Late-Course Lactate/Albumin Ratio and In-Hospital Mortality in CRRT-Treated Critically Ill Patients: A Retrospective Severity-Adjusted Analysis

Background/Objectives: The lactate/albumin ratio (LAR) integrates tissue hypoperfusion and nutritional reserve into a single dimensionless index. As a late-course measurement (last available value before the primary outcome), it may reflect terminal metabolic deterioration rather than early prognostic signal. We examined the association of the final LAR with in-hospital mortality after adjustment for validated severity scores (APACHE II, SOFA) in critically ill patients receiving CRRT, and evaluated its incremental value beyond lactate alone. Methods: A single-centre retrospective observational study was conducted, including 110 adult patients treated with CRRT/CVVHDF at Konya City Hospital (January 2024–January 2026). Multivariable logistic regression was performed across four hierarchically adjusted models. APACHE II was designated the primary adjustment covariate. ROC analyses compared the final LAR with final lactate and albumin alone. Results: In-hospital mortality was 70.9% (78/110); complete paired final lactate and albumin values were available for 109/110 patients (78 non-survivors, 31 survivors). After APACHE II adjustment, each 0.1-unit increase in the late-course LAR remained significantly associated with in-hospital mortality (OR 1.70; 95% CI 1.16–2.49; p = 0.007), a finding replicated in the ICU-stay sensitivity cohort (n = 100, 73 events; OR 1.63; 95% CI 1.09–2.42; p = 0.016). The LAR trajectory (ΔLAR; AUC 0.725) was significantly greater in non-survivors (+0.104 vs. −0.060; p < 0.001); however, ΔLAR did not significantly outperform the lactate trajectory alone (Δlactate; AUC 0.706; p = 0.212), and lost statistical significance after severity-score adjustment. Furthermore, the association was attenuated and non-significant after SOFA adjustment (OR 1.35; p = 0.077), and the final LAR did not provide statistically significant discriminative advantage over lactate alone (AUC 0.768 vs. 0.761; LRT p = 0.435). Conclusions: The late-course LAR remained significantly associated with mortality after APACHE II adjustment. Three qualifications temper this finding: the association was attenuated after SOFA adjustment, the final LAR did not outperform lactate alone, and the measurement design is susceptible to reverse causality. The LAR trajectory (ΔLAR; AUC 0.725) showed comparable discrimination to the final LAR but did not significantly outperform the lactate trajectory alone (Δlactate; AUC 0.706; p = 0.212) and lost statistical significance after severity-score adjustment, so it should likewise be regarded as hypothesis-generating rather than confirmatory; prospective evaluation with time-stamped serial measurements and landmark analyses in multi-centre CRRT cohorts is warranted.