A novel model for identifying infections in patients with acute-on-chronic liver disease (AoCLD): A nationwide, multicenter, prospective cohort study
Hui Zhou, Hai Li, Guohong Deng, Xianbo Wang, Xin Zheng, Jinjun Chen, Zhongji Meng, Yubao Zheng, Yanhang Gao, Zhiping Qian, Feng Liu, Xiaobo Lu, Yu Shi, Jia Shang, Yan Huang, Ruochan ChenAbstract
Background and aims
To establish an early and quick model for diagnosing infections in patients with acute-on-chronic liver disease (AoCLD).
Approach
This study analyzed 3,949 patients from two multicenter prospective cohorts of the Chinese Acute-on-Chronic Liver Failure (CATCH-LIFE) study. The dataset was randomly divided into training and validation cohorts in a 7:3 ratio. In the training cohort, logistic regression and least absolute shrinkage and selection operator regression analyses were used to identify predictive risk factors for infection in patients with AoCLD, and a simple nomogram was established. Two different cutoff values were determined to stratify infection risk in AoCLD patients.
Results
The developed diagnostic model included six variables: cirrhosis, ascites, neutrophil count (N), and total bilirubin, C-reactive protein (CRP), and blood sodium levels. The area under the receiver operating characteristic curve for the training and validation cohorts were 0.818 and 0.809, respectively, significantly higher than using CRP, procalcitonin, or N alone. Additionally, in the training cohort, we set a low cutoff value of 0.2028, resulting in a sensitivity of 80.15%, specificity of 68.25%, and a negative predictive value of 92.7% for rule-out diagnosis. A high cutoff value of 0.4045 resulting in a specificity of 90.1%, sensitivity of 52.7%, and a positive predictive value of 64% for rule-in diagnosis. These cutoff values were validated in the validation cohort.
Conclusions
We established a nomogram model to assist clinicians in diagnosing infections in patients with AoCLD, effectively improving the accuracy and timeliness of diagnosis.