Application of a machine learning-assisted prescription rationality prediction model in perioperative rational drug use management
10.12173/j.issn.1005-0698.202409026
- VernacularTitle:机器学习辅助处方合理性预测模型在围手术期合理用药管理中的应用
- Author:
Lijuan FAN
1
;
Zhiqi ZHANG
;
Xiaojun CHENG
;
Xiunan YUE
;
Haiyan CHENG
;
Nan SHANG
Author Information
1. 山西医科大学附属运城市中心医院药学部(山西运城 044000)
- Keywords:
Prescription rationality;
Machine learning;
Perioperative period;
Clinical pharmacy;
Prescription review
- From:
Chinese Journal of Pharmacoepidemiology
2024;33(11):1219-1228
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the influencing factors of rational perioperative drug use,and to establish a rationality prediction model based on machine learning to assist pharmacists in prescription review.Methods A retrospective analysis was conducted on the perioperative prescription data of neurosurgery patients from a tertiary hospital and a central hospital in Shanxi Province between March 2021 and March 2023.Univariate analysis and multivariate Logistic regression were initially used to identify factors influencing rational drug use,followed by Lasso regression and multicollinearity analysis to select important variables.The data was split into a training set and test set at a ratio of 7∶3,and decision tree(DT),multi-layer perceptron(MLP),extreme gradient boosting(XGBoost),support vector machine(SVM),and random forest(RF)learning models were constructed.Results A total of 1 500 prescriptions were included,of which 668 were classified as rational and 832 as irrational.In both the training and test sets,the AUC values of the DT,XGBoost,and RF models exceeded 0.9.The DT model showed the highest sensitivity(0.81),while the RF model demonstrated the highest specificity(0.90).In the RF model,the number of comorbidities,preoperative waiting time,total hospitalization cost,prescribing physician's title,and adverse reaction occurrence negatively impacted prescription rationality,whereas the number of drugs,age,and administration route positively influenced rationality.Conclusion The machine learning-based rational drug use prediction model demonstrates strong predictive performance,effectively assisting pharmacists in prescription review and helping to reduce the incidence of irrational drug use.
