Dissatisfaction-considered waiting time prediction for outpatients with interpretable machine learning.

Area of Science:

• Health Services Research
• Machine Learning in Healthcare
• Operations Research

Background:

• Long waiting times in outpatient departments significantly contribute to patient dissatisfaction.
• Accurate prediction of waiting times is essential for managing patient expectations and improving healthcare operational efficiency.
• Existing predictive models may not adequately address the consequences of underestimating waiting times.

Purpose of the Study:

• To develop and validate a novel framework for estimating outpatient waiting times that explicitly considers patient dissatisfaction.
• To improve the accuracy and reliability of waiting time predictions by preventing underestimation.
• To provide interpretable explanations of predicted waiting times for patients and healthcare administrators.

Main Methods:

• Implementation of a machine learning framework utilizing asymmetric loss functions to penalize underestimation more heavily.
• Proposal and application of a dissatisfaction-aware asymmetric error score (DAES) for optimal model selection.
• Utilization of Shapley additive explanation (SHAP) for interpreting model predictions and identifying key factors influencing waiting times.

Main Results:

• The proposed framework effectively prevents the underestimation of waiting times through the use of asymmetric loss functions.
• The DAES provides a balanced approach to model selection, optimizing the trade-off between accuracy and underestimation.
• SHAP analysis revealed key operational factors, such as queue length, as major determinants of waiting time, enabling actionable insights.

Conclusions:

• The developed framework offers a robust solution for predicting outpatient waiting times, significantly reducing patient dissatisfaction.
• The integration of asymmetric loss functions and DAES enhances predictive accuracy and model interpretability.
• This approach facilitates practical applications in hospitals for real-time patient notifications and overall operational improvement in healthcare services.