Predicting non-attendance in hospital outpatient appointments using deep learning approach

Area of Science:

• Health Informatics
• Machine Learning
• Artificial Intelligence in Healthcare

Background:

• Hospital outpatient non-attendance presents a significant financial challenge for healthcare systems.
• Non-attendance stems from a complex interplay of various patient-related and external factors.
• Accurate prediction of patient no-shows is crucial for optimizing hospital resource allocation and reducing financial losses.

Purpose of the Study:

• To develop an advanced predictive model for forecasting hospital outpatient non-attendance.
• To integrate heterogeneous data sources, including electronic health records and external data, for comprehensive analysis.
• To identify and model the spectrum of factors contributing to patient non-attendance.

Main Methods:

• Proposed a novel non-attendance prediction model utilizing deep neural networks and machine learning.
• Employed sparse stacked denoising autoencoders (SDAEs) for data dimensionality reduction and feature representation learning.
• Evaluated the model's performance against established machine learning algorithms using real-world hospital data.

Main Results:

• The proposed deep learning approach, incorporating SDAEs, demonstrated superior performance in predicting non-attendance.
• The model achieved higher accuracy compared to several well-known and scalable machine learning models.
• Integration of a softmax layer and logistic regression further enhanced the predictive power of the proposed method.

Conclusions:

• The developed deep learning model offers a robust solution for predicting hospital outpatient non-attendance.
• The approach effectively leverages heterogeneous data and advanced feature learning techniques.
• This predictive capability can aid hospitals in mitigating the financial impact of missed appointments and improving operational efficiency.