BACKGROUND: Dialysis access steal syndrome (DASS) is a severe complication of vascular access surgery. This study aimed to identify key risk factors and develop an interpretable prediction model for early risk assessment in patients with end-stage renal disease (ESRD). METHODS: This retrospective study analyzed 324 ESRD patients (March 2023-June 2025). Feature selection was performed using LASSO regression combined with SHapley Additive exPlanations (SHAP). Independent risk factors were identified via multivariable logistic regression, with robustness confirmed by sensitivity analysis and E-values. Restricted cubic splines (RCS) explored non-linear associations, while BMI-stratified and interaction analyses evaluated effect modification. Model performance was validated using AUC, 1000-sample bootstrapping, and 10-fold cross-validation. RESULTS: DASS occurred in 86 patients (26.5%). The Fried Frailty Score (FFS) emerged as the most significant contributor. Multivariable analysis identified FFS, BMI, CKD duration, sarcopenia, distal arterial pressure (DAP), and surgical experience as independent predictors. RCS analysis identified critical thresholds: BMI 23.3 kg/m2, DAP 66 mmHg, and CKD duration 8.2 years. Notably, BMI exerted significant effect modification (P interaction < 0.05); FFS, sarcopenia, DAP, and CKD duration were significantly associated with DASS only in the low/normal BMI group. The final model demonstrated excellent discrimination (AUC = 0.934) and stability (Brier score = 0.085; C-index = 0.931). High-risk patients showed a significantly higher DASS incidence than low-risk patients (50.00% vs. 3.09%, p < 0.001). CONCLUSION: FFS, sarcopenia, BMI, DAP, and CKD duration are core predictors of DASS. BMI acts as a key effect modifier, particularly influencing the impact of functional and hemodynamic indicators. This high-performance model provides a scientific basis for personalized preoperative screening and clinical intervention.