Efficient Prediction of Missed Clinical Appointment Using Machine Learning.

Area of Science:

• Health Informatics
• Machine Learning in Healthcare
• Public Health Resource Management

Background:

• Optimizing health resource utilization is critical for societal well-being, especially during pandemics.
• Patient appointment no-shows negatively impact healthcare systems and patient health outcomes.
• Analyzing large-scale patient data can reveal patterns in appointment behavior.

Purpose of the Study:

• To predict patient appointment no-show behavior using machine learning algorithms.
• To evaluate the effectiveness of ten different machine learning models on a large dataset.
• To identify key factors and feature combinations for accurate prediction of appointment status.

Main Methods:

• Processed over six million patient appointment records.
• Employed data cleaning and feature extraction techniques.
• Utilized data balancing methods including Synthetic Minority Oversampling Technique (SMOTE), Adaptive Synthetic Sampling Method (Adasyn), and random undersampling (RUS).
• Applied ten machine learning algorithms: random forest, decision tree, logistic regression, XGBoost, gradient boosting, Adaboost, Naive Bayes, stochastic gradient descent, multilayer perceptron, and Support Vector Machine.
• Evaluated model performance using recall, accuracy, precision, F1-score, area under the curve (AUC), and mean square error (MSE).

Main Results:

• Achieved high performance metrics: 94% recall, 86% accuracy, 83% precision, 87% F1-score, 92% AUC, and 0.106 MSE.
• Demonstrated the effectiveness of data cleaning and feature selection in improving predictive accuracy.
• Confirmed that combining multiple features yields superior prediction of patient appointment status compared to individual features.

Conclusions:

• Machine learning models can effectively predict patient no-show behavior.
• The study highlights the importance of comprehensive data analysis and feature engineering for healthcare applications.
• Improved prediction of appointment status can lead to better resource allocation and reduced healthcare costs.