OBJECTIVES: Our objectives were (1) to characterize the distribution of serum potassium levels at ICU admission, (2) to examine the relationship between dyskalemia at ICU admission and occurrence of cardiac events, and (3) to study both the association between dyskalemia at ICU admission and dyskalemia correction by day 2 on 28-day mortality. DESIGN: Inception cohort study from the longitudinal prospective French multicenter OUTCOMEREA database (1999-2014) SETTING: 22 French OUTCOMEREA network ICUs PATIENTS: Patients were classified into six groups according to their serum potassium level at admission: three groups of hypokalemia and three groups of hyperkalemia defined as serious hypokalemia [K+] \textless 2.5 and serious hyperkalemia [K+] \textgreater 7 mmol/L, moderate hypokalemia 2.5 \textless/= [K+] \textless 3 mmol/L and moderate hyperkalemia 6 \textless [K+] \textless/= 7 mmol/L, and mild hypokalemia 3 \textless/= [K+] \textless 3.5 mmol/L and mild hyperkalemia 5 \textless [K+] \textless/= 6 mmol/L. We sorted evolution at day 2 of dyskalemia into three categories: balanced, not-balanced, and overbalanced. INTERVENTION: None MEASUREMENTS AND MAIN RESULTS: Of 12,090 patients, 2108 (17.4%) had hypokalemia and 1445 (12%) had hyperkalemia. Prognostic impact of dyskalemia and its correction was assessed using multivariate Cox models. After adjustment, hypokalemia and hyperkalemia were independently associated with a greater risk of 28-day mortality. Mild hyperkalemic patients had the highest mortality (hazard ratio (HR) 1.29, 95% confidence interval (CI) [1.13-1.47], p \textless 0.001). Adjusted 28-day mortality was higher if serum potassium level was not-balanced at day 2 (aHR = 1.51, 95% CI [1.30-1.76], p \textless 0.0001) and numerically higher but not significantly different if serum potassium level was overbalanced at day 2 (aHR = 1.157, 95% CI [0.84-1.60], p = 0.38). Occurrence of cardiac events was evaluated by logistic regression. Except for patients with serious hypokalemia at admission, the depth of dyskalemia was associated with increased risk of cardiac events. CONCLUSIONS: Dyskalemia is common at ICU admission and associated with increased mortality. Occurrence of cardiac events increased with dyskalemia depth. A correction of serum potassium level by day 2 was associated with improved prognosis.