Background The appropriateness of antibiotics is the basis for improving the survival of patients with sepsis. Objective This study aimed to determine the appropriateness of empirical antibiotics, reasons for non-appropriate empirical antibiotics, risk factors of mortality, length of stay in intensive care unit (ICU-LOS) and Acute Physiology And Chronic Health Evaluation II (APACHE II) score predictors in adult patients with sepsis. Setting An adult ICU of a tertiary hospital in  Malaysia. Methods A retrospective cohort study was conducted amongst patients with sepsis. Data were retrieved from the patients’ files and computer system. Each case was reviewed for the appropriateness of empirical antibiotics based on ICU local guidelines, bacterial sensitivity, dose, frequency, creatinine clearance and time of administration of empirical antibiotics. Multivariable logistic and Cox regression modelling were performed to compute the adjusted association of receiving appropriate or inappropriate empirical antibiotics with ICU mortality. Multivariable linear regression modelling was performed using ICU-LOS and APACHE II scores. Main outcome measures were ICU mortality, severity score (APACHE II scores) and ICU-LOS. Results The total mortality rate amongst the 228 adult ICU patients was 84.6%. Males showed a higher mortality rate (119 [52.2%]) than females (74 [32.5%]). Inappropriate empirical antibiotics were significantly associated with mortality and ICU-LOS (P < 0.005). Results from multivariable logistic regression showed that the appropriateness of empirical antibiotics model was a potential predictor for survival (OR 0.395, 95% CI 0.184-0.850, P < 0.005). Results from simple linear regression indicated that the appropriateness of empirical antibiotics model was a remarkable predictor of decreasing ICU-LOS (R = 0.055, 95% CI - 7.184 to - 2.114, P < 0.001). Results from simple Cox regression suggested that the appropriateness of empirical antibiotics was a protective factor for ICU mortality (HR 0.610, 95% CI 0.433-0.858, P = 0.005). Multivariable Cox regression revealed that the administration of antibiotics exceeding the recommended dose based on creatinine clearance was a protective factor (HR 0.186, 95% CI 0.040-0.868, P = 0.032). Conclusion The appropriateness of empirical antibiotics is a good predictor for improving survival and decreasing ICU-LOS. Effective appropriateness of empirical antibiotics use and close adherence to the recommended dose can prevent the early mortality of patients with sepsis and acute renal failure.