Nomogram Model to Predict Acute Kidney Injury in Hospitalized Patients with Heart Failure

Abstract

Background: Acute kidney injury (AKI) is a common complication of acute heart failure (HF) that can prolong hospitalization time and worsen the prognosis. The objectives of this research were to ascertain independent risk factors of AKI in hospitalized HF patients and validate a nomogram risk prediction model established using those factors. Methods: Finally, 967 patients hospitalized for HF were included. Patients were randomly assigned to the training set (n = 677) or test set (n = 290). Least absolute shrinkage and selection operator (LASSO) regression was performed for variable selection, and multivariate logistic regression analysis was used to search for independent predictors of AKI in hospitalized HF patients. A nomogram prediction model was then developed based on the final identified predictors. The performance of the nomogram was assessed in terms of discriminability, as determined by the area under the receiver operating characteristic (ROC) curve (AUC), and predictive accuracy, as determined by calibration plots. Results: The incidence of AKI in our cohort was 19%. After initial LASSO variable selection, multivariate logistic regression revealed that age, pneumonia, D-dimer, and albumin were independently associated with AKI in hospitalized HF patients. The nomogram prediction model based on these independent predictors had AUCs of 0.760 and 0.744 in the training and test sets, respectively. The calibration plots indicate a strong concordance between the estimated AKI probabilities and the observed probabilities. Conclusions: A nomogram prediction model based on pneumonia, age, D-dimer, and albumin can help clinicians predict the risk of AKI in HF patients with moderate discriminability.

Keywords

• heart failure
• acute kidney injury
• nomogram
• risk prediction