Pre- to post-dialysis potassium gradient (ΔK) has arrhythmogenic effects; however, its effect on mortality remains unclear. The relationship between ΔK and mortality was assessed. All patients undergoing hemodialysis in Beijing in 2014 were eligible for inclusion. The low (≤1.2 mmol/L), median (1.2-1.8 mmol/L), and high (>1.8 mmol/L) ΔK groups were matched by sex, age, diabetes, and dialysis time for enrollmen. The primary and secondary outcomes were all-cause and cardiovascular death within the follow-up. Cox regression analysis was performed to evaluate the effect of ΔK on mortality. We also analyzed the associations of combinations of ΔK and pre-dialysis potassium with mortality. We enrolled 2181 patients in three matched groups (n = 727 per group). The median follow-up was 72.0 (interquartile range, 53.7-72.0) months. All-cause mortality occurred in 215/727 (29.6%), 95/727 (13.1%), and 198/727 (27.2%) patients in the low-, median-, and high-ΔK groups, respectively. After adjusting for multiple factors, the median ΔK group had better survival than the low- (hazard ratio (HR), 1.91; 95% confidence interval [95% CI], 1.45-2.52; p < 0.001) and high-ΔK groups (HR, 2.17; 95% CI, 1.57-2.99; p < 0.001). Further analysis based on pre-dialysis potassium revealed that when maintaining a level of 4.5-5.5 mmol/L and ΔK of 1.2-1.8 mmol/L, patients had the lowest risk of mortality, whereas the highest risk was observed when pre-dialysis potassium was >5.5 mmol/L and ΔK was >1.8 mmol/L. Maintaining serum potassium within a appropriate range and reducing potassium fluctuations during dialysis may help to reduce the mortality risk of maintenance hemodialysispatients. These findings provide important data support for the quality control of hemodialysis.