Systematic Review of 40 ML Models for Aortic Dissection Risk

Forty studies on machine learning models for predicting adverse outcomes in aortic dissection are systematically reviewed and meta-analyzed. The review synthesizes performance metrics like C-statistics for outcomes including early mortality, long-term mortality, and acute kidney injury. It was conducted by Yijun Mao following PRISMA guidelines, with data extracted from six databases up to September 2025.

Use Cases

• Analyze the pooled C-statistic performance for early mortality (0.891) and long-term mortality (0.847) across included studies
• Compare the application of ML algorithms like random forest, SVM, and XGBoost for predicting complications such as AKI and neurological events
• Assess methodological quality using the PROBAST risk-of-bias assessment results, which found 100% of studies at high or unclear risk
• Evaluate reporting completeness against TRIPOD standards, focusing on unreported areas like predictor definitions (62.5% unreported) and sample size justification (82.5% unreported)
• Investigate the prevalence of external validation, noted in only 5 of the 40 studies, as a feature of model robustness

Strengths

• Meta-analysis of 40 included studies provides a quantitative synthesis of model performance
• Reports specific pooled C-statistics with confidence intervals, such as 0.891 for early mortality prediction
• Assesses risk of bias for all 40 studies using the PROBAST tool
• Evaluates adherence to TRIPOD reporting standards, calculating an average score of 78.7%

Limitations

• Significant heterogeneity in meta-analyses, with I² values ranging from 61.3% to 78.8%
• All 40 reviewed studies were assessed as having high or unclear risk of bias, primarily in the analysis domain
• Incomplete reporting is common, with 72.5% of studies not providing full model specifications

Provenance

Source

Systematic review and meta-analysis by Yijun Mao, sourced from figshare.

Collection Method

Literature search across six databases (PubMed, Web of Science, Cochrane Library, Embase, CNKI, Wanfang) following PRISMA, CHARMS, and TRIPOD guidelines.

Time Range

Studies from database inception to September 30, 2025.

Freshness

Includes literature searched up to September 30, 2025, and was last updated in March 2026.

Geography

null