透析并发不宁腿综合征风险列线图预测模型的构建及验证 #br#

目的构建透析并发不宁腿综合征（RLS）风险列线图预测模型并进行外部验证。 方法收集透析患者524例为建模组，构建透析并发RLS风险列线图预测模型。另收集透析患者101例作为验证组对模型进行外部验证。2组均根据是否并发RLS分为并发组和未并发组。 结果建模组透析患者RLS并发率为11.64%。与未并发组相比，并发组糖尿病史、脑梗死史比例，透析龄、甲状旁腺激素（PTH）水平较高（P＜0.05），血红蛋白（Hb）、血清铁较低（P＜0.05）。建模组多因素Logistic回归分析显示糖尿病史、脑梗死史、透析龄≥48个月、PTH≥500 ng/L及Hb＜100 g/L是透析并发RLS的独立危险因素。据此构建的透析并发RLS风险的列线图预测模型采用Bootstrap内部验证，H-L检验示χ2=7.541（P=0.563），Calibration 校准曲线拟合良好，ROC 曲线下面积（AUC）及 C-index 指数均为 0.842（95%CI：0.787～0.896）。外部验证结果显示Calibration校准曲线的预测概率与实际概率接近，有良好的一致性，AUC为0.854（95%CI：0.738～0.969）。 结论构建的透析并发RLS的风险列线图预测模型具有较好的校准度和区分度。

Construction and verification of a nomogram prediction model for the risk of restless legs syndrome during dialysis

Objective To construct a nomogram prediction model of the risk of dialysis complicated with restless legs syndrome (RLS) and conduct external verification. Methods A total of 524 dialysis patients were collected as the modeling group, and a nomogram prediction model of the risk of RLS complicated by dialysis was constructed. In addition, 101 dialysis patients were collected as the verification group for external verification of the model. Both groups were divided into concurrent group and non-concurrent group according to whether concurrent RLS. Results The complication rate of RLS in dialysis patients was 11.64% in the modeling group. Compared with the uncomplicated group, the proportion of diabetes history, cerebral infarction history, dialysis age and parathyroid hormone (PTH) levels were higher in the complicated group (P＜0.05), and hemoglobin (Hb) and serum iron were lower (P＜0.05). Multivariate Logistic regression analysis showed that diabetes history, cerebral infarction history, dialysis age≥48 months, PTH ≥500 ng/L and Hb＜100 g/L were independent risk factors for dialysis complicated by RLS in the modeling group. Based on this, the risk nomogram prediction model of dialysis concurrent RLS was verified by Bootstrap internal verification. The H-L test showed that χ2=7.541 (P=0.563). The calibration curve fited well. The area under the ROC curve (AUC) and the C-index both were 0.842 (95%CI: 0.787 - 0.896). The external verification results showed that the predicted probability of the calibration curve was in good agreement with the actual probability, and the AUC was 0.854 (95%CI: 0.738 - 0.969). Conclusion The constructed risk nomogram prediction model of dialysis concurrent RLS has good calibration and discrimination.