Serum potassium as a predictor of 28-day and in-hospital mortality in patients with sepsis: A U-shaped association

To the Editor: Sepsis, the leading cause of intensive care unit (ICU) admissions, has a mortality rate ranging from 10% to 50%.[1] Despite significant advances in understanding sepsis, it continues to be associated with acute organ failure and poses a global health challenge. Its heterogeneous nature complicates the standardization of diagnosis, treatment, and prognosis, highlighting the importance of predicting patient outcomes. Systemic inflammation frequently results in electrolyte imbalances in ICU patients. Potassium (K+), essential for cellular metabolism and membrane potential, plays a crucial role in the proper functioning of the cardiovascular, visceral, and nervous systems. Both hypokalemia and hyperkalemia can lead to severe complications. Additional research is required to elucidate the relationship between serum K+ levels and sepsis prognosis. Therefore, this study retrospectively analyzed clinical data from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database to investigate the correlation between serum potassium levels and 28-day and in-hospital mortality in sepsis, aiming to provide insights and evidence to inform future clinical practice. This study adhered to the regulations outlined in the Declaration of Helsinki. Access to the MIMIC-IV database was approved by the review committees of Beth Israel Deaconess Medical Centre and Massachusetts Institute of Technology. As a result, the Ethical Committee of the Affiliated Kunshan Hospital of Jiangsu University waived the requirement for informed consent and ethical approval. Participants were categorized into three groups based on potassium tertiles (T1–T3). Continuous variables were compared using t-tests or Kruskal–Wallis tests, while categorical variables were compared using Fisher’s exact test. Mortality rates were calculated for each tertile. Multivariate Cox regression analysis, adjusted for confounding factors, were used to construct models and evaluate the predictive value of K+ levels. Nonlinear relationships were examined with restricted cubic splines and two-piecewise Cox models. Stratified analyses were performed to evaluate consistency across subgroups. Statistical significance was set at P