Abstract Delirium is a common neuropsychiatric complication in intensive care unit (ICU) patients and is associated with poor outcomes. This retrospective cohort study analyzed 2946 ICU patients from the MIMIC-IV database to examine whether the stress hyperglycemia ratio (SHR) is associated with delirium. SHR was classified into quartiles: Q1 (≤ 0.85), Q2 (0.86–1.01), Q3 (1.02–1.24), and Q4 (≥ 1.25). Delirium was assessed by confusion assessment method for the ICU (CAM-ICU). Logistic regression analyses with sequential adjustment for covariates were conducted across four models. Dose–response relationships were examined using restricted cubic splines (RCS), and subgroup analyses were performed to test interactions. After the first 24 h of being in the ICU, 21% (n = 619) of patients developed delirium for the first time. Higher SHR was associated with greater delirium risk in the unadjusted and partially adjusted models. After full adjustment, the continuous estimate narrowly missed statistical significance (Model 4: OR = 1.24, 95% CI 0.99–1.56; p = 0.057). When SHR was divided into quartiles, compared to Q1, patients in Q3 (OR = 1.58, 95% CI 1.14–2.2) and Q4 (OR = 1.59, 95% CI 1.15–2.21) had higher risk, with a significant trend ( p for trend < 0.001). RCS analysis indicated a nonlinear relationship, and threshold-effect analysis identified an inflection point at SHR = 1.192. Subgroup analyses detected no significant interactions, with the positive association between SHR and delirium risk consistent across all subgroups. Four sensitivity analyses further supported the robustness of the SHR–delirium association. Overall, SHR showed a positive, nonlinear relationship with delirium in ICU patients, with an inflection point at 1.192, and may serve as a pragmatic predictor to inform delirium prevention in critical illness.
Wang et al. (2026) studied this question.