子痫前期预测模型的建立与评价

Establishment and evaluation of predictive model for preeclampsia

目的了解子痫前期（PE）发病的预测指标,为预防PE提供依据。方法对确诊的PE患者210例（病例组）和正常妊娠200例（对照组）孕妇的临床资料进行回顾性调查,分析PE的影响因素,建立多因素Logistic模型,并绘制ROC曲线检测Logistic模型的预测效果。结果 1PE组的年龄≥30岁、孕检〈4次、水肿、尿蛋白阳性、高血压家族史、孕前BMI≥24 kg/m2的构成比明显高于对照组（P〈0.05）。2多因素回归分析显示,水肿（OR=20.240）、尿蛋白（OR=9.062）、高血压家族史（OR=4.605）、孕前BMI（OR=4.550）是预测PE的独立因素,预报概率（P）：P=1/[1＋EXP（2.334-3.008水肿-2.204尿蛋白-1.527高血压家族史-1.515孕前BMI）]。3模型预测PE的ROC曲线下面积为0.899（95%CI：0.869-0.930）,敏感性、特异性和一致率分别是0.810、0.850和82.95%。而水肿[0.793（95%CI：0.747-0.838）]、尿蛋白[0.667（95%CI：0.614-0.719）]、孕前BMI[0.654（95%CI：0.601-0.707）]和高血压家族史[0.565（95%CI：0.510-0.621）]的ROC曲线下面积均小于模型（P〈0.05）。结论以水肿、尿蛋白、孕前BMI和高血压家族史4个指标联合预测PE的效能高于任何单一指标。

Objective To understand the predictive index of preeclampsia, so as to provide evidence for the prevention of preeclampsia. Methods 210 patients with preeclampsia （case group） and 200 cases with normal pregnancy （control group） were retrospectively investigated. The influence factors of preeclampsia were analyzed, the logistic regression model was established and ROC curve was applied to evaluate predictive power of model. Results （ 1 ） Compared with control group, case group had higher proportion in age ≥30 year, prenatal examination 〈 4 times, edema, urinary protein positive, positive family history of hypertension, pro - pregnant body mass index ≥ 24 kg/m2 （ P 〈 0.05 ）. （ 2 ） Multiple regression analysis showed that edema （ OR = 20.240）, urinary protein （ OR = 9. 062）, positive family history of hypertension（ OR = 4. 605 ） and pre - pregnant body mass index （ OR = 4.550） were independent predictors of preeclampsia. Probabilities （P） = 1/[ 1 ＋ EXP （ 2. 334 - 3. 008 edema - 2. 204 urinary protein - 1. 527 positive family history of hypertension- 1. 515 pre- pregnant body mass index）]. （3）The area under the ROC curve predicted preeelampsia using the model was 0. 899 （95% CI： 0. 869 -0. 930 ）, the sensitivity, specificity and consistency of the model were 0. 810,0. 850 and 82. 95 % , respectively. And the area under the ROC curve of edema [ 0.793 （ 95 % CI： 0.747 - 0. 838 ） ] , urinary protein [ 0.667 （ 95 % CI： 0. 614 -0. 719 ） ], positive family history of hypertension [0. 565 （95% CI： 0.510 -0. 621 ） ] and pre - pregnant body mass index [0.654（95% CI：0. 601 -0. 707 ） ] was lower than that of model,P 〈 0.05. Conclusion Using the combined predictors of edema, urinary protein, positive family history of hypertension and pre - pregnant body mass index can improve the predictive power.