基于聚类分析的不良事件类型对医疗纠纷的预测效能研究

The Predictive Effect of Adverse Event Types on Medical Disputes Based on Cluster Analysis

目的探索医疗不良事件成因和级别的类型特点及其对医疗纠纷的预测效能。方法采用频数、系统聚类、Logistic回归和ROC曲线分析两家三级医院463例医疗不良事件数据。结果18类不良事件可聚合为4类,第Ⅰ~Ⅳ类的成因类型特点为诊疗技术、医患沟通、医院管理和其他原因。一到四级不良事件可聚合为2类,第Ⅰ和Ⅱ类的级别类型特点为有错误事实或损害和无错误事实及损害。成因类型为第Ⅰ类(OR=6.872)和第Ⅲ类(OR=4.383)、级别类型为第Ⅰ类(OR=4.137)的不良事件发生纠纷的风险更高(P<0.05)。医疗纠纷个案预测模型建立:P=1/{1+e^(-[-3.428+1.928 X1(1)+0.948 X1(2)+1.478 X1(3)+1.420 X2])},最佳截点值为0.098,敏感度0.870,特异性0.676,准确率83.4%,曲线下面积(AUC)为0.819(P<0.05)。结论聚类分析的不良事件成因和级别类型能体现客观实际,对医疗纠纷有良好的预测效能。医疗机构应特别重视技术水平和标准化管理的建设,并充分发挥不良事件管理制度对医疗纠纷的预防作用。

Objective To explore the cause and classification type characteristics of medical adverse events and their effects on medical disputes.Methods Frequency,systematic clustering,Logistic regression and ROC curve were used to analyze the data of 463 medical adverse events in two tertiary hospitals.Results The 18 types of adverse events could be grouped into four types,and the causes of the first to fourth types were characterized by diagnosis and treatment technology,doctor-patient communication,hospital management and other reasons.Level 1 to 4 adverse events could be grouped into 2 categories,with level types 1 and 2 characterized by false fact or injury and no false fact or injury.Genetic type was class 1(OR=6.872)and class 3(OR=4.383),and the level type was class 1(OR=4.137)of adverse events had a higher risk of disputes(P<0.05).Medical dispute individual prediction model:P=1/{1+e^(-[-3.428+1.928 X1(1)+0.948 X1(2)+1.478 X1(3)+1.420 X2])},the best cut-off value was 0.098,with sensitivity of 0.870,specificity 0.676,accuracy 83.4%,and AUC=0.819(P<0.05).Conclusion The causes and level types of adverse events in cluster analysis can reflect the objective reality and have a good predictive effect on medical disputes.Medical institutions should pay special attention to the construction of technical level and standardized mana-gement,and give full play to the adverse event management system for the prevention of medical disputes.