Machine Learning-Based Predictive Models for Patients with Venous Thromboembolism: A Systematic Review.
Vasiliki Danilatou1,2, Dimitrios Dimopoulos3, Theodoros Kostoulas3
1School of Medicine, European University of Cyprus, Nicosia, Cyprus.
Thrombosis and Haemostasis
|April 4, 2024
Summary
Machine learning clinical prediction models (ML-CPMs) show promise for improving venous thromboembolism (VTE) care. Further research is needed to standardize reporting and validate these advanced models for real-world application.
Area of Science:
- Medical Informatics
- Artificial Intelligence in Medicine
- Clinical Decision Support
Background:
- Venous thromboembolism (VTE) poses a significant health and economic burden.
- Existing VTE clinical prediction models (CPMs) have limitations in decision-making.
- This review explores machine learning (ML) to enhance CPMs using electronic health record data.
Approach:
- Systematic review of studies using structured data from PubMed, Google Scholar, and IEEE.
- Inclusion criteria focused on ML applications in VTE risk stratification, outcome prediction, diagnosis, and treatment.
- Excluded non-English, non-human, NLP, image processing, and studies on pregnant women, cancer patients, or children.
Key Points:
- ML-CPMs generally outperformed traditional CPMs in receiver operating area under the curve.
- Most included studies were retrospective, monocentric, and lacked external validation.
- Identified research gaps include standardized reporting, reproducibility, and model comparison.
Conclusions:
- ML-CPMs demonstrate potential for enhanced VTE risk assessment and personalized treatment.
- Urgent need for standardized reporting, methodology, external validation, and prospective studies.
- Interventional studies are required to assess the real-world impact of AI in VTE management.
