on the blood exam features achieved a ROC-AUC of 0.64±0.03. Our results suggest that features related to training habits and the menstrual cycle could potentially be more relevant to bone health than blood-related ones for elite female athletes. However, a larger sample size is required to make any conclusive affirmations.