1. Peripherally inserted central-related upper extremity deep vein thrombosis and machine learning.
- Author
-
Hu H, Wu Z, and Zhao J
- Subjects
- Humans, Middle Aged, Male, Female, Risk Factors, Risk Assessment, Aged, Adult, China epidemiology, Neoplasms, Central Venous Catheters, Decision Support Techniques, Catheterization, Peripheral adverse effects, Catheterization, Peripheral instrumentation, Machine Learning, Upper Extremity Deep Vein Thrombosis diagnostic imaging, Upper Extremity Deep Vein Thrombosis etiology, Catheterization, Central Venous adverse effects, Catheterization, Central Venous instrumentation, Predictive Value of Tests
- Abstract
Objective: To establish a prediction model of upper extremity deep vein thrombosis (UEDVT) associated with peripherally inserted central catheter (PICC) based on machine learning (ML), and evaluate the effect., Methods: 452 patients with malignant tumors who underwent PICC implantation in West China Hospital from April 2021 to December 2021 were selected through convenient sampling. UEDVT was detected by ultrasound. Machine learning models were established using the least absolute contraction and selection operator (LASSO) regression algorithm: Seeley scale model (ML-Seeley-LASSO) and ML model. The information of patients with and without UEDVT was randomly allocated to the training set and test set of the two models, and the prediction effect of machine learning and existing prediction tools was compared., Results: Machine learning training set and test set were better than Seeley evaluation results, and ML-Seeley-LASSO performance in training set was better than ML-LASSO. The performance of ML-LASSO in the test set is better than that of ML-Seeley-LASSO. The use of ML model (ML-LASSO and ML-Seeley-LASSO) in PICC-related UEDVT shows good effectiveness (the area under the subject's working characteristic curve is 0.856, 0.799), which is superior to the currently used Seeley assessment tool., Conclusion: The risk of PICC-related UEDVT can be estimated and predicted relatively accurately by using the method of ML modeling, so as to effectively reduce the incidence of PICC-related UEDVT in the future., Competing Interests: Declaration of conflicting interestsThe author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
- Published
- 2024
- Full Text
- View/download PDF