Predictive model for totally implanted venous access ports‐related long‐term complications in patients with lung cancer
Jian Jia,Xutong Fan,Wenhong Zhang,Zhiyang Xu,Mian Wu,Yiyang Zhan,Boqiang Fan
DOI: https://doi.org/10.3892/ol.2024.14459
2024-05-17
Oncology Letters
Abstract:Totally implanted venous access ports (TIVAPs), which are typically used in oncological chemotherapy and parenteral nutritional support, are convenient and safe, and thus offer patients a higher quality of life. However, insertion or removal of the device requires a minor surgical operation. Long‐term complications (>30 days post insertion), such as catheter migration, catheter‐related thrombosis and infection, are major reasons for TIVAP removal and are associated with a number of factors such as body mass index and hemoglobin count. Since management of complications is typically time‐consuming and costly, a predictive model of such events may be of great value. Therefore, in the present study, a predictive model for long‐term complications following TIVAP implantation in patients with lung cancer was developed. After excluding patients with a large amount of missing data, 902 patients admitted to The First Affiliated Hospital with Nanjing Medical University (Nanjing, China) were ultimately included in the present study. Of the included patients, 28 had complications, indicating an incidence rate of 3.1%. Patients were randomly divided into training and test cohorts (7:3), and three machine learning‐based anomaly detection algorithms, namely, the Isolation Forest, one‐class Support Vector Machines (one‐class SVM) and Local Outlier Factor, were used to construct a model. The performance of the model was initially evaluated by the Matthew's correlation coefficient (MCC), area under curve (AUC) and accuracy. The one‐class SVM model demonstrated the highest performance in classifying the risk of complications associated with the use of the intracavitary electrocardiogram method for TIVAP implantation in patients with lung cancer (MCC, 0.078; AUC, 0.62; accuracy, 66.0%). In conclusion, the predictive model developed in the present study may be used to improve the early detection of TIVAP‐related complications in patients with lung cancer, which could lead to the conservation of medical resources and the promotion of medical advances.
oncology