Predicting major adverse cardiac events in diabetes and chronic kidney disease: a machine learning study from the Silesia Diabetes-Heart Project
Authors' conclusion
Ten features-based ML models, especially the LGBM model, had acceptable performance in predicting CVEs in persons with DM and CKD. A decrease in eGFR, aging, and elevated inflammatory markers significantly enhanced the predictive capability of the model. Future external validation of our model is required prior to implementation in a clinical environment.
Study context
- Population
- People with diabetes mellitus (DM) and chronic kidney disease (CKD) from the Silesia Diabetes-Heart Project (1,116 of 3,056 with DM); median age 67 [IQR 57–76] years; 57% menof 3,056 with DM
- Participants
- 1,116 people
- Comparison
- Five machine learning (ML) models for predicting cardiovascular events
- Outcome
- Prediction of CVEs: composite of nonfatal MI, new onset HF, nonfatal stroke, incident AF, PCI or CABG, hospitalisation or death due to CVD
- Follow-up
- Median 3.1 years
Methodological detailsroutine standard of care dataset, five ML models compared, SHAP interpretability
- Routine standard of care dataset
- Five ML models compared
- SHAP interpretability
Performance
As reported in the abstract
Prediction of CVEs (AUC)
| Test or model | Value | AUC | Recall | F1-score | G-mean |
|---|---|---|---|---|---|
| LGBM | 0.740 (95% CI 0.738–0.743) | 0.740 | 0.739 | 0.820 | 0.826 |
| LR | 0.621 (95% CI 0.618–0.623) | 0.621 | |||
| RF | 0.707 (95% CI 0.704–0.709) | 0.707 | |||
| SVC | 0.707 (95% CI 0.704–0.710) | 0.707 | |||
| XGBM | 0.710 (95% CI 0.707–0.713) | 0.710 |
Value
LGBM0.740 (95% CI 0.738–0.743)
LR0.621 (95% CI 0.618–0.623)
RF0.707 (95% CI 0.704–0.709)
SVC0.707 (95% CI 0.704–0.710)
XGBM0.710 (95% CI 0.707–0.713)
AUC
LGBM0.740
LR0.621
RF0.707
SVC0.707
XGBM0.710
Recall
LGBM0.739
LR— not stated in the abstract
RF— not stated in the abstract
SVC— not stated in the abstract
XGBM— not stated in the abstract
F1-score
LGBM0.820
LR— not stated in the abstract
RF— not stated in the abstract
SVC— not stated in the abstract
XGBM— not stated in the abstract
G-mean
LGBM0.826
LR— not stated in the abstract
RF— not stated in the abstract
SVC— not stated in the abstract
XGBM— not stated in the abstract
Further findings
Incidence of cardiovascular events (CVEs) · median 3.1 years
- CVE incidence
- 14.1% (157/1,116)
SHAP feature importance (LGBM)
- Estimated glomerular filtration rate (eGFR), age and triglyceride glucose index were the three most important features for predicting CVEs
Original publication
Predicting major adverse cardiac events in diabetes and chronic kidney disease: a machine learning study from the Silesia Diabetes-Heart Project. Cardiovasc Diabetol. Published 15 February 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →