Classification of patients with chronic disease by activation level using machine learning methods.

Area of Science:

• Health Informatics
• Machine Learning in Healthcare
• Patient Engagement

Background:

• Patient Activation Measure (PAM) assesses patient engagement in chronic disease management.
• Low PAM levels correlate with poorer adherence, health outcomes, and higher costs.
• Identifying patients with low PAM is crucial for targeted interventions.

Purpose of the Study:

• To evaluate machine learning algorithms for classifying low Patient Activation Measure (PAM) levels.
• To identify key patient characteristics predicting low PAM in chronic conditions.
• To compare the performance of various machine learning models and feature sets.

Main Methods:

• Eight machine learning algorithms were tested: Logistic Regression, Lasso, Ridge, Random Forest, Gradient Boosted Trees, SVM, Decision Trees, and Neural Networks.
• Data from 431 adult patients with Diabetes Mellitus (DM) or Hypertension (HT) in Istanbul, Turkey.
• Analysis included various feature sets to determine the importance of patient information.

Main Results:

• Logistic Regression achieved the highest classification performance with an Area Under the Curve (AUC) of 0.72.
• Key predictors for low PAM included self-monitoring, smoking, exercise habits, education, and socio-economic status.
• Domain knowledge-driven feature selection outperformed automated methods.

Conclusions:

• Machine learning, particularly Logistic Regression, can effectively identify patients with low activation levels.
• Patient self-monitoring, lifestyle, education, and socio-economic status are significant indicators of activation.
• Integrating domain expertise with machine learning enhances predictive accuracy in healthcare.