We assessed the relationship between red blodd cell distribution width (RDW) and postoperative cognitive dysfunction (POCD) after coronary artery bypass grafting (CABG) in patients who usually had obvious hemodynamic changes. We enrolled 362 coronary heart disease patients who received CABG. POCD was assessed through neuropsychological examination at 21 days after operation. Demographics, history of diseases, blood biochemical parameters and perioperative data were collected. The receiver operating characteristic (ROC) curve was used to find the best cut-off value of RDW for diagnosis of POCD. Logistic regression was used to explore the relationship between RDW and POCD. The 21-day incidence of POCD in patients receiving CABG was 27.1% (98/362). The RDW of POCD patients was significantly higher than in the non-POCD patients (17.4 vs. 13.2). The sensitivity and specificity of RDW for predicting POCD were 82.7 and 64.8%, respectively. The POCD patients also tended to be older and had higher fasting plasma glucose, hypersensitive c-reactive protein, tumor necrosis factor-α, white blood cell levels and longer surgery time. No significant differences were found in other parameters. The 21-day neuropsychological test results were better in the POCD patients than the non-POCD patients. After adjustment of potential factors, the preoperative high RDW was still associated with an increased risk of POCD (odds ratio (OR) = 2.52, 95% confidence interval (CI): 1.28–4.31). Our study indicates that preoperative RDW is significantly elevated in POCD patients receiving CABG. The elevated preoperative RDW is associated with an increased risk of POCD and preoperative RDW can be an independent predictor of POCD.