Association between the red blood cell distribution width-to-platelet count ratio and in-hospital mortality among critically ill patients with leukemia: A retrospective study and machine learning with external validation
Huiqiang Wu, Yuwei Chen, Zhiyin Cai, Wanyi Liu, Zechuan Wang, Meiling Li, Weihuang ZhuangObjective
To explore the association between the red blood cell distribution width-to-platelet ratio (RPR) and the risk of in-hospital mortality in critically ill patients with leukemia.
Methods
Data were obtained from 807 critically ill patients with leukemia in the MIMIC-IV database. Multivariate logistic regression and restricted cubic spline (RCS) analysis were used to examine the association between RPR and in-hospital mortality. Furthermore, machine learning models were developed to evaluate predictive performance and were externally validated using the eICU-CRD.
Results
Higher RPR was consistently associated with increased in-hospital mortality across all adjustment models (all P < 0.001). Compared with the lowest quartile, higher RPR quartiles showed progressively increased mortality, with a significant trend. RCS analysis further demonstrated a nonlinear association between RPR and in-hospital mortality. ROC analysis demonstrated RPR outperformed RDW and platelet count in distinguishing in-hospital mortality. The Extreme Gradient Boosting(XGBoost)model exhibited good predictive performance in both internal validation and external eICU-CRD validation.
Conclusion
RPR was independently associated with in-hospital mortality in critically ill patients with leukemia and showed higher discriminatory ability than RDW and platelet count. The XGBoost model demonstrated promising predictive performance, suggesting that RPR may serve as a readily available marker for early mortality risk stratification.