精神分裂症患者自杀尝试风险预测模型的构建与评价

Construction and evaluation of a suicide risk prediction model for schizophrenia patients

目的:基于精神分裂症患者自杀尝试危险因素分析,构建风险预测模型,并且验证模型的预测价值。方法:按TRIPOS规范,选取2018年1月—2020年9月在常州市德安医院精神卫生中心就诊的精神分裂症患者,根据7:3原则,将先入组的436例患者纳入建模组,后入组的187例患者纳入验证组,收集患者的一般资料、临床特征资料、社会学资料等,采用二元Logiistic回归分析筛选出独立影响因素,建立风险预测模型,利用ROC曲线下面积评价模型的预测效应,并在验证组中进行实践验证。结果:性别、成长环境、过量饮酒、阳性症状、抑郁、自知力、支持利用度是精神分裂症患者自杀尝试的独立影响因素;建立的精神分裂症患者自杀尝试风险预测模型为Logit(P)=-13.018+0.841X1(性别)+3.694×X2(成长环境)+3.842×X3(过量饮酒)+0.149×X4(阳性症状)-0.329×X5(自知力)+0.207×X6(抑郁)-0.323×X7(支持利用度),ROC曲线下面积为0.936[95%CI(0.914~0.958),P<0.001],最大约登指数为0.740,以此时为最佳截断值0.294,预测敏感度为89.1%,预测特异度为84.9%;验证组ROC曲线下面积为0.880[95%CI(0.820~0.920),P<0.001],预测敏感度为89.4%,预测特异度为85.7%,准确度为86.6%。结论:女性、成长在农村、过量饮酒以及伴明显的阳性症状、抑郁情绪的精神分裂症患者易发生自杀尝试,自知力、支持利用度是自杀尝试的保护因素。该模型在实践验证中表现出较高的预测价值,并且模型中相关数据的获取也较为简便,可为精神分裂症患者自杀尝试的预防、监管、干预等工作提供参考依据。

Objective:To construct a risk prediction model based on the risk factors analysis of suicide attempts in patients with schizophrenia,and to verify the predictive value of the model.Methods:According to the TRIPOS specification,patients with schizophrenia of Mental health center in Changzhou DeAn hospital who were selected from January 2018 to September 2020,according to the principle,and will be into the first group of 436 patients were included in the model group,after the group of 187 patients were included in the validation group,general information,clinical features in patients with information collection,sociology etc,adopt dual Logiistic regression analysis select independent factors of risk prediction model is established,using the area under the ROC curve evaluation model prediction effect,and verifies the practice in the validation set.Results:Gender,growth environment,excessive alcohol consumption,positive symptoms,depression,self-knowledge and utilization of support were the independent factors influencing suicide attempts in patients with schizophrenia.The established risk prediction model for suicide attempt in schizophrenia patients was Logit(P)=-13.018+0.841X1(gender)+3.694×X2(growing environment)+3.842×X3(excessive drinking)+0.149×X4(positivesymptoms)-0.329×X5(self-awareness)+0.207×X6(depression)-0.323×X7(support utilization).The area under the ROC curve was 0.936[95%CI(0.914-0.958),P<0.001],and the maximum dANG index was 0.740.The critical value was 0.294,the predictive sensitivity was 89.1%,and the predictive specificity was 84.9%.The area under the ROC curve of the verification group was 0.880[95%CI(0.820~0.920),P<0.001],the prediction sensitivity was 89.4%,the prediction specificity was 85.7%,and the accuracy was 86.6%.Conclusion:Women,rural residents,schizophrenia patients with excessive alcohol consumption,obvious positive symptoms and depression were more likely to have suicide attempts.Self-knowledge and utilization of support were protective factors for suicide attempts.The model shows high predictive value in practice verification,and the acquisition of relevant data in the model is relatively simple,which can provide reference for the prevention,supervision and intervention of suicide attempts in patients with schizophrenia.