Development and Validation of a Risk Prediction Model for New-Onset Atrial Fibrillation in Sepsis

Figures

• 

  Figure 1. Flow diagram of study design.

• 

  Figure 2. Variable selection for the LASSO regression model. (A) Coefficient profile plot was constructed based on the log(λ) sequence. (B) By deriving the optimal lambda, 5 variables with non-zero coefficients were selected. After validating the optimal λ in the LASSO model, we plotted the partial likelihood deviance (binomial deviance) curve against log(lambda), with dashed vertical lines drawn according to the 1-standard error criterion.

• 

  Figure 3. Independent risk factors-HR, BUN, and Log IL-6 for nomogram prediction model. Annotation: The red circle in the figure indicates the scoring process for an example patient: Draw a vertical line downward from the first red dot on the “Points” axis to intersect with the red dot on the “BUN” axis, indicating that when BUN = 5, the Points is −0.43; Draw a vertical line downward from the second red dot on the “Points” axis to intersect with the red dot on the “Log IL-6” axis, indicating that when Log IL-6 = 0.47, the corresponding Points is 1.56; Draw a vertical line downward from the third red dot on the “Points” axis to intersect with the red dot on the “HR” axis, indicating that when HR = 131.25, the corresponding points is 2.06; Sum up the points of the three variables: the Total points = 3.19. Draw a vertical line downward from the total points to intersect with the “Risk of atrial fibrillation” axis, showing that a total points of 3.19 corresponds to a predicted risk of atrial fibrillation of 81.11%.

• 

  Figure 4. ROC of the NF risk nomogram prediction. (A) The blue line represents the performance of the nomogram in the training cohort. (B) The red line represents the performance of the nomogram in the validation cohort.

• 

  Figure 5. Calibration curve of risk of AF. The y-axis represents actual diagnosed cases of AF, the x-axis represents the predicted risk of AF. (A) training cohort; (B) validation cohort.

• 

  Figure 6. DCA curves and CIC of the nomogram prediction model for the risk of AF in sepsis patients. (A) DCA curves for training cohort; (B) DCA curves for validation cohort; (C) CIC for training cohort; (D) CIC for validation cohort.