Abstract
BACKGROUND: Fluid resuscitation in acute pancreatitis (AP) patients requires precise titration because both excess and insufficient volumes may worsen outcomes. This study aimed to develop a weight-normalized fluid balance index (FBI) and assess its association with in-hospital mortality in critically ill AP patients. METHODS: This retrospective cohort study utilized data from the MIMIC-IV 3.0 database and the emergency intensive care unit (EICU) of our hospital (validation cohort) and was based on inclusion and exclusion criteria. Using the R package cutoff, an FBI of 145 mL/kg was identified as the optimal risk stratification threshold. The primary outcome was in-hospital all-cause mortality. Machine learning was used to screen covariates for inclusion in multivariable Cox models. Cox regression and restricted cubic spline (RCS) models were used to evaluate the relationship between FBI and mortality. Propensity score matching (PSM) was applied to minimize baseline confounding. After PSM, Kaplan-Meier survival curves were generated, and the results were validated via data from our center. RESULTS: In this study, 547 AP patients from the MIMIC-IV database and 156 from the EICU of our hospital were included. In the MIMIC-IV cohort, the overall in-hospital mortality rate was 8.96%. Patients with FBI ≥145 mL/kg had significantly higher in-hospital mortality than did those with FBI <145 mL/kg (P<0.05). High-risk classification remained an independent predictor of death after full adjustment (hazard ratio [HR] 1.99, 95% confidence interval [95% CI]: 1.08-3.69). Post-PSM Kaplan-Meier analysis confirmed significantly higher in-hospital mortality in the high-risk group (P<0.05). This result was corroborated by our validation cohort. RCS analysis further demonstrated a non-linear increase in in-hospital mortality with increasing FBI values. CONCLUSION: An FBI ≥145 mL/kg may be associated with increased in-hospital mortality in critically ill AP patients.