Development and validation of a risk prediction model for hypoproteinemia after adult cardiac valve surgery: implications for clinical care

Figures

• 

  Figure 1. Results of LASSO regression analysis. (A) Illustrates the path of standardized coefficients for different variables in LASSO regression as the regularization parameter λ varies. The x-axis represents the logarithmic scale of the regularization parameter λ, while the y-axis shows the range of standardized coefficient values from −0.5 to 0.5. Each line depicts the trajectory of a variable’s coefficient; when λ is small (close to 0.01), all 24 coefficients are non-zero. As λ increases, some coefficients are shrunk to zero, demonstrating the sparsity characteristic of LASSO regression. (B) Illustrates the cross-validation error as a function of the regularization parameter λ. The x-axis represents the logarithmic scale of the regularization parameter λ, while the y-axis shows the cross-validation function value, indicating the model’s performance on the validation set. Two key points are annotated: λCV: The value of λ corresponding to the minimum cross-validation error, which is λ = 0.01, associated with 24 non-zero coefficients. λSE: The value of λ selected based on the standard error rule, which is λ = 0.081, associated with 9 non-zero coefficients.The shaded area indicates the…

• 

  Figure 2. Nomogram model for the risk of postoperative hypoproteinemia in subjects. This nomogram integrates various clinical factors to predict the risk of postoperative hypoproteinemia. Variables included in the model are hypertension, body mass index (BMI), red blood cell count at admission, platelet count at admission, albumin level at admission, chest infection, pre-operative heart failure, and hemodilution on cardiopulmonary bypass (CPB). Each variable is assigned a score based on its contribution to the risk, and the total score is calculated by summing the individual scores. The total score is then used to estimate the risk of postoperative hypoproteinemia. Specifically, a vertical line drawn upward from the specific value on the “Total score” axis corresponds to the precise risk value of postoperative hypoproteinemia on the “Risk of postoperative hypoproteinemia” scale, with higher scores indicating a higher risk. This model provides a quantitative tool for clinicians to assess and manage the risk of hypoproteinemia in surgical patients.

• 

  Figure 3. DCA curve of the nomogram model. This figure presents the decision curve analysis (DCA) of the nomogram model for predicting postoperative hypoproteinemia. The DCA curve compares the net benefit of using the nomogram model to the strategies of treating all patients or treating none. The x-axis represents the threshold probability, which is the probability at which a patient would choose treatment. The y-axis shows the net benefit, which is the difference between the true positive rate and the false positive rate, weighted by the threshold probability. The nomogram model demonstrates a higher net benefit compared to the “treat all” and “treat none” strategies across a range of threshold probabilities, indicating its clinical utility in guiding treatment decisions for postoperative hypoproteinemia.

• 

  Figure 4. ROC curve and calibration curve of the nomogram model. (A) The receiver operating characteristic (ROC) curve illustrates the diagnostic accuracy of the nomogram model for predicting postoperative hypoproteinemia. The area under the curve (AUC) is 0.823 (95% CI [0.774–0.873]), indicating good discrimination ability. (B) This curve compares the predicted probabilities of postoperative hypoproteinemia with the observed outcomes before internal verification. The Hosmer–Lemeshow chi-square test (Chi-squared = 5.190, P-value = 0.737) and Brier score (0.170) suggest good calibration. The Lowess smoothing line (solid) and 95% confidence intervals (dashed) are shown. (C) This curve compares the predicted probabilities with the observed outcomes after internal verification. The calibration metrics include an expected-to-observed (E:O) ratio of 1.001, a slope of 0.861, and a calibration-in-the-large (CITL) value of −0.003. The Lowess smoothing line (solid) and 95% confidence intervals (dashed) are shown. The calibration curves demonstrate the agreement between predicted and observed risks, with the nomogram showing good overall performance (Brier score scaled (%) = 24.8) and discrimination…