Churn Analysis in the Healthcare Sector using Machine Learning Methods

Abstract

Ensuring patient loyalty and minimizing patient churn are crucial for the sustainability of hospitals and providing better healthcare services to patients. Predicting patient churn helps hospitals take strategic actions to enhance patient satisfaction. Data-driven approaches and analytical methods play a key role in improving service quality and understanding patient behaviors. In this research, recency, frequency, monetary, and length of Relationship (RFML) analysis was used for patient churn prediction. This approach allows for a dynamic churn threshold to be set based on each medical department. The patients in the lowest quartile were labeled as churn based on their weighted RFML scores. The weights were defined by expert opinion. Machine learning methods were utilized to predict patient churn following the labeling process. Model performance was analyzed using different evaluation metrics such as precision, recall, F1-score, and accuracy. The findings from this study can support hospitals in making strategic decisions to improve healthcare services and improve patient loyalty.