Skip to main content
. 2025 Jul 4;13:1588041. doi: 10.3389/fpubh.2025.1588041

Figure 1.

Flowchart depicting the study process using NHANES data from 2011-2018. Initial participants numbered 39,156, with exclusions leading to a final group of 1,743. Exclusions included age under 18 (15,331), missing skeletal muscle mass and height (12,029), and missing heavy metal exposure indicators (10,053). The population was split into a training set (70%, 1,220) and a validation set (30%, 523). Variable screening involved Boruta and Lasso methods. Model validation included GBDT, Decision Tree, AdaBoost, NB, CatBoost, and SVM, leading to an optimal model. SHAP analysis was used for model explanation.

Flowchart of the study design.