Comparative predictive value of nine inflammation-derived haematological indices for 28-day mortality in patients with sepsis: a multicentre retrospective cohort study

Background: Sepsis is associated with high mortality, and early risk stratification is essential. Inflammation-derived hematological indices from routine complete blood count testing are readily available, inexpensive, and reproducible, but their comparative prognostic value in sepsis remains unclear. Methods: This multicenter retrospective cohort study used the MIMIC-IV database as the development cohort and the First Affiliated Hospital of Guangxi Medical University as the external validation cohort. Adult patients meeting Sepsis-3 criteria with an intensive care unit length of stay >24 h were included. Nine inflammation-derived hematological indices were evaluated: neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, monocyte-to-lymphocyte ratio, neutrophil-to-monocyte ratio, neutrophil-to-platelet ratio, monocyte-to-platelet ratio, systemic immune-inflammation index, systemic inflammation response index, and aggregate index of systemic inflammation. Cox regression, restricted cubic spline analysis, receiver operating characteristic curves, calibration curves, decision curve analysis, and Sequential Organ Failure Assessment combined analyses assessed their prognostic value for 28-day all-cause mortality after the 24 h landmark time. Results: The development cohort included 26,512 patients with sepsis, among whom 2,669 died within 28 days. All nine indices were significantly associated with 28-day mortality after the landmark time. In the fully adjusted model, monocyte-to-lymphocyte ratio, neutrophil-to-platelet ratio, neutrophil-to-lymphocyte ratio, and systemic inflammation response index showed relatively larger effect estimates. Unadjusted discrimination was generally moderate, with the neutrophil-to-lymphocyte ratio showing the highest area under the curve. Restricted cubic spline analysis showed non-linear associations between all indices and mortality risk, although inflection points should not be interpreted as clinical thresholds. Fully adjusted and Sequential Organ Failure Assessment combined analyses suggested supplementary prognostic information, particularly from neutrophil-to-lymphocyte ratio, systemic inflammation response index, and monocyte-to-lymphocyte ratio. The external validation cohort included 850 patients, among whom 182 died within 28 days. After full adjustment, none of the indices remained significantly associated with 28-day mortality. Conclusion: Inflammation-derived hematological indices were associated with 28-day mortality and showed non-linear risk patterns in the development cohort. However, their standalone discriminatory ability was limited, and externally validated associations were not consistent. These indices may serve as supplementary markers alongside conventional severity scores rather than independent mortality prediction tools. Further prospective multicentre validation is warranted.