基于诺莫图的溃疡性结肠炎合并巨细胞病毒血症诊断预测模型建立与评估  

Nomogram-based diagnostic prediction model for ulcerative colitis complicated with cytomegalovirus viremia

作　　者：王杨 郭亚慧 宋佳[1] 张晓岚[1] Wang Yang;Guo Yahui;Song Jia;Zhang Xiaolan(Department of Gastroentology,the Second Hospital of Hebei Medical University,Shijiazhuang 050035,China)

机构地区：[1]河北医科大学第二医院消化内科,石家庄050035

出　　处：《国际病毒学杂志》2021年第1期42-47,共6页International Journal of Virology

摘　　要：目的:为便于基层医院早期识别溃疡性结肠炎(ulcerative colitis,UC)合并巨细胞病毒(cytomegalovirus,CMV)血症患者,构建简便易行的诺莫图形式诊断预测模型。方法:回顾性分析近4年于我科住院治疗的UC患者临床资料,以2:1随机构建训练组和验证组,运用logistic回归分析训练组相关指标,筛选UC合并CMV血症的独立预测因素,通过R软件的rms包构建预测模型及诺莫图,采用C指数和Hosmer-Lemeshow检验对模型进行区分度和校准度评价,ROC曲线分析确定cut-off值,并利用验证组数据交叉验证评估模型效能。结果:纳入242例UC患者中162例为训练组,80例为验证组,两组UC合并CMV血症发生率分别为27.2%、20.0%,组间各项指标均无统计学差异(P>0.001);单因素logistic回归分析筛选显示年龄、对激素反应、发热、腹痛、体重下降、近一个月有无使用激素或免疫抑制剂、血红蛋白(hemoglobin,Hb)、血沉(erythrocyte sedimentation rate,ESR)、超敏C反应蛋白(hypersensitive C-reactive protein,hs-CRP)、血清白蛋白(albumin,ALB)为UC合并CMV血症的独立相关预测因素;为增强模型预测效能,纳入更多有意义临床指标,设定多因素logistic回归筛选P值临界点为0.2,筛选年龄(OR=1.025,95%CI:0.991~1.062,P=0.156)、发热(OR=12.093,95%CI:3.826~44.152,P<0.001)、腹痛(OR=9.204,95%CI:2.130~67.460,P=0.009)、近一个月使用激素或免疫抑制剂(OR=17.326,95%CI:3.770~96.423,P<0.001)均为具有统计学相关性的危险因素,以此构建的预测模型C指数为0.909,Hosmer-Lemeshow检验P=0.173,ROC曲线分析示cut-off值为0.179时,特异度为82.2%,敏感度为93.2%。相同cut-off值设定下,验证组准确率可达81.3%。结论:年龄、发热、腹痛、近一个月使用激素或免疫抑制剂对于UC合并CMV血症结局具有显著危险相关性,以此拟合的模型及诺莫图在临床使用中具有一定价值,其准确率可达81.3%。Objective To construct a simple and feasible diagnosis and predictive model based on nomogram,so as to facilitate the early identification of patients with ulcerative colitis(UC)complicated with cytomegalovirus(CMV)viremia in primary hospitals.Methods The clinical data of UC patients hospitalized in past 4 years were analyzed retrospectively.The training group and validation group were randomly constructed at a ratio of 2:1.Logistic regression analysis was used to analyze the relevant indicators of the training group,and the independent predictors of UC with CMV viremia were screened.The prediction model and nomogram were constructed by the rms package of R software.The discrimination and calibration of the model were evaluated by C-index and Hosmer-Lemeshow test.The cut-off value was determined by ROC curve analysis.The cross validation of the validation group was used to assess the model efficacy.Results Among 242 patients with UC,162 patients were in the training group and 80 patients were in the validation group.The incidence rates of UC with CMV viremia in both groups were 27.2%and 20.0%,respectively.There was no statistical difference in various indicators(P>0.001).Univariate logistic regression analysis showed that age,response to hormones,fever,abdominal pain,weight loss,presence or absence of hormones or immunosuppressive agents in the previous month,hemoglobin(Hb),erythrocyte sedimentation rate(ESR),hypersensitive C-reactive Protein(hs-CRP),and serum albumin(ALB)were the independent predictors of UC with CMV viremia.In order to enhance the predictive power of the model,more significant clinical indicators were included,and multivariate logistic regression screening P value cut-off point was set at 0.2.The screening age(OR=1.025,95%CI:0.991-1.062,P=0.156),fever(OR=12.093,95%CI:3.826-44.152,P<0.001),abdominal pain(OR=9.204,95%CI:2.130-67.460,P=0.009),and use of hormones or immunosuppressive agents in the past month(OR=17.326,95%CI:3.770-96.423,P<0.001)were all risk factors with statistical correlation.In

关 键 词：溃疡性结肠炎 巨细胞病毒 诊断预测模型 诺莫图 LOGISTIC回归 

分 类 号：R574.62[医药卫生—消化系统] R511[医药卫生—内科学]