LIU Jin1, WU Jinchun2, GUO Huijie1, Dawazhuoma3, ZHANG Xiaofei2, BAI Yuting2, WEI Xiaojuan2, LIU Yanmin2, SU Xiaoling2
Objective To analyze the clinical characteristics and risk factors for stroke in non-valvular atrial fibrillation (NVAF) patients residing in plateau region, and to construct a relevant risk prediction model. Methods Clinical data of 1 062 NVAF patients in the Qinghai region were retrospectively selected, and divided into a stroke group (n=102) and a non-stroke group (n=960) based on the presence or absence of stroke. Univariate analysis was performed to screen variables across relevant indicators between the two groups. Multivariate Logistic regression analysis was used to identify independent risk factors for stroke in NVAF patients in the Qinghai region. A nomogram-based prediction model was constructed, and its prediction performance was evaluated using methods such as ROC curve and calibration curve analysis. ResultsStatistically significant differences existed in age, BMI, systolic blood pressure, HAS-BLED and CHA2DS2-VASc scores, right ventricular diameter, INR, BNP, sex, and hypertension history between the stroke and nonstroke groups (all P<0.05). Following initial variable screening by univariate Logistic analysis, multivariate Logistic regression analysis identified 7 independent risk factors: age (OR=1.173, 95%CI 1.090-1.261), BMI (OR=1.304, 95%CI 1.131-1.503), HAS-BLED score (OR=2.695, 95%CI 1.740-4.176), CHA2DS2-VASc score (OR=2.378, 95%CI 1.791-3.157), BNP (OR=1.001, 95%CI 1.001-1.002), sex (OR=3.671, 95%CI 1.820-7.406), and hypertension (OR=2.071, 95%CI 1.079-3.977) for predicting stroke in NVAF patients. The nomogrambased prediction model incorporating these 7 influencing factors demonstrated excellent discrimination (C-index=0.93, 95%CI 0.90-0.95) and outstanding predictive accuracy (AUC=0.96, 95%CI 0.93-0.99). It indicated excellent model fit, with the calibration curve closely approximating the ideal curve. Conclusion The primary independent risk factors for stroke in NVAF patients in the Qinghai region are age, BMI, HAS-BLED score, CHA2DS2-VAS score, BNP, sex, and hypertension. The model constructed using these factors shows high accuracy in predicting stroke risk among this population.