子痫前期不良妊娠结局预测模型的构建与验证

Construction and Validation of A Predictive Model for Adverse Pregnancy Outcomes in Preeclampsia

目的:分析子痫前期(preeclampsia)发生围产期不良妊娠结局的相关影响因素,构建风险预测模型并加以验证。方法:回顾性分析2018年1月1日—2022年12月31日于中国人民解放军北部战区总医院(我院)分娩并诊断为子痫前期患者的临床资料,采用多因素Logistic回归分析筛选子痫前期围产期不良妊娠结局的独立危险因素,利用R语言构建风险预测列线图模型。根据受试者工作特征(receiver operator characteristic,ROC)曲线下面积(area under the curve,AUC)及Hosmer-Lemeshow检验评价该模型的预测性能与拟合优度,校准曲线评估其准确性,决策分析曲线评估模型的临床使用价值。选取2023年1月1日—2023年6月30日于我院分娩的子痫前期患者进行外部验证。结果:纳入1057例子痫前期患者进行建模,按7∶3的比例分为训练集(739例)和验证集(318例);纳入125例子痫前期患者进行外部验证。多因素Logistic回归分析显示,发病孕周≤34周、平均动脉压≥120 mmHg(1 mmHg=0.133 kPa)、胎儿生长受限、纤维蛋白原≤4 g/L、尿蛋白定性++以上、血清白蛋白≤30 g/L、乳酸脱氢酶≥263 U/L均是子痫前期患者出现不良妊娠结局的独立危险因素(均P<0.05)。据此建立Logistic风险预测模型。该模型的AUC为0.941(95%CI:0.925~0.958),以0.382为截断值,特异度87.8%,敏感度85.1%,校准曲线显示预测子痫前期不良妊娠结局发生的概率与实际发生的概率具有较好的一致性,内部验证校准曲线显示模型一致性良好。外部验证校准曲线显示模型校准度良好,Hosmer-Lemeshow检验显示模型的预测概率与实际观测概率差异无统计学意义(χ^(2)=12.164,P=0.144)。临床决策分析曲线表明列线图有一定的临床实用性。结论:构建的子痫前期不良妊娠结局预测模型具有良好的准确度,所选用指标简单易得,列线图应用简便,可以为临床医生评估子痫前期患者的母婴结局提供一定的参考依据。

Objective:To analyze the risk factors associated with the occurrence of adverse perinatal pregnancy outcomes in preeclampsia,construct a risk prediction model and validate it.Methods:A retrospective analysis was conducted on clinical data of patients who delivered and diagnosed with preeclampsia at General Hospital of Northern Theater Command(our hospital)from January 1st,2018 to December 31st,2022.Multivariable logistic regression analysis was used to screen for independent risk factors for adverse perinatal pregnancy outcomes in preeclampsia.The R language was utilized to construct the risk prediction columnar graphical model.The predictive performance and goodness of fit of the model were evaluated based on the area under the receiver operator characteristic curve(AUC)and the Hosmer-Lemeshow test,the calibration curves were assessed for accuracy,decision analysis curve was used to assess the clinical use of the model.Patients with preeclampsia who delivered at our hospital from January 1st,2023 to June 30th,2023 were selected for external validation.Results:A total of 1057 patients with preeclampsia were included for modelling,divided into a training set(739 patients)and a validation set(318 patients)in a 7∶3 ratio.125 patients with preeclampsia were included for external validation.Multifactorial logistic regression analysis showed that the following factors were independent risk factors for adverse pregnancy outcomes in patients with preeclampsia:gestational week of onset≤34 weeks,mean arterial pressure≥120 mmHg(1 mmHg=0.133 kPa),fetal growth restriction,fibrinogen≤4 g/L,urine protein qualification of++or higher,serum albumin≤30 g/L,and lactate dehydrogenase≥263 U/L(all P<0.05).Logistic risk prediction model was established accordingly.The AUC of the model was 0.941(95%CI:0.925-0.958),with Jordon index 0.382,specificity 87.8%,sensitivity 85.1%,and calibration curves show good agreement between the probability of predicting the occurrence of adverse pregnancy outcomes in preeclampsia and the probability of actual occurrence, internal validation calibration curves showed good model agreement. The external validation calibration curve showed that the model was well calibrated, and the Hosmer-Lemeshow test showed that the difference between the predicted probability of the model and the actual observed probability was not statistically significant ( χ^(2)=12.164, P=0.144), clinical decision analysis curves indicated the clinical utility of nomograms to some extent. Conclusions: The prediction model of adverse pregnancy outcome in preeclampsia constructed has good accuracy, the selected indicators are simple and easy to obtain, and the nomograms are easy to apply, which can provide a certain reference basis for clinicians to assess the maternal and infant outcomes of patients with preeclampsia.