Potential Predictors for Deterioration of Renal Function After Transfusion.

Area of Science:

• Nephrology
• Hematology
• Data Science in Medicine

Background:

• Packed red blood cell (pRBC) transfusions carry inherent risks, including potential renal complications.
• Predicting and mitigating post-transfusion adverse events like acute kidney injury (AKI) remains a clinical challenge.

Purpose of the Study:

• To develop and validate a machine learning model for predicting AKI following individual pRBC transfusions.
• To identify recipient and donor characteristics associated with increased risk of post-transfusion AKI.

Main Methods:

• A retrospective analysis of 3366 pRBC transfusion episodes was conducted.
• Random forest models were trained using recipient/donor demographics, lab values, and pRBC age.
• Machine learning models predicted AKI based on modified KDIGO criteria.

Main Results:

• The model achieved an AUC-ROC of 0.73 ± 0.02 for AKI prediction.
• Recipient characteristics were significantly more predictive of AKI than donor features.
• A base rate of 17.4% for post-transfusion AKI was observed.

Conclusions:

• Recipient factors alone were decisive in predicting post-transfusion AKI.
• The influence of specific pRBC selection on AKI risk appears less significant than recipient-specific variables.
• Future research may focus on recipient-centric approaches to minimize transfusion-related renal injury.