基于围产期保健流程的多时间窗、多维度产后抑郁症风险预测研究

Postpartum depression (PPD) is a mental disorder caused by multiple factors such as to demography, social psychology, biology and obstetrics. PPD can severely affect the health of pregnant women and their offspring. To predict, identify and then intervene in time are key measures to prevent and treat PPD. So far most studies have only focused on the time points of late pregnancy and puerperium, and mainly studied the relationship between one or two dimensional factors and PPD. Nevertheless, no risk prediction models that consider multiple factors and different time windows, or their interactions have been published yet. The key prediction factors and high risk exposure time window associated to PPD are still unclear. In this proposal, we aim to collect potential risk factors from a cohort of early pregnant women and will follow them till 6 weeks after postpartum. We will collect potential risk factors at four time windows, namely early, middle, late pregnancy and puerperium. Edinburgh Postnatal Depression Scale (EPDS) will be used to screen PPD, and DSM-5 will be used for diagnosis of PPD. Finally, we will develop an optimal PPD risk prediction model that will be validated based on the collected data, explore key impact factors and high risk exposure time windows, which will not only provide optimum opportunities for early prediction, early intervention, timely reducing and mitigating the risks of PPD, but also will provide theoretical justification for development of strategies of preventing and treating PPD during the perinatal health care work.

产后抑郁症（PPD）是女性孕育过程中由人口学、社会心理学、生物学、产科等多维因素所致的精神疾病，严重影响母婴健康。目前研究主要关注孕晚期和产褥期单维度或两个维度因素与PPD的关系，未全面关注围产期时间窗及多维因素交互累积作用并进行预测，PPD关键影响因素和高危暴露时间窗不明确，影响PPD的及时预测、识别和干预。本项目拟招募孕早期妇女，建立孕妇队列并随访至产后6周，结合围产期保健流程在孕早期、孕中期、孕晚期和产褥期 4个时间窗测量危险因素，产后6周内用中文版爱丁堡产后抑郁量表筛查，DSM-5确诊PPD，以有无PPD为结局变量，采用机器学习技术构建多时间窗、多维度PPD风险预测模型，筛选出关键预测因素和PPD高危暴露时间窗及最优PPD预测模型。本项目的实施将为PPD的早期预测提供工具，也为完善围产期保健工作中PPD的早期干预和防治措施提供理论依据。

产后抑郁症是女性孕育过程中由人口学、社会心理学、生物学、产科等多维因素所致的精神疾病，严重影响母婴健康。目前研究主要关注孕晚期和产褥期单维度或两个维度因素与产后抑郁的关系，未全面关注围产期时间窗及多维因素交互累积作用并进行预测，产后抑郁关键影响因素和高危暴露时间窗不明确，影响产后抑郁的及时预测、识别和干预。.本项目从人口经济学、心理社会学、生物学、产科4个维度收集数据，采用Logistic回归、支持向量机和随机森林分别建立孕早期、孕中期、孕晚期和产褥期最佳预测模型。结果显示，产后抑郁的关键预测因素是产前抑郁、产后经济担忧程度、性格特征、母婴同室、多巴胺、孕晚期血糖、晚期高密度脂蛋白等，其中孕晚期生物学指标是影响产后抑郁最为关键的预测因素；其次，基于孕早、中、晚期和产褥期分时间节点和次要关键因素对产后抑郁进行预测，发现孕晚期数据对产后抑郁的预测准确率最高，可以推测出孕晚期可能是产后抑郁的高危暴露时间窗。最后，通过对模型的评价发现，在不分时间节点对产后抑郁预测时，Logistic回归模型的预测效果较好；在分时间节点和次要关键因素筛选时，随机森林效果更佳。.完成了项目计划任务指标，达到了预期目标，明确了产后抑郁的关键预测因子和高危暴露时间窗，建立了产后抑郁的最佳预测模型，为产后抑郁的及时预测、识别和干预高危人群提供了一种有效的评估工具。发表SCI论文2篇，CSCD论文4篇。培养4名硕士研究生毕业。