Model for end-stage liver disease and pneumonia: An improved scoring model for critically ill cirrhotic patients with pneumonia

Background/aims Critically ill patients with cirrhosis with pneumonia are at an increased risk for mortality. Only a few accurate predictive models are existing specific to these patients. The aim of the present study was to compare the existing prognostic models and to develop an improved mortality risk model for patients with cirrhosis and pneumonia. Materials and methods A total of 231 patients were enrolled in our study (70% training and 30% validation cohorts). All participants were followed up for at least 21 days. Model for End-stage Liver Disease and Pneumonia (MELD-P) was derived by the Cox proportional hazards model. The performances of prognostic scoring systems were compared by calculation of the area under the receiver operating characteristic (AUROC) curve. Results MELD-P showed better discriminative capabilities than existing scoring systems. Four clinical variables, including loge bilirubin (hazard ratio (HR) 1.29, 95% confidence interval (CI) 1.01-1.73), loge international normalized ratio (HR 3.57, 95% CI 1.30-9.78), loge pulse oxygen saturation/fraction of inspired oxygen (HR 0.38, 95% CI 0.14-0.99), and vasopressors used (HR 3.72, 95% CI 1.85-7.49), were considered as independent prognostic values associated with 21-day mortality. MELD-P had AUROC curve values of 0.78 (95% CI 0.71-0.84) in predicting in-hospital mortality, 0.78 (95% CI 0.70-0.84) at 21-day, 0.88 (95% CI 0.82-0.93) at 14-day, and 0.87 (95% CI 0.81-0.92) at 7-day. A similar result was obtained in validation cohort. Conclusion MELD-P, as the first model specifically designed to evaluate the risk of mortality in critically ill patients with cirrhosis and pneumonia, performs well on the mortality assessment of short-term mortality.