[1]肖正国,刘效栓,李喜香,等.一种简便的临床决策模型——注射用血栓通不良反应危险因素分析[J].西部中医药,2025,38(03):36-41.[doi:10.12174/j.issn.2096-9600.2025.03.08]
　XIAO Zhengguo,LIU Xiaoshuan,LI Xixiang,et al.A Simple Clinical Decision Model: Analysis of Adverse Reaction Risk Factors of Xueshuantong Injection[J].Western Journal of Traditional Chinese Medicine,2025,38(03):36-41.[doi:10.12174/j.issn.2096-9600.2025.03.08]
点击复制

一种简便的临床决策模型——注射用血栓通不良反应危险因素分析()
分享到：

《西部中医药》[ISSN:2096-9600/CN:62-1204/R]

卷:
38
期数:
2025年03期
页码:
36-41
栏目:
调查分析
出版日期:
2025-03-15

文章信息/Info

Title:
A Simple Clinical Decision Model: Analysis of Adverse Reaction Risk Factors of Xueshuantong Injection
作者:
肖正国1, 刘效栓1, 李喜香1, 刘军刚1, 张民1, 王炯1, 李莉1, 豆金彦1, 季文莹2
1.甘肃省中医院，甘肃 兰州 730050
2.兰州大学第二医院，甘肃 兰州 730000
Author(s):
XIAO Zhengguo1, LIU Xiaoshuan1, LI Xixiang1, LIU Jun′gang1, ZHANG Min1, WANG Jiong1, LI Li1, DOU Jinyan1, JI Wenying2
1.Gansu Provincial Hospital of Traditional Chinese Medicine, Lanzhou 730050, China
2.Lanzhou University Second Hospital, Lanzhou 730000, China
关键词:
注射用血栓通; 倾向性匹配; 决策树; 随机森林; 单位体表面积给药剂量; 不良反应; 危险因素
Keywords:
injection; propensity score matching; decision tree; random forest; single dose per body surface area; adverse reaction; risk factors
分类号:
R285.6
DOI:
10.12174/j.issn.2096-9600.2025.03.08
文献标志码:
A
摘要:
目的利用在兰州地区收集的数据，探讨引起注射用血栓通（Xueshuantong injection，XSTI）不良反应的潜在危险因素，并开发一种简单的工具以预测和控制其人为因素，为临床合理用药提供依据。 方法在兰州地区收集1 229例使用注射用血栓通的住院病例，其中23例发生了不良反应。运用倾向性匹配（propensity score matching，PSM）对性别、年龄、民族、身高、体质量等混杂因素按照（1∶4）匹配后，得到共115个病例，用决策树（decision tree，DT）分析建模，用随机森林模型（random forest，RF）计算出各自变量的重要性。 结果重要性排名前四的变量为合并用药种类（number of combined drugs，NOCD）、单位体表面积单次给药量（single dose per body surface area，DPBSA）、静滴速度（IV drip rate，IV rate）、溶媒种类（solution type，S type），其平均Gini指数减少量＞3.7，是最重要的危险因素。最终临床预测模型中输出的变量只有NOCD、DPBSA，是密切相关因素。 结论Decision Tree模型是一种简单的不良反应风险预测工具；合并用药种类、单位体表面积给药剂量与注射用血栓通不良反应发生率密切相关；注射用血栓通给药剂量应按体表面积法计算，单位体表面积给药剂量应＜289.2 mg/m2。
Abstract:
ObjectiveTo discuss the potential risk factors of adverse reaction of Xueshuantong injection (XSTI) using the data collected in Lanzhou region, and to develop a simple tool for the prediction and management of its human factors, so as to provide the reference for clinical rational medication. MethodsA total of 1229 inpatient cases of applying XSTI were gathered in Lanzhou region, among them, 23 cases suffered adverse reactions. 115 medical cases were obtained after using propensity score matching (PSM) to match the confounder including gender, age, ethnic groups, body height and body mass in accordance with 1∶4, and these medical cases were analyzed using decision tree (DT) for modeling, and the importance of the respective variables was calculated using the random forest (RF) model. ResultsThe top four variables in terms of importance were number of combined drugs (NOCD), single dose per body surface area (DPBSA), IV drip rate (IV rate) and solution type (S type), and the mean Gini index reduction>3.7, was the most important risk factor. The variables exported from the final clinical prediction model were NOCD and DPBSA, which were closely related factors. ConclusionDecision Tree model is a simple tool for predicting the risk of adverse reactions; NOCD and DPBSA are closely related to the incidence of adverse reaction of XSTI; The dose of XSTI administered should be calculated according to the body surface area method, and the dose administered per unit of body surface area should be <289.2 mg/m2.

备注/Memo

备注/Memo:
肖正国（1966—），男，主任药师。研究方向：医院药学。2022年中央转移支付医疗服务与保障能力提升（中医药事业传承与发展部分）——中医药人才培养重点学科建设项目（甘财社〔2022〕48号）。
更新日期/Last Update: 2025-03-15