A multifactor model for predicting mortality in critically ill patients: A multicenter prospective cohort study.

PURPOSE: The objective of this study was to develop a model using a combination of routine clinical variables to predict mortality in critically ill patients.

METHODS: A cohort of 500 patients recruited from eight university hospital intensive care units (ICUs) was used to develop a model via logistic regression analyses. Discrimination and calibration analyses were performed to assess the model.

RESULTS: The model included the lactate level (odds ratio [OR]=1.11, 95% confidence interval [CI] 1.01 to 1.22, P=0.029), neutrophil-to-lymphocyte ratio (OR=1.03, 95% CI 1.01 to 1.04, P=0.002), acute physiology score (OR=1.11, 95% CI 1.06 to 1.15, P<0.001), Charlson comorbidity index (OR=1.36, 95% CI 1.15 to 1.60, P<0.001) and surgery type (OR: selective=Ref, no surgery=8.04, 95% CI 3.74 to 17.30, P<0.001, emergency=3.66, 95% CI 1.60 to 8.36, P=0.002). The model showed good discrimination (area under receiver operating characteristic curve: 0.84, 95% CI: 0.80 to 0.87) and calibration (Hosmer-Lemeshow test P=0.137) for predicting in-hospital mortality.

CONCLUSION: The developed multifactor model can be used to effectively predict mortality in critically ill patients at ICU admission.