[Predictive value of ten scoring systems for 28-day mortality in adult patients with sepsis in intensive care unit]

To evaluate the predictive value of ten commonly used clinical scoring systems for 28-day mortality in adult sepsis patients in the intensive care unit (ICU), and to identify the optimal scoring system for assessing short-term outcomes in adult sepsis patients.

Clinical data were extracted from the Medical Information Mart for Intensive Care-IV (MIMIC-IV, 3.0 database) for sepsis patients aged 18-90 years on their first ICU admission, including demographic information, vital signs, comorbidities, laboratory indicators, assessment scales, and prognostic outcomes. Patients were divided into survival and death groups based on 28-day outcome after ICU admission, and clinical data were compared between the two groups. Lasso regression analysis was used for preliminary screening for variables associated with the prognosis of sepsis patients, and factors affecting the prognosis of sepsis patients were analyzed through univariate and multivariate Logistic regression analysis. The predictive value of ten scoring systems for 28-day mortality in sepsis patients was assessed using receiver operator characteristic curve (ROC curve).

A total of 7 631 adults with sepsis were included in the MIMIC-IV database, of which 5 575 survived for 28 days and 2 056 died. Compared with the survival group, patients in the death group were older, had higher proportions of type 2 diabetes and heart failure, exhibited worse respiratory, circulatory, and metabolic conditions, carried a higher infection risk, and had significantly higher scores on the Sequential Organ Failure Assessment (SOFA), Acute Physiology Score III (APS III), Acute Physiology and Chronic Health Evaluation II (APACHE II), Simplified Acute Physiology Score II (SAPS II), Oxford Acute Severity of Illness Score (OASIS), Charlson comorbidity index (CCI; all P<0.05). Lasso regression analysis was initially applied to screen 49 characteristic variables, identifying 32 that were preliminarily associated with sepsis prognosis. These 32 variables were subsequently included in univariate and multivariate Logistic regression analyses. The results revealed that 18 variables were associated with 28-day mortality in sepsis patients (all P<0.05). Among these, an increase in APS III score [odds ratio (OR)=1.089, 95% confidence interval (95%CI) was 1.081-1.096, P<0.001], SAPS II score (OR=1.052, 95%CI was 1.048-1.056, P<0.001), Glasgow Coma Scale (GCS; OR=0.976, 95%CI was 0.962-0.991, P<0.001), and CCI score (OR=1.195, 95%CI was 1.175-1.215, P<0.001) were identified as independent risk factors for 28-day mortality in sepsis patients. ROC curve analysis demonstrated that among the ten scoring systems evaluated, the APS III and SAPS II scores showed superior predictive performance compared to the others. Their areas under the curve (AUC) with 95%CI were 0.721 (0.708-0.734) and 0.701 (0.687-0.714), respectively.

Among the ten commonly used clinical scoring systems, the APS III score demonstrates a relatively favorable performance in predicting the risk of 28-day mortality in adult sepsis patients. It can provide a reliable basis for timely clinical intervention, thereby reducing patient mortality risk.