Risk factors analysis and prediction model establishment of acute kidney injury after heart valve replacement in patients with normal renal function

Abstract

Background: The study aimed to develop a risk prediction model through screening preoperative risk factors for acute kidney injury (AKI) after heart valve replacement in patients with normal renal function.

Methods: A total of 608 patients with normal renal function who underwent heart valve replacement from November 2013 to June 2022 were analyzed retrospectively. The Lasso regression was used to preliminarily screen potential risk factors, which were entered into the multivariable logistic regression analysis to identify preoperative independent risk factors for postoperative AKI. Based on the results, a risk prediction model was developed, and traditional and dynamic nomograms were constructed. The risk prediction model was evaluated using receiver operating characteristic (ROC), calibration curve, and decision curve analysis (DCA).

Results: 220 patients (36.2%) developed AKI after surgery. Current smoker, hypertension, heart failure, previous myocardial infarction, cerebrovascular disease, CysC, and NT-proBNP were selected as independent risk factors for AKI. A risk prediction model, a traditional and a dynamic nomogram were developed based on the above factors. The area under the curve (AUC) of the ROC for predicting the risk of postoperative AKI was 0.803 (95% CI 0.769-0.836), with sensitivity and specificity of 84.9% and 63.4%, respectively. The calibration curve slope was close to 1, and the DCA showed that the model produced better clinical benefits when the probability threshold was set at 10%-82%.

Conclusions: We developed a preoperative risk prediction model for AKI after heart valve replacement in patients with normal renal function, which demonstrated satisfactory discrimination and calibration.

Keywords: acute kidney injury; dynamic nomogram; heart valve replacement; normal renal function; prediction model.

Figure 1

Flow chart presenting patient selection for the study. eGFR, estimated glomerular filtration rate.

Figure 2

Confounder identification using lasso technique. (A) Select the optimal parameter (λ) in the LASSO model using 10-fold cross-validation on the basis of minimum criteria. (B) LASSO coefficient profiles of the 23 features. The coefficient profiles are drawn as a function of log (Lambda).

Figure 3

(A) Traditional nomogram of the AKI risk prediction model after heart valve replacement in patients with normal renal function. For instance, a patient was a current non-smoker, with preoperative NT-proBNP 112 pg/ml, CysC 1.15 mg/L, cerebrovascular disease, heart failure, and hypertension, without previous myocardial infarction. The total point was calculated as 83.5, which corresponded to the probability of 70.1%. (B) Dynamic nomogram of the AKI risk prediction model after heart valve replacement in patients with normal renal function. For instance, a patient was a current smoker, with preoperative NT-proBNP 2,609 pg/ml, CysC 1.00 mg/L, without hypertension, heart failure, previous myocardial infarction and cerebrovascular disease. The web page demonstrated that the probability of postoperative AKI was 27.6% (95% CI: 17.7%–40.2%).

Figure 4

The receiver operating characteristic (ROC) curve (A), calibration curve (B), and decision curve analysis (DCA) (C) of the preoperative risk prediction model for AKI after heart valve replacement in patients with normal renal function.