Abstract
INTRODUCTION: Conventional risk scoring systems for patients with heart failure are insufficient for accurately predicting outcomes in older patients. In this study, we aimed to develop a prognostic prediction model specifically for this population. METHODS: A total of 5690 patients aged ≥80 years (median age: 86 years, 41.8% were men) from the JROADHF (The Japanese Registry of Acute Decompensated Heart Failure) were followed up for 3 years. A randomly selected 70% of the cohort (derivation cohort, n = 3983) was used to develop the risk prediction model, while the remaining 30% (validation cohort, n = 1707) was employed to evaluate its discriminative ability and calibration. Hazard ratios were estimated using the Cox proportional hazards model. Variables were selected using the backward elimination method (threshold: P < .001). The discrimination of the model was assessed by Harrell's C statistic, and the calibration was assessed by a calibration plot. RESULTS: During the follow-up period, a total of 2382 patients died. In the multivariable model, 11 variables, i.e. age, male sex, Barthel index, history of heart failure, systolic blood pressure, haemoglobin, albumin, blood urea nitrogen, b-type natriuretic peptide, sodium levels, and use of renin-angiotensin system inhibitors were selected from 31 potential risk factors. The developed model for predicting mortality at 3 years demonstrated acceptable discriminative ability (Harrell's C-statistic = 0.68, 95% confidence interval: 0.66-0.70) and calibration (Greenwood-Nam-D'Agostino test, P = .30). CONCLUSION: The prognostic model developed in this study (JROADHF over 80 Score) demonstrated satisfactory performance in predicting mortality in a cohort of older Japanese patients with heart failure. Estimating prognosis based on factors obtainable in routine clinical practice has the potential for widespread implementation in clinical settings.