Application of Bayesian network and regression method in treatment cost prediction

Area of Science:

• Health economics
• Medical informatics
• Computational statistics

Background:

• Accurate disease-based medical cost forecasting is crucial for healthcare reform, resource allocation, and reducing patient expenses.
• Predicting treatment costs aids in understanding influencing factors and managing medical resources effectively.

Purpose of the Study:

• To develop a robust method for predicting single disease treatment costs using electronic medical records, especially when data is scarce.
• To improve the accuracy of cost prediction compared to traditional regression models.

Main Methods:

• A data preprocessing pipeline involving text-based medical record conversion and weighted interpolation for missing values.
• A Bayesian network model for patient treatment process classification.
• Integration of local weight regression with the Bayesian network for enhanced cost prediction.

Main Results:

• The proposed Bayesian network combined with regression analysis achieved higher prediction accuracy on hospital electronic medical record data.
• The methodology effectively handles small datasets and improves upon traditional regression models for cost forecasting.

Conclusions:

• The developed model offers a promising approach for accurate medical cost prediction, supporting efficient healthcare management and resource allocation.
• This method is valuable for optimizing medical insurance mechanisms and mitigating financial burdens on patients.