基于数据挖掘的脑卒中就医延迟风险预测模型建立与风险管理机制研究

Study demonstrated that approximately 1–3% of patients with acute ischemic stroke receive thrombolysis. Treatment delay is the main reason for insufficient and inefficient thrombolysis therapy in acute stroke. Our preliminary study focused on the patients delay. However, treatment delay involves multiple links and multiple influence factors, in addition to the patients delay, also includes transfer delay and in-hospital delay. Therefore, the risk factors identification, prediction and risk management for treatment delay of stroke are the urgent and key problems need to solve. . The purpose of our study is to study the related concept and build a multidimensional influence factors analysis framework according to the system theory; then deeply to analyze the risk factors and the key link of stroke treatment delay, based on the framework of risk management procedures and in-depth interview and focus group discussion. The study intends to establish risk prediction models of stroke treatment delay using a variety of statistical recognition method of data mining. In addition, delphi, risk matrix, and borda count methods will be used for analysis of the risk factors, and stroke treatment delay risk matrix will be builded. Finally, we will establish the multi-agent risk management framework and brush stroke treatment delay risk management strategy.. This study will help to screen high-risk population and reveal risk factors and the key link of stroke treatment delay. Furthermore,the study will provide the basis for the prevention, effective intervention and control strategy for stroke treatment delay, and provide theoretical basis for risk management research.

调查显示，脑卒中发病时，仅有1～3%的缺血性脑卒中患者及时接受了溶栓治疗，溶栓率低是导致其高死亡和残疾率的重要祸根。而导致溶栓率低的主要原因是就医延迟。因此，对就医延迟的风险因素识别、预测及对就医延迟的风险管理是目前迫切需要解决的两大关键问题。本课题在系统理论研究基础上，界定就医延迟的相关概念内涵，构建多维影响要素的分析框架；以风险管理程序为框架，通过深入访谈、专题小组讨论等深入剖析脑卒中就医延迟的风险因素及关键环节，通过调查并利用数据挖掘的统计识别方法，构建脑卒中就医延迟风险预测模型及组合模型；并综合应用Delphi 法、风险矩阵法和Borda序值法对风险因素分析和评价，率先构建脑卒中就医延迟风险矩阵；构建多主体风险管理框架，梳理脑卒中就医延迟风险管理策略。据此对脑卒中高危人群就医延迟进行筛选，并揭示脑卒中就医延迟风险因素及关键环节，为实施综合有效的干预和风险防范措施提供依据。

近年来脑卒中已高居全国居民死因首位。就医延迟是导致脑卒中患者再灌注治疗不足和低效率的主要原因，尽早开通阻塞血管，挽救缺血半暗带是治疗缺血性脑卒中的关键。在我国，溶栓率远低于其他国家，目前国内外尚缺乏相应的脑卒中院前延迟风险评估模型。本研究通过现场调研并结合文献分析法、定性研究等方法识别脑卒中患者就医延迟的风险因素，在此基础上利用数据挖掘技术构建脑卒中院前延迟风险预测模型，明确脑卒中院前延迟预测变量，构建脑卒中就医延迟风险矩阵，筛选了就医延迟的关键风险因素，并构建了风险管理策略。调研结果显示：12.7%的急性缺血性脑卒中患者在发病后3小时内到达医院，仅6.0%的患者接受了rt-PA溶栓治疗，7.1%的患者接受了尿激酶溶栓治疗。院前延迟数据挖掘结果显示贝叶斯网络、支持向量机和逻辑回归模型均能较好地区分入院及时和院前延迟患者，模型都表现良好（范围平均AUC：0.800-0.846），通过逻辑回归向前逐步选择法筛选的预测变量按重要性排序为：患者对脑卒中静脉溶栓时间窗的了解、症状首次出现时患者的初始反应、转院、居住地、知道身边有人患脑卒中、年龄、子女数。就医延迟风险矩阵的构建结果显示：有6个风险指标的是目前脑卒中就医延迟最关键的风险因素，分别为：“居住地为农村”、“患者发病时身边无人，没有得到帮助”、“患者及家属对脑卒中救治紧急性认识不足”、“患者及家属不知道脑卒中需溶栓及溶栓时间窗”、“发病后患者不认为症状严重，自行服药/没当回事，等待症状缓解”和“患者和家属治疗决策时间过长”。根据以上研究结果提出了风险管理机制：加强全民脑卒中健康教育、提高公众脑卒中急救意识、加强医院溶栓流程进行监控和优化管理等。本研究构建的预测模型和关键风险研究为脑卒中就医延迟高危人群筛选和精准干预策略构建奠定了实证基础。