Machine Learning Predictions on Outpatient No-Show Appointments in a Malaysia Major Tertiary Hospital

Area of Science:

• Health Informatics
• Machine Learning in Healthcare
• Predictive Analytics

Background:

• Patient no-show appointments disrupt treatment continuity and waste resources.
• Machine learning offers a promising approach to predict and mitigate no-show incidents.

Purpose of the Study:

• To develop a predictive model for patient no-show appointments at Hospital Kuala Lumpur (HKL).
• To apply machine learning algorithms for analyzing outpatient appointment data.

Main Methods:

• Utilized 246,943 outpatient appointment records from HKL's Patient Management System (2019).
• Employed seven machine learning algorithms: logistic regression, decision tree, k-NN, Naïve Bayes, random forest, gradient boosting, and MLP.
• Conducted descriptive and predictive analyses on 13 appointment attributes.

Main Results:

• Descriptive analysis revealed a 28% no-show rate, influenced by appointment month and patient gender.
• The Gradient Boosting model demonstrated the highest performance with 78% accuracy, 0.76 F1 score, and 0.65 AUC.
• Key factors influencing no-shows were identified.

Conclusions:

• The developed predictive model can inform targeted interventions to reduce patient no-shows.
• Implementing this model can enhance patient care quality and optimize resource allocation in healthcare settings.