Predicting Medical Claim Denial Using Logistic Regression and Decision Tree Algorithm

This study utilizes decision tree and logistic regression models to explore the factors contributing to medical claim denials and identify areas for improvement. We adapt undersampling technique in data preprocessing to balance the dataset. To enhance model accuracy, we employ augmented backward elimination to remove insignificant variables, and grid search cross-validation in tree pruning. Our findings suggest that decision trees are superior in predicting claim denials, though they underperform in predicting claim approvals compared to logistic regression. The analysis also highlights factors such as the insurance category, hospital specialty, and the interval between service date and claim submission date significantly influence denial rates. These insights are crucial for healthcare denial management, enabling targeted interventions and enhanced training for clinic operation to focus on key denial factors. Finally, the findings suggest the potential of predictive models to improve denial processes and enhance the efficiency of healthcare billing systems.