Cardiometabolic & kidneyDiagnostic or prognostic study
Clinical risk prediction models for worsening heart failure events and all-cause mortality in adults with mild-to-moderate chronic kidney disease
- development and internal validation of electronic health record (EHR)-based machine-learning models
- 80:20 train-test split
- >500 EHR-derived covariates
- extreme gradient boosting (XGBoost)
Authors' conclusion
Electronic health record-based machine-learning models accurately predicted WHF and mortality in mild-to-moderate CKD with strong calibration across key subgroups. These models are positioned for EHR deployment to support risk-stratified cardiovascular-kidney-metabolic care-prioritizing guideline-directed therapies and care pathways for those at highest risk.
Study context
- Population
- Adults with mild-to-moderate chronic kidney disease (CKD) [estimated glomerular filtration rate (eGFR) 30–59 ml/min/1.73 m² or eGFR ≥60 with albuminuria] receiving care in a large health system from 2012 to 2021
- Participants
- 375,495 adults
- Outcome
- (i) WHF events (outpatient encounters, emergency department (ED)/observation stays, and hospitalizations) and (ii) all-cause mortality — model performance
- Follow-up
- Outcomes ascertained through 31 December 2022
Performance
As reported in the abstract
Worsening heart failure (WHF) model AUC (test set)
0.887 (95% CI 0.879-0.893)
WHF model calibration slope
0.955
Mortality model AUC (test set)
0.875 (95% CI 0.868-0.883)
Mortality model calibration slope
0.914
Further findings
Performance across subgroups by age, sex, race and ethnicity, and CKD stage
- Performance was consistent across age, sex, and race and ethnicity, with a slight decrement as CKD stage worsened
Original publication
Summarised from the published abstract; every number is checked against it