Heart valve disease is one of the important factors leading to heart failure and cardiovascular death. Double valve replacement (DVR) and tricuspid valve plasty (TVP) have become important surgical approaches for treating severe valve lesions. However, combining with other surgeries may increase perioperative risk and have an impact on the long-term prognosis of patients. To address the complexity of postoperative outcomes in cardiac surgery, this study employs a combination of traditional statistical methods and machine learning techniques to assess risk factors. The primary aim was to investigate the effects of DVR + TVP and combined surgery on postoperative survival and adverse outcomes. Patients who underwent DVR + TVP surgery, they were divided into 4 groups: DVR + TVP, MAZE + TVP, coronary artery bypass grafting (CABG) + TVP, or ascending aortic surgery (AAS) + TVP. Kaplan–Meier survival analysis was used to evaluate the impact of different surgical approaches on postoperative survival rate, Cox proportional hazards regression model was used to analyze contribution of postoperative complications and reoperation to the mortality risk. A neural network model was used to identify factors affecting postoperative mortality risk of patients, to evaluate role of perioperative biomarkers in predicting postoperative mortality risk. The survival rate of patients in AAS + TVP group was the lowest (2.5%), while that in TVP group was the highest (78.8%). Postoperative complications and reoperation were independent predictors of postoperative death. The mortality risk of patients with complications was 2.164 times that of patients without complications (hazard ratio (HR) = 2.164, 95% confidence interval (CI): 1.275–3.671, P = .004), underwent reoperation had a 2.6-fold increased risk of mortality (HR = 2.599, 95% CI: 1.221–5.532, P = .013). Postoperative biomarkers (lactate dehydrogenase (LDH), D-dimer) were significantly associated with postoperative mortality risk. When using neural network model to evaluate the postoperative mortality risk, age (2.0783) and length of stay in the intensive care unit (ICU) (2.0135) were the most important predictors, the area under the curve value of the model was 0.79. Different surgical approaches have a significant impact on postoperative survival rate and the incidence of complications in patients undergoing DVR + TVP. Complications and reoperation are independent factors for poor prognosis. Perioperative biomarkers (LDH, D-dimer) have important value in predicting postoperative mortality risk. The machine learning model based on neural networks can effectively predict postoperative adverse outcomes.
Li et al. (Fri,) studied this question.