Developing and validating a multivariable prediction model for predicting the cost of colon surgery.

Area of Science:

• Medical Informatics
• Health Services Research
• Machine Learning in Healthcare

Background:

• Hospitals face challenges in accurately predicting patient treatment costs.
• Comorbidities complicate traditional cost prediction, impacting hospital revenue.
• Efficient cost management is crucial for financial sustainability in healthcare.

Purpose of the Study:

• To evaluate machine learning (ML) algorithms for predicting cost factors in colon surgery patients.
• To develop a predictive tool to aid hospitals in cost management and operational efficiency.
• To establish a foundation for a multicenter ML-based cost prediction tool.

Main Methods:

• Utilized machine learning algorithms to forecast healthcare costs.
• Incorporated multiple patient and treatment predictors into the forecasting model.
• Validated the model's predictive accuracy using mean absolute percentage error (MAPE).

Main Results:

• The ML model achieved a mean absolute percentage error (MAPE) of 18%-25.6% in cost prediction.
• Demonstrated decent accuracy in forecasting costs across various diagnoses and surgical approaches.
• Identified potential for ML in predicting costs associated with specific complications.

Conclusions:

• Machine learning shows promise in predicting healthcare costs for colon surgery.
• Accurate cost prediction can enhance hospital financial efficiency.
• Further investigation into ML for predicting costs in complex cases, such as anastomotic leakage, is urgently required to minimize hospital expenditures.