The Models for End-stage Liver Disease as prognostic assessment and risk stratification tools in sepsis: a study based on MIMIC-Ⅳ database

Background: The Model for End-stage Liver Disease (MELD) and its modifications have been used to predict mortality and stratify risk for a variety of non-hepatic diseases with good stability. The purpose of this study was to evaluate the value of the MELD and its modifications in evaluating the prognosis of patients with sepsis. Methods: This study is based on the MIMIC-Ⅳ database. A total of 15,882 patients were included. The correlation between the three models and the mortality rate of patients with sepsis was evaluated, and the optimal cut-off values were calculated. Then, further subgroup analysis was performed to seek better stratification criteria. Finally, stratification was performed according to comorbidities to observe the predictive value of the MELDs in patients with different comorbidities. Results: MELD, MELD including Na (MELD-Na) and MELD excluding INR (MELD-XI) were all independent predictors of in-hospital mortality, and the optimal cut-off values were 22.5, 22.5 and 19.5, respectively. When grouped by cut-off values, high score groups were significantly associated with increased in-hospital mortality. Further subgroup analysis based on lactate revealed that patients with high MELD score and lactate level (> 4 mmol/L) had higher in-hospital mortality. Conclusions: The MELDs can effectively predict the in-hospital mortality of sepsis patients and stratify their risk. The MELDs combined with lactate can provide convenient risk stratification for sepsis patients, thus guiding clinicians to better intervene in sepsis patients at an early stage.