Nonlinear association of early serum albumin with 28-day ICU mortality in adults with an admission diagnosis of sepsis: a multicenter eICU cohort study.

Background: Serum albumin is frequently measured early in intensive care, but its relation to short-term outcome in adults admitted with sepsis has varied across observational studies. We examined the shape and consistency of the association between early serum albumin and 28-day ICU mortality in a multicenter cohort. Methods: We conducted a retrospective multicenter cohort study using original eICU Collaborative Research Database version 2.0 tables. Adults with admission-diagnosis text containing sepsis were eligible. The exposure was the first valid albumin measurement obtained within 24 h of ICU admission and before ICU discharge; the endpoint was ICU death within 28 days. Hospital-clustered logistic regression and restricted cubic splines were used. Prespecified analyses addressed measurement timing, missing data, selective albumin testing, short stays, repeated admissions, and discharge alive as a competing event. Results: Among 23,124 adults with an admission diagnosis containing sepsis, 13,634 (59.0%) had an eligible early albumin measurement. The primary cohort included 11,424 ICU stays and 1,161 deaths. In the primary complete-case model ( Conclusion: Early serum albumin was nonlinearly associated with 28-day ICU mortality and may provide complementary information for early risk assessment, particularly at lower concentrations. The clinical utility of particular concentration ranges and the effects of albumin administration require prospective evaluation.