We thank Kok et al for their careful review of our report. There is no question that extreme hypokalemia and hyperkalemia are potentially life threatening. However, despite our hypothesis that changes in potassium during hospitalization for worsening heart failure would predict outcomes, we could not demonstrate such changes to be associated with short-term and long-term morbidity and mortality across the spectrum of serum potassium concentrations. Although it is of interest to understand if larger decreases, for example, >0.5 mEq/l, are associated with outcomes, such cutoffs will be arbitrary and applicable to only a small group of patients. However, we did investigate changes in serum potassium as a continuous variable and did not see strong evidence for adverse effect of potassium changes across the range available in this data set. Second, our prespecified hypothesis was focused on the change in serum potassium levels during hospitalization, and we did not explore discharge potassium levels as a risk stratifier. Also, most participants included in the analysis did indeed have a discharge potassium level used in the analysis.

We appreciate the careful review of the analysis by quartiles. The total sample size for the manuscript was 1,907. The reason that the quartiles are not equal is because many inhospital change values were the same, and we could not divide a group of patients with the same value for change in potassium. The N provided in Table 1 and Figure 1 is for patients with available data in serum potassium. If baseline potassium values were missing, day 1 values were used for the change in potassium analysis, resulting in differences in the total number of patients described. In Figure 2, the total number of observations is again 1,907 not the 1,753 number quoted. The 153 missing were not included in the total number of observations of 1,907. We hope that this clarifies the differences in observations noted.

Given the ongoing and prevalent use of agents during hospitalization for decompensated heart failure that reduce potassium (loop and thiazide diuretics) and concomitantly increase potassium (angiotensin-converting enzyme inhibitors, aldosterone receptor antagonists), this line of investigations is certainly not concluded and needs further investigation. However, our report is the first and remains the only to define and investigate changes in serum potassium levels in a large cohort of hospitalized patients with heart failure.