输尿管结石进展为尿脓毒血症预测模型的优化与可视化呈现

Optimization and visualization of predictive model for the risk of ureteral calculi developed into urosepsis

目的对已建立的输尿管结石进展为尿脓毒血症的预测模型相关危险因素进一步筛选和优化,并利用列线图方法对预测模型进行可视化呈现。方法再次对南海医院泌尿外科2013年1月-2015年12月收治747例输尿管结石患者的临床资料进行回顾性分析。选取输尿管结石进展为尿脓毒血症患者62例为观察组,随机抽取同期本院住院并且未进展为无尿脓毒血症的输尿管结石患者685例为对照组。剔除既往预测模型中影响早期诊断的相关危险因素尿液细菌培养,并通过单因素和多因素Logistic回归分析评价输尿管结石进展为尿脓毒血症的独立危险因素。应用R语言建立预测模型的列线图,利用C-index参数评价预测模型的区分度,通过拟合优度检验(H-L)进行判断预测模型的校准度。结果多因素Logistic回归分析结果显示,患者性别、尿常规WBC计数、尿亚硝酸盐、肾积液平均CT值和功能性孤立肾5个危险因素是输尿管结石进展为尿脓毒血症的独立危险因素。列线预测模型的初始C-index为0.913,经过500次自抽样的模型内部验证,C-index校准为0.907,提示新的列线图预测模型拥有很好的区分度。H-L检验提示X^2=7.887,P=0.343,提示新的列线图预测模型拟合度好,预测校准度强。结论经过优化的个体化列线图预测模型有助于提高输尿管结石进展为尿脓毒血症高危患者的早期识别和筛选能力。

Objective To further screen and optimize the risk factors of prediction model for ureteral calculi developing into urosepsis,and to visualize the prediction model by using nomogram.Methods The clinical data of 747 patients with ureteral calculi who were admitted from January 2013 to December 2015 were selected.62 patients with urosepsis were included in the case group,and 685 cases without urosepsis were randomly selected into the control group.The risk factor of urine bacterial culture which impact early diagnosis was eliminated in the previous prediction model,and the other factors remained of ureteral calculi developing into urosepsis were screened using univariate and multivariate logistic regression analysis.The corresponding nomogram prediction model was drawn according to the regression coefficients.The C-index and Hosmer-Lemeshow goodness of fit test were used to evaluate the discrimination and calibration of the prediction model,respectively.Results Multivariate logistic regression analysis showed that gender,mean CT attenuation value of hydronephrosis,urine WBC count,urine nitrite and functional solitary kidney were the independent risk factors for ureteral calculi developing into urosepsis(P<0.05).The initial C-index of the nomogram prediction model was 0.913 and it was reduced to 0.907 after internal validation by bootstrap,suggested that the new nomogram prediction model had good discrimination capacity.Hosmer and Lemeshow test showed a good fitting of the new prediction model(χ^2=7.887,P=0.343).Conclusions The new optimized individualized nomogram prediction model can be used to improve the early identification and screening of high risk patients of ureteral calculi developing into urosepsis.