急性主动脉夹层早期诊断模型的初步建立与评价

Preliminary establishment and evaluation of a model for early diagnosis of acute aortic dissection

目的基于胸痛中心数据库,构建急性主动脉夹层(acute aortic dissection,AAD)的早期诊断模型。方法回顾性收集2020年1-12月于济宁医学院附属医院急诊科胸痛中心就诊的患者临床资料,按照是否诊断为AAD将患者分为AAD组和非AAD组,对两组患者的临床相关指标进行比较,将两组间有统计学差异的研究指标纳入多因素Logistic回归分析,并建立AAD早期诊断列线图模型;绘制模型的受试者工作特征(receiver operating characteristic,ROC)曲线以评估预测准确度,利用Homser-Lemeshow统计量对模型进行拟合优度检验,另收集2021年1-3月就诊的胸痛患者资料共630例对模型进行外部验证。正态分布的计量资料组间比较采用独立样本的t检验,偏态分布的计量资料组间比较采用非参数检验;计数资料组间比较采用χ^(2)检验。结果共纳入2738例患者,AAD患者占4.09%(112/2738)。单因素分析显示AAD组患者男性发病率[74.11%(83/112)]、高血压病史患者占比[70.54%(79/112)]、主动脉疾病史患者占比[10.71%(12/112)]、主动脉疾病家族史患者占比[4.46%(5/112)]、症状突发患者占比[76.79%(86/112)]、撕裂样疼痛患者占比[38.39%(43/112)]、伴背痛患者占比[66.07%(74/112)]、伴腹痛患者占比[16.96%(19/112)]、收缩压[(159.44±30.94)mmHg]、双侧血压/脉搏不对称[23.21%(26/112)]、合并神经体征发生率[7.14%(8/112)]以及D-二聚体[3.57(2.10,6.62)mg/L]均明显高于非AAD组患者[59.56%(1564/2626)、46.23%(1214/2626)、0.23%(6/2626)、0.08%(2/2626)、35.99%(945/2626)、0.08%(2/2626)、3.08%(81/2626)、3.81%(100/2626)、(142.46±27.90)mmHg、0.15%(4/2626)、0.27%(7/2626)、0.31(0.20,0.50)mg/L],年龄[(57.95±14.35)岁]、CK-MB[1.50(0.90,3.25)μg/L]均明显低于非AAD组患者[(61.94±15.77)岁、2.50(1.24,4.81)μg/L],差异均有统计学意义(统计量值分别为χ^(2)=9.47、χ^(2)=25.46、χ^(2)=180.80、χ^(2)=81.11、χ^(2)=76.17、χ^(2)=975.60、χ^(2)=798.00、χ^(2)=44.72、t=6.28、χ^(2)=527.20、χ^(2)=93.22、Z=14.09、t=2.61、Z=3.51;P值分别为0.002、<0.001、<0.001、<0.001、<0.001、<0.001、<0.001、<0.001、<0.001、<0.001、<0.001、<0.001、0.009、<0.001)。多因素分析显示,高血压病史(OR=3.088,95%CI:1.294~7.374)、主动脉疾病史(OR=20.771,95%CI:2.132~202.361)、主动脉疾病家族史(OR=266.425,95%CI:17.610~4030.851)、症状突发(OR=3.538,95%CI:1.643~7.619)、撕裂样疼痛(OR=1771.971,95%CI:204.048~15387.935)、伴背痛(OR=61.550,95%CI:27.987~135.367)、伴腹痛(OR=12.325,95%CI:4.201~36.161)、收缩压(OR=1.026,95%CI:1.013~1.039)、双侧血压/脉搏不对称(OR=338.357,95%CI:60.704~1885.949)和D-二聚体(OR=1.241,95%CI:1.176~1.309)是胸痛患者诊断AAD的独立影响因素(P值分别为0.011、0.009、<0.001、0.001、<0.001、<0.001、<0.001、<0.001、<0.001、<0.001),进一步构建了列线图模型,ROC曲线分析显示,曲线下面积为0.976(P<0.01),特异度为94.52%,灵敏度为91.96%。采用Homser-Lemeshow进行拟合优度检验,表明该模型可以很好的拟合(χ^(2)=2.928,P=0.939)。利用外部验证数据对预测模型进行验证,ROC曲线下面积为0.934(P<0.01),认为该模型有较好的预测性能。结论高血压史、主动脉疾病史、主动脉疾病家族史、症状突发、撕裂样疼痛、伴背痛、伴腹痛、收缩压、双侧血压/脉搏不对称和D二聚体是急性胸痛患者诊断AAD的独立因素,基于上述因素构建的AAD早期诊断列线图模型有较好的预测性能。

