Non‐Linear Threshold Effect of Dynamic Estimated Plasma Volume Status on Revealing In‐Hospital Mortality Risk in Sepsis

This retrospective cohort study evaluated the association between estimated plasma volume status (ePVS) and in-hospital mortality in adult patients with sepsis from the MIMIC-IV (v2.2) database. After excluding patients with early death, readmission, red blood cell transfusion, malignancy, or missing ePVS data, 17,403 patients meeting Sepsis-3 criteria were included. ePVS was calculated using hemoglobin and hematocrit. Early ePVS (ePVS1), late-ICU ePVS (ePVS2), and ePVS change were analyzed using multivariable logistic regression, restricted cubic splines, threshold analysis, prediction metrics, Kaplan-Meier analysis, subgroup analysis, and sensitivity analyses. In-hospital mortality was 13.2%. ePVS2 and ePVS change were significantly higher in non-survivors, whereas the difference in ePVS1 was small. After adjustment, ePVS2 (adjusted odds ratio [OR] = 1.14, 95% confidence interval [CI]: 1.11-1.18, p < 0.001) and ePVS change (adjusted OR = 1.04, 95% CI: 1.02-1.05, p < 0.001) remained independently associated with in-hospital mortality, while ePVS1 was not significant. Non-linear and threshold analyses suggested lower risk at ePVS2 of approximately 6-8 dL/g and ePVS change of -4.5 to 2, with increased risk at ePVS2 ≥ 8 dL/g or ePVS change ≥ 2. Late-ICU ePVS2 and ePVS change may supplement volume-status assessment and short-term prognostic stratification in sepsis, but prospective validation is needed.