Predicting Blood Donor Retention Using the Random Forest Classification Algorithm: A Machine Learning Approach

Area of Science:

• Health Services Research
• Biomedical Informatics
• Behavioral Science

Background:

• Maintaining a stable blood supply is challenging due to donor participation influenced by behavioral and psychological factors.
• Identifying predictors of blood donor retention is crucial for effective donor management and sustaining blood services.

Purpose of the Study:

• To develop and evaluate a Random Forest machine learning model for predicting blood donor retention.
• Utilize demographic, behavioral, and psychological data to enhance donor retention prediction.

Main Methods:

• Retrospective cohort study with a cross-sectional component involving 612 donors' records and 100 prospective donors' surveys.
• Analyzed demographic data, donation history (frequency, recency, tenure, count), and psychological factors (motivation, attitudes, satisfaction, barriers).
• Employed Principal Component Analysis, K-means clustering, and Random Forest modeling, evaluating performance with accuracy, F1 score, and ROC-AUC.

Main Results:

• Behavioral indicators (donation frequency, recency, tenure, cumulative count) were the strongest predictors of retention.
• Psychological analysis revealed distinct donor profiles based on motivation, satisfaction, and perceived barriers.
• The Random Forest model achieved high predictive performance: 0.989 accuracy, 0.994 F1 score, and 0.994 ROC-AUC.

Conclusions:

• Blood donor retention is primarily influenced by behavioral engagement and psychological factors, not demographics or biology.
• The Random Forest model accurately identifies donors likely to return, supporting targeted engagement and resource allocation for blood services.