Routine Troponin I Assessment Enhances Risk Stratification in Hospitalized Patients with Seasonal Influenza

Figures

• 

  Figure 1. Baseline characteristics and distribution of troponin I in patients with influenza. (A) Incidence of the composite primary adverse endpoint (death, ICU admission, or mechanical ventilation) among influenza patients. During the 30-day follow-up period, 60 individuals (21.7%) experienced a composite primary adverse event. (B) Distribution of influenza A and B subtypes among the patient cohort. (C) Log10-transformed distribution of high-sensitivity troponin I (hsTnI) levels at admission. Histogram colors indicate increasing hsTnI concentrations, from low (blue) to high (yellow). Myocardial injury was defined as hs-TnI > 99th percentile (>37 ng/L for women and >57 ng/L for men). (D) Baseline hsTnI levels across did not differ among influenza virus variants (e.g., H1N1, H3N2, influenza B). (E) Correlation matrix displaying the association of hsTnI levels with key clinical parameters. INR, International Normalized Ratio; NT-proBNP, N-terminal pro–B-type natriuretic peptide; CK, creatine kinase; LDH, lactate dehydrogenase; AST, aspartate aminotransferase; ALT, alanine aminotransferase; LVEF, left ventricular ejection fraction. (F) Distribution of critical laboratory markers (e.g.,…

• 

  Figure 2. Association of hsTnI elevation with clinical outcomes in influenza patients. (A) Kaplan–Meier survival curves demonstrating an increased incidence of the composite endpoint in influenza patients with elevated hsTnI (p < 0.0001). (B) Quantitative analysis of hsTnI concentrations reveals increased myocardial injury in patients who experienced adverse events during follow-up. (C) Longitudinal analysis of hsTnI in patients with serial measurements (n = 68), demonstrating dynamic changes over time and hinting at acute myocardial injury (>20% change) in patients with influenza and adverse events. (D) A significant (p = 0.026) decline in hsTnI concentrations was observed among influenza patients with adverse outcomes indicating a temporal pattern of myocardial injury contributing to poor prognosis.

• 

  Figure 3. Prognostic relevance of isolated troponin I elevation in influenza. (A) Nelson–Aalen cumulative hazard curves for the composite endpoint in patients with elevated vs. normal hsTnI and in those with dynamic vs. stable hsTnI profiles. Indidivual Cox proportional hazards were adjusted for age, gender, creatinine, and CRP. Log-rank test indicates significant group differences (χ2 = 7.8, p < 0.05) with a critically enhanced risk of influenza patients with elevated hsTnI. (B) Predominant detection of elevated hsTnI among those patients who died, required mechanical ventilation, or were admitted to the ICU, supporting hsTnI was associated with specific adverse outcomes. (C) ROC curve of multivariable regression analysis identifying hsTnI as the only significant (p < 0.0001) predictor of the composite endpoint after adjustment for age, gender, influenza subtypes, cardiovascular risk factors, and comorbidities.

• 

  Figure 4. Prognostic modeling of adverse outcomes in influenza using hsTnI-based machine learning. (A) Feature importance ranking in the XGBoost model, identifying hsTnI as the strongest predictor of the composite endpoint. (B) SHapley Additive exPlanations (SHAP) plot visualizing the impact of individual predictors. Positive SHAP values indicate an increased likelihood for the composite endpoint, while negative SAHP values indicate lower likelihood of the outcome, with hsTnI showing the strongest association with adverse outcomes. (C) Receiver operating characteristic (ROC) analysis of the XGBoost algorithm to predict the composite endpoint by integration of laboratory data, including hsTnI. The supervised cross-validated machine learning model demonstrated a high diagnostic accuracy based on area under the curve (AUC). The model was trained and validated using a 10-fold cross-validation loop. ROC curves demonstrating excellent predictive performance of the XGBoost model in both training (AUC = 1.00) and test (AUC = 0.99) cohorts. (D) Based on the cross-validated model, patients were stratified according to the transformed likelihood for adverse outcomes utilizing the TROP-ID risk…