Risk factors of vancomycin-related acute kidney injury in elderly patients based on machine learning
10.12173/j.issn.1005-0698.202412129
- VernacularTitle:基于机器学习的老年患者万古霉素相关急性肾损伤危险因素分析
- Author:
Xinyu GUO
1
;
Libo DAI
;
Hongxin YANG
Author Information
1. 内蒙古科技大学包头医学院(内蒙古包头 014040);内蒙古自治区人民医院药学处(呼和浩特 010017)
- Publication Type:Journal Article
- Keywords:
Vancomycin;
Acute kidney injury;
Elderly patients;
Risk factor;
Machine learning
- From:
Chinese Journal of Pharmacoepidemiology
2025;34(9):1032-1041
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the risk factors for vancomycin-related acute kidney injury(VA-AKI)in elderly patients.Methods Clinical data of elderly inpatients who used vancomycin at the Inner Mongolia Autonomous Region People's Hospital from January 2021 to June 2024 were retrospectively collected.The incidence of VA-AKI and the situation of treatment drug monitoring(TDM)were statistically analyzed.LASSO regression was used for feature selection,and this process was repeated 10,000 times.In each iteration,75％of the training samples were randomly selected,and the frequency of each feature being selected was counted.Finally,the features with higher frequency in multiple iterations were selected for model training.The data were divided into training set and test set at an 8∶2 ratio.Four machine learning prediction models,including Logistic regression,random forest,extreme gradient boosting(XGBoost),and support vector machine(SVM),were established.The accuracy and area under the receiver operating characteristic curve(AUC)of the above prediction models were calculated in the test set.The minimum depth distribution was used to visualize the importance of the characteristics of the model.Results A total of 305 elderly patients receiving vancomycin were included,among which 49 cases(16.07％)developed VA-AKI.LASSO regression analysis selected 7 characteristic variables to build 4 machine learning models,and finally selected the random forest model as the risk prediction model.The random forest model has an AUC value of 0.91,an accuracy of 0.89,an accuracy of 0.88,a recall rate of 0.98,and an F1 value of 0.93.The predictor importance ranking was in order of post-treatment creatinine level,C-reactive protein(CRP),albumin(Alb),respiratory failure,cardiac insufficiency,trough concentration time,and dose.Conclusion Post-treatment creatinine level,respiratory weakness,trough concentration time,cardiac insufficiency,Alb,CRP,and dosage were the risk factors for VA-AKI.The random forest model is the most effective in predicting the risk of VA-AKI in elderly patients,providing a reference for rational use of vancomycin in elderly patients.
