Efficacy and validation of a clinical model to predict acute kidney injury in severe pneumonia requiring mechanical ventilation in the elderly patients: A multi center retrospective observational analysis

Li Yao^1*Jingjing Zhao¹Xiang Fang¹Di Ma²Wenjing Ding³Ting Chen¹Jingyu Li⁴Yu Fu⁵Yuan Zhan¹Gaoqiang Ling¹Wei Wang⁶

• ¹Intensive Care Unit, The Second People’s Hospital of Hefei, Hefei Hospital Affiliated to Anhui Medical University, Hefei, China
• ²Intensive Care Unit, First Affiliated Hospital of Anhui University of Chinese Medicine, Hefei, China
• ³The Second Clinical Medical School, Anhui University of Chinese Medicine, Hefei, China
• ⁴Institute of Artificial Intelligence, Hefei Comprehensive National Science Center, Hefei, China
• ⁵School of Medicine, Xian Jiaotong University, Xi'an, China
• ⁶Department of Colorectum, First Affiliated Hospital of Anhui University of Chinese Medicine, Hefei, China

Background The objective of our retrospective multi center analysis was to identify risk factors and construct a statistical model for predicting acute kidney injury (AKI) among patients who already have severe pneumonia requiring mechanical ventilation (SPRMV) in the elderly patients with in different intensive care units (ICU). AKI SPRMV Methods We utilized a multi-center retrospective analysis, including 353 cases of SPRMV patients diagnosed and treated in the ICU of the Hefei Second People's Hospital between May 2018 and February 2025 as training dataset, and 151 participates were admitted the ICU of the First Affiliated Hospital of Anhui Medical University between June 2020 and March 2025 considered as validation dataset. Both univariate and multivariate logistic regression analysis were utilized to investigate the risk factors of SPRMV with AKI. After that, our predictive model was evaluated by nomogram, receiver operating characteristic (ROC) curve for discrimination of the predictive model, calibration curves and decision curve analysis (DCA) curves for clinical validity. Results Our multivariate logistic regression analysis indicated that CREA, SOFA, APACHE II, driving pressure, mechanical kinetic energy, CRP/ALB and MAP are independent risk factors of SPRMV in the elderly patients with AKI. A nomogram of SPRMV in the elderly patients with AKI was constructed. The ROC curve revealed that our predictive model showed the good predictive efficacy with an area under curve (AUC) of 0.920 (95% confidence intervals (CI) = 0.892-0.948) with a specificity of 0.993 and sensitivity of 0.763 in training dataset, and an AUC value of 0.938 (95%CI = 0.899-0.977) with a specificity of 0.952 and sensitivity of 0.854 in validation dataset. Moreover, calibration and DCA curves demonstrated that our predictive model had a good fit, better net benefit and predictive efficiency of SPRMV in the elderly patients with AKI. Conclusions Our predictive model demonstrated that CREA, SOFA, APACHE II, driving pressure, mechanical kinetic energy, CRP/ALB and MAP were the independent risk factors of AKI in SPRMV in the elderly patients with high accuracy and good calibration.