Lung Transplant Mortality Risk Model Performance Metrics

A novel topological feature engineering model achieved 87.4% accuracy, 84.1% sensitivity, and 89.6% specificity for predicting post-operative mortality in lung transplant recipients. The model, developed by Alexy Tran-Dinh and published on figshare in 2026, integrates static and time-dependent clinical variables to outperform traditional risk scores. It demonstrated an absolute AUC gain of 0.08 over the best non-topological baseline.

Use Cases

• Benchmark predictive models for post-transplant mortality based on reported accuracy and AUC metrics.
• Study the impact of topological feature extraction on clinical risk prediction based on the described methodology.
• Analyze the importance of dynamic variables like oxygenation trends and inflammatory markers based on Shapley analysis results.
• Compare model performance across subgroups like age and transplant type based on the study findings.

Strengths

• Model performance metrics are explicitly reported: 87.4% accuracy, 84.1% sensitivity, 89.6% specificity.
• Comparative analysis shows a statistically significant AUC improvement of 0.08 over the best baseline (p < 0.001).
• Interpretability is assessed using Shapley Additive explanations, identifying key predictive features.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is very small (9.5 KB), indicating limited scope, likely containing only summary performance metrics.

Provenance

Source

figshare

Collection Method

Retrospective cohort study using clinical data from lung transplant recipients.

Freshness

Last updated 2026-05-05 18:00:39; freshness should be verified.