Development and validation of a machine learning algorithm-based risk prediction model of esomeprazole-associated acute kidney injury
10.3760/cma.j.cn114015-20231220-00920
- VernacularTitle:基于机器学习的艾司奥美拉唑相关急性肾损伤风险预测模型的构建与验证
- Author:
Pei ZHANG
1
;
Jiahui LAO
;
Zhaoyang CHEN
;
Shixian CHEN
;
Xiao LI
;
Xin HUANG
Author Information
1. 山东第一医科大学第一附属医院（山东省千佛山医院）临床药学科，济南 250014
- Publication Type:Journal Article
- Keywords:
Risk factors;
Artificial intelligence;
Acute kidney injury;
Esomeprazole;
Prediction model;
Machine learning
- From:
Adverse Drug Reactions Journal
2024;26(7):405-411
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To analyze the influencing factors on the occurrence of acute kidney injury (AKI) in hospitalized patients treated with esomeprazole and to construct a risk prediction model to predict the occurrence of esomeprazole-associated AKI.Methods:The study was designed as a retrospective study. The subjects were selected from patients who were hospitalized in the First Affiliated Hospital of Shandong First Medical University from January 2018 to December 2020 and received treatment with esomeprazole. The clinical data of patients, including basic information, operations, intervention measures, medication, and laboratory test results, was collected through the hospital′s electronic medical record system. Patients were divided into AKI and non-AKI groups according to the occurrence of esomeprazole-associated AKI, and the clinical characteristics between the 2 groups were compared. The least absolute shrinkage and selection operator (LASSO regression) was used to analyze the influencing factors of esomeprazole-associated AKI. Patients were randomly divided into the training set and the test set at a 8∶2 ratio. Based on data in the training set, 5 machine learning algorithms were used to build esomeprazole-associated AKI prediction models, including logistic regression, random forest, gradient boosting machine (GBM), extreme gradient boosting, and light gradient boosting machine. Based on data in the test set, the performance of 5 models was validated through the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and accuracy.Results:A total of 5 436 patients were enrolled in the study, including 3 231 males and 2 205 females, with an age of 61(51, 70) years. Esomeprazole-associated AKI occurred in 393 patients, with an incidence of 7.23%. The results of LASSO regression analysis identified 24 variables closely related to esomeprazole-associated, such as hepatic insufficiency, chronic renal insufficiency, hypoproteinemia. Based on data in the training set (4 349 patients), the esomeprazole-associated AKI risk prediction models were constructed and their predictive performance was good (all AUC>0.900). The predictive performance validation was conducted using the data in the test set (1 087 patients), and the results showed that the GBM model has the highest AUC (0.922) and relatively stable performance, with small differences in various indicators between the training and the test sets.Conclusions:The use of esomeprazole is significantly associated with AKI, and the risk is influenced by factors such as baseline renal function, comorbidities, and combined medications. The risk prediction model based on GBM algorithm is helpful for early assessment of the risk of esomeprazole-related AKI in clinical practice.
