Predicting 1-year heart failure hospitalization and mortality post-discharge from the intensive cardiac care unit

Abstract

Aims: Despite the high risk of rehospitalization for heart failure (HF) and death among patients admitted to the intensive cardiac care unit (ICCU), no accurate prediction score for these outcomes exists. We aimed to develop a risk score to predict unplanned HF hospitalization and death 1-year post-discharge in an unselected cohort of patients admitted to the ICCU.

Methods: Based on a national, multicentre study, we included all consecutive patients admitted to the ICCUs in 39 French centres from 7 to 22 April 2021. We randomly selected a training cohort of 21 centres (n = 1008) to develop the ICCU-HF score and a validation cohort of eight other centres (n = 463). The primary composite outcome was unplanned hospitalization for HF and cardiovascular death at 1-year follow-up after discharge. Using the score, patients were stratified into three risk groups to evaluate the prognostic value.

Results: Using a least absolute shrinkage and selection operator (LASSO) regression approach, we identified seven predictors: left ventricular ejection fraction, significant valvular disease grade 2+, Killip score >1, NT-proBNP, creatinine level, previous ventricular arrhythmia and use of inotropes during hospitalization. In 1471 patients (63 ± 15 years, 70% men), 99 (6.7%) experienced the primary outcome. The ICCU-HF score outperformed NT-proBNP, the strongest individual predictor (area under the curve [AUC] 0.77, 95% CI [0.71-0.83] vs. AUC 0.72, 95% CI [0.66-0.79], P = 0.008), demonstrating excellent performance with an AUC of 0.83 (95% CI: 0.77-0.89) to predict outcomes in the validation cohort. Compared with the low-risk group, the intermediate-risk and high-risk groups had significantly higher risks of the composite outcome (HR 4.09, 95% CI [2.23-7.50], P < 0.001 and 12.69, 95% CI [7.02-22.95], P < 0.001), proving strong risk stratification capability of the ICCU-HF score.

Conclusions: The ICCU-HF score showed good performance in predicting the 1-year risk of unplanned HF hospitalization and death in a large cohort of unselected patients admitted to the ICCU, with excellent results in the validation cohort. This score effectively stratifies patients into risk groups, enhancing its utility in clinical decision-making.

Keywords: Death; Heart failure; Intensive cardiac care unit (ICCU); Outcomes; Risk score.

Figure 1

Flowchart of study patients. In total, 1904 patients were eligible for the ADDICT‐ICCU study, 1575 were included, and 1471 patients were analysed for the development and validation of the ICCU‐HF score. ICCU, intensive cardiac care unit.

Figure 2

Performance of the ICCU‐HF score in the training cohort and the validation cohort. Receiver‐operating characteristic (AUROC) for prediction of 1‐year HF outcomes. (A) AUROC of the ICCU‐HF score in the training cohort compared with the strongest individual predictor, NT‐proBNP, showed superior performance of the score (AUC 0.77, 95% CI [0.71–0.83] vs. 0.72, 95% CI [0.66–0.79], P = 0.008). B: AUROCs in the validation cohort showed better predictive performance of the ICCU‐HF score than the ACUTE HF score (AUC 0.83, 95% CI [0.77–0.89] vs. 0.69, 95% CI [0.59–0.79], P = 0.006). AUC, area under the curve; HF, heart failure; ICCU, intensive cardiac care unit; NT‐proBNP, N‐terminal pro‐brain natriuretic peptide.

Figure 3

Cumulative event curves of 1‐year HF outcomes. Based on the ICCU‐HF score, three risk categories were identified. Green, low risk; yellow, intermediate risk; red, high‐risk. Cumulative event curves are shown for the (A) training cohort (n = 1008) and (B) validation cohort (n = 463). Event curves were compared with the log‐rank test. HF, heart failure; ICCU, intensive cardiac care unit.