A Predictive Model for Venous Thromboembolism Based on Multi-View Clustering

Background: Venous thromboembolism (VTE) is a common chronic disease with a considerable risk of recurrence, and early screening for VTE risk assessment and comprehensive data decision-making analyses can reduce its incidence and harm to a certain extent. We proposed a clustering model based on multi-view learning for VTE incidence risk prediction. Methods: This is a retrospective study of 15,856 orthopedic surgical patients who met the diagnosis in a hospital information system between 1992-2017. The risk of developing venous thromboembolism was analyzed and predicted from multiple views using multi-view learning to improve the accuracy of disease prediction. Results: Five multi-view clustering algorithms were selected as comparison algorithms and the performance of these models was evaluated using metrics such as accuracy, purity, and F-score. After comparing the ACC values for each algorithm, it was found that the proposed algorithm had a significantly higher ACC value (0.9172) than the other comparison algorithms (0.6481, 0.6242, 0.7740, 0.8306, and 0.7844, respectively). Conclusions: The proposed algorithm has high effectiveness for VTE risk prediction. The model can assist healthcare professionals to improve the accuracy and timeliness of VTE risk assessment and identify the risk of VTE in patients as early as possible.