Abstract / Summary
Abstract Dysnatremia is common in critically ill patients, but the prognostic relevance of its duration remains uncertain. This retrospective cohort study used MIMIC-IV data from 2008 to 2022. Of 94,458 ICU records screened, 9,799 adults with a first ICU stay of at least 7 days were included. The primary outcome was 28-day all-cause mortality; no secondary outcome was prespecified. Associations were evaluated using concentration–duration heatmaps, restricted cubic splines, and Fine–Gray models treating discharge alive as a competing event. MIMIC-IV was approved by the review boards of the Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center, with consent waived for deidentified data. Within 28 days, 2,002 patients (20.4%) died. In fully adjusted analyses, each 1-percentage-point increase in monitored time at < 125 mmol/L was associated with a higher subdistribution hazard of death (sHR 1.016, 95% CI 1.006–1.026), whereas time at 135–140 mmol/L was associated with a lower hazard (sHR 0.996, 95% CI 0.994–0.997). Positive associations were also observed at 150–155 mmol/L (sHR 1.013, 95% CI 1.009–1.018) and ≥ 155 mmol/L (sHR 1.021, 95% CI 1.017–1.026). Time-weighted analyses showed a U-shaped association, with the lowest estimated mortality probability near 140.6 mmol/L. These findings may inform risk assessment but do not establish causality or treatment targets.