Regional Validation and Recalibration of Clinical Predictive Models for Patients With Acute Heart Failure

Area of Science:

• Cardiology
• Health Informatics
• Clinical Epidemiology

Background:

• Clinical practice guidelines advocate for validated clinical predictive models (CPMs) in heart failure management.
• The generalizability of existing heart failure CPMs across diverse populations remains largely unexamined.

Purpose of the Study:

• To assess the performance and generalizability of North American-derived acute heart failure CPMs in international cohorts.
• To evaluate the impact of different recalibration strategies on CPM performance across global regions.

Main Methods:

• Independent external validation of three CPMs predicting in-hospital, 60-day, and 1-year mortality in an international acute heart failure trial (N=4133).
• Assessment of CPM discrimination (Area Under the Receiver Operating Curve) and calibration across distinct world regions.
• Comparison of global versus region-specific recalibration techniques.

Main Results:

• CPM discrimination significantly decreased in all international validation cohorts, with a median reduction in AUC of -0.09.
• Regional calibration exhibited high variability, ranging from <1% to >50% in performance metrics.
• Global recalibration failed to adequately improve calibration, whereas region-specific recalibration substantially enhanced CPM performance.

Conclusions:

• The predictive accuracy and calibration of acute heart failure CPMs differ significantly across global regions.
• Region-specific recalibration methods are crucial for optimizing the clinical utility and reliability of CPMs in diverse international settings.