LDAR is superior to other albumin-derived indices in predicting 28-day ICU mortality in critically ill patients with intracerebral hemorrhage: a two-cohort study.
Yang Ping P, Peng Daizhou D, Huang Zhigui Z, Luo Dengjian D et al.
Spontaneous intracerebral hemorrhage (ICH) carries a high risk of poor outcomes, making early identification of high-risk patients crucial for improving prognosis. Albumin-derived composite indices can reflect the systemic inflammatory, nutritional, and metabolic status, but their predictive performance in the ICH population lacks systematic comparison. This study aimed to evaluate the prognostic value of six albumin-derived composite indices for 28-day ICU mortality in critically ill patients with ICH. This study included 1,638 adult patients with first-time ICH admission from the MIMIC-IV database as the derivation cohort, and 493 patients with ICH from the People's Hospital of Xingyi City as the external validation cohort. Clinical data within 24 h of admission were collected. Six indices were calculated: red blood cell distribution width to albumin ratio (RAR), creatinine to albumin ratio (CAR), anion gap to albumin ratio (AGAR), total bilirubin to albumin ratio (TAR), blood urea nitrogen to albumin ratio (UAR), and lactate dehydrogenase to albumin ratio (LDAR, log-transformed for analysis). Cox regression, restricted cubic splines (RCS), and Kaplan-Meier (KM) curves were used to analyze the association between each index and 28-day ICU mortality. Predictive performance was compared using receiver operating characteristic (ROC) curves. The incremental predictive value of log(LDAR) beyond traditional critical care scores was also assessed. In the derivation cohort, 295 patients (18.0%) died within 28 days. After multivariable adjustment, log(LDAR) (HR = 1.68, 95%CI: 1.35-2.10), RAR (HR = 1.09, 95%CI: 1.01-1.19), and TAR (HR = 1.27, 95%CI: 1.16-1.40) were independent risk factors for 28-day mortality. RCS analysis revealed nonlinear associations with mortality risk for log(LDAR), TAR, and UAR (all P-non-linear < 0.05), whereas RAR showed a linear positive correlation (P-non-linear = 0.467). ROC curve analysis demonstrated that log(LDAR) had the highest predictive efficacy (AUC = 0.695), significantly outperforming the other indices (all DeLong test p < 0.05). Adding log(LDAR) to five traditional severity scores, including APACHE II and SOFA, significantly improved their predictive ability (AUC increase range: 0.016-0.039, all p < 0.05). Subgroup analyses confirmed the robustness of log(LDAR)'s predictive effect across different populations (all p for interaction > 0.05). These findings were validated in the external cohort. Among six albumin-derived composite indices, log(LDAR) offers superior predictive value for 28-day ICU mortality in critically ill ICH patients. It significantly enhances the risk stratification capability of traditional critical illness scores and may serve as a useful tool for early clinical screening.