产后创伤后应激障碍的预测模型构建与验证

Construction and validation of predictive models for postpartum post-traumatic stress disorder

目的探讨产后创伤后应激障碍(PTSD)的风险因素,构建列线图预测模型。方法选取2021年6月至2023年6月该院收治的产妇610例作为研究对象,采用简单随机化方式分为建模组(458例)和验证组(152例),收集产妇临床资料,建模组根据是否发生产后PTSD分为无产后PTSD组(403例)与产后PTSD组(55例)。采用logistic多因素回归模型分析发生产后PTSD的危险因素;采用R3.6.1软件绘制预测产后PTSD影响因素的列线图模型,采用Bootstrap法对列线图模型进行内部验证,以Hosmer-Lemeshow拟合优度检验分析一致性;绘制受试者工作特征(ROC)曲线评估列线图模型区分度,校准曲线评估准确度。结果建模组产妇中初产妇、不良孕产史比例与验证组比较,差异均有统计学意义(P<0.05),其余各指标比较,差异均无统计学意义(P>0.05)。产后PTSD组产妇中有精神创伤史、新生儿并发症、剖宫产/辅助分娩、妊娠期并发症、睡眠质量差、有产后抑郁、心理一致感水平低比例均明显高于无产后PTSD组,差异均有统计学意义(P<0.05)。精神创伤史、辅助分娩/剖宫产、睡眠质量差、有产后抑郁、心理一致感水平低是发生产后PTSD的独立危险因素,差异均有统计学意义(P<0.05)。基于结果建立的列线图模型建模组ROC曲线的曲线下面积为0.865(95%可信区间0.830~0.895),灵敏度为70.91%,特异度为86.85%,验证组ROC曲线的曲线下面积为0.838(95%可信区间0.801~0.878)。列线图模型的区分度、一致性较高。结论基于产后PTSD风险因素构建的列线图预测模型具有良好的区分度与一致性,有利于临床尽早制定干预策略。

Objective To analyze the risk factors of postpartum post-traumatic stress disorder(PTSD)and construct a nomogram prediction model.Methods A total of 610 pregnant women admitted to our hospital from June 2021 to August 2023 were selected as the study objects,and they were divided into the modeling group(458 cases)and the validation group(152 cases)according to simple randomization.Clinical data were collected,and the modeling group was divided into a group without postnatal PTSD(403 cases)and a group without postnatal PTSD(55 cases)according to whether PTSD occurred.Logistic multivariate regression was used to analyze the risk factors of postpartum PTSD.R 3.6.1 software was used to draw a nomogram model for predicting the influencing factors of postpartum PTSD.Bootstrap method was used to internally verify the nomogram model,and Hosmer-Lemeshow goodness of fit test was used to analyze the consistency.The receiver operating curve(ROC)curve was drawn to evaluate the model differentiation of the nomogram,and the accuracy of curve evaluation was calibrated.Results Compared with the verification group,the proportion of primipara and bad pregnancy history in the modeling group was statistically significant(P<0.05),while the other indexes were not statistically significant(P>0.05).The proportion of patients with history of mental trauma,neonatal complications,cesarean section/assisted delivery,complications during pregnancy,poor sleep quality,postpartum depression and low level of psychological unity in postpartum PTSD group were significantly higher than those without postpartum PTSD group(P<0.05).Logistic multivariate regression analysis showed that history of mental trauma,assisted delivery/cesarean section,poor sleep quality,postpartum depression and low level of psychological congruence were independent risk factors for postpartum PTSD(P<0.05).In the nomogram model established based on the results,the area under the curve of ROC curve in the modeling group was 0.865(95%CI 0.830-0.895),the sensitivity was 70.91%,and the specificity was 86.85%,while the area under the curve of ROC curve in the verification group was 0.838(95%CI 0.801-0.878).The nomogram model has high discrimination and consistency.Conclusion The nomogram prediction model based on postpartum PTSD risk factors has good differentiation and consistency,which is conducive to making intervention strategies as soon as possible in clinic.