To compare the predictive value of sequential organ failure assessment (SOFA) scores at different time points for hospital mortality of patients in intensive care unit (ICU) and to provide research evidence for the rational selection of SOFA scores in actual clinical work.

Adult ICU patients with a length of hospital stay greater than 72 h were included from the American Critical Care Database. The basic information and related indicators were extracted and SOFA scores at different time points were calculated. Hospital mortality was chosen as the outcome and multivariable logistic regression analysis was performed to assess the associations between SOFA scores at different time points and the outcome. ROC curve analysis was also conducted and the area under the curve was calculated to evaluate their prognostic value.

A total of 11 968 patients were included finally, of which male patients accounted for 56.15% with an average age of (64.75±16.63) years old and a hospital mortality rate of 10.41% (1246/11 968). Multivariable logistic regression analysis showed that SOFA scores at different time points were all closely related to hospital mortality (P<0.0001). ROC curve analysis showed that SOFA scores at different time points had different predictive value for hospital mortality and T₇₂ (AUC=0.7246, 95%CI: 0.7101-0.7391) had the highest AUC.

For adult ICU patients whose length of hospital stay is greater than 72 h, SOFA scores at 72 h after admission may have better prognostic value.