The red cell distribution width (RDW) predicts mortality among many populations. RDW is calculated as the standard deviation (SD) of the red blood cell (RBC) volume divided by mean corpuscular volume (MCV). Because higher MCV also predicts mortality, we hypothesized that the RDW numerator (one SD of RBC volume or 1SD‐RDW) predicts mortality more strongly than the RDW.

Adult subjects hospitalized during a contemporary clinical era (10/2005‐1/2014, N = 135 963) and a historical era (1/1999‐9/2005, N = 119 530) were studied. The RDW was obtained from the complete blood count (CBC), while 1SD‐RDW was calculated (RDW multiplied by MCV and divided by 100).

In univariable Cox regression (2005–2014 cohort), 1SD‐RDW (quintile 5 vs. 1: hazard ratio [HR] = 8·38, 95% confidence interval [CI] = 7·94, 8·85; P < 0·001) was a superior predictor of mortality compared to RDW (quintile 5 vs. 1: HR = 4·78, CI = 4·57, 5·00; P < 0·001). This superiority remained after adjustment for age, sex, basic metabolic profile components and other CBC factors excluding MCV (1SD‐RDW: HR = 2·41, CI = 2·28, 2·55; RDW: HR = 2·01, CI = 1·92, 2·11). Further adjustment for MCV strengthened the RDW association (HR = 2·14, CI = 2·04, 2·24; P < 0·001), becoming indistinct from 1SD‐RDW (HR = 2·20, CI = 2·08, 2·33; P < 0·001). Findings were similar for the 1999–2005 cohort.

The 1SD‐RDW predicted mortality more strongly than RDW, suggesting that 1SD‐RDW is superior to RDW as an individual risk predictor. Further, these results indicate that the dispersion of RBC volume and its mean are independent risk markers. Further research is required to understand the clinical value and mechanistic basis of these associations.