Cardiovascular Classification in Middle‐Aged and Elderly Chronic Kidney Disease (CKD) Patients: A Machine Learning Approach in China and the United States
Authors' conclusion
CVD-associated feature profiles differ substantially between US and Chinese middle-aged and elderly CKD patients, supporting the necessity of population-calibrated classification models rather than applying a unified model across countries.
Study context
- Population
- Middle-aged and elderly CKD patients in China and the US
- Participants
- 1,057 vs 1,109
- Exposure
- Machine learning (ML) models for classifying cardiovascular disease (CVD) in US vs Chinese CKD patients
- NHANESn = 1,057
- CHARLSn = 1,109
- Outcome
- CVD classification performance (test AUC) of best model
Methodological detailsretrospective, machine learning, bidirectional cross-population external validation, two nationally representative cohorts
- Retrospective
- Machine learning
- Bidirectional cross-population external validation
- Two nationally representative cohorts
Main finding
As reported in the abstract
| Group | XGBoost test AUC | Logistic Regression test AUC |
|---|---|---|
| NHANES | 0.753 | — not stated in the abstract |
| CHARLS | — not stated in the abstract | 0.769 |
XGBoost test AUC
NHANES0.753
CHARLS— not stated in the abstract
Logistic Regression test AUC
NHANES— not stated in the abstract
CHARLS0.769
— not stated in the abstract
Observational data: associations, not effects.
Further findings
Bidirectional cross-population external validation
- Both models showed a decline in discriminative performance when deployed across populations, indicating limited cross-population transportability
SHAP feature contributions
- Age was the top contributing feature in the US cohort; hypertension was the primary contributing feature in the Chinese cohort
Original publication
Cardiovascular Classification in Middle‐Aged and Elderly CKD Patients: A Machine Learning Approach in China and the United States. Diabetes Obes Metab. Published 27 September 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →