Objective:To compare the predictive ability of Sequential Organ Failure Assessment(SOFA)scores for the prognosis of sepsis between elderly and non-elderly patients.Methods:Medical information of patients with sepsis based on the diagnostic criteria for sepsis were screened from the database of critical care units of the Beth Israel Deaconess Medical Center in the United States during 2001 to 2012.Hospital mortality was chosen as the primary outcome for the prognosis of sepsis, and receiver operating characteristics(ROC)curve analysis was performed to compare the differences in predictive ability of SOFA scores for the prognosis of sepsis between the elderly(≥60 years)and the non-elderly on admission.Parameters related to age, heart rate and body temperature were added into the SOFA scoring systems, and the differences in predictive ability of SOFA scores versus the modified SOFA scores for sepsis were compared by using ROC curve analysis.Multivariable regression models were used to assess the associations of the modified SOFA scores with clinical outcomes in the elderly sepsis patients.Results:A total of 10 206 patients with sepsis were included in this study, including 6 928(67.88%)elderly patients and 3 278(32.12%)non-elderly patients.ROC curve analysis showed that the power of admission SOFA score for predicting hospital mortality in elderly sepsis patients(AUC=0.67, 95% CI: 0.65-0.68)was lower than in non-elderly sepsis patients(AUC=0.75, 95% CI: 0.72-0.78). AS compared with the SOFA scores, the modified SOFA scores after increasing the parameters related to age, heart rate and body temperature significantly improved the predictive power for hospital mortality of sepsis patients, only in elderly(AUC: 0.67 vs.0.69, P<0.01), but not in non-elderly patients(AUC: 0.75 vs.0.75, P=0.77). After adjusting for potential confounders and in elderly sepsis patients, multivariate regression analysis showed that the modified SOFA scores were significantly associated with multiple clinical outcomes, including hospital mortality( OR=1.17, 95% CI: 1.15-1.19, P<0.01), 28-day mortality( HR=1.12, 95% CI: 1.10-1.13, P<0.01), 90-day mortality( HR=1.10, 95% CI: 1.08-1.11, P<0.01). Conclusions:The predictive ability of SOFA scores for the prognosis of sepsis is lower in elderly patients than in non-elderly patients.The predictive value of SOFA scores for the prognosis of elderly sepsis patients can be improved by increasing parameters related to age, heart rate and body temperature.