Glucocorticoid use and 28-day mortality in sepsis: A retrospective analysis of the MIMIC-III database

The therapeutic role of glucocorticoids in sepsis management continues to be debated. This study aimed to investigate the relationship between glucocorticoid administration and 28-day mortality in septic patients admitted to the intensive care unit (ICU).

Adult patients with sepsis were identified from the MIMIC-III database. Exposure was defined as systemic glucocorticoid use during ICU stay. In crude logistic regression, glucocorticoid use was associated with increased 28-day mortality (OR = 1.27, 95 % CI: 1.11-1.46, p < 0.001). However, inverse probability weighted Cox models indicated reduced mortality risk (HR = 0.88, 95 % CI: 0.79-0.98, p = 0.016). In contrast, time-dependent Cox regression suggested increased hazard associated with glucocorticoid exposure (HR = 1.22, 95 % CI: 1.11-1.34, p < 0.001), while piecewise Cox models showed consistently elevated hazard during both early (0-14 days) and late (14-28 days) follow-up periods (HR = 1.26, 95 % CI: 1.09-1.44).

Glucocorticoid use was associated with increased mortality in unadjusted models, while weighted Cox models indicated a potential survival benefit. However, time-dependent and piecewise Cox analyses consistently showed increased hazard following glucocorticoid exposure. These findings underscore substantial heterogeneity across analytical strategies and highlight the need for prospective trials to clarify the impact of glucocorticoids in sepsis.