Association between the lactate dehydrogenase-to-albumin ratio and 28-day mortality in septic patients with malignancies: analysis of the MIMIC-IV database

BACKGROUND: Sepsis remains a leading cause of mortality in critically ill patients, particularly those with malignancies who face heightened risks due to immunosuppression and metabolic dysregulation. This study aimed to evaluate the prognostic value of the lactate dehydrogenase-to-albumin ratio (LDAR) for predicting 28-day ICU mortality in septic patients with malignancies. METHODS: A retrospective cohort analysis was conducted using data from 1,635 septic patients with malignancies in the MIMIC-IV (3.1) database. Participants were stratified into quartiles based on LDAR values. The primary outcome was 28-day ICU mortality, with secondary outcomes including in-hospital and ICU mortality. Multivariable logistic regression, restricted cubic spline (RCS) analysis, and machine learning models were employed to assess associations between LDAR and outcomes. Subgroup analyses and feature importance evaluations were performed to validate robustness. The Shapley additive explanations method was used to enhance model interpretability and assess individual predictor contributions. RESULTS: Higher LDAR is independently associated with increased 28-day ICU mortality (OR: 3.441, 95% CI: 2.497-4.741), ICU mortality (OR: 3.478, 95% CI: 2.396-5.049), and in-hospital mortality (OR: 3.747, 95% CI: 2.688-5.222), even after adjustment, highlighting its potential as a prognostic marker in ICU patients. RCS analysis revealed a nonlinear relationship, with mortality risk escalating sharply beyond log₂(LDAR) = 6.940. Metastatic cancer patients had higher median LDAR (135.0 vs. 118.5, P = 0.004) and mortality rates (52.0% vs. 36.4%, P < 0.001). Boruta feature selection showed that LDAR as the top predictor of mortality. Nine machine learning model with 20 variables were built, with random forest model performing best, achieving an AUC of 0.751 (0.708-0.794) in validation and 0.727 (0.682- 0.772) in text cohort. CONCLUSIONS: LDAR is a robust, independent prognostic biomarker for 28-day ICU mortality in septic patients with malignancies, outperforming traditional scoring systems. The identified threshold (log₂(LDAR) ≥ 6.940) may aid early risk stratification and clinical decision-making. Prospective studies are warranted to validate these findings and explore dynamic LDAR monitoring in diverse populations.