Utilizing Life's Crucial 9 for Rheumatoid Arthritis Risk Prediction: A Machine Learning Approach Based on NHANES Data

To investigate the association between the cardiovascular health metric Life's Crucial 9 (LC9) and the risk of rheumatoid arthritis (RA), and to establish a predictive model for RA risk. Data were obtained from the U.S. National Health and Nutrition Examination Survey (NHANES) 2011-2018, including 16,154 adults aged ≥ 20 years. RA status was determined by self-report (RA group: 836 cases; non-RA group: 15,318 cases). Weighted multivariable logistic regression was used to analyze the association between LC9 score (as both a continuous variable and quartiles) and RA, with restricted cubic splines applied to test for nonlinearity. LC9 score together with 10 covariates (age, sex, race, education, marital status, PIR group, AST, ALT, HDL cholesterol, and alcohol consumption) were incorporated into multiple machine learning algorithms (random forest, support vector machine, K-nearest neighbor, XGBoost, LightGBM, and naive Bayes) to construct prediction models. The dataset was randomly divided into training and testing sets in a 7:3 ratio, with five-fold cross-validation for parameter optimization. Each 1-unit increase in LC9 score was associated with a 27% reduction in RA risk (OR = 0.73, 95% CI: 0.68-0.80, p < 0.0001). Compared with the lowest quartile (Q1), the highest quartile (Q4) had a 57% lower risk (OR = 0.43, 95% CI: 0.33-0.55, p < 0.0001). A linear dose-response relationship was observed (P for trend < 0.0001; P for nonlinearity = 0.159). Among machine learning models, XGBoost performed best (test set AUC = 0.987, accuracy = 96.20%, sensitivity = 95.06%), followed by LightGBM (AUC = 0.982). Random forest (AUC = 0.817) and naive Bayes (AUC = 0.735) showed relatively weaker performance. SHAP analysis indicated that AST, age, ALT, and HDL were the top four predictors, while LC9, alcohol consumption, and PIR group had comparable importance, all significantly higher than other variables. LC9 is an independent protective factor against RA risk. An XGBoost-based predictive model integrating LC9 demonstrated excellent diagnostic performance and may serve as a valuable tool for risk stratification in high-risk RA populations.