Predicting no-shows at outpatient appointments in internal medicine using machine learning models

Area of Science:

• Healthcare Informatics
• Machine Learning in Medicine
• Clinical Operations

Background:

• Patient absenteeism in medical appointments causes significant healthcare delivery delays and operational inefficiencies.
• Addressing no-shows is critical in internal medicine, which manages complex chronic conditions.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting patient absenteeism risk in an internal medicine department.
• To identify key factors influencing medical appointment no-shows.

Main Methods:

• Statistical analysis of institutional data to identify absenteeism predictors.
• Evaluation of seven machine learning models, including data processing and class imbalance techniques (SMOTE).
• Hyperparameter optimization and model selection using Bagging RandomForest, with SHAP for interpretability.

Main Results:

• The Bagging RandomForest model, optimized and balanced with SMOTE, achieved 84.80% predictive accuracy.
• Key predictors included previous absences, appointment timing, and diagnosed diseases.
• SHAP analysis provided interpretability, highlighting influential variables.

Conclusions:

• Machine learning effectively predicts internal medicine patient absenteeism risk.
• This approach enables proactive interventions, optimizing resource allocation and improving care quality.
• The methodology supports reduced operational inefficiencies and enhanced patient outcomes.