Objectives: Decreasing in albumin levels are widely used to predict mortality. We evaluate the overall 28-day mortality prognostic value of percentage changes in albumin levels during first week of admission (%∆ ALB 1wk) in comparison to either albumin level at admission (ALB 1) or albumin level after 1 week of admission (ALB 2) taking into consideration the instability of critically ill patients. Methods: We retrospectively obtained serial measurements of albumin levels of patients admitted to our adult Intensive Care Unit (ICU) between April 2017 and Sep 2018. Patients will be excluded if they discharged or died before 1 week of ICU admission. A receiver operating characteristic (ROC) analysis was conducted to compare the prognostic efficacy of the three tested prognosticators. Results: A total of 163 critically ill patients were finally included in this study. The mean age was 58.37±9.96 years, and 112 (68.71%) were male. The overall 28-day ICU mortality rate was 39.26% (64 patients). The %∆ ALB 1wk was significantly higher in nonsurvivors than in survivors (-11.47%±8.23% vs 0.69%±4.61%; P<0.05). The prognostic value of the % ΔALB 1wk (0.929) was significantly greater than that of ALB 1 and ALB 2 (0.266and 0.655, respectively; P<0.001). Conclusion In comparison to Alb 1 , the %ΔALB 1wk and Alb 2 have a significantly prognostication which is significantly higher in % ΔALB 1wk than ALB 2 with a sensitivity and performance of (93.94% and 75.19% vs 59.59% and 29.91%, respectively; P<0.001). The % Δ ALB 1wk may serve a novel independent prognosticating indicator for critically ill patient's mortality who are taking intravenous human albumin.