目的 构建老年糖尿病肾病（diabetic kidney disease，DKD）患者血液透析低血糖列线图风险模型并验证其预测效果，为临床制订个体化的低血糖防控措施提供参考。方法 采用便利抽样法选取2022年6月至2023年6月在南宁市某三级甲等医院A院区血液净化室进行血液透析治疗的老年DKD患者273例建模，同法选取2023年7-12月该院B院区的老年DKD患者65例进行模型验证。以Lasso回归筛选特征变量，采用Logistic回归构建模型；以受试者工作特征曲线、校准曲线和临床决策曲线评价模型预测效能。结果 338例研究对象低血糖发生率为43.5%。体质量指数、文化程度、自我管理、社会支持、DKD病程、透析过程中进食、过去1年低血糖史、透析前血糖是老年DKD患者透析低血糖的影响因素（均P＜0.05）。构建的预测模型的受试者工作特征曲线下面积为0.828，校准曲线显示预测结果与实际有较好一致性（P＞0.05），经验证该模型预测的总体准确率为80.0%，临床决策曲线阈值概率为7%~95%时，模型临床净获益高。结论 基于Lasso-Logistic回归构建的列线图模型在临床中应用有较好的预测价值。

Objective To construct and validate a prediction model of hypoglycemia in hemodialysis for elderly patients with diabetic kidney disease (DKD),and to provide reference for the clinical formulation of individualized hypoglycemia prevention and control measures.Methods From June 2022 to June 2023,273 elderly DKD patients who underwent hemodialysis in the blood purification room of A Branch of a tertiary A hospital in Nanning City were selected for modeling by the convenience sampling method,and 65 elderly DKD patients in B Branch from July to December 2023 were selected for model validation by the convenience sampling method.Lasso regression was used to select characteristic variables,and Logistic regression was used to construct the model.The receiver operating characteristic curve,calibration curve and clinical decision curve were used to evaluate the prediction performance of the model.Results The incidence of hypoglycemia in 338 subjects was 43.5%.Body mass index,educational level,self-management,social support,duration of DKD,eating during dialysis,history of hypoglycemia in the past year,and blood glucose before dialysis were the influencing factors for dialysis hypoglycemia in elderly DKD patients (all P＜0.05).The area under the receiver operating characteristic curve of the constructed prediction model was 0.828,and the calibration curve showed that the prediction results were in good agreement with the reality (P＞0.05).It was verified that the overall accuracy of the prediction model was 80.0%,the threshold probability of the clinical decision curve was 7%-95%,and the model had a high clinical net benefit.Conclusion The nomogram model based on Lasso-logistic regression has good predictive value in clinical practice.

R473.58

广西中医药重点学科建设项目（GZXK-Z-20-56）；广西研究生教育创新计划项目（YCSY2022043）