Using Interpretable Machine Learning with SHAP to Assess Dynapenic Abdominal Obesity as a Stroke Risk Predictor: A Prospective Cohort Study.

Dynapenic abdominal obesity was significantly associated with increased stroke risk over a 4-year follow-up period.

• Adjusted OR = 1.58 (95% CI: 1.21–2.06) for stroke among participants with DAO compared to those without.
• Over the 4-year follow-up, 210 (1.9%) of 11,207 participants experienced a stroke.
• The study used prospective data from the China Health and Retirement Longitudinal Study (CHARLS), including adults aged ≥ 45 years.
• Stroke events were identified via self-reported physician diagnoses.

DAO was defined by the combination of dynapenia and abdominal obesity using sex-specific thresholds.

• Dynapenia was defined as handgrip strength ≤ 28 kg for men and ≤ 18 kg for women.
• Abdominal obesity was defined as waist circumference ≥ 90 cm for men and ≥ 80 cm for women.
• DAO required the co-presence of both conditions simultaneously.

Subgroup analyses demonstrated consistent associations between DAO and stroke risk across all examined subgroups.

• All interaction p-values were > 0.05, indicating no statistically significant effect modification by subgroup.
• This consistency suggests the DAO-stroke association is robust across different demographic and clinical subpopulations.

XGBoost achieved the highest predictive performance among the machine learning models evaluated.

• XGBoost demonstrated an AUC of 0.92 and an accuracy of 0.84.
• Multiple machine learning models were employed alongside logistic regression to assess the association and evaluate robustness.
• Other machine learning models were tested but XGBoost outperformed them on predictive metrics.

SHAP analysis ranked DAO as the fourth most important predictor of stroke in the XGBoost model.

• The three predictors ranked above DAO in importance were age, BMI, and residence.
• Shapley additive explanations (SHAP) were used to provide interpretability for the machine learning model's predictions.
• SHAP values allowed assessment of each feature's contribution to individual stroke risk predictions.

Stroke carries a particularly high burden in China, providing context for this study's population focus.

• Stroke is described as a major cause of mortality and disability worldwide.
• The study used a nationally representative cohort of middle-aged and older Chinese adults from CHARLS.
• The sample included 11,207 participants aged ≥ 45 years.