Predictors of 1-Year Mortality Among Patients With Heart Failure With Preserved Ejection Fraction

Accepted heart failure (HF) with preserved ejection fraction (EF) prognostic scores rely on limited variables and linear assumptions that are likely to miss complex risk patterns.

The objectives of the study were to develop, compare, and internally validate prediction models for 1-year all-cause mortality after first hospitalization for decompensated HF with preserved EF.

We performed a retrospective cohort study using electronic medical records from a large academic health system, including adults with EF ≥50% admitted for a first-time HF exacerbation. Variables spanned demographics, comorbidities, laboratory tests, echocardiographic variables, medications, and outcomes. Data were split into training (80%) and test (20%) sets with stratification by outcome. Missing values were handled with multiple imputation by chained equations. Two tree-based classifiers (Extreme Gradient Boosting and Light Gradient Boosting) were tuned with cross-validation and evaluated by area under the receiver operating characteristic curve (AUROC) and calibration. Time-to-event models included Cox proportional hazards, random survival forest (RSF), and gradient boosting survival (GBS) with concordance index and calibration assessment. Global and local (patient-level) explainability was extracted from each model, with cross-model predictor ranking and comparison.

We analyzed 7,840 admissions; the mean age was 78 years with 55.6% women. One-year mortality was 31.5%. Test-set AUROC was 0.751 (95% CI: 0.727-0.775) for Extreme Gradient Boosting and 0.749 for (95% CI: 0.721-0.776) Light Gradient Boosting with acceptable calibration. GBS achieved the highest concordance index (0.718; 95% CI: 0.696-0.740), followed by RSF (0.711; 95% CI: 0.690-0.734) and Cox (0.704; 95% CI: 0.680-0.728). The 12-month time-dependent AUROCs for survival models were GBS 0.759 (95% CI: 0.716-0.799), RSF: 0.750 (95% CI 0.708-0.789), and Cox: 0.735 (95% CI 0.692-0.777). Lower albumin, older age, higher N-terminal pro-B-type natriuretic peptide, renal dysfunction, and lower hemoglobin were the most consistent risk signals.

Our transparent risk tool using routinely available admission data appears feasible, allowing for patient-level, precision health risk assessment.