Prediction model for cardiovascular disease risk in hemodialysis patients

Purpose: To derive and validate a prediction score for cardiovascular disease (CVD) risk in hemodialysis patients in China.

Methods: Three hundred and eighty-eight patients with regular hemodialysis for more than 3 months were recruited from January 1, 2015 to September 30, 2019 and followed up till May 31, 2020. We derived a prediction score using all participants as a training data set and validated using a bootstrap validation data set. Discriminatory ability of the prediction score was assessed by the area under the receiver operating characteristic curve (AUC).

Results: Of 388 patients without CVD at baseline, 132 developed first CVD events during an average follow-up of 3.27 (inter-quartile range = 3.08) years. Of 26 clinical parameters, age, hypertension, diabetes and abnormal white blood cell (WBC) count were identified as significant predictors and included in the prediction model. Compared to those without any of these risk factors, those with one, two, and three to four points showed increased risks of CVD, with the adjusted hazards ratio and 95% confidence interval (CI) being 3.29 (1.17-9.26), 7.42 (2.68-20.51) and 15.43 (5.44-43.75), respectively. The score showed satisfactory discriminatory ability in both training and validation data set (AUC = 0.7025, 95% CI 0.6520-0.7530, and 0.6876, 95% CI 0.6553-0.7200, respectively).

Conclusion: We derived and validated a prediction score for CVD risk in hemodialysis patients in China. Given there is a rapid increase in the number of hemodialysis patients, this simple point score can be used to identify high-risk individuals in clinical practice for more precise and efficient personalized treatment.