Predictors and outcomes of deep venous thrombosis in patients with acute ischemic stroke: results from the Chinese Stroke Center Alliance

CAI, Weixin; ZHANG, Ran; WANG, Yongjun; LI, Zixiao; LIU, Liping; GU, Hongqiu; YANG, Kaixuan; YANG, Xin; WANG, Chunjuan; WANG, Anxin; SUN, Weige; XIONG, Yunyun

doi:10.23736/s0392-9590.23.05077-0

Cited by 1 publication

(1 citation statement)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A critical concern in the aftermath of an ischemic stroke is the development of venous thromboembolism (VTE), a severe complication typically arising within the first two weeks post-event, with its highest risk noted in the initial seven days(2, 3). The occurrence of VTE is closely linked with an escalated risk of mortality within three months following a stroke, highlighting its importance as a preventable aspect of post-stroke care management(4, 5).…”

Section: Introductionmentioning

confidence: 99%

Leveraging Machine Learning for Enhanced and Interpretable Risk Prediction of Venous Thromboembolism in Acute Ischemic Stroke Care

Jiang,

Li,

et al. 2024

Preprint

View full text Add to dashboard Cite

Background: Stroke is the second leading cause of death globally, with acute ischemic strokes constituting the majority. Venous thromboembolism (VTE) poses a significant risk during the acute phase post-stroke, and early recognition is critical for preventive intervention of VTE. Methods: Utilizing data from the Shenzhen Neurological Disease System Platform to develop multiple machine learning models that included variables such as demographics, clinical data, and laboratory results. Advanced technologies such as K nearest neighbor and synthetic minority oversampling technique are used for data preprocessing, and algorithms such as gradient boosting machine and support vector machine are used for model development.Feature analysis of optimal models using SHapley Additive exPlanations interpretable algorithm. Results: In our study of 1,632 participants, in which women were more prevalent, the median age of patients with VTE was significantly older than that of non-VTE individuals. Data analysis showed that key predictors such as age, alcohol consumption, and specific medical conditions were significantly associated with VTE outcomes. The AUC of all prediction models is above 0.7, and the GBM model shows the highest prediction accuracy with an AUC of 0.923. These results validate the effectiveness of this model in identifying high-risk patients and demonstrate its potential for clinical application in post-stroke VTE risk management. Conclusion: This study presents an innovative, machine learning-based approach to predict VTE risk in acute ischemic stroke patients, offering a tool for personalized patient care. Future research could explore integration into clinical decision systems for broader application.

show abstract