Systematic Review of Machine Learning Models for Deep Vein Thrombosis Prediction

17 studies developing machine learning models to predict deep vein thrombosis in hospitalized patients were systematically reviewed. The pooled AUC of 10 validated models was 0.85, indicating good discrimination, but substantial heterogeneity was present. The review, authored by An Liu and published on figshare in 2026, assessed model quality, performance, and clinical applicability.

Use Cases

• Benchmarking new DVT prediction models based on reported performance metrics like AUC
• Identifying common predictive features for DVT based on frequently reported predictors like D-dimer and age
• Assessing methodological quality of clinical ML studies based on the PROBAST risk-of-bias framework

Strengths

• Analysis includes 17 studies identified through a systematic search of four major databases
• Meta-analysis quantitatively synthesizes performance data from 10 validated models
• Methodological quality is assessed using the PROBAST tool, a standard for prediction model reviews

Limitations

• The dataset is a 165.1 KB document summarizing review findings, not the underlying patient-level data
• Column-level documentation for the models' data is absent; predictor details must be extracted from the text
• The file format is DOCX, which may require conversion for automated data extraction

Provenance

Source

figshare

Collection Method

Systematic literature review and meta-analysis following PRISMA guidelines.

Time Range

Literature search covered from November 1, 2015, to November 1, 2025.

Freshness

Last updated 2026-05-26 04:39:01