围生期孕产妇红细胞输注风险预测评分模型的构建与验证

Construction and validation of predictive risk-scoring model for red blood cells transfusion of pregnant women during perinatal period

目的构建围生期孕产妇红细胞输注风险的预测评分模型,并验证其效能。方法选择2016年1月至2019年1月,于北大医疗鲁中医院行常规产前检查及分娩的4593例孕产妇为研究对象。孕产妇的年龄为(28.1±3.3)岁,孕次为(1.8±0.5)次;初产妇为3164例,经产妇为1429例。根据孕产妇围生期是否接受红细胞输注治疗,将其分为输注组(n=96)及未输注组(n=4497)。采用回顾性研究方法,收集2组孕产妇的一般临床和分娩相关资料。单因素分析中,对孕产妇年龄、妊娠前人体质量指数(BMI)及孕次等计量资料的比较,采用独立样本t检验;对经产妇比例、妊娠合并症发生率等计数资料的比较,采用χ^2检验或Fisher确切概率法。根据单因素分析结果,以及既往研究结果和临床经验,将可能影响孕产妇围生期红细胞输注的因素纳入多因素非条件logistic回归分析。采用R软件绘制围生期孕产妇红细胞输注风险列线图,并构建围生期孕产妇红细胞输注风险预测评分模型。采用Bootstrap抽样法进行重复自抽样500次,对围生期孕产妇红细胞输注风险预测评分模型进行内部验证。根据风险预测评分模型,绘制围生期孕产妇红细胞输注率的校正曲线,评价二者的一致性。本研究遵循的程序符合2013年修订版《世界医学协会赫尔辛基宣言》要求。结果①对2组围生期孕产妇红细胞输注影响因素的单因素分析结果显示,输注组孕产妇的年龄,妊娠前BMI,以及经产妇、产前重度贫血、剖宫产术分娩、前置胎盘、胎盘早剥、子痫前期、多胎妊娠、分娩巨大儿的比例,均高于未输注组,并且差异均有统计学意义(t=4.950,P<0.001;t=2.291,P=0.022;χ^2=4.140,P=0.042;P<0.001;χ^2=20.348,P<0.001;P<0.001;P<0.001;P<0.001;χ^2=7.519,P=0.006;P=0.011)。②多因素非条件logistic回归分析的结果显示,孕产妇年龄增加(OR=1.183,95%CI:1.113~1.258,P<0.001),产前重度贫血(OR=12.167,95%CI:1.269~116.608,P=0.030),剖宫产术分娩(OR=6.118,95%CI:1.099~34.042,P=0.039),前置胎盘(OR=9.204,95%CI:1.207~70.198,P=0.032),胎盘早剥(OR=10.714,95%CI:1.125~102.044,P=0.039),子痫前期(OR=4.003,95%CI:1.097~14.606,P=0.036)及分娩巨大儿(OR=1.183,95%CI:1.113~1.285,P<0.001),均为影响围生期孕产妇红细胞输注的独立危险因素。③围生期孕产妇红细胞输注风险预测评分模型列线图显示,风险预测评分模型对围生期孕产妇红细胞输注的一致性指数为0.881(95%CI:0.805~0.932)。校正曲线结果显示,围生期孕产妇红细胞的预测输注率与实际输注率的平均绝对误差为0.011,二者一致性良好。结论根据影响围生期孕产妇红细胞输注的独立危险因素构建的围生期孕产妇红细胞输注风险预测评分模型,可有效预测围生期孕产妇红细胞输注风险。但是,该评分模型对于不同地区围生期孕产妇红细胞输血风险的预测价值,尚需要多中心、大样本研究进一步证实。

Objective To construct a scoring model for predicting red blood cells transfusion risk of pregnant women during perinatal period and verify its efficacy.Methods From January 2016 to January 2019,a total of 4593 pregnant women who underwent routine prenatal examination and delivery in Peking University Medical Luzhong Hospital were selected as the subjects.The age of pregnant women was(28.1±3.3)years,and the gravidity was(1.8±0.5)times.There were 3164 cases of primiparas,and 1429 cases of postpartums.According to whether these pregnant women received red blood cells transfusion during perinatal period,they were divided into transfusion group(n=96)and non-transfusion group(n=4497).Retrospective study method was used to collect the general clinical and delivery related data of the pregnant women in two groups.In univariate analysis,the independent-samples t test was used to compare measurement data of pregnant women between two groups,such as age,pre-pregnancy body mass index(BMI)and gravidity etc..And chi-square test or Fisher′s exact probabilities was used to compare enumeration data,such as the proportion of pregnant women and the incidence of complications during pregnancy etc..According to the results of univariate analysis,previous research results and clinical experience,the factors influencing red blood cells transfusion of pregnant women during perinatal period were included in the multivariate unconditional logistic regression analysis.R software was used to draw the risk nomogram of red blood cells transfusion of pregnant women during perinatal period,and the predictive risk-scoring model for red blood cells transfusion of pregnant women during perinatal period was established.The Bootstrap sampling method was used to repeat self-sampling 500 times,in order to internally verified the predictive risk-scoring model for red blood cells transfusion of pregnant women.According to the risk-predictive scoring model,draw the calibration curve for red blood cells infusion rates of pregnant women during perinatal period,in order to evaluate the consistency of the predicted and actual red blood cells transfusion rates.The procedures followed in this study were in line with the requirements of the World Medical Association Declaration of Helsinki revised 2013.Results①In this study,the results of univariate analysis of factors influencing the red blood cells transfusion of pregnant women during perinatal period in two groups showed that the age,pre-pregnancy BMI,proportion of multipara,prenatal severe anemia,cesarean section,placenta previa,placental abruption,preeclampsia,multifetal pregnancy,macrosomia of pregnant women in transfusion group,were all higher than those of non-transfusion group,and the differences were statistically significant(t=4.950,P<0.001;t=2.291,P=0.022;χ^2=4.14,P=0.042;P<0.001;χ^2=20.348,P<0.001;P<0.001;P<0.001;P<0.001;χ^2=7.519,P=0.006;P=0.011).②The results of multivariate unconditional logistic regression analysis showed that increasing maternal age(OR=1.183,95%CI:1.113-1.258,P<0.001),prenatal severe anemia(OR=12.167,95%CI:1.269-116.608,P=0.030),cesarean section(OR=6.118,95%CI:1.099-34.042,P=0.039),placenta previa(OR=9.204,95%CI:1.207-70.198,P=0.032),placental abruption(OR=10.714,95%CI:1.125-102.044,P=0.039),preeclampsia(OR=4.003,95%CI:1.097-14.606,P=0.036)and macrosomia(OR=1.183,95%CI:1.113-1.285,P<0.001)were the independent risk factors influencing red blood cells transfusion of pregnant women during perinatal period.③According to the nomogram of the predictive risk-scoring model of red blood cells transfusion of pregnant women during perinatal period,the consistency index of predictive risk-scoring model for red blood cells transfusion of pregnant women during perinatal period was 0.881(95%CI:0.805-0.932).The results of calibration curve showed that the average absolute deviation between the predicted and actual red blood cells transfusion rate of pregnant women during perinatal period was 0.011,which proved that the predicted and actual red blood cells transfusion rates were in good consistence.Conclusions Predictive risk-scoring model for red blood cells transfusion of pregnant women during perinatal period was constructed based on the independent risk factors influencing red blood cells transfusion of pregnant women during perinatal period could effectively predict the risk of red blood cells transfusion of pergnant women during perinatal.However,the value of this scoring model for predicting risk of red blood cells transfusion of pregnant women during perinatal period in different regions needs to be further clarified by multi-center and large-sample studies.