Machine learning-based prediction of the post-thrombotic syndrome: Model development and validation study

Area of Science:

• Vascular Medicine
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Post-thrombotic syndrome (PTS) is a significant complication following deep vein thrombosis (DVT).
• Effective prevention strategies are crucial for reducing PTS incidence.
• Accurate prediction models are needed to identify high-risk individuals.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting PTS occurrence within 24 months post-acute DVT.
• To assess the predictive accuracy and generalization capability of different ML algorithms.

Main Methods:

• Utilized clinical data from the Acute Venous Thrombosis: Thrombus Removal with Adjunctive Catheter-Directed Thrombolysis study and an external Chinese cohort.
• Included 23 clinical variables and applied four ML algorithms (e.g., logistic regression, gradient boosting).
• Evaluated model performance using F scores and Area Under the Curve (AUC) for discrimination and calibration.

Main Results:

• Models were built using 555 DVT patients and validated on 117 patients.
• Logistic regression with gradient descent and L1 regularization achieved an AUC of 0.83 in external validation.
• Gradient boosting models demonstrated stable generalization across derivation and validation cohorts, categorizing patients into risk groups.

Conclusions:

• Machine learning models exhibit accurate prediction and stable generalization for PTS.
• These models can enhance clinical decision-making for PTS management.
• Implications include improved patient selection for interventions like endovascular surgery.