Development and internal validation of a prediction model for early identification of sepsis-associated acute kidney injury based on admission serum biomarkers: a retrospective cohort study.

OBJECTIVE

Sepsis-associated acute kidney injury (SA-AKI) is a critical complication that substantially increases intensive care unit mortality. Early identification is paramount for timely intervention. This study aimed to develop and internally validate a prediction model relying exclusively on the first serum laboratory indicators after hospital admission to predict SA-AKI risk at the earliest available laboratory assessment.

METHODS

Clinical data of 1,573 sepsis patients admitted to West China Hospital of Sichuan University (January 2024-December 2025) were retrospectively analyzed. Patients were divided into SA-AKI and non-SA-AKI groups per 2012 KDIGO criteria, and randomly split into training (70%, n = 1,102) and validation (30%, n = 471) cohorts. LASSO regression screened potential predictors exclusively from first serum laboratory indicators after admission, and multivariate logistic regression constructed the model. A nomogram was constructed to facilitate individualized risk estimation. Subgroup analyses were performed across age and sex strata. Model performance was evaluated via ROC curves, calibration curves, and decision curve analysis (DCA).

RESULTS

LASSO regression selected 10 serum indicators, and multivariate logistic regression confirmed 8 independent risk factors: myoglobin (MYO), alanine aminotransferase (ALT), phosphorus (PO4), sodium (Na), potassium (K), carbon dioxide combining power (CO2-CP), platelet (PLT), and neutrophil (NEUT) (all P < 0.05). The model exhibited favorable discrimination (training cohort AUC = 0.839 [95% CI: 0.814-0.864]; validation cohort AUC = 0.832 [95% CI: 0.791-0.874]), acceptable graphical calibration, and significant clinical utility across a wide threshold probability range. Subgroup analyses showed stable discriminative performance across age and sex strata (all AUCs > 0.81).

CONCLUSIONS

This study developed and internally validated a promising predictive model for estimating SA-AKI risk in sepsis patients using solely first routine serum laboratory indicators after admission. A nomogram is provided for individualized bedside risk estimation. This tool may support early risk stratification of high-risk individuals. External validation in multi-center, diverse cohorts is warranted before broader clinical implementation.