Why the study?
Current ICU severity scoring systems are complex and time-consuming, limiting their utility for rapid clinical decision-making.
Do novel simplified prediction models using readily available biomarkers accurately predict in-hospital mortality compared to SOFA and APACHE II scores in adult ICU patients?
Do novel simplified prediction models using readily available biomarkers accurately predict in-hospital mortality compared to SOFA and APACHE II scores in adult ICU patients?
Simplified prediction models using readily available biomarkers like LAR and NPAR perform comparably to complex traditional scoring systems (SOFA, APACHE II) for predicting in-hospital mortality in ICU patients.
May support rapid ICU mortality prediction with simple biomarkers; hypothesis-generating and requires prospective validation before any practice change.
Background: Current severity scoring systems in intensive care units (ICUs) are complex and time-consuming, limiting their utility for rapid clinical decision-making. This study aimed to develop and validate simplified prediction models using readily available biomarkers for assessing in-hospital mortality risk. Methods: We analyzed 19,720 adult ICU patients in this retrospective study. Three prediction models were developed: a basic model using lactate-to-albumin ratio (LAR) and neutrophil percent-to-albumin ratio (NPAR) and two enhanced models incorporating mechanical ventilation and continuous renal replacement therapy. Model performance was evaluated against Sequential Organ Failure Assessment (SOFA) score and Acute Physiology and Chronic Health Evaluation (APACHE) II score using machine learning approaches and validated through comprehensive subgroup analyses. Results: Among individual biomarkers, SOFA score showed the highest discriminatory power (area under these curves [AUC] 0.931), followed by LAR (AUC 0.830), CAR (AUC 0.749), and NPAR (AUC 0.748). Our enhanced Model 3 demonstrated exceptional predictive performance (AUC 0.929), statistically comparable to SOFA (p = 0.052), and showed a trend toward superiority over APACHE II (AUC 0.900, p = 0.079). Model 2 performed comparably to APACHE II (AUC 0.913, p = 0.430), while Model 1, using only LAR and NPAR, achieved robust performance (AUC 0.898) despite its simplicity. Subgroup analyses across different ICU types demonstrated consistent performance of all three models, supporting their broad clinical applicability. Conclusions: This study introduces novel, simplified prediction models that rival traditional scoring systems in accuracy while offering significantly faster implementation. These findings represent a crucial step toward more efficient and practical risk assessment in critical care, potentially enabling earlier clinical interventions and improved patient outcomes.
No takes yet. Share an insight, caveat, or question.
Chung et al. (2025) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: