# Factors influencing postoperative hyperbilirubinemia in valvular heart disease and establishment of a predictive model

## TL;DR

This study identifies risk factors for post-surgery high bilirubin levels in heart valve disease patients and builds a model to predict who is likely to experience this complication.

## Contribution

A novel predictive nomogram model for postoperative hyperbilirubinemia in valvular heart disease patients is developed and validated.

## Key findings

- Surgery type, MELD score, CPB time, and blood transfusion volume are significant risk factors for postoperative hyperbilirubinemia.
- The nomogram model achieved high predictive accuracy with AUCs of 0.901 and 0.904 in modeling and validation groups.
- Calibration curves confirmed strong consistency between predicted and actual outcomes in the model.

## Abstract

To explore the influencing factors of postoperative hyperbilirubinemia (HB) in patients with valvular heart disease (VHD) and establish a predictive model based on these factors.

Clinical data of VHD patients who underwent surgical treatment in Qingdao Cardiovascular Hospital from March 2022 to February 2024 were retrospectively collected. The patients were divided into a modeling group (n = 215) and a validation group (n = 54) in an 8:2 ratio. The modeling group was further divided into an HB group (n = 73) and a non-HB group (n = 142) based on whether HB occurred within one week after surgery. Multivariable logistic regression analysis was used to analyze the risk factors for HB in VHD patients. Risk prediction nomogram models were established using R3.6.1 software, and receiver operating characteristic (ROC) curves and calibration curves were plotted to evaluate the predictive performance and accuracy of the nomogram models.

The results of multivariable logistic regression analysis showed that the type of surgery (OR = 4.959, 95% CI: 2.592–9.487), preoperative MELD score (OR = 4.332, 95% CI: 2.061–9.105), CPB time (OR = 3.851, 95% CI: 1.591–9.321), aortic cross-clamp time (OR = 3.667, 95% CI: 1.521–8.841), intraoperative total blood transfusion volume (OR = 4.125, 95% CI: 1.982–8.586), and mechanical ventilation time (OR = 4.089, 95% CI: 2.000–8.362) were risk factors for the occurrence of HB in the modeling group (P < 0.05). The ROC analysis results showed that the area under the curve (AUC) of the nomogram model for predicting HB in the modeling group and validation group was 0.901 and 0.904, respectively. The calibration curve analysis results showed good consistency between the predicted and actual occurrence of HB in the predictive model, with Hosmer-Lemeshow chi-square statistics of 4.32 and 1.95 and corresponding P-values of 0.821 and 0.199.

The type of surgery, preoperative MELD score, CPB time, aortic cross-clamp time, intraoperative total blood transfusion volume, and mechanical ventilation time are risk factors for the occurrence of HB in patients with VHD. The nomogram model constructed based on these risk factors has good predictive value and accuracy.