The relationship between serum potassium levels and mortality in acute medical admissions is uncertain. In particular, the relevance of minor abnormalities in potassium level or variations within the normal range remains to be determined. We performed a retrospective cohort study of all emergency medical admissions to St James's Hospital (Dublin, Ireland) between 2002 and 2012. We used a stepwise logistic regression model to predict in-hospital mortality, adjusting risk estimates for major predictor variables. There were 67,585 admissions in 37,828 patients over 11 years. After removing long-stay patients, 60,864 admissions in 35,168 patients were included in the study. Hypokalaemia was present in 14.5% and hyperkalaemia in 4.9%. In-hospital mortality was 3.9, 5.0, and 18.1% in the normokalaemic, hypokalaemic and hyperkalaemic groups respectively. Hypokalaemic patients had a univariate odds ratio (OR) of 1.29 for in-hospital mortality (95% confidence interval (CI) 1.16-1.43; p<0.001). Hyperkalaemic patients had a univariate OR for in-hospital mortality of 5.2 (95% CI 4.7-5.7; p<0.001). The ORs for an in-hospital death for potassium between 4.3 and 4.7 mmol/l, and 4.7 and 5.2 mmol/l, were 1.73 (95% CI 1.51-1.99) and 2.97 (95% CI 2.53-3.50) respectively. Hyperkalaemia and hypokalaemia are associated with increased mortality. [ABSTRACT FROM AUTHOR]