A simple risk score for prediction of the need of rescue ultrafiltration at admission in patients hospitalized for acute decompensated heart failure episodes. Development and initial validation
P Klin, F Nutinez, R M Fortunato, P M Raffaele, F R Klein, L FavaloroAbstract
Background
Diuretic strategies in patients (pts) hospitalized for acute heart failure (AHF) have evolved in last years, resulting in more efficacious decongestion and shorter length of stay (LOS). However, rescue ultrafiltration (r-UF) remains a last barrier for diuretic resistant pts despite remnant congestion. Although several risk factors for r-UF have been identified, up-to-date information regarding risk prediction at admission is scarce.
Objectives
We sought to develop a simple risk score of the need of r-UF in patients pts hospitalized for AHF at the time of admission.
Methods
A total of 2011 pts were admitted for AHF between July, 2011, and December, 2025. They were randomly assigned to a development and a validation dataset (80/20%). Baseline clinical characteristics of the 1608 pts in the development dataset were considered as candidate univariate predictors of r-UF. Multivariate logistic regression was then used to identify independent predictors of r-UF. Based on the coefficients, 6 identified variables (age, left ventricular ejection fraction, clinical predominant heart failure phenotype, admission serum creatinine, admission serum sodium, and admission T3) were assigned to a weighted integer; the sum of the integers was a total risk score for each patient.
Results
The overall occurrence of r-UF was 4.1%. Score ranged between 0-7; Low scores included 0-4, while high scores 5-7 points. Development set r-UF rate was 3.1% for a low and 31.6% for a high risk score, respectively. It increased exponentially with increasing risk score (Cochran Armitage chi-square, p 0.0001). In the 403 pts of the validation dataset, the model demonstrated a high discriminative (C statistic 0.90) and calibration power (accuracy 0.96). Increasing risk score strongly associated with r-UF (range 3% to 38% for a low and high risk score, p< 0.001) and LOS (7 vs 9 days, p= 0.005), with no impact in 1-year mortality (p= 0.2) or readmission rates (p= 0.3), respectively.
Conclusions
The risk for the need of rescue UF in AHF hospitalizations can be simply assessed using readily available information at admission, with a high and internally validated performance model. Introduced risk score was linked to length of stay, not to mortality or readmission rates.Rescue UF risk scoreModel discrimination. ROC Curve