Development and validation of a clinical prediction model for sepsis-induced cardiomyopathy

Figures

• 

  Figure 1. Flowchart of patient selection. This diagram illustrates the inclusion and exclusion process for patients enrolled in the SICM study.1: Total ICU Admissions (94,458), 2: Age ≥18 years (94,458), 3: Meet Sepsis Criteria (41,295), 4: Not Meet Sepsis Criteria (53,163), 5: Missing Data (32,318), 6: Complete Data (8,977), 7: CAD/Cardiomyopathy (7,163). 8: No CAD/Cardiomyopathy (1,814), 9: EF < 50% (482), 10: EF ≥ 50% (1,332).

• 

  Figure 2. Comparison of baseline characteristic distributions before and after propensity score matching (PSM). The histogram (or density plot) demonstrates the balance of baseline characteristics between the SICM group (treatment/exposed group) and the non-SICM group (control group) before and after PSM. After matching, the distributions in the matched cohort (matched treatment and matched control groups) were more balanced, as evidenced by improved overlap in the histograms (SICM, sepsis-induced cardiomyopathy; PSM, propensity score matching).

• 

  Figure 3. Selection of predictors using LASSO regression. (A) Displays the coefficient path plot of the LASSO regression, and (B) Presents the results of 10-fold cross-validation for variable selection.

• 

  Figure 4. Nomogram for predicting sepsis-induced cardiomyopathy. The nomogram incorporates predictors recorded within 24 h of ICU admission: serum phosphate level (phosphate_icu24h), neutrophil percentage (neu_icu24h), troponin level (troponin_icu24h), heart rate (heartrate_icu24h), and Charlson Comorbidity Index (charlson). SICM, sepsis-induced cardiomyopathy.

• 

  Figure 5. ROC curves for the prediction model. The ROC curves are presented for the (A) derivation cohort, (B) internal validation cohort, and (C) external validation cohort, demonstrating the discriminative performance of the model incorporating independent risk variables.

• 

  Figure 6. Calibration curves for the prediction model. The plots depict the agreement between predicted probabilities and observed frequencies of sepsis-induced cardiomyopathy in the (A) derivation cohort, (B) internal validation cohort, and (C) external validation cohort.

• 

  Figure 7. Decision curve analysis for predicting SICM. The net benefit of the prediction model is plotted across a range of threshold probabilities for the (A) derivation cohort, (B) internal validation cohort, and (C) external validation cohort, demonstrating its potential clinical utility.