Machine learning-based prediction models for patients no-show in online outpatient appointments

• DOI: 10.1016/j.dsm.2021.06.002
• Publication Date: 2021-06-25T00:19:30Z
• Authors: Guorui Fan; Zhaohua Deng; Qing Ye; Bin Wang
• Keywords: Boosting; Predictive modelling; Ensemble Learning

With the development of information and communication technologies, all public tertiary hospitals in China began to use online outpatient appointment systems. However, phenomenon patient no-shows appointments is becoming more serious. The objective this study design a prediction model for no-shows, thereby assisting making relevant decisions, reducing probability no-show behavior. We used 382,004 original records, divided data set into training (N1 ​= ​286,503), validation (N2 ​95,501). machine learning algorithms such as logistic regression, k-nearest neighbor (KNN), boosting, decision tree (DT), random forest (RF) bagging models appointments. rate was 11.1% (N ​42,224). From set, had highest area under ROC curve AUC value, which 0.990, followed by boosting models, were 0.987 0.976, respectively. In contrast, compared with previous values tree, neighbors lower at 0.597, 0.499 0.843, This demonstrates possibility using from multiple sources predict no-shows. results can provide basis reduce medical resource waste, develop effective policies, optimize operations.