Independent external validation of the LIFE-HF clinical prediction model using the GUIDE-IT trial

Accurate prognostication is central to decision-making in heart failure (HF). The recently-developed LIFE-HF models offer promise, but their performance has not been independently and externally validated. The aim of this study was to assess the external validity of the LIFE-HF models in a contemporary North American population.

We externally validated the LIFE-HF models for 1-year all-cause death and the composite of death or HF hospitalization using patient-level data from the GUIDE-IT trial. All LIFE-HF predictors were available; missing data were handled using multiple imputation. Predicted risks were calculated using original LIFE-HF model equations. We assessed model discrimination using time-dependent area under the receiver operating characteristic curve (AUROC), calibration [observed-to-expected (O/E) ratios, calibration slopes and curves], overall prediction error, and net benefit using decision curve analysis.

The validation cohort included 801 participants (median age 64 years, 69% male). The mortality model demonstrated good discrimination [AUROC 0.77, 95% confidence interval (CI): 0.72-0.82], but underprediction (O/E 1.28, 95% CI: 1.01-1.54) and underfitting (calibration slope 1.58, 95% CI: 1.20-1.95). The composite model showed moderate discrimination (AUROC 0.69, 95% CI: 0.64-0.73), underprediction (O/E 1.40, 95% CI: 1.25-1.54), and overfitting (calibration slope 0.82, 95% CI: 0.63-1.00). Decision curve analysis showed net clinical benefit for the mortality model, but not the composite model, over a broad range of risk thresholds.

In a contemporary, high-risk North American HF with reduced ejection fraction cohort, the LIFE-HF models showed good discrimination but systematically underpredicted 1-year risk. The mortality model may support risk-informed decision-making, whereas the composite model requires further validation.