目的 构建中青年糖尿病患者视网膜病变(diabetic retinopathy，DR) 风险模型并验证其预测能力，为相关人员风险筛查提供依据。方法 2022年9月至2023年2月，采用便利抽样法选取就诊于天津市某2所医院的糖尿病患者为研究对象，采用随机数字表法按7∶3将患者分为训练组(n=327)和验证组(n=140)。通过单因素分析和多因素Logistic回归分析，确定中青年糖尿病患者视网膜病变的独立危险因素，并构建列线图风险模型。采用受试者工作特征曲线（receiver operator characteristic，ROC）、校正图、Hosmer-Lemeshow检验和决策曲线分析(decision curve analyses，DCA)等不同方法评估模型的性能。结果 单因素分析及多因素Logistic回归分析确定，文化程度、医疗支付方式、体质量指数、病程、糖尿病治疗、药物依从性、身体活动水平、饮食依从性是中青年糖尿病患者发生DR的独立危险因素(P＜0.05)。训练组和验证组的ROC曲线下面积分别为0.911(95%CI 0.880~0.942)和0.905(95%CI 0.854~0.956)。两个队列的校准曲线在预测结果和实际结果之间表现出高度的一致性。Hosmer-Lemeshow检验显示训练组和验证组P值分别为0.161和0.415。结论 构建的风险模型可预测中青年糖尿病患者视网膜病变风险，其在筛查中青年糖尿病视网膜病变高危人群中具有应用前景。

Objective To construct a risk model of diabetic retinopathy (DR) in young and middle-aged patients and verify its predictive ability，and to provide evidence for risk screening.Methods From September 2022 to February 2023，diabetic patients who were treated in 2 hospitals in Tianjin were selected by the convenience sampling method and were divided into the training group (n=327) and the validation group (n=140) according to the random number table method according to 7∶3.Univariate analysis and multivariate logistic regression analysis were used to determine the independent risk factors for retinopathy in young and middle-aged diabetic patients，and a nomogram risk model was constructed.Receiver operator characteristic (ROC) curve，correction plot，Hosmer-Lemeshow test and decision curve analyses （DCA） were used to evaluate the performance of the model.Results Univariate analysis and multivariate logistic regression analysis showed that education level，medical payment method，body mass index，course of disease，diabetes treatment，medication compliance，physical activity level and diet compliance were independent risk factors for DR in young and middle-aged diabetic patients(P＜0.05).The area under the ROC curve of the training group and the validation group were 0.911(95%CI: 0.880-0.942) and 0.905 (95%CI 0.854-0.956)，respectively.The calibration curves for both cohorts show a high degree of agreement between the predicted and actual results.The Hosmer-Lemeshow test shows a P value of 0.161 for training and 0.415 for validation.Conclusions The constructed risk model can predict the risk of retinopathy in young and middle-aged diabetic patients，and it has application prospects in screening young and middle-aged people at high risk of diabetic retinopathy.

R473.58

国家自然科学基金项目（82003092；82000949）；天津市重点医学学科（专业）建设项目（TJYXZDXK-016A）