胡蜂蜇伤致急性呼吸窘迫综合征风险预测模型的构建

Construction of a prediction model for risk of acute respiratory distress syndrome caused by wasp stings

关键词: 胡蜂蜇伤  急性呼吸窘迫综合征(ARDS)  列线图  预测模型

Keywords: wasp sting  acute respiratory distress syndrome (ARDS)  column chart  prediction model

摘 要: 目的 分析胡蜂蜇伤致急性呼吸窘迫综合征(acute respiratory distress syndrome,ARDS)的危险因素并构建列线图风险预测模型。方法 收集 2020—2023 年杭州市中医院病房及重症监护室收治的胡蜂蜇伤患者,根据胡蜂蜇伤后是否发生 ARDS 分为 ARDS 组和非 ARDS 组,分别采用单因素和多因素 logistic 回归分析筛选 ARDS 发生的独立危险因素, 构建列线图预测模型。结果 本研究共纳入符合入组标准的病例 236 例,其中 ARDS 组 93 例、非 ARDS 组 143 例,单因素及多因素 logistic 回归分析显示,蜂蜇伤≥20 针、肉眼血尿或无尿、活化部分凝血活酶时间(APTT)增加以及天门冬氨酸氨基转移酶(AST)升高,为胡蜂蜇伤并发 ARDS 的独立危险因素(OR>1);基于此 4 项独立危险因素构建胡蜂蜇伤致 ARDS 的预测模型并绘制列线图。预测模型特征曲线下面积(area under the curve,AUC)为 0.800(95%CI 0.742~0.858),提示模型有很好的区分度;采用 Bootstrap 法重复抽样 1 000 次对模型进行内部验证并绘制校准曲线,结果显示平均 AUC 为0.806(95%CI 0.746~0.859),校准曲线与理想曲线基本重合,Hosmer-Lemeshow 检验示 P>0.05,说明预测模型与理想模型有较好拟合度。结论 基于蜂蜇伤≥20 针、肉眼血尿或无尿、APTT 增加、AST升高 4 项危险因素构建的胡蜂蜇伤致 ARDS 风险预测模型具有良好的区分度、校准度和临床实用性,有助于临床早期评估 ARDS 的发生风险。

Abstract: Objective To analyze the risk factors of acute respiratory distress syndrome (ARDS) caused by wasp stings and construct a column chart risk prediction model.Methods Collect the data of patients with wasp stings admitted to the wards or ICU of Hangzhou Municipal Hospital of Traditional Chinese Medicine from 2020 to 2023,divided them into ARDS group and non-ARDS group based on whether ARDS occurred after wasp stings,then using single factor and multiple factor logistic regression analyses respectively to screen independent risk factors for ARDS occurrence and construct a column chart prediction model.Results A total of 236 cases that met the inclusion criteria were included, including 93 cases in the ARDS group and 143 cases in the non-ARDS group,univariate and multivariate logistic regression analysis suggested that stings ≥20 times,gross hematuria or anuria,prolonged activated partial thromboplastin time (APTT) prolonged, and elevated aspartate aminotransferase(AST) were the independent risk factors for ARDS caused by wasp stings (OR>1),based on these four independent risk factors,a prediction model for ARDS caused by wasp stings was constructed and the column chart was drawn as well.The area under the curve (AUC) was 0.800 (95%CI 0.742~0.858),indicating that the model had good discriminability.The Bootstrap method was used for 1000 times repeated sampling to internally validate the model and draw a calibration curve,the results showed that the average AUC verified by the Bootstrap method was 0.806 (95% CI 0.746~0.859).The calibration curve basically coincident with the ideal curve,the Hosmer-Lemeshow test showed the P>0.05,indicating that the prediction model had a good fit with the ideal model.Conclusion The results suggested that the prediction model for ARDS caused by wasp stings based on four risk factors including ≥20 stings,gross hematuria or anuria,proloned APTT,and rised AST had good discrimination,calibration, and clinical practicality,which were helpful for early assessment of the risk of ARDS in clinical practice.