Meta-Analysis of 40 Machine Learning Models for Aortic Dissection Risk

40 studies on machine learning models for predicting adverse outcomes in aortic dissection were systematically reviewed and meta-analyzed. The analysis, registered with PROSPERO, synthesized performance metrics like C-statistics and assessed risk of bias using PROBAST. It covers outcomes including early mortality, long-term mortality, and acute kidney injury.

Use Cases

• Compare pooled C-statistics for early mortality (0.891) and long-term mortality (0.847) across different ML algorithms.
• Analyze the heterogeneity (I² = 61.3–78.8%) and risk of bias findings from the PROBAST assessment of the 40 included studies.
• Evaluate reporting completeness against TRIPOD standards, focusing on unreported predictor definitions (62.5%) and sample size justifications (82.5%).
• Assess the frequency of external validation, noted in only 5 of the 40 studies, as a key methodological limitation.

Strengths

• Systematic review of 40 studies following PRISMA, CHARMS, and TRIPOD guidelines.
• Meta-analysis provides pooled performance metrics, such as a C-statistic of 0.891 for early mortality prediction.
• Comprehensive risk of bias assessment using PROBAST, applied to all 40 included studies.

Limitations

• Significant methodological heterogeneity (I² = 61.3–78.8%) and high/unclear risk of bias in 100% of studies limits generalizability.
• Critical reporting gaps exist, with 82.5% of studies omitting sample size justification and 72.5% not providing full model specifications.
• The dataset is a 20.8 KB document summarizing the review, not the underlying primary data from the 40 studies.

Provenance

Source

figshare, authored by Yijun Mao.

Collection Method

Systematic review and meta-analysis of studies from six databases (PubMed, Web of Science, Cochrane Library, Embase, CNKI, Wanfang).

Time Range

Literature search from database inception to September 30, 2025.

Freshness

Last updated March 26, 2026.

Geography

null