Predictors of Hyperkalemia and Hypokalemia in Individuals with Diabetes: a Classification and Regression Tree Analysis

Background Both hyperkalemia and hypokalemia can lead to cardiac arrhythmias and are associated with increased mortality. Information on the predictors of potassium in individuals with diabetes in routine clinical practice is lacking. Objective To identify predictors of hyperkalemia and hypokalemia in adults with diabetes. Design Retrospective cohort study, with classification and regression tree (CART) analysis. Participants 321,856 individuals with diabetes enrolled in four large integrated health care systems from 2012 to 2013. Main Measures We used a single serum potassium result collected in 2012 or 2013. Hyperkalemia was defined as a serum potassium ≥ 5.5 mEq/L and hypokalemia as