Age-specific predictive modelling of sepsis mortality: The diagnostic utility of CRP and serum creatinine.

BACKGROUND

Sepsis remains a leading cause of global mortality, with culture-negative cases posing major challenges to diagnosis and risk stratification. This study explores the utility of commonly used biomarkers in predicting sepsis-induced mortality and evaluates the performance of age-stratified binary logistic regression model.

METHODS

A prospective observational study was conducted over one year, including 960 patients with suspected sepsis. Clinical parameters including procalcitonin (PCT), C-reactive protein (CRP), Total Leucocyte count (TLC), blood culture & other relevant variables related to sepsis were recorded. Binary logistic regression models were developed separately for each age group (new-born/infants, adults, older adults) to predict the sepsis-induced mortality.

RESULTS

A statistically significant logistic regression model was achieved only in the new-born/infant group. CRP levels ≥100 mg/L emerged as a strong and independent predictor of mortality (AUC = 0.80, p = 0.02). Although serum creatinine demonstrated a significant association in univariate analysis, it did not retain significance in the multivariate model. PCT levels and blood culture positivity were not significant predictors of mortality in any age group. In adults and older adults, model performance was poor, showing a strong class imbalance and limited sensitivity.

CONCLUSIONS

The strong predictive value of CRP in the new-born/infant group serves as a revalidation of its clinical utility in sepsis, particularly in the context of the poor performance of other commonly used biomarkers such as PCT and TLC. The results also highlight the importance of age-specific modelling and reinforce the need for a multi-marker approach in sepsis diagnostics.