Machine Learning–Augmented Traditional Analysis of Lactate vs Lactate-to-Albumin Ratio for Predicting Mortality Risk in Patients With Sepsis: Large-Scale Retrospective Study - Summary - MDSpire
Advertisement
Enhanced Traditional Analysis of Lactate and Lactate-to-Albumin Ratio for Assessing Mortality Risk in Sepsis Patients: A Comprehensive Retrospective Study Using Machine Learning Techniques
To rigorously investigate the association between lactate and lactate-to-albumin ratio (LAR) with 28-day mortality in patients with sepsis.
Approach:
Statistical Analysis: Utilized the DeLong test for paired AUC comparison, threshold effect analysis, restricted cubic spline modeling, and machine learning models with SHAP analyses to evaluate the prognostic utility of lactate and LAR.
Key Findings:
LAR outperformed lactate alone in predicting hospital mortality (AUC 0.74 vs 0.70).
LAR demonstrated significant prognostic value in predicting 28-day mortality.
The study formally tested the statistical significance of LAR's superiority over lactate.
Interpretation:
Limitations:
The cohort represents patients with critical illness requiring ICU admission, which may not fully align with Sepsis-3 definitions.
Potential biases in data collection and patient selection inherent in retrospective studies.
Conclusion:
This study is the first to formally verify LAR's superiority over lactate in predicting mortality in a large multicenter sepsis cohort.
