Development of A Predictive Model for Adverse Inhalation Risk in COPD Inhaler Therapy Using Machine Learning Algorithms
10.3870/j.issn.1004-0781.2024.09.028
- VernacularTitle:基于机器学习算法构建慢性阻塞性肺疾病吸入剂治疗患者不良吸入风险预警模型
- Author:
Lijuan ZHOU
1
;
Xianxiu WEN
;
Haiyan WU
;
Rong JIANG
;
Xuan WANG
;
Li GOU
;
Qin LYU
;
Dingding ZHANG
;
Qian HUANG
;
Xingwei WU
Author Information
1. 电子科技大学附属医院·四川省人民医院护理部/护理研究中心,成都 610072
- Keywords:
Chronic obstructive pulmonary disease(COPD);
Inhalant;
Adverse inhalation;
Machine learning;
Predictive models
- From:
Herald of Medicine
2024;43(9):1509-1518
- CountryChina
- Language:Chinese
-
Abstract:
Objective To construct and validate a risk prediction model for poor inhalation in chronic obstructive pulmonary disease(COPD)patients receiving inhaler therapy,providing a decision support tool for personalized prevention of poor inhalation.Methods A cross-sectional study was conducted to collect data related to COPD patients receiving inhaler therapy,forming a dataset.The dataset was randomly divided into a training set and a test set in a ratio of 4∶1.Four different methods for missing value imputation,3 methods for variable feature selection,and 18 machine learning algorithms were employed to successfully construct 216 models on the training set.The monte carlo simulation method was used for resampling in the test set to validate the models,with the area under curve(AUC),accuracy,precision,recall,and F1 score used to evaluate model performance.The optimal model was selected to build the poor inhalation prediction platform.Results A study involving 308 patients with COPD found that 135(43.8%)were at risk of adverse inhalation.Using 33 predictor variables,216 risk prediction models were developed.Of these models,the ensemble learning algorithm yielded the highest average AUC of 0.844,with a standard deviation of 0.058[95%CI=(0.843,0.845)].The differences in predictive performance among the 216 models were statistically significant(P＜0.01).Under the ensemble learning algorithm,adherence to inhaler use(38.087 4%),inhaler satisfaction(25.680 1%),literacy(24.031 3%),number of inhalers(5.482 3%),age(4.204 5%)and number of acute exacerbations in the past year(2.184 7%)contributed most to the predictive model.The model exhibited superior performance,with an AUC of 0.869 3,an accuracy of 83.87%,a precision of 86.96%,a recall of 74.07%,and an F1 score of 0.8.Conclusion This study has developed a predictive model for poor inhalation risk in COPD inhaler therapy patients using machine learning algorithms,which exhibits strong predictive capabilities and holds potential clinical application value.
