Prediction of Venous Thromboembolism in Patients With Cancer Using Machine Learning Approaches: A Systematic Review...

Area of Science:

• Oncology
• Medical Informatics
• Biostatistics

Background:

• Venous thromboembolism (VTE) is a significant complication in cancer patients.
• Accurate prediction of VTE risk is crucial for timely intervention and improved patient outcomes.
• Machine learning (ML) approaches are increasingly explored for complex medical predictions.

Approach:

• A systematic literature search was conducted across PubMed, Web of Science, and EMBASE.
• Seven studies encompassing 12,249 cancer patients were included in this meta-analysis.
• Various ML models were evaluated for their performance in predicting VTE events.

Key Points:

• The pooled analysis demonstrated strong performance of ML models in VTE prediction.
• In the training set, pooled sensitivity was 0.87, specificity 0.87, and AUC 0.91.
• In the test set, pooled sensitivity was 0.65, specificity 0.84, and AUC 0.80.

Conclusions:

• ML models exhibit good predictive performance for VTE in cancer patients.
• The findings highlight the potential of ML in clinical decision support for VTE risk stratification.
• External validation is essential to confirm the generalizability and reproducibility of these predictive models.