摘要
目的系统评价2型糖尿病患者发生视网膜病变的风险预测模型。方法检索PubMed、Embase、Cochrane Library、中国知网、万方数据库、维普网、中国生物医学文献数据库中发表的有关2型糖尿病患者发生视网膜病变预测模型的研究,检索时限为建库至2023年3月17日,由2名研究者独立筛选文献并提取数据,应用预测模型研究的偏倚风险评估工具分析纳入文献的偏倚风险和适用性。结果共纳入18篇文献,其中9篇研究采用Logistic回归,4篇研究采用Cox比例风险回归,2篇研究采用Lasso回归,1篇研究采用半参数回归,2篇研究采用机器学习的方法。2篇研究同时开展了内、外部验证,3篇研究采取内部验证,1篇研究采取外部验证。12篇研究提及预测模型的曲线下面积(AUC)为0.715~0.966,且AUC均>0.7;4篇研究提及C指数,为0.770~0.848;2篇研究未提及模型预测性能。年龄、2型糖尿病病程、尿蛋白、糖化血红蛋白水平、是否应用胰岛素、高血压病、入院次数为多变量模型重复报告的独立预测因子。所有研究均存在一定偏倚性,但适用性较好。结论现有2型糖尿病患者发生视网膜病变预测模型有较好的预测性能,但整体偏倚性较高。未来需进一步对各个模型性能进行验证,同时仍需进一步开发适合不同人群(如老年患者、女性、不同种族或地区等)的风险预测模型。
Objective To systematically review the risk prediction model of retinopathy in type 2 diabetes patients.Methods Research on the prediction model of retinopathy in patients with type 2 diabetes was retrieved in PubMed,Embase,Cochrane Library,China National Knowledge Infrastructure,WanFang Data,VIP,and China Biology Medicine.The search period was from database establishment to March 17,2023.Two researchers independently screened the literature and extracted data,and used the prediction model risk of bias assessment tool to analyze the bias risk and applicability of the included literature.Results A total of 18 articles were included,of which nine studies used Logistic regression,four studies used Cox proportional hazards regression,two studies used Lasso regression,one study used semi-parametric regression,and two studies used machine learning methods.Two studies simultaneously conducted internal and external validation,three studies conducted internal validation,and one study conducted external validation.A total of 12 studies mentioned that the area under the curve(AUC)of predictive models ranged from 0.715 to 0.966,and all AUCs were greater than 0.7.Four studies mentioned the C index,which was 0.770 to 0.848,while two studies did not mention the predictive performance of the model.Age,course of type 2 diabetes,urinary protein,glycosylated hemoglobin,insulin use,hypertension,and time of hospital admissions were independent predictors of repeated reporting in the multivariate model.All studies had a certain risk of bias,but their applicability was good.Conclusions The existing prediction models for retinopathy in type 2 diabetes patients have good prediction performance,but the overall risk of bias is high.Further validation of the performance of each model is needed,while further development of risk prediction models suitable for different populations(such as elderly patients,women,different races or regions)is still needed.
作者
董旭辉
陈德凤
李蓓
彭婉琳
Dong Xuhui;Chen Defeng;Li Bei;Peng Wanlin(School of Nursing,Guangxi University of Chinese Medicine,Nanning 530001,China;Department of Gastrointestinal,Hernia,and Intestinal Fistula Surgery,the People's Hospital of Guangxi Zhuang Autonomous Region,Nanning 530016,China)
出处
《中华现代护理杂志》
2024年第5期644-650,共7页
Chinese Journal of Modern Nursing
关键词
糖尿病
2型
视网膜病变
预测模型
系统评价
Diabetes mellitus,type 2
Diabetic retinopathy
Predictive model
Systematic review
