A study on the factors influencing mortality risk in sepsis-induced acute kidney injury based on analysis of the MIMIC database

Chongyang Ye, Chunyan Zhu, Shijing Hu, Yulin Mei, Tianjun YangClinical 2025 · DOI

Abstract

Sepsis-induced acute kidney injury (SA-AKI) significantly increases mortality and healthcare burdens. Identifying key mortality risk factors is crucial for improving patient outcomes. This study aims to identify the primary factors affecting mortality in SA-AKI patients using the MIMIC-III database. A retrospective analysis was conducted on 4,868 SA-AKI patients. Clinical data from the first 24 h of ICU admission were analyzed using logistic regression to identify mortality predictors. Key mortality predictors included advanced age (OR = 1.015), severe AKI stages (OR = 1.470), low serum albumin (OR = 0.606), delayed antibiotics (OR = 1.001), high AST (OR = 1.035), and bilirubin (OR = 1.055). The area under the curve (AUC) of the combined predictors for mortality risk was 0.796, indicating high predictive accuracy. Conclusions: Early intervention and monitoring of identified risk factors such as age, AKI stage, albumin levels, and antibiotic timeliness can enhance survival rates in SA-AKI patients.

Introduction

Sepsis is a life-threatening systemic inflammatory response that frequently leads to multiple organ failure. Among its most prevalent and lethal complications is Sepsis-Induced Acute Kidney Injury (SA-AKI), which is associated with prolonged hospitalization and high mortality rates. While individual biomarkers for AKI are well-documented, there is a critical need for integrated models that combine clinical, biochemical, and treatment-related variables to predict outcomes. Specifically, clinicians require better tools to identify patients at risk of "late death"—mortality occurring after the initial 48 hours of ICU admission, often due to secondary infections or progressive organ failure. This study utilizes the MIMIC-III database to establish a multifactorial predictive framework for SA-AKI mortality.

Methods and Patient Cohort

This retrospective study utilized the Medical Information Mart for Intensive Care (MIMIC-III) v1.4 database, which contains de-identified data from over 40,000 patients. The initial cohort consisted of 7,359 patients diagnosed with sepsis according to ICD-9 codes.

Patient selection flowchart illustrating the derivation of the final study cohort (n=4,868) from the initial MIMIC-III sepsis population.

As shown in the selection process, patients were excluded if they did not meet the KDIGO criteria for AKI, were under 18 years of age, or had an ICU stay of less than 48 hours. The latter exclusion was designed to focus the analysis on factors contributing to sustained risk and late mortality. After accounting for 58 patients lost to follow-up, the final analysis included 4,868 patients, of whom 956 (19.6%) died.

Statistical analysis involved univariate screening followed by Cox and logistic regression to identify independent risk factors. Bayesian multiple imputation was employed to address missing laboratory values, and the model's performance was evaluated using the C-statistic and Area Under the Receiver Operating Characteristic (ROC) curve.

Key Results

The study identified several independent predictors of mortality. Advanced age (HR = 1.015 per year) and severe AKI (Stage 2, HR = 1.470) were significant baseline risks. Biochemical markers also played a crucial role: low serum albumin (HR = 0.606) was strongly associated with increased mortality, while elevated Blood Urea Nitrogen (BUN, HR = 1.016) and total bilirubin (HR = 1.055) were independent predictors of poor outcomes.

ROC curves comparing the predictive performance of the combined multi-factor model (AUC = 0.796) against individual clinical and biochemical predictors.

The integrated model achieved an AUC of 0.796, demonstrating superior discriminatory power compared to individual markers like BUN (AUC 0.674) or AKI Stage (AUC 0.665). Notably, the timing of treatment was a critical factor: each one-hour delay in antibiotic administration was associated with a 0.1% increase in mortality risk.

Discussion: Clinical Insights

The findings highlight the complex interplay between organ systems in sepsis. The significance of AST and bilirubin as independent predictors suggests a "liver-kidney crosstalk" where hepatic dysfunction exacerbates the progression of SA-AKI. Furthermore, the role of hypoalbuminemia likely reflects both nutritional status and the severity of systemic inflammation and capillary leak syndrome.

The quantifiable impact of antibiotic delays reinforces the "Golden Hour" concept in sepsis management, suggesting that even marginal improvements in delivery speed can impact survival. Additionally, the identification of decreased eosinophil counts as a risk factor points toward profound immune dysregulation as a hallmark of high-risk SA-AKI.

Conclusion

This study provides a robust predictive model for mortality in SA-AKI patients, emphasizing the importance of early intervention and multi-organ monitoring. While limited by its retrospective, single-center design, the results offer a foundation for developing AI-driven Clinical Decision Support Systems. Future prospective, multicenter studies are needed to validate these findings and integrate them into real-time ICU management protocols.

Summary updated

Abstract

Sepsis-induced acute kidney injury (SA-AKI) significantly increases mortality and healthcare burdens. Identifying key mortality risk factors is crucial for improving patient outcomes. This study aims to identify the primary factors affecting mortality in SA-AKI patients using the MIMIC-III database. A retrospective analysis was conducted on 4868 SA-AKI patients from the MIMIC-III database. Clinical data from the first 24 h of ICU admission were analyzed using logistic regression to identify mortality predictors. Key mortality predictors included advanced age (OR = 1.015, 95% CI: 1.006-1.024), severe AKI stages (OR = 1.470, 95% CI: 1.285-1.676), low serum albumin (OR = 0.606, 95% CI: 0.506-0.722), delayed antibiotics (OR = 1.001, 95% CI: 1.000-1.002), high AST (OR = 1.035, 95% CI: 1.027-1.083), and bilirubin (OR = 1.055, 95% CI: 1.037-1.083). The area under the curve (AUC) of the combined predictors for mortality risk was 0.796, indicating high predictive accuracy. Conclusions: Early intervention and monitoring of identified risk factors such as age, AKI stage, albumin levels, and antibiotic timeliness can enhance survival rates in SA-AKI patients.
BibTeX
@inproceedings{ye2025study,
  author = {Chongyang Ye and Chunyan Zhu and Shijing Hu and Yulin Mei and Tianjun Yang},
  title = {{A study on the factors influencing mortality risk in sepsis-induced acute kidney injury based on analysis of the MIMIC database}},
  booktitle = {Clinical},
  year = {2025},
  doi = {10.1007/s10238-025-01681-4},
}
Figure