Construction of Guideline-Based Decision Tree for Medication Recommendation

Area of Science:

• Medical Informatics
• Clinical Decision Support Systems
• Computational Medicine

Background:

• Clinical decision support systems (CDSS) are vital for evidence-based treatment recommendations.
• Accurate patient partitioning is crucial for personalized medicine but often limited by incomplete clinical guidelines.
• Existing methods struggle to reconcile detailed guidelines with diverse patient data.

Purpose of the Study:

• To develop an integrated approach for patient partitioning and treatment recommendation.
• To enhance the adherence to clinical guidelines and interpretability of CDSS.
• To address the limitations of purely data-driven or guideline-driven approaches.

Main Methods:

• A nested decision tree model was constructed by combining clinical guidelines with real-world patient medical data.
• The model was designed to partition patients into clinically similar groups and recommend divergent treatments.
• The approach was validated using a case study involving hyperthyroidism patients.

Main Results:

• The proposed model demonstrated improved guideline adherence compared to pure data-driven decision trees.
• Interpretability of treatment recommendations was significantly enhanced.
• Successful application in a real-world hyperthyroidism case study confirmed the model's efficacy.

Conclusions:

• Combining clinical guidelines with medical data offers a robust method for patient partitioning and treatment recommendation.
• The nested decision tree approach improves CDSS by balancing guideline adherence and data-driven insights.
• This integrated strategy holds promise for advancing personalized medicine and clinical decision support.