Hematocrit–albumin gap linked to sepsis risk in ICU adults in retrospective analysis

BackgroundSepsis is a major global health concern. Prior research has identified hematocrit and albumin levels as potential prognostic indicators for sepsis. Nevertheless, the specific relationship between the hematocrit–albumin gap (HAG) and susceptibility to sepsis remains underexplored. To address this study, we conducted a retrospective cross-sectional analysis using two independent datasets: (1) the MIMIC-IV database (derivation cohort) and (2) a retrospectively collected database from the Medical Intensive Care Unit (MICU) of The First People's Hospital of Yunnan Province, China (validation cohort).ObjectiveThis study aims to elucidate the relationship between HAG and sepsis risk.MethodsThis retrospective cross-sectional study utilized two independent datasets: (1) a derivation cohort of 19,134 adults with first-recorded ICU admissions during initial hospitalizations from the MIMIC-IV database and (2) an independent single-center validation cohort of 630 consecutive patients retrospectively enrolled from the Medical Intensive Care Unit (MICU) at The First People's Hospital of Yunnan Province, China (2022–2023). We collected data on hematocrit, albumin levels, and sepsis status. The study employed the logistic regression analysis and smooth curve fitting to analyze the data, with subgroup analyses conducted across different parameters and comorbidity statuses.ResultsThe investigation revealed an overall sepsis prevalence of 54.5%, disaggregated by gender: 56.1% among men and 43.9% among women. A multivariable-adjusted regression analysis determined the odds ratio (95% confidence interval) for sepsis—associated with HAG (per 7.848 increment)—to be 1.43 (95% CI: 1.37–1.48). Furthermore, a multivariate restricted cubic spline regression analysis revealed an S-shaped non-linear relationship between HAG and the incidence of sepsis, characterized by significant non-linearity (P < 0.05). When the HAG is below −9.162, the odds ratio (OR) for sepsis was 0.938, indicating a negative correlation. Between −9.162 and 17.27, a significant positive correlation exists with an OR of 1.051. Above 17.27, a relative saturation effect is observed, decreasing the OR to 1.023. Additionally, a significant statistical interaction was noted between HAG and conditions such as anemia, liver disease, and metastatic solid tumors in the prediction of sepsis (P < 0.05). Sensitivity analyses reinforced the reliability of the primary outcomes, supporting the study's conclusions.ConclusionAn S-shaped association was observed between the HAG and sepsis. Notably, the risk increased when the range of this difference was between −9.162 and 17.27. While HAG calculation is simple and rapid, its diagnostic utility requires prospective validation.