Objective An early diagnosis model of acute aortic dissection(AAD)was established based on chest pain center database.Methods The clinical data of patients who attended Chest Pain Center of Department of Emergency in Affiliated Hospital of Jining Medical University of Shandong Province from January 2020 to December 2020 were retrospectively collected.Patients were divided into AAD and non-AAD groups according to whether or not AAD was diagnosed.The clinical related indicators of the two groups were compared.The research indicators with statistical differences between the two groups were included in multivariate Logistic regression analysis,and the early diagnosis of AAD nomogram model was established.The receiver operating characteristic(ROC)curve of the model was used to evaluate the prediction accuracy,and the Homser-Lemeshow statistics were used to test the goodness of fit for the model.A total of 630 patients with chest pain who visited the hospital from January 2021 to March 2021 were also collected for external validation of the model.The t-test of independent samples was used to compare the measurement data of normal distribution,nonparametric test was used to compare the measurement data of skewness distribution,andχ^(2) test was used to compare the counting data between groups.Results A total of 2738 patients were included,of which 4.09%(112/2738)were AAD patients.Univariate analysis showed that in AAD group,male morbidity(74.11%(83/112)),hypertension history(70.54%(79/112)),aortic disease history(10.71%(12/112)),family history of aortic disease(4.46%(5/112)),sudden onset of symptoms(76.79%(86/112)),percentage of patients with laceration pain(38.39%(43/112)),patients with back pain(66.07%(74/112)),patients with abdominal pain(16.96%(19/112)),systolic blood pressure((159.44±30.94)mmHg),bilateral blood pressure/pulse asymmetry(23.21%(26/112)),incidence of complicated neurological signs(7.14%(8/112))and D-dimer(3.57(2.10,6.62)mg/L)were significantly higher than those in non-AAD group(59.56%(1564/2626),46.23%(1214/2626),0.23%(6/2626),0.08%(2/2626),35.99%(945/2626),0.08%(2/2626),3.08%(81/2626),3.81%(100/2626),(142.46±27.90)mmHg,0.15%(4/2626),0.27%(7/2626),0.31(0.20,0.50)mg/L).Age((57.95±14.35)years old)and CK-MB(1.50(0.90,3.25)μg/L)were significantly lower than those in the non-AAD group((61.94±15.77)years,2.50(1.24,4.81)μg/L).The differences were statistically significant(the statistical values wereχ^(2)=9.47,χ^(2)=25.46,χ^(2)=180.80,χ^(2)=81.11,χ^(2)=76.17,χ^(2)=975.60,χ^(2)=798.00,χ^(2)=44.72,t=6.28,χ^(2)=527.20,χ^(2)=93.22,Z=14.09,t=2.61,and Z=3.51,respectively;P values were 0.002,<0.001,<0.001,<0.001,<0.001,<0.001,<0.001,<0.001,<0.001,<0.001,<0.001,<0.001,0.009,and<0.001,respectively).Multivariate analysis showed that history of hypertension(OR=3.088,95%CI:1.294-7.374),history of aortic disease(OR=20.771,95%CI:2.132-202.361),family history of aortic disease(OR=266.425,95%CI:17.610-4030.851),sudden onset of symptoms(OR=3.538,95%CI:1.643-7.619),laceration pain(OR=1771.971,95%CI:204.048-15387.935),back pain(OR=61.550,95%CI:27.987-135.367),abdominal pain(OR=12.325,95%CI:4.201-36.161),systolic blood pressure(OR=1.026,95%CI:1.013-1.039),bilateral blood pressure/pulse asymmetry(OR=338.357,95%CI:60.704-1885.949)and D-dimer(OR=1.241,95%CI:1.176-1.309)were independent factors for the diagnosis of AAD in patients with chest pain(P values were 0.011,0.009,<0.001,0.001,<0.001,<0.001,<0.001,<0.001,<0.001,and<0.001,respectively).Furthermore,the nomogram model was constructed.ROC curve analysis showed that the area under the curve was 0.976(P<0.01),the specificity was 94.52%,and the sensitivity was 91.96%.The statistics of Homser-lemeshow was used to test the goodness of fit,which shows that the model can be fitted well(χ^(2)=2.928,P=0.939).The prediction model was verified by external validation data,and the area under the ROC curve was 0.934(P<0.01),indicating that the model had good prediction performance.Conclusions History of hypertension,history of aortic disease,family history of aortic disease,sudden onset of symptoms,laceration pain,back pain,abdominal pain,systolic blood pressure,bilateral blood pressure/pulse asymmetry and D-dimer were independent factors for the diagnosis of AAD in patients with acute chest pain.The AAD early diagnosis nomogram model based on the above factors has good predictive performance.