[Prediction of 6-year incidence risk of chronic kidney disease in the elderly aged 65 years and older in 8 longevity areas in China].

Objective: To establish a prediction model for 6-year incidence risk of chronic kidney disease (CKD) in the elderly aged 65 years and older in China. Methods: In this prospective cohort study, we used the data of 3 742 participants collected during 2008/2009-2014 and during 2012-2017/2018 from Healthy Aging and Biomarkers Cohort Study, a sub-cohort of the Chinese Longitudinal Healthy Longevity Survey. Two follow up surveys for renal function were successfully conducted for 1 055 participants without CKD in baseline survey. Lasso method was used for the selection of risk factors. The risk prediction model of CKD was established by using Cox proportional hazards regression models and visualized through nomogram tool. Bootstrap method (1 000 resample) was used for internal validation, and the performance of the model was assessed by C-index and calibration curve. Results: The mean age of participants was (80.8±11.4) years. In 4 797 person years of follow up, CKD was found in 262 participants (24.8%). Age, BMI, sex, education level, marital status, having retirement pension or insurance, hypertension prevalence, blood uric acid, blood urea nitrogen and total cholesterol levels and estimated glomerular filtration rate in baseline survey were used in the model to predict the 6-year incidence risk of CKD in the elderly. The corrected C-index was 0.766, the calibration curve showed good consistence between predicted probability and observed probability in high risk group, but relatively poor consistence in low risk group. Conclusion: The incidence risk prediction model of CKD established in this study has a good performance, and the nomogram can be used as visualization tool to predict the 6-year risk of CKD in the elderly aged 65 years and older in China.目的： 建立适用于中国≥65岁老年人慢性肾脏病（CKD）的发生风险预测工具。 方法： 基于前瞻性队列研究，收集来自"中国老年健康影响因素跟踪调查"子队列-"老年健康生物标志物队列研究"2008/2009年至2014年和2012年至2017/2018年3 742名研究对象的数据，对基线未患CKD的1 055人追踪2次肾功能结局，使用Lasso方法筛选预测因素，采用Cox比例风险回归模型构建CKD发生风险预测模型并通过列线图工具实现模型可视化。采取bootstrap 1 000次重复抽样的方法进行内部验证，并使用C指数和校准曲线评估模型的性能。 结果： 研究对象年龄为（80.8±11.4）岁。经过4 797人年随访，262例（24.8%）研究对象发生CKD。年龄、BMI、性别、文化程度、婚姻状况、退休金或保险、高血压、血尿酸、血尿素氮、TC和基线肾小球滤过率因素纳入6年CKD发生风险预测模型，模型校正C指数为0.766。校准曲线显示预测CKD发生概率和实际发生概率在高风险组中一致性较高，但在低风险组人群中一致性相对较差。 结论： 本研究建立的CKD发生风险预测模型性能较好，列线图可作为可视化工具用于中国≥65岁老年人的6年CKD发生风险预测。.