Correlation between complete blood count-derived inflammatory markers and sepsis-associated delirium in older patients: a retrospective analysis of the MIMIC-IV database.

Abstract

To assess the correlation between complete blood count (CBC)-derived inflammatory markers and sepsis-associated delirium (SAD) risk in older intensive care unit (ICU) patients. Retrospective cohort study. ICUs at Beth Israel Deaconess Medical Center (2008-2019), using the Medical Information Mart for Intensive Care IV V.3.0 database. 3412 critically ill patients aged ≥65 years with sepsis. repeated ICU admission, death/discharge within 24 hours, missing delirium assessment or pre-sepsis delirium from non-septic aetiologies. SAD was diagnosed by Confusion Assessment Method for the ICU. Incidence of sepsis-associated delirium. Among 3412 older sepsis patients, 2092 (61.3%) developed SAD. Significant differences in platelet-to-lymphocyte ratio, neutrophil-to-lymphocyte ratio, monocyte-to-lymphocyte ratio (MLR), systemic immune-inflammation index, systemic inflammation response index, pan-immune-inflammation value and neutrophil-monocyte-to-lymphocyte ratio were observed between SAD and non-SAD groups (all p<0.001). After multivariable adjustment, elevated levels of these markers were associated with increased SAD risk, with MLR showing the strongest association (Q4 vs Q1 adjusted OR=2.66, 95% CI 2.14 to 3.30). Restricted cubic spline analysis revealed non-linear associations between inflammatory marker levels and SAD risk (p for non-linear <0.001). The results of most subgroup analyses were consistent, indicating that the association between inflammatory markers and SAD had high stability (p for interaction >0.05). Receiver operating characteristic and random forest analyses demonstrated predictive utility; incorporating markers into a baseline model significantly improved discrimination, with MLR providing the largest gain (area under the curve (AUC)=0.716 vs 0.703; ΔAUC=0.013, DeLong test, p<0.001). CBC-derived inflammatory markers, particularly MLR, are associated with increased SAD risk in older adults and enhance the performance of a clinical prediction model in this population. Further research is needed to better understand the pathophysiological mechanisms underlying these associations.