Development and validation of a prediction model for in-hospital mortality in patients with intra-abdominal sepsis: a dual-database study using MIMIC-IV and eICU databases.

OBJECTIVES

To develop and validate a predictive model for assessing in-hospital mortality in patients with intra-abdominal sepsis (IAS), a leading cause of sepsis.

DESIGN

Secondary analysis of two retrospective critical care databases.

SETTING

Data extracted from the Intensive Care Medicine Information Marketplace IV (MIMIC-IV) and the eICU Collaborative Research Database.

PARTICIPANTS

Patients with IAS from MIMIC-IV (2008-2019; 1300 patients, 264 deaths) for model training and internal validation, and eICU (2014-2015; 149 patients, 33 deaths) for external validation.

INTERVENTIONS: Clinical data were used for constructing a predictive model. Variable selection was performed using least absolute shrinkage and selection operator regression, followed by model development with multivariable logistic regression. The model was visualised as a nomogram.

PRIMARY AND SECONDARY OUTCOME MEASURES: The primary outcome was in-hospital mortality. Secondary outcomes were model performance metrics, including the area under the receiver operating characteristic curve (AUC), calibration curves, decision curve analysis and clinical impact curves.

RESULTS

Six predictors (lactate, age, activated partial thromboplastin time, blood urea nitrogen, total bilirubin and platelets) were identified. The predictive model showed good performance with an AUC of 0.795 (95% CI 0.758 to 0.831) in the training set (n=910) and 0.846 (95% CI 0.772 to 0.919) in the external validation set (n=149).

CONCLUSION

A robust predictive model was developed to estimate the risk of in-hospital mortality in patients with IAS. This tool may assist clinicians in enhancing patient management and decision-making.