基于大数据挖掘下银屑病患者心血管疾病风险评估的价值

The value of cardiovascular disease risk assessment of psoriasis patients based on Big data mining

目的通过大数据回顾性分析银屑病患者心血管疾病(CVD)的发生,建立大数据风险模型评估银屑病患者CVD发病风险。方法收集2014年1月-2018年1月新疆医科大学附属中医医院皮肤科治疗银屑病患者2500例的临床资料,根据随访5年后患者是否发生CVD分为CVD组和无CVD组,对比2组患者临床及实验室资料;采用多因素Logistic回归分析银屑病患者发生CVD的危险因素;通过随机森林算法建立大数据风险模型,应用决策曲线分析大数据风险模型用于银屑病患者CVD发病风险的评估价值。结果随访5年,2500例银屑病患者新发CVD 350例。CVD组年龄、PASI评分、FRS评分、CRP、IL-17、IL-22、TNF-α和IgG水平高于无CVD组,CD4^(+)水平低于无CVD组(χ^(2)/t=9.467,10.512,12.158,15.492,10.677,13.496,15.023,16.002,8.194,P均<0.001);多因素Logistic回归分析显示,年龄大、PASI评分高、FRS评分高、CRP高、IL-17高、IL-22高、TNF-α高、IgG高是银屑病罹患CVD的独立危险因素,CD4^(+)低水平为银屑病罹患CVD的独立保护因素[OR(95%CI)=1.051(1.035~1.068),1.083(1.061~1.106),1.245(1.176~1.318),1.429(1.334~1.532),1.142(1.106~1.179),1.170(1.133~1.209),1.370(1.218~1.464),1.601(1.469~1.745),0.947(0.929~0.965)];大数据模型中各变量的重要程度依次为:CRP、IgG、TNF-α、FRS评分、IL-17、PASI评分、IL-22、CD4^(+)、年龄;ROC曲线显示,基于随机森林算法构建的大数据模型预测银屑病患者新发CVD风险的AUC为0.988(95%CI 0.955~0.999,P<0.001);决策曲线分析显示,与FRS评分相比,基于随机森林算法构建的大数据模型对银屑病患者新发CVD风险具有较高的预测能力。结论银屑病患者具有较高的新发CVD风险,基于随机森林算法建立大数据风险模型可提高对银屑病患者发生CVD风险早期评估的准确性。

Objective To analyze the occurrence of cardiovascular diseases(CVD)in psoriasis patients through big data and to establish a big data risk model to assess the risk of CVD development in psoriasis patients.Methods The clinical data of 2500 patients with psoriasis who were treated in the Dermatology Department of the Affiliated Hospital of Traditional Chinese Medicine of Xinjiang Medical University from January 2014 to January 2018 were collected.The patients were divided into CVD group and non-CVD group according to whether they had CVD after 5 years of follow-up.The clinical and laboratory data of the patients in the two groups were compared;Using multivariate logistic regression analysis to identify the risk factors for CVD in psoriasis patients;The Big data risk model was established through Random Forest algorithm,and the decision curve was used to analyze the evaluation value of big data risk model for CVD risk of psoriasis patients.Results Following a 5-year follow-up,350 new cases of CVD were found in 2500 psoriasis patients.Based on this grouping,clinical data were compared,and the age,PASI score,FRS score,CRP,IL-17,IL-22,TNF of the CVD group were determined-αAnd IgG levels were higher than those in the non CVD group,while CD4^(+)levels were lower than those in the non CVD group(χ^(2)/t=9.467,10.512,12.158,15.492,10.677,13.496,15.023,8.194,all P<0.001);Multivariate logistic regression analysis showed that older age,higher PASI score,FRS score,CRP,IL-17,IL-22,TNF-αHigh and IgG levels are independent risk factors for CVD in psoriasis,while low CD4^(+)levels are independent protective factors for CVD in psoriasis[OR(95%CI)=1.051(1.035-1.068),1.083(1.061-1.106),1.245(1.176-1.318),1.429(1.334-1.532),1.142(1.106-1.179),1.170(1.133-1.209),1.370(1.218-1.464),1.601(1.469-1.745),0.947(0.929-0.965)];The importance of each variable in the Big data model is in the order of CRP,IgG,TNF-α、FRS score,IL-17,PASI score,IL-22,CD4^(+),age;Receiver operating characteristic shows that the AUC of Big data model based on Random forest algorithm to predict the risk of new CVD in psoriasis patients is 0.988(95%CI=0.955~0.999,P<0.001);The analysis of the decision curve shows that,compared with the FRS score,the Big data model based on Random forest algorithm has a higher predictive ability for the risk of new CVD in psoriasis patients.Conclusion Psoriasis patients have a high risk of new CVD.Building a Big data risk model based on Random forest algorithm can improve the accuracy of early assessment of CVD risk of psoriasis patients.