What type of study is this?

This is a Cohort Study study (also classified as: Validation Study).

What question did this study set out to answer?

The aim is to develop and validate a machine learning prognostic model for predicting mortality in patients with coronary artery disease and diabetes.

January 9, 2026Frontiers in Cardiovascular MedicineOpen Access

Explainable machine learning model of coronary artery disease combined with diabetes: development and validation study

Q: What does this research mean for the field?

A multimodal prognostic model accurately predicts all-cause mortality with an AUC of 0.824 in patients with comorbid coronary artery disease and diabetes. Novelty: ClaimNovelty.CONFIRMATORY. Consensus alignment: ConsensusAlignment.SUPPORTS_CONSENSUS.

View Full Paper

Ask AI

Bookmark

Key result

A multimodal prognostic model accurately predicted all-cause mortality with an AUC of 0.824 in patients with comorbid coronary artery disease and diabetes.

Authors

XWX. H. Wang SWShipeng Wang XLXuhui Liu

Discussion

Member takes

Implication

Retrospective cohort study develops a machine learning model predicting mortality in CAD and diabetes patients, indicating clinical value.

Key Points

The aim is to develop and validate a machine learning prognostic model for predicting mortality in patients with coronary artery disease and diabetes.
Conducted a retrospective cohort study of 389 patients with CAD and diabetes.
Used LASSO regression for feature selection and backward stepwise Cox regression analysis.
Developed a nomogram based on eight independent predictors for clinical use.
Assessed model performance through AUC analysis, calibration plots, and decision curve analysis.
Utilized random survival forest analysis for validating clinical relevance of predictors.
The final model showed AUC values of 0.846 in training and 0.824 in validation cohorts for 3-year mortality prediction.
Significant survival stratification was achieved using individual predictors in Kaplan–Meier analysis.
The comprehensive model outperformed a 9-variable model across all evaluated timepoints.
Five key predictors (hemoglobin, INR, albumin, NT-proBNP, age) were prioritized in random survival forest analysis.
The model demonstrated strong reliability and utility in clinical decision-making.

Cite This Study

Wang et al. (2026) studied this question. A multimodal prognostic model accurately predicted all-cause mortality with an AUC of 0.824 in patients with comorbid coronary artery disease and diabetes.

synapsesocial.com/papers/69612fd330ef6c21f6853ceb https://doi.org/10.3389/fcvm.2025.1674287

View Full Paper

Ask AI

Bookmark

Implication

Retrospective cohort study develops a machine learning model predicting mortality in CAD and diabetes patients, indicating clinical value.