BACKGROUND RDW is a prognostic biomarker and associated with mortality in cardiovascular disease, stroke and metabolic syndrome. For elderly patients, malnutrition and multiple comorbidities exist, which could affect the discrimination ability of RDW in sepsis. The main purpose of our study was to evaluate the prognostic value of RDW in sepsis among elderly patients. METHODS This was a retrospective cohort study conducted in emergency department intensive care units (ED-ICU) between April 2015 and November 2015. Elderly patients (≥65years old) who were admitted to the ED-ICU with a diagnosis of severe sepsis and/or septic shock were included. The demographic data, biochemistry data, qSOFA, and APACHE II score were compared between survivors and nonsurvivors. RESULTS A total of 117 patients was included with mean age 81.5±8.3years old. The mean APACHE II score was 21.9±7.1. In the multivariate Cox proportional hazards model, RDW level was an independent variable for mortality (hazard ratio: 1.18 [1.03-1.35] for each 1% increase in RDW, p=0.019), after adjusting for CCI, any diagnosed malignancy, and eGFR. The AUC of RDW in predicting mortality was 0.63 (95% confidence interval [CI]: 0.52-0.74, p=0.025). In subgroup analysis, for qSOFA <2, nonsurvivors had higher RDW levels than survivors (17.0±3.3 vs. 15.3±1.4%, p=0.044). CONCLUSIONS In our study, RDW was an independent predictor of in-hospital mortality in elderly patients with sepsis. For qSOFA scores <2, higher RDW levels were associated with poor prognosis. RDW could be a potential parameter used alongside the clinical prediction rules.