# Development and validation of a prediction model for in-hospital mortality in intensive care unit patients with cirrhosis and sepsis: a multicentre retrospective cohort study

## TL;DR

This study created a model to predict in-hospital deaths for ICU patients with cirrhosis and sepsis, using data from multiple hospitals to train and validate it.

## Contribution

A novel prediction model for ICU mortality in cirrhosis and sepsis patients, validated across multiple databases and time periods.

## Key findings

- The model achieved an AUROC of 0.822 in the training cohort and 0.810 in the temporal validation cohort.
- External validation showed AUROCs of 0.777 in eICU-CRD and 0.763 in SZPH-ICU.
- The model outperformed existing prognostic scores for mortality prediction in this patient group.

## Abstract

This study aimed to develop and validate a novel prediction model for in-hospital mortality among patients with cirrhosis and sepsis admitted to the intensive care unit (ICU).

Data were obtained from three multicentre databases: the Medical Information Mart for Intensive Care IV (MIMIC-IV v3.1), the eICU Collaborative Research Database (eICU-CRD v2.0), and the Shenzhen People's Hospital ICU (SZPH-ICU). The MIMIC-IV cohort was chronologically divided into a training set (2008–2016) and a temporal validation set (2017–2022), whereas the eICU-CRD and SZPH-ICU cohorts were used for external validation. Variable selection was performed using the least absolute shrinkage and selection operator (LASSO) regression. A multivariable logistic regression model was constructed and visualized as a nomogram. Model performance was assessed using the area under the receiver operating characteristic curve (AUROC), Brier score, calibration plots, and decision curve analysis. A web-based calculator was developed to facilitate clinical implementation.

A total of 2,052 adult ICU patients with cirrhosis and sepsis from the MIMIC-IV database were included. The training cohort (2008–2016; n = 1,328) had a 24.0% in-hospital mortality rate, whereas the temporal validation cohort (2017–2022; n = 724) had a 35.9% in-hospital mortality rate. In the external validation cohorts, in-hospital mortality was 25.9% in the eICU-CRD (n = 657) and 38.2% in the SZPH-ICU (n = 131). The final model comprised 13 predictors: age, respiratory rate, body temperature, oxygen saturation, heart rate, total bilirubin, lactate, creatinine, white blood cell count, international normalized ratio (INR), vasopressor use, urine output, and the Glasgow Coma Scale (GCS) score. The model achieved an AUROC of 0.822 (95% confidence interval [CI]: 0.797–0.847) in the training cohort and 0.810 (95% CI: 0.777–0.843) in the temporal validation cohort. External validation yielded AUROCs of 0.777 (95% CI: 0.734–0.821) in the eICU-CRD cohort and 0.763 (95% CI: 0.680–0.846) in the SZPH-ICU cohort. The proposed model demonstrated superior discriminative performance compared with existing prognostic scores.

This validated multivariable prediction model accurately estimates in-hospital mortality in ICU patients with cirrhosis and sepsis, supporting early risk stratification and more efficient allocation of clinical resources.